Devolution, de-evolution, or backward evolution (not to be confused with dysgenics) is the notion that species can revert to supposedly more primitive forms over time. The concept relates to the idea that evolution is purposeful (teleology) and thus progressive (orthogenesis), for example that feet might be better than hooves, or lungs than gills. However, evolutionary biology makes no such assumptions, and natural selection shapes adaptations
with no foreknowledge or foresights of any kind regarding the outcome.
It is possible for small changes (such as in the frequency of a single
gene) to be reversed by chance or selection, but this is no different
from the normal course of evolution and as such de-evolution is not
compatible with a proper understanding of evolution due to natural
selection.
In the 19th century, when belief in orthogenesis was widespread, zoologists such as Ray Lankester and Anton Dohrn and palaeontologists Alpheus Hyatt and Carl H. Eigenmann advocated the idea of devolution. The concept appears in Kurt Vonnegut's 1985 novel Galápagos, which portrays a society that has evolved backwards to have small brains.
Dollo's law of irreversibility, first stated in 1893 by the palaeontologist Louis Dollo, denies the possibility of devolution. The evolutionary biologist Richard Dawkins explains Dollo's law as being simply a statement about the improbability of evolution's following precisely the same path twice.
Context
Lamarck's theory of evolution involved a complexifying force that progressively drives animal body plans towards higher levels, creating a ladder of phyla, as well as an adaptive force that causes animals with a given body plan to adapt to circumstances. The idea of progress in such theories permits the opposite idea of decay, seen in devolution.
The idea of devolution is based on the presumption of orthogenesis, the view that evolution has a purposeful direction towards increasing complexity. Modern evolutionary theory, beginning with Darwin at least, poses no such presumption; further, the concept of evolutionary change is independent of either
any increase in complexity of organisms sharing a gene pool, or any
decrease, such as in vestigiality or in loss of genes. Earlier views that species are subject to "cultural decay", "drives to
perfection", or "devolution" are practically meaningless in terms of
current (neo-)Darwinian theory. Early scientific theories of transmutation of species such as Lamarckism
perceived species diversity as a result of a purposeful internal drive
or tendency to form improved adaptations to the environment. In
contrast, Darwinian evolution and its elaboration in the light of
subsequent advances in biological research, have shown that adaptation through natural selection
comes about when particular heritable attributes in a population happen
to give a better chance of successful reproduction in the reigning
environment than rival attributes do. By the same process less
advantageous attributes are less "successful"; they decrease in
frequency or are lost completely. Since Darwin's time it has been shown
how these changes in the frequencies of attributes occur according to
the mechanisms of genetics and the laws of inheritance originally investigated by Gregor Mendel. Combined with Darwin's original insights, genetic advances led to what has variously been called the modern evolutionary synthesis or the neo-Darwinism of the 20th century. In these terms evolutionary
adaptation may occur most obviously through the natural selection of
particular alleles. Such alleles may be long established, or they may be new mutations. Selection also might arise from more complex epigenetic or other chromosomal changes, but the fundamental requirement is that any adaptive effect must be heritable.
The concept of devolution on the other hand, requires that there
be a preferred hierarchy of structure and function, and that evolution
must mean "progress" to "more advanced" organisms. For example, it could
be said that "feet are better than hooves" or "lungs are better than gills",
so their development is "evolutionary" whereas change to an inferior or
"less advanced" structure would be called "devolution". In reality an
evolutionary biologist defines all heritable changes to relative
frequencies of the genes or indeed to epigenetic states in the gene pool as evolution. All gene pool changes that lead to increased fitness in terms of
appropriate aspects of reproduction are seen as (neo-)Darwinian
adaptation because, for the organisms possessing the changed structures,
each is a useful adaptation to their circumstances. For example, hooves
have advantages for running quickly on plains, which benefits horses,
and feet offer advantages in climbing trees, which some ancestors of
humans did.
The concept of devolution as regress from progress relates to the
ancient ideas that either life came into being through special creation
or that humans are the ultimate product or goal of evolution. The
latter belief is related to anthropocentrism,
the idea that human existence is the point of all universal existence.
Such thinking can lead on to the idea that species evolve because they
"need to" in order to adapt to environmental changes. Biologists refer
to this misconception as teleology, the idea of intrinsic finality
that things are "supposed" to be and behave a certain way, and
naturally tend to act that way to pursue their own good. From a
biological viewpoint, in contrast, if species evolve it is not a
reaction to necessity, but rather that the population contains
variations with traits that favour their natural selection.
This view is supported by the fossil record which demonstrates that
roughly ninety-nine percent of all species that ever lived are now
extinct.
People thinking in terms of devolution commonly assume that
progress is shown by increasing complexity, but biologists studying the evolution of complexity
find evidence of many examples of decreasing complexity in the record
of evolution. The lower jaw in fish, reptiles and mammals has seen a
decrease in complexity, if measured by the number of bones. Ancestors of
modern horses had several toes on each foot; modern horses have a
single hooved toe. Modern humans may be evolving towards never having wisdom teeth, and already have lost most of the tail found in many other mammals - not to mention other vestigial structures, such as the vermiform appendix or the nictitating membrane. In some cases, the level of organization of living creatures can also “shift” downwards (e.g., the loss of multicellularity in some groups of protists and fungi).
A more rational version of the concept of devolution, a version
that does not involve concepts of "primitive" or "advanced" organisms,
is based on the observation that if certain genetic changes in a
particular combination (sometimes in a particular sequence as well) are
precisely reversed, one should get precise reversal of the evolutionary
process, yielding an atavism or "throwback", whether more or less complex than the ancestors where the process began. At a trivial level, where just one or a few mutations are involved,
selection pressure in one direction can have one effect, which can be
reversed by new patterns of selection when conditions change. That could
be seen as reversed evolution, though the concept is not of much
interest because it does not differ in any functional or effective way
from any other adaptation to selection pressures.
The concept of degenerative evolution was used by scientists in the 19th century; at this time it was believed by most biologists that evolution had some kind of direction.
In 1857 the physician Bénédict Morel, influenced by Lamarckism, claimed that environmental factors such as taking drugs or alcohol would produce social degeneration in the offspring of those individuals, and would revert those offspring to a primitive state. Morel, a devout Catholic,
had believed that mankind had started in perfection, contrasting modern
humanity to the past. Morel claimed there had been "Morbid deviation
from an original type". His theory of devolution was later advocated by some biologists.
According to Roger Luckhurst:
Darwin soothed readers that evolution was progressive,
and directed towards human perfectibility. The next generation of
biologists were less confident or consoling. Using Darwin's theory, and
many rival biological accounts of development then in circulation,
scientists suspected that it was just as possible to devolve, to slip back down the evolutionary scale to prior states of development.
One of the first biologists to suggest devolution was Ray Lankester, he explored the possibility that evolution by natural selection may in some cases lead to devolution, an example he studied was the regressions in the life cycle of sea squirts. Lankester discussed the idea of devolution in his book Degeneration: A Chapter in Darwinism
(1880). He was a critic of progressive evolution, pointing out that
higher forms existed in the past which have since degenerated into
simpler forms. Lankester argued that "if it was possible to evolve, it
was also possible to devolve, and that complex organisms could devolve
into simpler forms or animals".
Anton Dohrn also developed a theory of degenerative evolution based on his studies of vertebrates. According to Dohrn many chordates are degenerated because of their environmental conditions. Dohrn claimed cyclostomes such as lampreys
are degenerate fish as there is no evidence their jawless state is an
ancestral feature but is the product of environmental adaptation due to parasitism. According to Dohrn if cyclostomes would devolve further then they would resemble something like an Amphioxus.
The historian of biology Peter J. Bowler has written that devolution was taken seriously by proponents of orthogenesis and others in the late 19th century who at this period of time firmly believed that there was a direction in evolution. Orthogenesis was the belief that evolution travels in internally directed trends and levels. The paleontologistAlpheus Hyatt discussed devolution in his work, using the concept of racial senility as the mechanism of devolution. Bowler defines racial senility as "an evolutionary retreat back to a state resembling that from which it began."
Hyatt who studied the fossils of invertebrates believed that up to a point ammonoids
developed by regular stages up until a specific level but would later
due to unfavourable conditions descend back to a previous level, this
according to Hyatt was a form of lamarckism as the degeneration was a
direct response to external factors. To Hyatt after the level of
degeneration the species would then become extinct, according to Hyatt
there was a "phase of youth, a phase of maturity, a phase of senility or
degeneration foreshadowing the extinction of a type". To Hyatt the devolution was predetermined by internal factors which
organisms can neither control or reverse. This idea of all evolutionary
branches eventually running out of energy and degenerating into extinction was a pessimistic view of evolution and was unpopular amongst many scientists of the time.
Carl H. Eigenmann an ichthyologist wrote Cave vertebrates of America: a study in degenerative evolution (1909) in which he concluded that caveevolution was essentially degenerative. The entomologistWilliam Morton Wheeler and the LamarckianErnest MacBride (1866–1940) also advocated degenerative evolution. According to Macbride invertebrates were actually degenerate vertebrates, his argument was based on the idea that "crawling on the seabed was inherently less stimulating than swimming in open waters."
Johann Friedrich Blumenbach and other monogenists such as Georges-Louis Leclerc, Comte de Buffon
were believers in the "Degeneration theory" of racial origins. The
theory claims that races can degenerate into "primitive" forms.
Blumenbach claimed that Adam and Eve were white
and that other races came about by degeneration from environmental
factors such as the sun and poor diet. Buffon believed that the
degeneration could be reversed if proper environmental control was taken
and that all contemporary forms of man could revert to the original Caucasian race.
Blumenbach claimed Negroid pigmentation arose because of the result of the heat of the tropical sun, cold wind caused the tawny colour of the Eskimos and the Chinese were fair skinned compared to the other Asian stocks because they kept mostly in towns protected from environmental factors.
I have allotted the first place to the Caucasian because this stock displays the most beautiful race of men.
According to Blumenbach the other races are supposed to have
degenerated from the Caucasian ideal stock. Blumenbach denied that his
"Degeneration theory" was racist; he also wrote three essays claiming non-white peoples are capable of excelling in arts and sciences in reaction against racialists of his time who believed they couldn't.
In literature and popular culture
Jonathan Swift's 1726 novel Gulliver's Travels contains a story about Yahoos, a kind of human-like creature turned into a savage, animal-like the state of society in which the Houyhnhnms—descendants of horses—are the dominant species.
H. G. Wells' 1895 novel, The Time Machine,
describes a future world where humanity has degenerated into two
distinct branches who have their roots in the class distinctions of
Wells' day. Both have sub-human intelligence and other putative dysgenic
traits.
Similarly, Helena Blavatsky, founder of Theosophy,
believed, contrary to standard evolutionary theory, that apes had
devolved from humans rather than the opposite, through affected people
"putting themselves on the animal level".
Cyril M. Kornbluth's 1951 short story "The Marching Morons" is an example of dysgenic pressure
in fiction, describing a man who accidentally ends up in the distant
future and discovers that dysgenics has resulted in mass stupidity.
Similarly, Mike Judge's 2006 film Idiocracy has the same premise, with the main character the subject of a military hibernation
experiment that goes awry, taking him 500 years into the future. While
in "The Marching Morons", civilization is kept afloat by a small group
of dedicated geniuses, in Idiocracy, voluntary childlessness among high-IQ couples leaves only automated systems to fill that role.
T. J. Bass's novels Half Past Human and The Godwhale describe humanity becoming cooperative and "low-maintenance" to the detriment of all other traits.
The American new wave band Devo
derived both their name and overarching philosophy from the concept of
"de-evolution" and used social satire and humor to espouse the idea that
humanity had actually regressed over time. According to music critic Steve Huey, the band "adapted the theory to
fit their view of American society as a rigid, dichotomized instrument
of repression ensuring that its members behaved like clones, marching
through life with mechanical, assembly-line precision and no tolerance
for ambiguity."
The 1998 song "Flagpole Sitta" by Harvey Danger
finds lighthearted humor in dysgenics with the lines "Been around the
world and found/That only stupid people are breeding/The cretins cloning
and feeding/And I don't even own a tv".
LEGO's 2009 Bionicle
sets include Glatorian and Agori. One of the six tribes includes The
Sand Tribe, which the Glatorian and Agori of that tribe are turned into scorpion-like beasts—the Vorox and the Zesk—by their creators, The Great Beings; whom are also of the same species as Glatorian and Agori.
A scenario is a set of related concepts pertinent to the origin of life (abiogenesis), such as the iron-sulfur world. Many alternative abiogenesis scenarios
have been proposed by scientists in a variety of fields from the 1950s
onwards in an attempt to explain how the complex mechanisms of life
could have come into existence. These include hypothesized ancient
environments that might have been favourable for the origin of life, and
possible biochemical mechanisms.
Environments
Many environments have been proposed for the origin of life.
Fluctuating salinity: dilute and dry-down
Harold Blum noted in 1957 that if proto-nucleic acid chains spontaneously form duplex structures, then there is no way to dissociate them.
The Oparin-Haldane hypothesis
addresses the formation, but not the dissociation, of nucleic acid
polymers and duplexes. However, nucleic acids are unusual because, in
the absence of counterions (low salt) to neutralize the high charges on
opposing phosphate groups, the nucleic acid duplex dissociates into
single chains. Early tides, driven by a close moon, could have generated rapid cycles
of dilution (high tide, low salt) and concentration (dry-down at low
tide, high salt) that exclusively promoted the replication of nucleic
acids through a process dubbed tidal chain reaction (TCR). This theory has been criticized on the grounds that early tides may not have been so rapid, although regression from current values requires an Earth–Moon
juxtaposition at around two Ga, for which there is no evidence, and
early tides may have been approximately every seven hours. Another critique is that only 2–3% of the Earth's crust may have been
exposed above the sea until late in terrestrial evolution.
The tidal chain reaction theory has mechanistic advantages over
thermal association/dissociation at deep-sea vents because it requires
that chain assembly (template-driven polymerization) takes place during
the dry-down phase, when precursors are most concentrated, whereas
thermal cycling needs polymerization to take place during the cold
phase, when the rate of chain assembly is lowest and precursors are
likely to be more dilute.
Hot freshwater lakes
Jack W. Szostak
suggested that geothermal activity provides greater opportunities for
the origination of life in open lakes where there is a buildup of
minerals. In 2010, based on spectral analysis of sea and hot mineral
water, Ignat Ignatov and Oleg Mosin demonstrated that life may have
predominantly originated in hot mineral water. Hot mineral water that
contains hydrogen carbonate and calcium ions has the most optimal range. This case is similar to the origin of life in hydrothermal vents, but
with hydrogen carbonate and calcium ions in hot water. The main studies
were conducted in Rupite, Bulgaria, where a novel thermophylic bacterium
Anoxybacillus rupiences sp. Nov. and cyanobacteria were identified. At a pH of 9–11, the reactions can take place in seawater. According to Melvin Calvin,
certain reactions of condensation-dehydration of amino acids and
nucleotides in individual blocks of peptides and nucleic acids can take
place in the primary hydrosphere with pH 9–11 at a later evolutionary
stage.Some of these compounds like hydrocyanic acid (HCN) have been proven in the experiments of Miller. This is the environment in which the stromatolites have been created. David Ward described the formation of stromatolites in hot mineral water at the Yellowstone National Park. In 2011, Tadashi Sugawara created a protocell in hot water.
Geothermal springs
Bruce Damer and David Deamer argue that cell membranes cannot be formed in salty seawater, and must therefore have originated in freshwater environments like pools replenished by a combination of geothermal springs
and rainfall. Before the continents formed, the only dry land on Earth
would be volcanic islands, where rainwater would form ponds where lipids
could form the first stages towards cell membranes. During multiple
wet-dry cycles, biopolymers would be synthesized and are encapsulated in
vesicles after condensation. Zinc sulfide and manganese sulfide in
these ponds would have catalyzed organic compounds by abiotic
photosynthesis. Experimental research at geothermal springs successfully synthesized
polymers and were encapsulated in vesicles after exposure to UV light
and multiple wet-dry cycles. At temperatures of 60 to 80 °C at geothermal fields, biochemical reactions can occur. These predecessors of true cells are assumed to have behaved more like a superorganism
rather than individual structures, where the porous membranes would
house molecules which would leak out and enter other protocells. Only
when true cells had evolved would they gradually adapt to saltier
environments and enter the ocean.
6 of the 11 biochemical reactions of the rTCA cycle can occur in
hot metal-rich acidic water which suggests metabolic reactions might
have originated in this environment, this is consistent with the
enhanced stability of RNA phosphodiester, aminoacyl-tRNA bonds, and
peptides in acidic conditions. Cycling between supercritical and subcritical CO2
at tectonic fault zones might have led to peptides integrating with and
stabilizing lipid membranes. This is suggested to have driven membrane
protein evolution, as it shown that a selected peptide
(H-Lys-Ser-Pro-Phe-Pro-Phe-Ala-Ala-OH) causes the increase of membrane
permeability to water. David Deamer and Bruce Damer states that the prebiotic chemistry does
not require ultraviolet irradiation as the chemistry could also have
occurred under shaded areas that protected biomolecules from photolysis.
Deep sea alkaline vents
Nick Lane
believes that no known life forms could have utilized zinc-sulfide
based photosynthesis, lightning, volcanic pyrite synthesis, or UV
radiation as a source of energy. Rather, he instead suggests that deep
sea alkaline vents is more likely to have been a source energy for early
cellular life. Serpentinization at alkaline hydrothermal vents produce methane and ammonia. Mineral particles that have similar properties to enzymes at deep sea
vents would catalyze organic compounds out of dissolved CO2 within seawater. Porous rock might have promoted condensation reactions of biopolymers
and act as a compartment of membranous structures, however it is unknown
about how it could promote coding and metabolism. Acetyl phosphate, which is readily synthesized from thioacetate, can
promote aggregation of adenosine monophosphate of up to 7 monomers which
is considered energetically favored in water due to interactions
between nucleobases. Acetyl phosphate can stabilize aggregation of
nucleotides in the presence of Na+ and could possibly promote polymerization at mineral surfaces or lower water activity. An external proton gradient within a membrane would have been maintained between the acidic ocean and alkaline seawater. The descendants of the last universal common ancestor, bacteria and archaea, were probably methanogens and acetogens. The earliest microfossils, dated to be 4.28 to 3.77 Ga, were found at
hydrothermal vent precipitates. These microfossils suggest that early
cellular life began at deep sea hydrothermal vents. Exergonic reactions at these environments could have provided free
energy that promoted chemical reactions conducive to prebiotic
biomolecules.
Nonenzymatic reactions of glycolysis and the pentose phosphate
pathway can occur in the presence of ferrous iron at 70 °C, the
reactions produce erythrose 4-phosphate, an amino acid precursor and ribose 5-phosphate, a nucleotide precursor. Pyrimidines are shown to be synthesized from the reaction between
aspartate and carbamoyl phosphate at 60 °C and in the presence of
metals, it is suggested that purines could be synthesized from the catalysis of metals. Adenosine monophosphate are also shown to be synthesized from adenine,
monopotassium phosphate or pyrophosphate, and ribose at silica at 70 °C. Reductive amination and transamination reactions catalyzed by alkaline
hydrothermal vent mineral and metal ions produce amino acids. Long chain fatty acids can be derived from formic acid or oxalic acid during Fischer-Tropsch-type synthesis. Carbohydrates containing an isoprene skeleton can be synthesized from
the formose reaction. Isoprenoids incorporated into fatty acid vesicles
can stabilize the vesicles, which are suggested to have driven the
divergence of bacterial and archaeal lipids.
A scenario
Lane has proposed a possible scenario for the origin of life that
integrates much of the available evidence from biochemistry, geology,
phylogeny, and experimentation:
Iron-sulfur minerals like greigite catalyse the reduction of carbon dioxide in hydrothermal vents to make Krebs cycle intermediates.
Protocells in contact with a thin rock barrier in a hydrothermal vent get a free supply of energy from the pH gradient.
Protocells in a hydrothermal vent can grow by adding fatty acids to their membrane, other organics to their cytoplasm.
Nucleotides in a protocell in a hydrothermal
vent can polymerise into random strings of RNA. Any that have even
slight catalytic activity will favour the growth and replication of
their protocells, a start to natural selection.
A protocell away from a hydrothermal vent must create its own proton-motive force, such as by splitting hydrogen sulfide.
Ferredoxin
catalyses the splitting of hydrogen sulfide, its earliest repeating
amino acid sequence perhaps coded for by an incomplete genetic code.
Anoxygenic photosynthesis, using hydrogen sulfide, ended the need for scarce hydrogen.
Early heterotrophs used Krebs cycle respiration; then oxygenic photosynthesis gave full independence of volcanic energy.
Volcanic ash in the ocean
Geoffrey W. Hoffmann
has argued that a complex nucleation event as the origin of life
involving both polypeptides and nucleic acid is compatible with the time
and space available in the primary oceans of Earth. Hoffmann suggests that volcanic ash may provide the many random shapes
needed in the postulated complex nucleation event. This aspect of the
theory can be tested experimentally.
Gold's deep-hot biosphere
In the 1970s, Thomas Gold
proposed the theory that life first developed not on the surface of the
Earth, but several kilometers below the surface. It is claimed that the
discovery of microbial life below the surface of another body in our
Solar System would lend significant credence to this theory.
Radioactive beach hypothesis
Zachary Adam claims that tidal processes that occurred during a time
when the Moon was much closer may have concentrated grains of uranium and other radioactive elements at the high-water mark on primordial beaches, where they may have been responsible for generating life's building blocks. According to computer models, a deposit of such radioactive materials could show the same self-sustaining nuclear reaction as that found in the Oklo uranium ore seam in Gabon.
Such radioactive beach sand might have provided sufficient energy to
generate organic molecules, such as amino acids and sugars from acetonitrile in water. Radioactive monazite
material also has released soluble phosphate into the regions between
sand-grains, making it biologically "accessible." Thus amino acids,
sugars, and soluble phosphates might have been produced simultaneously,
according to Adam. Radioactive actinides, left behind in some concentration by the reaction, might have formed part of organometallic complexes. These complexes could have been important early catalysts to living processes.
John Parnell has suggested that such a process could provide part
of the "crucible of life" in the early stages of any early wet rocky
planet, so long as the planet is large enough to have generated a system
of plate tectonics which brings radioactive minerals to the surface. As
the early Earth is thought to have had many smaller plates, it might
have provided a suitable environment for such processes.
The hypercycle
In the early 1970s, Manfred Eigen and Peter Schuster examined the transient stages between the molecular chaos and a self-replicating hypercycle in a prebiotic soup. In a hypercycle, the information storing system (possibly RNA) produces an enzyme,
which catalyzes the formation of another information system, in
sequence until the product of the last aids in the formation of the
first information system. Mathematically treated, hypercycles could
create quasispecies, which through natural selection entered into a form of Darwinian evolution. A boost to hypercycle theory was the discovery of ribozymes
capable of catalyzing their own chemical reactions. The hypercycle
theory requires the existence of complex biochemicals, such as
nucleotides, which do not form under the conditions proposed by the
Miller–Urey experiment.
In the 1980s, Wächtershäuser and Karl Popper postulated the iron–sulfur world hypothesis
for the evolution of pre-biotic chemical pathways. It traces today's
biochemistry to primordial reactions which synthesize organic building
blocks from gases. Wächtershäuser systems have a built-in source of energy: iron sulfides
such as pyrite. The energy released by oxidising these metal sulfides
can support synthesis of organic molecules. Such systems may have
evolved into autocatalytic sets constituting self-replicating,
metabolically active entities predating modern life forms. Experiments with sulfides in an aqueous environment at 100 °C produced a small yield of dipeptides (0.4% to 12.4%) and a smaller yield of tripeptides (0.10%). However, under the same conditions, dipeptides were quickly broken down.
Several models postulate a primitive metabolism, allowing RNA replication to emerge later. The centrality of the Krebs cycle
(citric acid cycle) to energy production in aerobic organisms, and in
drawing in carbon dioxide and hydrogen ions in biosynthesis of complex
organic chemicals, suggests that it was one of the first parts of the
metabolism to evolve. Concordantly, geochemists Szostak and Kate Adamala demonstrated that non-enzymatic RNA replication in primitive protocells is only possible in the presence of weak cation chelators like citric acid. This provides further evidence for the central role of citric acid in primordial metabolism. Russell has proposed that "the purpose of life is to hydrogenate carbon
dioxide" (as part of a "metabolism-first", rather than a
"genetics-first", scenario). The physicistJeremy England has argued from general thermodynamic considerations that life was inevitable. An early version of this idea was Oparin's 1924 proposal for
self-replicating vesicles. In the 1980s and 1990s came Wächtershäuser's
iron–sulfur world theory and Christian de Duve's thioester models. More abstract and theoretical arguments for metabolism without genes include Freeman Dyson's mathematical model and Stuart Kauffman's
collectively autocatalytic sets in the 1980s. Kauffman's work has been
criticized for ignoring the role of energy in driving biochemical
reactions in cells.
The active site of the acetyl-CoA synthase enzyme, part of the acetyl-CoA pathway, contains nickel-iron-sulfur clusters.
A multistep biochemical pathway like the Krebs cycle did not just
self-organize on the surface of a mineral; it must have been preceded by
simpler pathways. The Wood–Ljungdahl pathway is compatible with self-organization on a metal sulfide surface. Its key enzyme unit, carbon monoxide dehydrogenase/acetyl-CoA synthase, contains mixed nickel-iron-sulfur clusters in its reaction centers and catalyzes the formation of acetyl-CoA. However, prebiotic thiolated
and thioester compounds are thermodynamically and kinetically unlikely
to accumulate in the presumed prebiotic conditions of hydrothermal
vents. One possibility is that cysteine and homocysteine may have reacted with nitriles from the Strecker reaction, forming catalytic thiol-rich polypeptides.
It has been suggested that the iron-sulfur world hypothesis and
RNA world hypothesis are not mutually exclusive as modern cellular
processes do involve both metabolites and genetic molecules.
Zinc world
Armen Mulkidjanian's zinc world (Zn-world) hypothesis extends Wächtershäuser's pyrite hypothesis. The Zn-world theory proposes that hydrothermal fluids rich in H2S
interacting with cold primordial ocean (or Darwin's "warm little pond")
water precipitated metal sulfide particles. Oceanic hydrothermal
systems have a zonal structure reflected in ancient volcanogenic massive sulfide ore deposits. They reach many kilometers in diameter and date back to the Archean. Most abundant are pyrite (FeS2), chalcopyrite (CuFeS2), and sphalerite (ZnS), with additions of galena (PbS) and alabandite
(MnS). ZnS and MnS have a unique ability to store radiation energy,
e.g. from ultraviolet light. When replicating molecules were
originating, the primordial atmospheric pressure was high enough
(>100 bar) to precipitate near the Earth's surface, and ultraviolet
irradiation was 10 to 100 times more intense than now; hence the
photosynthetic properties mediated by ZnS provided the right energy
conditions for the synthesis of informational and metabolic molecules
and the selection of photostable nucleobases.
The Zn-world theory has been filled out with evidence for the
ionic constitution of the interior of the first protocells. In 1926, the
Canadian biochemist Archibald Macallum noted the resemblance of body fluids such as blood and lymph to seawater; however, the inorganic composition of all cells
differ from that of modern seawater, which led Mulkidjanian and
colleagues to reconstruct the "hatcheries" of the first cells combining
geochemical analysis with phylogenomic
scrutiny of the inorganic ion requirements of modern cells. The authors
conclude that ubiquitous, and by inference primordial, proteins and
functional systems show affinity to and functional requirement for K+, Zn2+, Mn2+, and [PO 4]3− .
Geochemical reconstruction shows that this ionic composition could not
have existed in the ocean but is compatible with inland geothermal
systems. In the oxygen-depleted, CO2-dominated primordial atmosphere, the chemistry of water condensates near geothermal fields
would resemble the internal milieu of modern cells. Therefore,
precellular evolution may have taken place in shallow "Darwin ponds"
lined with porous silicate minerals mixed with metal sulfides and enriched in K+, Zn2+, and phosphorus compounds.
Clay
The clay hypothesis was proposed by Graham Cairns-Smith in 1985. It postulates that complex organic molecules arose gradually on
pre-existing, non-organic replication surfaces of silicate crystals in
contact with an aqueous solution. The clay mineralmontmorillonite has been shown to catalyze the polymerization of RNA in aqueous solution from nucleotide monomers, and the formation of membranes from lipids. In 1998, Hyman Hartman proposed that "the first organisms were
self-replicating iron-rich clays which fixed carbon dioxide into oxalic acid and other dicarboxylic acids. This system of replicating clays and their metabolic phenotype then evolved into the sulfide rich region of the hot spring acquiring the ability to fix nitrogen. Finally phosphate was incorporated into the evolving system which allowed the synthesis of nucleotides and phospholipids."
Biochemistry
Different forms of life with variable origin processes may have appeared quasi-simultaneously in the early Earth. The other forms may be extinct, having left distinctive fossils through their different biochemistry.
Metabolism-like reactions could have occurred naturally in early
oceans, before the first organisms evolved. Some of these reactions can
produce RNA, and others resemble two essential reaction cascades of
metabolism: glycolysis and the pentose phosphate pathway, that provide essential precursors for nucleic acids, amino acids and lipids.
In trying to uncover the intermediate stages of abiogenesis mentioned by Bernal, Sidney Fox in the 1950s and 1960s studied the spontaneous formation of peptide
structures under plausibly early Earth conditions. In one of his
experiments, he allowed amino acids to dry out as if puddled in a warm,
dry spot in prebiotic conditions: In an experiment to set suitable
conditions for life to form, Fox collected volcanic material from a cinder cone in Hawaii.
He discovered that the temperature was over 100 °C just 4 inches
(100 mm) beneath the surface of the cinder cone, and suggested that this
might have been the environment in which life was created—molecules
could have formed and then been washed through the loose volcanic ash
into the sea. He placed lumps of lava over amino acids derived from
methane, ammonia and water, sterilized all materials, and baked the lava
over the amino acids for a few hours in a glass oven. A brown, sticky
substance formed over the surface, and when the lava was drenched in
sterilized water, a thick, brown liquid leached out. He found that, as
they dried, the amino acids formed long, often cross-linked,
thread-like, submicroscopic polypeptides.
Protein amyloid
An origin-of-life theory based on self-replicating beta-sheet structures has been put forward by Maury in 2009. The theory suggest that self-replicating and self-assembling catalytic amyloids were the first informational polymers in a primitive pre-RNA world. The main arguments for the amyloid hypothesis
is based on the structural stability, autocatalytic and catalytic
properties, and evolvability of beta-sheet based informational systems.
Such systems are also error correcting and chiroselective.
First protein that condenses substrates during thermal cycling: thermosynthesis
Convection
cells in fluid placed in a gravity field are selforganizing and enable
thermal cycling of the suspended contents in the fluid such as
protocells containing protoenzymes that work on thermal cycling.
The thermosynthesis hypothesis considers chemiosmosis more basal than
fermentation: the ATP synthase enzyme, which sustains chemiosmosis, is
the currently extant enzyme most closely related to the first metabolic
process. The thermosynthesis hypothesis does not even invoke a pathway: ATP
synthase's binding change mechanism resembles a physical adsorption
process that yields free energy. The result would be convection which
would bring a continual supply of reactants to the protoenzyme. The described first protein may be simple in the sense that it requires
only a short sequence of conserved amino acid residues, a sequent
sufficient for the appropriate catalytic cleft.
Pre-RNA world: The ribose issue and its bypass
A different type of nucleic acid, such as peptide nucleic acid, threose nucleic acid or glycol nucleic acid, could have been the first to emerge as a self-reproducing molecule, later replaced by RNA.Larralde et al., say that "the generally accepted prebiotic synthesis of ribose, the formose reaction, yields numerous sugars without any selectivity". They conclude that "the backbone of the first genetic material could
not have contained ribose or other sugars because of their instability",
meaning that the ester linkage of ribose and phosphoric acid in RNA is
prone to hydrolysis.
Pyrimidine ribonucleosides and nucleotides have been synthesized
by reactions which by-pass the free sugars, and are assembled stepwise
using nitrogenous or oxygenous chemistries. Sutherland has demonstrated
high-yielding routes to cytidine and uridine ribonucleotides from small 2
and 3 carbon fragments such as glycolaldehyde, glyceraldehyde or glyceraldehyde-3-phosphate, cyanamide and cyanoacetylene. A step in this sequence allows the isolation of enantiopure ribose aminooxazoline if the enantiomeric excess of glyceraldehyde is 60% or greater. This can be viewed as a prebiotic purification step. Ribose
aminooxazoline can then react with cyanoacetylene to give alpha cytidine
ribonucleotide. Photoanomerization with UV light allows for inversion
about the 1' anomeric centre to give the correct beta stereochemistry. In 2009 they showed that the same simple building blocks allow access,
via phosphate controlled nucleobase elaboration, to 2',3'-cyclic
pyrimidine nucleotides directly, which can polymerize into RNA. Similar photo-sanitization can create pyrimidine-2',3'-cyclic phosphates.
Autocatalysis
Autocatalysts
are substances that catalyze the production of themselves and therefore
are "molecular replicators." The simplest self-replicating chemical
systems are autocatalytic, and typically contain three components: a
product molecule and two precursor molecules. The product molecule joins
the precursor molecules, which in turn produce more product molecules
from more precursor molecules. The product molecule catalyzes the
reaction by providing a complementary template that binds to the
precursors, thus bringing them together. Such systems have been
demonstrated both in biological macromolecules and in small organic molecules.
It has been proposed that life initially arose as autocatalytic chemical networks. Julius Rebek and colleagues combined amino adenosine and pentafluorophenyl esters
with the autocatalyst amino adenosine triacid ester (AATE). One product
was a variant of AATE which catalyzed its own synthesis. This
demonstrated that autocatalysts could compete within a population of
entities with heredity, a rudimentary form of natural selection.
Synthesis based on hydrogen cyanide
A research project completed in 2015 by John Sutherland and others found that a network of reactions beginning with hydrogen cyanide and hydrogen sulfide,
in streams of water irradiated by UV light, could produce the chemical
components of proteins and lipids, as well as those of RNA, while not producing a wide range of other compounds. The researchers used the term "cyanosulfidic" to describe this network of reactions.
Simulated chemical pathways
In 2020, chemists described possible chemical pathways from nonliving prebiotic chemicals to complex biochemicals that could give rise to living organisms, based on a new computer program named AllChemy.
Viral origin
Evidence for a "virus first" hypothesis, which may support theories of the RNA world, was suggested in 2015. One of the difficulties for the study of the origins of viruses is
their high rate of mutation; this is particularly the case in RNA
retroviruses like HIV. A 2015 study compared protein fold
structures across different branches of the tree of life, where
researchers can reconstruct the evolutionary histories of the folds and
of the organisms whose genomes
code for those folds. They argue that protein folds are better markers
of ancient events as their three-dimensional structures can be
maintained even as the sequences that code for those begin to change. Thus, the viral protein repertoire retain traces of ancient evolutionary history that can be recovered using advanced bioinformatics approaches. Those researchers think that "the prolonged pressure of genome and particle size reduction eventually reduced virocells
into modern viruses (identified by the complete loss of cellular
makeup), meanwhile other coexisting cellular lineages diversified into
modern cells." The data suggest that viruses originated from ancient cells that
co-existed with the ancestors of modern cells. These ancient cells
likely contained segmented RNA genomes.
A computational model (2015) has shown that virus capsids may have originated in the RNA world and served as a means of horizontal transfer
between replicator communities. These communities could not survive if
the number of gene parasites increased, with certain genes being
responsible for the formation of these structures and those that favored
the survival of self-replicating communities. The displacement of these ancestral genes between cellular organisms
could favor the appearance of new viruses during evolution. Viruses retain a replication module inherited from the prebiotic stage since it is absent in cells. So this is evidence that viruses could originate from the RNA world and
could also emerge several times in evolution through genetic escape in
cells.
Encapsulation without a membrane
Polyester droplets
Tony Jia and Kuhan Chandru have proposed spontaneously-forming
membraneless polyester droplets in early cellularization before the
innovation of lipid vesicles. Protein function within and RNA function
in the presence of certain polyester droplets was shown to be preserved
within the droplets. The droplets have scaffolding ability, by allowing
lipids to assemble around them; this may have prevented leakage of
genetic materials.
Proteinoid microspheres
Fox observed in the 1960s that proteinoids could form cell-like structures named "proteinoid microspheres". The amino acids had combined to form proteinoids, which formed small
globules. These were not cells; their clumps and chains were reminiscent
of cyanobacteria, but they contained no functional nucleic acids or other encoded information. Colin Pittendrigh
stated in 1967 that "laboratories will be creating a living cell within
ten years", a remark that reflected the typical contemporary naivety
about the complexity of cell structures.
Jeewanu protocell
A further protocell model is the Jeewanu.
First synthesized in 1963 from simple minerals and basic organics while
exposed to sunlight, it is reported to have some metabolic
capabilities, the presence of a semipermeable membrane, amino acids, phospholipids, carbohydrates and RNA-like molecules. However, the nature and properties of the Jeewanu remains to be
clarified. Electrostatic interactions induced by short, positively
charged, hydrophobic peptides containing 7 amino acids in length or
fewer can attach RNA to a vesicle membrane, the basic cell membrane.
RNA-DNA world
In 2020, coevolution of a RNA-DNA mixture based on diamidophosphate was proposed. The mixture of RNA-DNA sequences, called chimeras, have weak affinity and form weaker duplex structures. This is advantageous in an abiotic scenario and these chimeras have
been shown to replicate RNA and DNA – overcoming the "template-product"
inhibition problem, where a pure RNA or pure DNA strand is unable to
replicate non-enzymatically because it binds too strongly to its
partners. This could lead to an abiotic cross-catalytic amplification of RNA and DNA.
A continuous chemical reaction network in water and under high-energy radiation can generate precursors for early RNA.
In 2022, evolution experiments of self-replicating RNA showed how RNA may have evolved to diverse complex molecules in RNA world
conditions. The RNA evolved to a "replicator network comprising five
types of RNAs with diverse interactions" such as cooperation for
replication of other members (multiple coexisting host and parasite
lineages).
Artificial intelligence was founded as an academic discipline in 1956, and the field went through multiple cycles of optimism throughout its history, followed by periods of disappointment and loss of funding, known as AI winters. Funding and interest vastly increased after 2012 when graphics processing units started being used to accelerate neural networks, and deep learning outperformed previous AI techniques. This growth accelerated further after 2017 with the transformer architecture. In the 2020s, an ongoing period of rapid progress in advanced generative AI became known as the AI boom. Generative AI's ability to create and modify content has led to several unintended consequences and harms. Ethical concerns have been raised about AI's long-term effects and potential existential risks, prompting discussions about regulatory policies to ensure the safety and benefits of the technology.
Goals
The general problem of simulating (or creating) intelligence has been
broken into subproblems. These consist of particular traits or
capabilities that researchers expect an intelligent system to display.
The traits described below have received the most attention and cover
the scope of AI research.
Reasoning and problem-solving
Early researchers developed algorithms that imitated step-by-step
reasoning that humans use when they solve puzzles or make logical deductions. By the late 1980s and 1990s, methods were developed for dealing with uncertain or incomplete information, employing concepts from probability and economics.
Many of these algorithms
are insufficient for solving large reasoning problems because they
experience a "combinatorial explosion": They become exponentially slower
as the problems grow. Even humans rarely use the step-by-step deduction that early AI
research could model. They solve most of their problems using fast,
intuitive judgments. Accurate and efficient reasoning is an unsolved problem.
Knowledge representation
An ontology represents knowledge as a set of concepts within a domain and the relationships between those concepts.
Knowledge representation and knowledge engineering allow AI programs to answer questions intelligently and make deductions
about real-world facts. Formal knowledge representations are used in
content-based indexing and retrieval, scene interpretation, clinical decision support, knowledge discovery (mining "interesting" and actionable inferences from large databases), and other areas.
A knowledge base is a body of knowledge represented in a form that can be used by a program. An ontology is the set of objects, relations, concepts, and properties used by a particular domain of knowledge. Knowledge bases need to represent things such as objects, properties, categories, and relations between objects; situations, events, states, and time; causes and effects; knowledge about knowledge (what we know about what other people know); default reasoning (things that humans assume are true until they are told differently and will remain true even when other facts are changing); and many other aspects and domains of knowledge.
Among the most difficult problems in knowledge representation are the breadth of commonsense knowledge (the set of atomic facts that the average person knows is enormous); and the sub-symbolic form of most commonsense knowledge (much of what
people know is not represented as "facts" or "statements" that they
could express verbally). There is also the difficulty of knowledge acquisition, the problem of obtaining knowledge for AI applications.
Planning and decision-making
An "agent" is anything that perceives and takes actions in the world. A rational agent has goals or preferences and takes actions to make them happen. In automated planning, the agent has a specific goal. In automated decision-making,
the agent has preferences—there are some situations it would prefer to
be in, and some situations it is trying to avoid. The decision-making
agent assigns a number to each situation (called the "utility") that measures how much the agent prefers it. For each possible action, it can calculate the "expected utility":
the utility of all possible outcomes of the action, weighted by the
probability that the outcome will occur. It can then choose the action
with the maximum expected utility.
In classical planning, the agent knows exactly what the effect of any action will be. In most real-world problems, however, the agent may not be certain
about the situation they are in (it is "unknown" or "unobservable") and
it may not know for certain what will happen after each possible action
(it is not "deterministic"). It must choose an action by making a
probabilistic guess and then reassess the situation to see if the action
worked.
In some problems, the agent's preferences may be uncertain,
especially if there are other agents or humans involved. These can be
learned (e.g., with inverse reinforcement learning), or the agent can seek information to improve its preferences. Information value theory can be used to weigh the value of exploratory or experimental actions. The space of possible future actions and situations is typically intractably large, so the agents must take actions and evaluate situations while being uncertain of what the outcome will be.
A Markov decision process has a transition model
that describes the probability that a particular action will change the
state in a particular way and a reward function that supplies the
utility of each state and the cost of each action. A policy associates a decision with each possible state. The policy could be calculated (e.g., by iteration), be heuristic, or it can be learned.
Game theory
describes the rational behavior of multiple interacting agents and is
used in AI programs that make decisions that involve other agents.
Learning
Machine learning is the study of programs that can improve their performance on a given task automatically. It has been a part of AI from the beginning.
In supervised learning, the training data is labelled with the expected answers, while in unsupervised learning, the model identifies patterns or structures in unlabelled data.
There are several kinds of machine learning. Unsupervised learning analyzes a stream of data and finds patterns and makes predictions without any other guidance. Supervised learning requires labeling the training data with the expected answers, and comes in two main varieties: classification (where the program must learn to predict what category the input belongs in) and regression (where the program must deduce a numeric function based on numeric input).
In reinforcement learning,
the agent is rewarded for good responses and punished for bad ones. The
agent learns to choose responses that are classified as "good". Transfer learning is when the knowledge gained from one problem is applied to a new problem. Deep learning is a type of machine learning that runs inputs through biologically inspired artificial neural networks for all of these types of learning.
Modern deep learning techniques for NLP include word embedding (representing words, typically as vectors encoding their meaning), transformers (a deep learning architecture using an attention mechanism), and others. In 2019, generative pre-trained transformer (or "GPT") language models began to generate coherent text, and by 2023, these models were able to get human-level scores on the bar exam, SAT test, GRE test, and many other real-world applications.
Perception
Machine perception is the ability to use input from sensors (such as cameras, microphones, wireless signals, active lidar, sonar, radar, and tactile sensors) to deduce aspects of the world. Computer vision is the ability to analyze visual input.
Kismet, a robot head which was made in the 1990s; it is a machine that can recognize and simulate emotions.
Affective computing is a field that comprises systems that recognize, interpret, process, or simulate human feeling, emotion, and mood. For example, some virtual assistants
are programmed to speak conversationally or even to banter humorously;
it makes them appear more sensitive to the emotional dynamics of human
interaction, or to otherwise facilitate human–computer interaction.
However, this tends to give naïve users an unrealistic conception of the intelligence of existing computer agents. Moderate successes related to affective computing include textual sentiment analysis and, more recently, multimodal sentiment analysis, wherein AI classifies the effects displayed by a videotaped subject.
AI research uses a wide variety of techniques to accomplish the goals above.
Search and optimization
AI can solve many problems by intelligently searching through many possible solutions. There are two very different kinds of search used in AI: state space search and local search.
State space search
State space search searches through a tree of possible states to try to find a goal state. For example, planning algorithms search through trees of goals and subgoals, attempting to find a path to a target goal, a process called means-ends analysis.
Simple exhaustive searches are rarely sufficient for most real-world problems: the search space (the number of places to search) quickly grows to astronomical numbers. The result is a search that is too slow or never completes. "Heuristics" or "rules of thumb" can help prioritize choices that are more likely to reach a goal.
Adversarial search is used for game-playing programs, such as chess or Go. It searches through a tree of possible moves and countermoves, looking for a winning position.
Local search
Illustration of gradient descent for 3 different starting points; two parameters (represented by the plan coordinates) are adjusted in order to minimize the loss function (the height)
Gradient descent is a type of local search that optimizes a set of numerical parameters by incrementally adjusting them to minimize a loss function. Variants of gradient descent are commonly used to train neural networks, through the backpropagation algorithm.
Another type of local search is evolutionary computation, which aims to iteratively improve a set of candidate solutions by "mutating" and "recombining" them, selecting only the fittest to survive each generation.
Deductive reasoning in logic is the process of proving a new statement (conclusion) from other statements that are given and assumed to be true (the premises). Proofs can be structured as proof trees, in which nodes are labelled by sentences, and children nodes are connected to parent nodes by inference rules.
Given a problem and a set of premises, problem-solving reduces to
searching for a proof tree whose root node is labelled by a solution of
the problem and whose leaf nodes are labelled by premises or axioms. In the case of Horn clauses, problem-solving search can be performed by reasoning forwards from the premises or backwards from the problem. In the more general case of the clausal form of first-order logic, resolution
is a single, axiom-free rule of inference, in which a problem is solved
by proving a contradiction from premises that include the negation of
the problem to be solved.
Inference in both Horn clause logic and first-order logic is undecidable, and therefore intractable. However, backward reasoning with Horn clauses, which underpins computation in the logic programming language Prolog, is Turing complete. Moreover, its efficiency is competitive with computation in other symbolic programming languages.
Fuzzy logic assigns a "degree of truth" between 0 and 1. It can therefore handle propositions that are vague and partially true.
Many problems in AI (including reasoning, planning, learning,
perception, and robotics) require the agent to operate with incomplete
or uncertain information. AI researchers have devised a number of tools
to solve these problems using methods from probability theory and economics. Precise mathematical tools have been developed that analyze how an agent can make choices and plan, using decision theory, decision analysis, and information value theory. These tools include models such as Markov decision processes, dynamic decision networks, game theory and mechanism design.
Probabilistic algorithms can also be used for filtering,
prediction, smoothing, and finding explanations for streams of data,
thus helping perception systems analyze processes that occur over time
(e.g., hidden Markov models or Kalman filters).
Expectation–maximizationclustering of Old Faithful
eruption data starts from a random guess but then successfully
converges on an accurate clustering of the two physically distinct modes
of eruption.
Classifiers and statistical learning methods
The simplest AI applications can be divided into two types:
classifiers (e.g., "if shiny then diamond"), on one hand, and
controllers (e.g., "if diamond then pick up"), on the other hand. Classifiers are functions that use pattern matching to determine the closest match. They can be fine-tuned based on chosen examples using supervised learning. Each pattern (also called an "observation") is labeled with a certain predefined class. All the observations combined with their class labels are known as a data set. When a new observation is received, that observation is classified based on previous experience.
There are many kinds of classifiers in use. The decision tree is the simplest and most widely used symbolic machine learning algorithm. K-nearest neighbor algorithm was the most widely used analogical AI until the mid-1990s, and Kernel methods such as the support vector machine (SVM) displaced k-nearest neighbor in the 1990s. The naive Bayes classifier is reportedly the "most widely used learner" at Google, due in part to its scalability. Neural networks are also used as classifiers.
Artificial neural networks
A neural network is an interconnected group of nodes, akin to the vast network of neurons in the human brain.
An artificial neural network is based on a collection of nodes also known as artificial neurons, which loosely model the neurons
in a biological brain. It is trained to recognise patterns; once
trained, it can recognise those patterns in fresh data. There is an
input, at least one hidden layer of nodes and an output. Each node
applies a function and once the weight
crosses its specified threshold, the data is transmitted to the next
layer. A network is typically called a deep neural network if it has at
least 2 hidden layers.
Learning algorithms for neural networks use local search to choose the weights that will get the right output for each input during training. The most common training technique is the backpropagation algorithm. Neural networks learn to model complex relationships between inputs and outputs and find patterns in data. In theory, a neural network can learn any function.
In feedforward neural networks the signal passes in only one direction. The term perceptron typically refers to a single-layer neural network. In contrast, deep learning uses many layers. Recurrent neural networks (RNNs) feed the output signal back into the input, which allows short-term memories of previous input events. Long short-term memory networks (LSTMs) are recurrent neural networks that better preserve longterm dependencies and are less sensitive to the vanishing gradient problem. Convolutional neural networks (CNNs) use layers of kernels to more efficiently process local patterns. This local processing is especially important in image processing,
where the early CNN layers typically identify simple local patterns
such as edges and curves, with subsequent layers detecting more complex
patterns like textures, and eventually whole objects.
Deep learning uses several layers of neurons between the network's inputs and outputs. The multiple layers can progressively extract higher-level features from the raw input. For example, in image processing,
lower layers may identify edges, while higher layers may identify the
concepts relevant to a human such as digits, letters, or faces.
Deep learning has profoundly improved the performance of programs
in many important subfields of artificial intelligence, including computer vision, speech recognition, natural language processing, image classification, and others. The reason that deep learning performs so well in so many applications is not known as of 2021. The sudden success of deep learning in 2012–2015 did not occur because
of some new discovery or theoretical breakthrough (deep neural networks
and backpropagation had been described by many people, as far back as
the 1950s) but because of two factors: the incredible increase in computer power
(including the hundred-fold increase in speed by switching to GPUs) and the availability of vast amounts of training data, especially the giant curated datasets used for benchmark testing, such as ImageNet.
GPT
Generative pre-trained transformers (GPT) are large language models
(LLMs) that generate text based on the semantic relationships between
words in sentences. Text-based GPT models are pre-trained on a large corpus of text that can be from the Internet. The pretraining consists of predicting the next token
(a token being usually a word, subword, or punctuation). Throughout
this pretraining, GPT models accumulate knowledge about the world and
can then generate human-like text by repeatedly predicting the next
token. Typically, a subsequent training phase makes the model more
truthful, useful, and harmless, usually with a technique called reinforcement learning from human feedback (RLHF). Current GPT models are prone to generating falsehoods called "hallucinations". These can be reduced with RLHF and quality data, but the problem has been getting worse for reasoning systems. Such systems are used in chatbots, which allow people to ask a question or request a task in simple text.
The transistor density in integrated circuits has been observed to roughly double every 18 months—a trend known as Moore's law, named after the Intel co-founder Gordon Moore, who first identified it. Improvements in GPUs have been even faster, a trend sometimes called Huang's law, named after Nvidia co-founder and CEO Jensen Huang.
It has been suggested that AI can overcome discrepancies in funding allocated to different fields of research.
AlphaFold 2 (2021) demonstrated the ability to approximate, in hours rather than months, the 3D structure of a protein. In 2023, it was reported that AI-guided drug discovery helped find a
class of antibiotics capable of killing two different types of
drug-resistant bacteria. In 2024, researchers used machine learning to accelerate the search for Parkinson's disease drug treatments. Their aim was to identify compounds that block the clumping, or aggregation, of alpha-synuclein
(the protein that characterises Parkinson's disease). They were able to
speed up the initial screening process ten-fold and reduce the cost by a
thousand-fold.
Game playing programs have been used since the 1950s to demonstrate and test AI's most advanced techniques. Deep Blue became the first computer chess-playing system to beat a reigning world chess champion, Garry Kasparov, on 11 May 1997. In 2011, in a Jeopardy!quiz show exhibition match, IBM's question answering system, Watson, defeated the two greatest Jeopardy! champions, Brad Rutter and Ken Jennings, by a significant margin. In March 2016, AlphaGo won 4 out of 5 games of Go in a match with Go champion Lee Sedol, becoming the first computer Go-playing system to beat a professional Go player without handicaps. Then, in 2017, it defeated Ke Jie, who was the best Go player in the world. Other programs handle imperfect-information games, such as the poker-playing program Pluribus. DeepMind developed increasingly generalistic reinforcement learning models, such as with MuZero, which could be trained to play chess, Go, or Atari games. In 2019, DeepMind's AlphaStar achieved grandmaster level in StarCraft II, a particularly challenging real-time strategy game that involves incomplete knowledge of what happens on the map. In 2021, an AI agent competed in a PlayStation Gran Turismo competition, winning against four of the world's best Gran Turismo drivers using deep reinforcement learning. In 2024, Google DeepMind introduced SIMA, a type of AI capable of autonomously playing nine previously unseen open-world
video games by observing screen output, as well as executing short,
specific tasks in response to natural language instructions.
Mathematics
Large language models, such as GPT-4, Gemini, Claude, Llama or Mistral,
are increasingly used in mathematics. These probabilistic models are
versatile, but can also produce wrong answers in the form of hallucinations. They sometimes need a large database of mathematical problems to learn from, but also methods such as supervisedfine-tuning or trained classifiers with human-annotated data to improve answers for new problems and learn from corrections. A February 2024 study showed that the performance of some language
models for reasoning capabilities in solving math problems not included
in their training data was low, even for problems with only minor
deviations from trained data. One technique to improve their performance involves training the models to produce correct reasoning steps, rather than just the correct result. The Alibaba Group developed a version of its Qwen models called Qwen2-Math,
that achieved state-of-the-art performance on several mathematical
benchmarks, including 84% accuracy on the MATH dataset of competition
mathematics problems. In January 2025, Microsoft proposed the technique rStar-Math that leverages Monte Carlo tree search and step-by-step reasoning, enabling a relatively small language model like Qwen-7B to solve 53% of the AIME 2024 and 90% of the MATH benchmark problems.
Alternatively, dedicated models for mathematical problem solving
with higher precision for the outcome including proof of theorems have
been developed such as AlphaTensor, AlphaGeometry, AlphaProof and AlphaEvolve all from Google DeepMind, Llemma from EleutherAI or Julius.
When natural language is used to describe mathematical problems,
converters can transform such prompts into a formal language such as Lean to define mathematical tasks. The experimental model Gemini Deep Think accepts natural language prompts directly and achieved gold medal results in the International Math Olympiad of 2025.
Some models have been developed to solve challenging problems and
reach good results in benchmark tests, others to serve as educational
tools in mathematics.
Finance is one of the fastest growing sectors where applied AI tools
are being deployed: from retail online banking to investment advice and
insurance, where automated "robot advisers" have been in use for some
years.
According to Nicolas Firzli, director of the World Pensions & Investments Forum,
it may be too early to see the emergence of highly innovative
AI-informed financial products and services. He argues that "the
deployment of AI tools will simply further automatise things: destroying
tens of thousands of jobs in banking, financial planning, and pension
advice in the process, but I'm not sure it will unleash a new wave of
[e.g., sophisticated] pension innovation."
Various countries are deploying AI military applications. The main applications enhance command and control, communications, sensors, integration and interoperability. Research is targeting intelligence collection and analysis, logistics,
cyber operations, information operations, and semiautonomous and autonomous vehicles. AI technologies enable coordination of sensors and effectors, threat detection and identification, marking of enemy positions, target acquisition, coordination and deconfliction of distributed Joint Fires between networked combat vehicles, both human-operated and autonomous.
AI has been used in military operations in Iraq, Syria, Israel and Ukraine.
Generative AI
Vincent van Gogh in watercolour created by generative AI software
Companies in a variety of sectors have used generative AI, including those in software development, healthcare, finance, entertainment, customer service, sales and marketing, art, writing, and product design.
AI agents are software entities designed to perceive their
environment, make decisions, and take actions autonomously to achieve
specific goals. These agents can interact with users, their environment,
or other agents. AI agents are used in various applications, including virtual assistants, chatbots, autonomous vehicles, game-playing systems, and industrial robotics.
AI agents operate within the constraints of their programming,
available computational resources, and hardware limitations. This means
they are restricted to performing tasks within their defined scope and
have finite memory and processing capabilities. In real-world
applications, AI agents often face time constraints for decision-making
and action execution. Many AI agents incorporate learning algorithms,
enabling them to improve their performance over time through experience
or training. Using machine learning, AI agents can adapt to new
situations and optimise their behaviour for their designated tasks.
Web search
Microsoft introduced Copilot Search in February 2023 under the name Bing Chat, as a built-in feature for Microsoft Edge and Bing mobile app. Copilot Search provides AI-generated summaries and step-by-step reasoning based of information from web publishers, ranked in Bing Search. For safety, Copilot uses AI-based classifiers and filters to reduce potentially harmful content.
Google officially pushed its AI Search at its Google I/O event on 20 May 2025. It keeps people looking at Google instead of clicking on a search result. AI Overviews uses Gemini 2.5 to provide contextual answers to user queries based on web content.
Sexuality
Applications of AI in this domain include AI-enabled menstruation and
fertility trackers that analyze user data to offer predictions, AI-integrated sex toys (e.g., teledildonics), AI-generated sexual education content, and AI agents that simulate sexual and romantic partners (e.g., Replika). AI is also used for the production of non-consensual deepfake pornography, raising significant ethical and legal concerns.
There are also thousands of successful AI applications used to solve
specific problems for specific industries or institutions. In a 2017
survey, one in five companies reported having incorporated "AI" in some
offerings or processes. A few examples are energy storage, medical diagnosis, military logistics, applications that predict the result of judicial decisions, foreign policy, or supply chain management.
AI applications for evacuation and disaster
management are growing. AI has been used to investigate patterns in
large-scale and small-scale evacuations using historical data from GPS,
videos or social media. Furthermore, AI can provide real-time
information on the evacuation conditions.
In agriculture, AI has helped farmers to increase yield and
identify areas that need irrigation, fertilization, pesticide
treatments. Agronomists use AI to conduct research and development. AI
has been used to predict the ripening time for crops such as tomatoes,
monitor soil moisture, operate agricultural robots, conduct predictive analytics, classify livestock pig call emotions, automate greenhouses, detect diseases and pests, and save water.
Artificial intelligence is used in astronomy to analyze
increasing amounts of available data and applications, mainly for
"classification, regression, clustering, forecasting, generation,
discovery, and the development of new scientific insights." For example,
it is used for discovering exoplanets, forecasting solar activity, and
distinguishing between signals and instrumental effects in gravitational
wave astronomy. Additionally, it could be used for activities in space,
such as space exploration, including the analysis of data from space
missions, real-time science decisions of spacecraft, space debris
avoidance, and more autonomous operation.
During the 2024 Indian elections, US$50 million was spent on authorized AI-generated content, notably by creating deepfakes
of allied (including sometimes deceased) politicians to better engage
with voters, and by translating speeches to various local languages.
AI has potential benefits and potential risks. AI may be able to advance science and find solutions for serious problems: Demis Hassabis of DeepMind hopes to "solve intelligence, and then use that to solve everything else". However, as the use of AI has become widespread, several unintended consequences and risks have been identified. In-production systems can sometimes not factor ethics and bias into
their AI training processes, especially when the AI algorithms are
inherently unexplainable in deep learning.
Machine learning algorithms require large amounts of data. The techniques used to acquire this data have raised concerns about privacy, surveillance and copyright.
AI-powered devices and services, such as virtual assistants and
IoT products, continuously collect personal information, raising
concerns about intrusive data gathering and unauthorized access by third
parties. The loss of privacy is further exacerbated by AI's ability to
process and combine vast amounts of data, potentially leading to a
surveillance society where individual activities are constantly
monitored and analyzed without adequate safeguards or transparency.
Sensitive user data collected may include online activity records, geolocation data, video, or audio. For example, in order to build speech recognition algorithms, Amazon has recorded millions of private conversations and allowed temporary workers to listen to and transcribe some of them. Opinions about this widespread surveillance range from those who see it as a necessary evil to those for whom it is clearly unethical and a violation of the right to privacy.
AI developers argue that this is the only way to deliver valuable
applications and have developed several techniques that attempt to
preserve privacy while still obtaining the data, such as data aggregation, de-identification and differential privacy. Since 2016, some privacy experts, such as Cynthia Dwork, have begun to view privacy in terms of fairness. Brian Christian wrote that experts have pivoted "from the question of 'what they know' to the question of 'what they're doing with it'."
Generative AI is often trained on unlicensed copyrighted works,
including in domains such as images or computer code; the output is then
used under the rationale of "fair use".
Experts disagree about how well and under what circumstances this
rationale will hold up in courts of law; relevant factors may include
"the purpose and character of the use of the copyrighted work" and "the
effect upon the potential market for the copyrighted work". Website owners can indicate that they do not want their content scraped via a "robots.txt" file. However, some companies will scrape content regardlessbecause the robots.txt file has no real authority. In 2023, leading authors (including John Grisham and Jonathan Franzen) sued AI companies for using their work to train generative AI. Another discussed approach is to envision a separate sui generis system of protection for creations generated by AI to ensure fair attribution and compensation for human authors.
Fueled by growth in artificial intelligence, data centers' demand for power increased in the 2020s.
In January 2024, the International Energy Agency (IEA) released Electricity 2024, Analysis and Forecast to 2026, forecasting electric power use. This is the first IEA report to make projections for data centers and
power consumption for artificial intelligence and cryptocurrency. The
report states that power demand for these uses might double by 2026,
with additional electric power usage equal to electricity used by the
whole Japanese nation.
Prodigious power consumption by AI is responsible for the growth
of fossil fuel use, and might delay closings of obsolete,
carbon-emitting coal energy facilities. There is a feverish rise in the
construction of data centers throughout the US, making large technology
firms (e.g., Microsoft, Meta, Google, Amazon) into voracious consumers
of electric power. Projected electric consumption is so immense that
there is concern that it will be fulfilled no matter the source. A
ChatGPT search involves the use of 10 times the electrical energy as a
Google search. The large firms are in haste to find power sources – from
nuclear energy to geothermal to fusion. The tech firms argue that – in
the long view – AI will be eventually kinder to the environment, but
they need the energy now. AI makes the power grid more efficient and
"intelligent", will assist in the growth of nuclear power, and track
overall carbon emissions, according to technology firms.
A 2024 Goldman Sachs Research Paper, AI Data Centers and the Coming US Power Demand Surge,
found "US power demand (is) likely to experience growth not seen in a
generation...." and forecasts that, by 2030, US data centers will
consume 8% of US power, as opposed to 3% in 2022, presaging growth for
the electrical power generation industry by a variety of means. Data centers' need for more and more electrical power is such that they
might max out the electrical grid. The Big Tech companies counter that
AI can be used to maximize the utilization of the grid by all.
In 2024, the Wall Street Journal reported that big AI
companies have begun negotiations with the US nuclear power providers to
provide electricity to the data centers. In March 2024 Amazon purchased
a Pennsylvania nuclear-powered data center for US$650 million. Nvidia CEO Jensen Huang said nuclear power is a good option for the data centers.
In September 2024, Microsoft announced an agreement with Constellation Energy to re-open the Three Mile Island
nuclear power plant to provide Microsoft with 100% of all electric
power produced by the plant for 20 years. Reopening the plant, which
suffered a partial nuclear meltdown of its Unit 2 reactor in 1979, will
require Constellation to get through strict regulatory processes which
will include extensive safety scrutiny from the US Nuclear Regulatory Commission.
If approved (this will be the first ever US re-commissioning of a
nuclear plant), over 835 megawatts of power – enough for 800,000 homes –
of energy will be produced. The cost for re-opening and upgrading is
estimated at US$1.6 billion and is dependent on tax breaks for nuclear
power contained in the 2022 US Inflation Reduction Act. The US government and the state of Michigan are investing almost US$2 billion to reopen the Palisades Nuclear
reactor on Lake Michigan. Closed since 2022, the plant is planned to be
reopened in October 2025. The Three Mile Island facility will be
renamed the Crane Clean Energy Center after Chris Crane, a nuclear
proponent and former CEO of Exelon who was responsible for Exelon's spinoff of Constellation.
After the last approval in September 2023, Taiwan suspended the approval of data centers north of Taoyuan with a capacity of more than 5 MW in 2024, due to power supply shortages. Taiwan aims to phase out nuclear power by 2025. On the other hand, Singapore imposed a ban on the opening of data centers in 2019 due to electric power, but in 2022, lifted this ban.
Although most nuclear plants in Japan have been shut down after the 2011 Fukushima nuclear accident, according to an October 2024 Bloomberg
article in Japanese, cloud gaming services company Ubitus, in which
Nvidia has a stake, is looking for land in Japan near a nuclear power
plant for a new data center for generative AI. Ubitus CEO Wesley Kuo said nuclear power plants are the most efficient, cheap and stable power for AI.
On 1 November 2024, the Federal Energy Regulatory Commission (FERC) rejected an application submitted by Talen Energy for approval to supply some electricity from the nuclear power station Susquehanna to Amazon's data center. According to the Commission Chairman Willie L. Phillips,
it is a burden on the electricity grid as well as a significant cost
shifting concern to households and other business sectors.
In 2025, a report prepared by the International Energy Agency estimated the greenhouse gas emissions
from the energy consumption of AI at 180 million tons. By 2035, these
emissions could rise to 300–500 million tonnes depending on what
measures will be taken. This is below 1.5% of the energy sector
emissions. The emissions reduction potential of AI was estimated at 5%
of the energy sector emissions, but rebound effects (for example if people switch from public transport to autonomous cars) can reduce it.
YouTube, Facebook and others use recommender systems to guide users to more content. These AI programs were given the goal of maximizing user engagement (that is, the only goal was to keep people watching). The AI learned that users tended to choose misinformation, conspiracy theories, and extreme partisan
content, and, to keep them watching, the AI recommended more of it.
Users also tended to watch more content on the same subject, so the AI
led people into filter bubbles where they received multiple versions of the same misinformation. This convinced many users that the misinformation was true, and
ultimately undermined trust in institutions, the media and the
government. The AI program had correctly learned to maximize its goal, but the
result was harmful to society. After the U.S. election in 2016, major
technology companies took some steps to mitigate the problem.
In the early 2020s, generative AI
began to create images, audio, and texts that are virtually
indistinguishable from real photographs, recordings, or human writing, while realistic AI-generated videos became feasible in the mid-2020s. It is possible for bad actors to use this technology to create massive amounts of misinformation or propaganda; one such potential malicious use is deepfakes for computational propaganda. AI pioneer and Nobel Prize-winning computer scientist Geoffrey Hinton
expressed concern about AI enabling "authoritarian leaders to
manipulate their electorates" on a large scale, among other risks. The ability to influence electorates has been proved in at least one
study. This same study shows more inaccurate statements from the models
when they advocate for candidates of the political right.
AI researchers at Microsoft, OpenAI, universities and other organisations have suggested using "personhood credentials" as a way to overcome online deception enabled by AI models.
Machine learning applications can be biased if they learn from biased data. The developers may not be aware that the bias exists. Discriminatory behavior by some LLMs can be observed in their output. Bias can be introduced by the way training data is selected and by the way a model is deployed. If a biased algorithm is used to make decisions that can seriously harm people (as it can in medicine, finance, recruitment, housing or policing) then the algorithm may cause discrimination. The field of fairness studies how to prevent harms from algorithmic biases.
On 28 June 2015, Google Photos's
new image labeling feature mistakenly identified Jacky Alcine and a
friend as "gorillas" because they were black. The system was trained on a
dataset that contained very few images of black people, a problem called "sample size disparity". Google "fixed" this problem by preventing the system from labelling anything
as a "gorilla". Eight years later, in 2023, Google Photos still could
not identify a gorilla, and neither could similar products from Apple,
Facebook, Microsoft and Amazon.
COMPAS is a commercial program widely used by U.S. courts to assess the likelihood of a defendant becoming a recidivist. In 2016, Julia Angwin at ProPublica
discovered that COMPAS exhibited racial bias, despite the fact that the
program was not told the races of the defendants. Although the error
rate for both whites and blacks was calibrated equal at exactly 61%, the
errors for each race were different—the system consistently
overestimated the chance that a black person would re-offend and would
underestimate the chance that a white person would not re-offend. In 2017, several researchers showed that it was mathematically impossible for COMPAS to accommodate
all possible measures of fairness when the base rates of re-offense were
different for whites and blacks in the data.
A program can make biased decisions even if the data does not
explicitly mention a problematic feature (such as "race" or "gender").
The feature will correlate with other features (like "address",
"shopping history" or "first name"), and the program will make the same
decisions based on these features as it would on "race" or "gender". Moritz Hardt said "the most robust fact in this research area is that fairness through blindness doesn't work."
Criticism of COMPAS highlighted that machine learning models are
designed to make "predictions" that are only valid if we assume that the
future will resemble the past. If they are trained on data that
includes the results of racist decisions in the past, machine learning
models must predict that racist decisions will be made in the future. If
an application then uses these predictions as recommendations, some of these "recommendations" will likely be racist. Thus, machine learning is not well suited to help make decisions in areas where there is hope that the future will be better than the past. It is descriptive rather than prescriptive.
Bias and unfairness may go undetected because the developers are
overwhelmingly white and male: among AI engineers, about 4% are black
and 20% are women.
There are various conflicting definitions and mathematical models
of fairness. These notions depend on ethical assumptions, and are
influenced by beliefs about society. One broad category is distributive fairness,
which focuses on the outcomes, often identifying groups and seeking to
compensate for statistical disparities. Representational fairness tries
to ensure that AI systems do not reinforce negative stereotypes
or render certain groups invisible. Procedural fairness focuses on the
decision process rather than the outcome. The most relevant notions of
fairness may depend on the context, notably the type of AI application
and the stakeholders. The subjectivity in the notions of bias and
fairness makes it difficult for companies to operationalize them. Having
access to sensitive attributes such as race or gender is also
considered by many AI ethicists to be necessary in order to compensate
for biases, but it may conflict with anti-discrimination laws.
At the 2022 ACM Conference on Fairness, Accountability, and Transparency a paper reported that a CLIP‑based (Contrastive Language-Image Pre-training)
robotic system reproduced harmful gender‑ and race‑linked stereotypes
in a simulated manipulation task. The authors recommended robot‑learning
methods which physically manifest such harms be "paused, reworked, or
even wound down when appropriate, until outcomes can be proven safe,
effective, and just."
Many AI systems are so complex that their designers cannot explain how they reach their decisions. Particularly with deep neural networks, in which there are many non-linear relationships between inputs and outputs. But some popular explainability techniques exist.
It is impossible to be certain that a program is operating
correctly if no one knows how exactly it works. There have been many
cases where a machine learning program passed rigorous tests, but
nevertheless learned something different than what the programmers
intended. For example, a system that could identify skin diseases better
than medical professionals was found to actually have a strong tendency
to classify images with a ruler as "cancerous", because pictures of malignancies typically include a ruler to show the scale. Another machine learning system designed to help effectively allocate
medical resources was found to classify patients with asthma as being at
"low risk" of dying from pneumonia. Having asthma is actually a severe
risk factor, but since the patients having asthma would usually get much
more medical care, they were relatively unlikely to die according to
the training data. The correlation between asthma and low risk of dying
from pneumonia was real, but misleading.
People who have been harmed by an algorithm's decision have a right to an explanation. Doctors, for example, are expected to clearly and completely explain to
their colleagues the reasoning behind any decision they make. Early
drafts of the European Union's General Data Protection Regulation in 2016 included an explicit statement that this right exists. Industry experts noted that this is an unsolved problem with no
solution in sight. Regulators argued that nevertheless the harm is real:
if the problem has no solution, the tools should not be used.
DARPA established the XAI ("Explainable Artificial Intelligence") program in 2014 to try to solve these problems.
Several approaches aim to address the transparency problem. SHAP
enables to visualise the contribution of each feature to the output. LIME can locally approximate a model's outputs with a simpler, interpretable model. Multitask learning
provides a large number of outputs in addition to the target
classification. These other outputs can help developers deduce what the
network has learned. Deconvolution, DeepDream and other generative
methods can allow developers to see what different layers of a deep
network for computer vision have learned, and produce output that can
suggest what the network is learning. For generative pre-trained transformers, Anthropic developed a technique based on dictionary learning that associates patterns of neuron activations with human-understandable concepts.
A lethal autonomous weapon is a machine that locates, selects and engages human targets without human supervision. Widely available AI tools can be used by bad actors to develop
inexpensive autonomous weapons and, if produced at scale, they are
potentially weapons of mass destruction. Even when used in conventional warfare, they currently cannot reliably choose targets and could potentially kill an innocent person. In 2014, 30 nations (including China) supported a ban on autonomous weapons under the United Nations' Convention on Certain Conventional Weapons, however the United States and others disagreed. By 2015, over fifty countries were reported to be researching battlefield robots.
There are many other ways in which AI is expected to help bad
actors, some of which can not be foreseen. For example, machine-learning
AI is able to design tens of thousands of toxic molecules in a matter
of hours.
Economists have frequently highlighted the risks of redundancies from
AI, and speculated about unemployment if there is no adequate social
policy for full employment.
In the past, technology has tended to increase rather than reduce
total employment, but economists acknowledge that "we're in uncharted
territory" with AI. A survey of economists showed disagreement about whether the increasing
use of robots and AI will cause a substantial increase in long-term unemployment, but they generally agree that it could be a net benefit if productivity gains are redistributed. Risk estimates vary; for example, in the 2010s, Michael Osborne and Carl Benedikt Frey
estimated 47% of U.S. jobs are at "high risk" of potential automation,
while an OECD report classified only 9% of U.S. jobs as "high risk". The methodology of speculating about future employment levels has been
criticised as lacking evidential foundation, and for implying that
technology, rather than social policy, creates unemployment, as opposed
to redundancies. In April 2023, it was reported that 70% of the jobs for Chinese video
game illustrators had been eliminated by generative artificial
intelligence.
Unlike previous waves of automation, many middle-class jobs may be eliminated by artificial intelligence; The Economist
stated in 2015 that "the worry that AI could do to white-collar jobs
what steam power did to blue-collar ones during the Industrial
Revolution" is "worth taking seriously". Jobs at extreme risk range from paralegals
to fast food cooks, while job demand is likely to increase for
care-related professions ranging from personal healthcare to the clergy. In July 2025, Ford CEO Jim Farley predicted that "artificial intelligence is going to replace literally half of all white-collar workers in the U.S."
From the early days of the development of artificial intelligence, there have been arguments, for example, those put forward by Joseph Weizenbaum,
about whether tasks that can be done by computers actually should be
done by them, given the difference between computers and humans, and
between quantitative calculation and qualitative, value-based judgement.
Recent public debates in artificial intelligence have increasingly
focused on its broader societal and ethical implications. It has been
argued AI will become so powerful that humanity may irreversibly lose
control of it. This could, as physicist Stephen Hawking stated, "spell the end of the human race". This scenario has been common in science fiction, when a computer or
robot suddenly develops a human-like "self-awareness" (or "sentience" or
"consciousness") and becomes a malevolent character. These sci-fi scenarios are misleading in several ways.
First, AI does not require human-like sentience
to be an existential risk. Modern AI programs are given specific goals
and use learning and intelligence to achieve them. Philosopher Nick Bostrom argued that if one gives almost any goal to a sufficiently powerful AI, it may choose to destroy humanity to achieve it (he used the example of an automated paperclip factory that destroys the world to get more iron for paperclips). Stuart Russell
gives the example of household robot that tries to find a way to kill
its owner to prevent it from being unplugged, reasoning that "you can't
fetch the coffee if you're dead." In order to be safe for humanity, a superintelligence would have to be genuinely aligned with humanity's morality and values so that it is "fundamentally on our side".
Second, Yuval Noah Harari
argues that AI does not require a robot body or physical control to
pose an existential risk. The essential parts of civilization are not
physical. Things like ideologies, law, government, money and the economy are built on language; they exist because there are stories that billions of people believe. The current prevalence of misinformation suggests that an AI could use language to convince people to believe anything, even to take actions that are destructive. Geoffrey Hinton said in 2025 that modern AI
is particularly "good at persuasion" and getting better all the time.
He asks "Suppose you wanted to invade the capital of the US. Do you have
to go there and do it yourself? No. You just have to be good at
persuasion."
The opinions amongst experts and industry insiders are mixed,
with sizable fractions both concerned and unconcerned by risk from
eventual superintelligent AI. Personalities such as Stephen Hawking, Bill Gates, and Elon Musk, as well as AI pioneers such as Geoffrey Hinton, Yoshua Bengio, Stuart Russell, Demis Hassabis, and Sam Altman, have expressed concerns about existential risk from AI.
In May 2023, Geoffrey Hinton announced his resignation from
Google in order to be able to "freely speak out about the risks of AI"
without "considering how this impacts Google". He notably mentioned risks of an AI takeover, and stressed that in order to avoid the worst outcomes, establishing
safety guidelines will require cooperation among those competing in use
of AI.
In 2023, many leading AI experts endorsed the joint statement
that "Mitigating the risk of extinction from AI should be a global
priority alongside other societal-scale risks such as pandemics and
nuclear war".
Some other researchers were more optimistic. AI pioneer Jürgen Schmidhuber
did not sign the joint statement, emphasising that in 95% of all cases,
AI research is about making "human lives longer and healthier and
easier." While the tools that are now being used to improve lives can also be
used by bad actors, "they can also be used against the bad actors." Andrew Ng
also argued that "it's a mistake to fall for the doomsday hype on
AI—and that regulators who do will only benefit vested interests." Yann LeCun
", a Turing Award winner, disagreed with the idea that AI will
subordinate humans "simply because they are smarter, let alone destroy
[us]", "scoff[ing] at his peers' dystopian scenarios of supercharged misinformation and even, eventually, human extinction." In the early 2010s, experts argued that the risks are too distant in
the future to warrant research or that humans will be valuable from the
perspective of a superintelligent machine. However, after 2016, the study of current and future risks and possible solutions became a serious area of research.
Friendly AI are machines that have been designed from the beginning to minimize risks and to make choices that benefit humans. Eliezer Yudkowsky,
who coined the term, argues that developing friendly AI should be a
higher research priority: it may require a large investment and it must
be completed before AI becomes an existential risk.
Machines with intelligence have the potential to use their
intelligence to make ethical decisions. The field of machine ethics
provides machines with ethical principles and procedures for resolving
ethical dilemmas. The field of machine ethics is also called computational morality, and was founded at an AAAI symposium in 2005.
Active organizations in the AI open-source community include Hugging Face, Google, EleutherAI and Meta. Various AI models, such as Llama 2, Mistral or Stable Diffusion, have been made open-weight, meaning that their architecture and trained parameters (the "weights")
are publicly available. Open-weight models can be freely fine-tuned, which allows companies to specialize them with their own data and for their own use-case. Open-weight models are useful for research and innovation but can also
be misused. Since they can be fine-tuned, any built-in security measure,
such as objecting to harmful requests, can be trained away until it
becomes ineffective. Some researchers warn that future AI models may
develop dangerous capabilities (such as the potential to drastically
facilitate bioterrorism)
and that once released on the Internet, they cannot be deleted
everywhere if needed. They recommend pre-release audits and cost-benefit
analyses.
Frameworks
Artificial intelligence projects can be guided by ethical
considerations during the design, development, and implementation of an
AI system. An AI framework such as the Care and Act Framework, developed
by the Alan Turing Institute and based on the SUM values, outlines four main ethical dimensions, defined as follows:
Respect the dignity of individual people
Connect with other people sincerely, openly, and inclusively
Care for the wellbeing of everyone
Protect social values, justice, and the public interest
Other developments in ethical frameworks include those decided upon during the Asilomar Conference, the Montreal Declaration for Responsible AI, and the IEEE's Ethics of Autonomous Systems initiative, among others; however, these principles are not without criticism, especially regarding the people chosen to contribute to these frameworks.
Promotion of the wellbeing of the people and communities that
these technologies affect requires consideration of the social and
ethical implications at all stages of AI system design, development and
implementation, and collaboration between job roles such as data
scientists, product managers, data engineers, domain experts, and
delivery managers.
The UK AI Safety Institute
released in 2024 a testing toolset called 'Inspect' for AI safety
evaluations available under an MIT open-source licence which is freely
available on GitHub and can be improved with third-party packages. It
can be used to evaluate AI models in a range of areas including core
knowledge, ability to reason, and autonomous capabilities.
The first global AI Safety Summit was held in the United Kingdom in November 2023 with a declaration calling for international cooperation.
The regulation of artificial intelligence is the development of
public sector policies and laws for promoting and regulating AI; it is
therefore related to the broader regulation of algorithms. The regulatory and policy landscape for AI is an emerging issue in jurisdictions globally. According to AI Index at Stanford,
the annual number of AI-related laws passed in the 127 survey countries
jumped from one passed in 2016 to 37 passed in 2022 alone. Between 2016 and 2020, more than 30 countries adopted dedicated strategies for AI. Most EU member states had released national AI strategies, as had Canada,
China, India, Japan, Mauritius, the Russian Federation, Saudi Arabia,
United Arab Emirates, U.S., and Vietnam. Others were in the process of
elaborating their own AI strategy, including Bangladesh, Malaysia and
Tunisia. The Global Partnership on Artificial Intelligence
was launched in June 2020, stating a need for AI to be developed in
accordance with human rights and democratic values, to ensure public
confidence and trust in the technology. Henry Kissinger, Eric Schmidt, and Daniel Huttenlocher published a joint statement in November 2021 calling for a government commission to regulate AI. In 2023, OpenAI leaders published recommendations for the governance of
superintelligence, which they believe may happen in less than 10 years. In 2023, the United Nations also launched an advisory body to provide
recommendations on AI governance; the body comprises technology company
executives, government officials and academics. On 1 August 2024, the EU Artificial Intelligence Act entered into force, establishing the first comprehensive EU-wide AI regulation. In 2024, the Council of Europe created the first international legally binding treaty on AI, called the "Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law". It was adopted by the European Union, the United States, the United Kingdom, and other signatories.
In a 2022 Ipsos
survey, attitudes towards AI varied greatly by country; 78% of Chinese
citizens, but only 35% of Americans, agreed that "products and services
using AI have more benefits than drawbacks". A 2023 Reuters/Ipsos poll found that 61% of Americans agree, and 22% disagree, that AI poses risks to humanity. In a 2023 Fox News
poll, 35% of Americans thought it "very important", and an additional
41% thought it "somewhat important", for the federal government to
regulate AI, versus 13% responding "not very important" and 8%
responding "not at all important".
In November 2023, the first global AI Safety Summit was held in Bletchley Park in the UK to discuss the near and far term risks of AI and the possibility of mandatory and voluntary regulatory frameworks. 28 countries including the United States, China, and the European Union
issued a declaration at the start of the summit, calling for
international co-operation to manage the challenges and risks of
artificial intelligence. In May 2024 at the AI Seoul Summit, 16 global AI tech companies agreed to safety commitments on the development of AI.
In 2024, AI patents in China and the US numbered more than three-fourths of AI patents worldwide. Though China had more AI patents, the US had 35% more patents per AI patent-applicant company than China.
The study of mechanical or "formal" reasoning began with philosophers
and mathematicians in antiquity. The study of logic led directly to Alan Turing's theory of computation,
which suggested that a machine, by shuffling symbols as simple as "0"
and "1", could simulate any conceivable form of mathematical reasoning. This, along with concurrent discoveries in cybernetics, information theory and neurobiology, led researchers to consider the possibility of building an "electronic brain". They developed several areas of research that would become part of AI, such as McCulloch and Pitts design for "artificial neurons" in 1943, and Turing's influential 1950 paper 'Computing Machinery and Intelligence', which introduced the Turing test and showed that "machine intelligence" was plausible.
The field of AI research was founded at a workshop at Dartmouth College in 1956. The attendees became the leaders of AI research in the 1960s. They and their students produced programs that the press described as "astonishing": computers were learning checkers strategies, solving word problems in algebra, proving logical theorems and speaking English. Artificial intelligence laboratories were set up at a number of British
and U.S. universities in the latter 1950s and early 1960s.
Researchers in the 1960s and the 1970s were convinced that their methods would eventually succeed in creating a machine with general intelligence and considered this the goal of their field. In 1965 Herbert Simon predicted, "machines will be capable, within twenty years, of doing any work a man can do". In 1967 Marvin Minsky agreed, writing that "within a generation ... the problem of creating 'artificial intelligence' will substantially be solved". They had, however, underestimated the difficulty of the problem. In 1974, both the U.S. and British governments cut off exploratory research in response to the criticism of Sir James Lighthill and ongoing pressure from the U.S. Congress to fund more productive projects. Minsky and Papert's book Perceptrons was understood as proving that artificial neural networks would never be useful for solving real-world tasks, thus discrediting the approach altogether. The "AI winter", a period when obtaining funding for AI projects was difficult, followed.
In the early 1980s, AI research was revived by the commercial success of expert systems, a form of AI program that simulated the knowledge and analytical skills
of human experts. By 1985, the market for AI had reached over a billion
dollars. At the same time, Japan's fifth generation computer project inspired the U.S. and British governments to restore funding for academic research. However, beginning with the collapse of the Lisp Machine market in 1987, AI once again fell into disrepute, and a second, longer-lasting winter began.
Up to this point, most of AI's funding had gone to projects that used high-level symbols to represent mental objects
like plans, goals, beliefs, and known facts. In the 1980s, some
researchers began to doubt that this approach would be able to imitate
all the processes of human cognition, especially perception, robotics, learning and pattern recognition, and began to look into "sub-symbolic" approaches. Rodney Brooks rejected "representation" in general and focussed directly on engineering machines that move and survive.[x]Judea Pearl, Lotfi Zadeh,
and others developed methods that handled incomplete and uncertain
information by making reasonable guesses rather than precise logic. But the most important development was the revival of "connectionism", including neural network research, by Geoffrey Hinton and others. In 1990, Yann LeCun successfully showed that convolutional neural networks can recognize handwritten digits, the first of many successful applications of neural networks.
AI gradually restored its reputation in the late 1990s and early
21st century by exploiting formal mathematical methods and by finding
specific solutions to specific problems. This "narrow" and "formal" focus allowed researchers to produce verifiable results and collaborate with other fields (such as statistics, economics and mathematics). By 2000, solutions developed by AI researchers were being widely used,
although in the 1990s they were rarely described as "artificial
intelligence" (a tendency known as the AI effect). However, several academic researchers became concerned that AI was no
longer pursuing its original goal of creating versatile, fully
intelligent machines. Beginning around 2002, they founded the subfield
of artificial general intelligence (or "AGI"), which had several well-funded institutions by the 2010s.
Deep learning began to dominate industry benchmarks in 2012 and was adopted throughout the field. For many specific tasks, other methods were abandoned. Deep learning's success was based on both hardware improvements (faster computers, graphics processing units, cloud computing) and access to large amounts of data (including curated datasets, such as ImageNet). Deep learning's success led to an enormous increase in interest and funding in AI. The amount of machine learning research (measured by total publications) increased by 50% in the years 2015–2019.
The number of Google searches for the term "AI" accelerated in 2022.
In 2016, issues of fairness
and the misuse of technology were catapulted into center stage at
machine learning conferences, publications vastly increased, funding
became available, and many researchers re-focussed their careers on
these issues. The alignment problem became a serious field of academic study.
In the late 2010s and early 2020s, AGI companies began to deliver programs that created enormous interest. In 2015, AlphaGo, developed by DeepMind, beat the world champion Go player. The program taught only the game's rules and developed a strategy by itself. GPT-3 is a large language model that was released in 2020 by OpenAI and is capable of generating high-quality human-like text. ChatGPT,
launched on 30 November 2022, became the fastest-growing consumer
software application in history, gaining over 100 million users in two
months. It marked what is widely regarded as AI's breakout year, bringing it into the public consciousness. These programs, and others, inspired an aggressive AI boom,
where large companies began investing billions of dollars in AI
research. According to AI Impacts, about US$50 billion annually was
invested in "AI" around 2022 in the U.S. alone and about 20% of the new
U.S. Computer Science PhD graduates have specialized in "AI". About 800,000 "AI"-related U.S. job openings existed in 2022. According to PitchBook research, 22% of newly funded startups in 2024 claimed to be AI companies.
Philosophical debates have historically sought to determine the nature of intelligence and how to make intelligent machines. Another major focus has been whether machines can be conscious, and the associated ethical implications. Many other topics in philosophy are relevant to AI, such as epistemology and free will. For example, debates center on whether machines can genuinely
understand meaning, whether they possess autonomous decision-making
capabilities, and to what extent their actions can be considered
intentional rather than merely the result of algorithmic processes.
Rapid advancements have intensified public discussions on the philosophy
and ethics of AI.
Alan Turing wrote in 1950 "I propose to consider the question 'can machines think'?" He advised changing the question from whether a machine "thinks", to
"whether or not it is possible for machinery to show intelligent
behaviour". He devised the Turing test, which measures the ability of a machine to simulate human conversation. Since we can only observe the behavior of the machine, it does not
matter if it is "actually" thinking or literally has a "mind". Turing
notes that we can not determine these things about other people but "it is usual to have a polite convention that everyone thinks."
The Turing test can provide some evidence of intelligence, but it penalizes non-human intelligent behavior.
Russell and Norvig agree with Turing that intelligence must be defined in terms of external behavior, not internal structure. However, they are critical that the test requires the machine to imitate humans. "Aeronautical engineering texts", they wrote, "do not define the goal of their field as making 'machines that fly so exactly like pigeons that they can fool other pigeons.'" AI founder John McCarthy agreed, writing that "Artificial intelligence is not, by definition, simulation of human intelligence".
McCarthy defines intelligence as "the computational part of the ability to achieve goals in the world". Another AI founder, Marvin Minsky, similarly describes it as "the ability to solve hard problems". The leading AI textbook defines it as the study of agents that perceive
their environment and take actions that maximize their chances of
achieving defined goals. These definitions view intelligence in terms of well-defined problems
with well-defined solutions, where both the difficulty of the problem
and the performance of the program are direct measures of the
"intelligence" of the machine – and no other philosophical discussion is
required, or may not even be possible.
Another definition has been adopted by Google, a major practitioner in the field of AI. This definition stipulates the
ability of systems to synthesize information as the manifestation of
intelligence, similar to the way it is defined in biological
intelligence.
As a result of the many circulating definitions scholars have started to critically analyze and order the AI discourse itself including discussing the many AI narratives and myths to be found within societal, political and academic discourses. Similarly, in practice, some authors have suggested that the term 'AI'
is often used too broadly and vaguely. This raises the question of where
the line should be drawn between AI and classical algorithms, with many companies during the early 2020s AI boom using the term as a marketing buzzword, often even if they did "not actually use AI in a material way".
Aside from philosophical debate, there is also an ongoing legal and
policy debate about how to define AI in regulation and technical
standards without sweeping in non‑AI systems. The International Organization for Standardization
describes an “AI system” as a "an engineered system that generates
outputs such as content, forecasts, recommendations, or decisions for a
given set of human‑defined objectives, and can operate with varying
levels of automation". The EU AI Act
defines an AI system as "a machine-based system that is designed to
operate with varying levels of autonomy and that may exhibit
adaptiveness after deployment, and that, for explicit or implicit
objectives, infers, from the input it receives, how to generate outputs
such as predictions, content, recommendations, or decisions that can
influence physical or virtual environments". In the United States, influential but non‑binding guidance such as the National Institute of Standards and Technology's
AI Risk Management Framework describes an "AI system" as "an engineered
or machine-based system that can, for a given set of objectives,
generate outputs such as predictions, recommendations, or decisions
influencing real or virtual environments. AI systems are designed to
operate with varying levels of autonomy".
These overlapping but different legal and standards definitions
have raised practical questions about how broadly AI should be drawn,
and where to draw the line between AI systems and conventional
algorithms.
Evaluating approaches to AI
No established unifying theory or paradigm has guided AI research for most of its history.[aa]
The unprecedented success of statistical machine learning in the 2010s
eclipsed all other approaches (so much so that some sources, especially
in the business world, use the term "artificial intelligence" to mean
"machine learning with neural networks"). This approach is mostly sub-symbolic, soft and narrow. Critics argue that these questions may have to be revisited by future generations of AI researchers.
Symbolic AI and its limits
Symbolic AI (or "GOFAI") simulated the high-level conscious reasoning that people use when they
solve puzzles, express legal reasoning and do mathematics. They were
highly successful at "intelligent" tasks such as algebra or IQ tests. In
the 1960s, Newell and Simon proposed the physical symbol systems hypothesis: "A physical symbol system has the necessary and sufficient means of general intelligent action."
However, the symbolic approach failed on many tasks that humans solve easily, such as learning, recognizing an object or commonsense reasoning. Moravec's paradox
is the discovery that high-level "intelligent" tasks were easy for AI,
but low level "instinctive" tasks were extremely difficult. Philosopher Hubert Dreyfus had argued
since the 1960s that human expertise depends on unconscious instinct
rather than conscious symbol manipulation, and on having a "feel" for
the situation, rather than explicit symbolic knowledge. Although his arguments had been ridiculed and ignored when they were
first presented, eventually, AI research came to agree with him.
The issue is not resolved: sub-symbolic reasoning can make many of the same inscrutable mistakes that human intuition does, such as algorithmic bias. Critics such as Noam Chomsky argue continuing research into symbolic AI will still be necessary to attain general intelligence, in part because sub-symbolic AI is a move away from explainable AI:
it can be difficult or impossible to understand why a modern
statistical AI program made a particular decision. The emerging field of
neuro-symbolic artificial intelligence attempts to bridge the two approaches.
"Neats" hope that intelligent behavior is described using simple, elegant principles (such as logic, optimization, or neural networks).
"Scruffies" expect that it necessarily requires solving a large number
of unrelated problems. Neats defend their programs with theoretical
rigor, scruffies rely mainly on incremental testing to see if they work.
This issue was actively discussed in the 1970s and 1980s, but eventually was seen as irrelevant. Modern AI has elements of both.
Finding a provably correct or optimal solution is intractable for many important problems. Soft computing is a set of techniques, including genetic algorithms, fuzzy logic
and neural networks, that are tolerant of imprecision, uncertainty,
partial truth and approximation. Soft computing was introduced in the
late 1980s and most successful AI programs in the 21st century are
examples of soft computing with neural networks.
AI researchers are divided as to whether to pursue the goals of artificial general intelligence and superintelligence
directly or to solve as many specific problems as possible (narrow AI)
in hopes these solutions will lead indirectly to the field's long-term
goals. General intelligence is difficult to define and difficult to measure,
and modern AI has had more verifiable successes by focusing on specific
problems with specific solutions. The sub-field of artificial general
intelligence studies this area exclusively.
There is no settled consensus in philosophy of mind on whether a machine can have a mind, consciousness and mental states
in the same sense that human beings do. This issue considers the
internal experiences of the machine, rather than its external behavior.
Mainstream AI research considers this issue irrelevant because it does
not affect the goals of the field: to build machines that can solve
problems using intelligence. Russell and Norvig
add that "[t]he additional project of making a machine conscious in
exactly the way humans are is not one that we are equipped to take on." However, the question has become central to the philosophy of mind. It is also typically the central question at issue in artificial intelligence in fiction.
David Chalmers identified two problems in understanding the mind, which he named the "hard" and "easy" problems of consciousness. The easy problem is understanding how the brain processes signals,
makes plans and controls behavior. The hard problem is explaining how
this feels or why it should feel like anything at all, assuming
we are right in thinking that it truly does feel like something
(Dennett's consciousness illusionism says this is an illusion). While
human information processing is easy to explain, human subjective experience
is difficult to explain. For example, it is easy to imagine a
color-blind person who has learned to identify which objects in their
field of view are red, but it is not clear what would be required for
the person to know what red looks like.
Computationalism is the position in the philosophy of mind
that the human mind is an information processing system and that
thinking is a form of computing. Computationalism argues that the
relationship between mind and body is similar or identical to the
relationship between software and hardware and thus may be a solution to
the mind–body problem.
This philosophical position was inspired by the work of AI researchers
and cognitive scientists in the 1960s and was originally proposed by
philosophers Jerry Fodor and Hilary Putnam.
Philosopher John Searle characterized this position as "strong AI":
"The appropriately programmed computer with the right inputs and
outputs would thereby have a mind in exactly the same sense human beings
have minds." Searle challenges this claim with his Chinese room argument, which attempts to show that even a computer capable of perfectly simulating human behavior would not have a mind.
AI welfare and rights
It is difficult or impossible to reliably evaluate whether an advanced AI is sentient (has the ability to feel), and if so, to what degree. But if there is a significant chance that a given machine can feel and
suffer, then it may be entitled to certain rights or welfare protection
measures, similarly to animals. Sapience (a set of capacities related to high intelligence, such as discernment or self-awareness) may provide another moral basis for AI rights. Robot rights are also sometimes proposed as a practical way to integrate autonomous agents into society.
In 2017, the European Union considered granting "electronic
personhood" to some of the most capable AI systems. Similarly to the
legal status of companies, it would have conferred rights but also
responsibilities. Critics argued in 2018 that granting rights to AI systems would downplay the importance of human rights,
and that legislation should focus on user needs rather than speculative
futuristic scenarios. They also noted that robots lacked the autonomy
to take part in society on their own.
Progress in AI increased interest in the topic. Proponents of AI
welfare and rights often argue that AI sentience, if it emerges, would
be particularly easy to deny. They warn that this may be a moral blind spot analogous to slavery or factory farming, which could lead to large-scale suffering if sentient AI is created and carelessly exploited.
However, technologies cannot improve exponentially indefinitely, and typically follow an S-shaped curve, slowing when they reach the physical limits of what the technology can do.
Robot designer Hans Moravec, cyberneticist Kevin Warwick and inventor Ray Kurzweil have predicted that humans and machines may merge in the future into cyborgs that are more capable and powerful than either. This idea, called transhumanism, has roots in the writings of Aldous Huxley and Robert Ettinger.
Isaac Asimov introduced the Three Laws of Robotics in many stories, most notably with the "Multivac" super-intelligent computer. Asimov's laws are often brought up during lay discussions of machine ethics; while almost all artificial intelligence researchers are familiar with
Asimov's laws through popular culture, they generally consider the laws
useless for many reasons, one of which is their ambiguity.
.jpg)
![Iron-sulfur minerals like greigite catalyse the reduction of carbon dioxide in hydrothermal vents to make Krebs cycle intermediates.[39]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_1.svg)
![Protocells in contact with a thin rock barrier in a hydrothermal vent get a free supply of energy from the pH gradient.[40]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_2.svg)
![Protocells in a hydrothermal vent can grow by adding fatty acids to their membrane, other organics to their cytoplasm.[41]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_3.svg)
![Nucleotides in a protocell in a hydrothermal vent can polymerise into random strings of RNA. Any that have even slight catalytic activity will favour the growth and replication of their protocells, a start to natural selection.[42]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_4.svg)
![A protocell away from a hydrothermal vent must create its own proton-motive force, such as by splitting hydrogen sulfide.[43]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_5.svg)
![Ferredoxin catalyses the splitting of hydrogen sulfide, its earliest repeating amino acid sequence perhaps coded for by an incomplete genetic code.[44]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_6.svg)
![Anoxygenic photosynthesis, using hydrogen sulfide, ended the need for scarce hydrogen.[44]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_7.svg)
![Early heterotrophs used Krebs cycle respiration; then oxygenic photosynthesis gave full independence of volcanic energy.[44]](https://en.wikipedia.org/wiki/File:Abiogenesis_Scenario_8.svg)
_on_Google_Trends.svg)
