The intelligence explosion is a possible outcome of humanity building artificial general intelligence (AGI). AGI would be capable of recursive self-improvement leading to rapid emergence of ASI (artificial superintelligence), the limits of which are unknown. An intelligence explosion would be associated with a technological singularity.
The notion of an "intelligence explosion" was first described by Good (1965), who speculated on the effects of superhuman machines, should they ever be invented:
Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an ‘intelligence explosion,’ and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.
Although technological progress has been accelerating, it has been limited by the basic intelligence of the human brain, which has not, according to Paul R. Ehrlich, changed significantly for millennia. However, with the increasing power of computers and other technologies, it might eventually be possible to build a machine that is more intelligent than humans. If a superhuman intelligence were to be invented—either through the amplification of human intelligence or through artificial intelligence—it would bring to bear greater problem-solving and inventive skills than current humans are capable of. Such an AI is referred to as Seed AI because if an AI were created with engineering capabilities that matched or surpassed those of its human creators, it would have the potential to autonomously improve its own software and hardware or design an even more capable machine. This more capable machine could then go on to design a machine of yet greater capability. These iterations of recursive self-improvement could accelerate, potentially allowing enormous qualitative change before any upper limits imposed by the laws of physics or theoretical computation set in. It is speculated that over many iterations, such an AI would far surpass human cognitive abilities.
Computer scientist Vernor Vinge said in his 1993 essay "The Coming Technological Singularity" that this would signal the end of the human era, as the new superintelligence would continue to upgrade itself and would advance technologically at an incomprehensible rate.
Most proposed methods for creating superhuman or transhuman minds fall into one of two categories: intelligence amplification of human brains and artificial intelligence. The means speculated to produce intelligence augmentation are numerous, and include bioengineering, genetic engineering, nootropic drugs, AI assistants, direct brain–computer interfaces and mind uploading. The existence of multiple paths to an intelligence explosion makes a singularity more likely; for a singularity to not occur they would all have to fail.
Hanson (1998) is skeptical of human intelligence augmentation, writing that once one has exhausted the "low-hanging fruit" of easy methods for increasing human intelligence, further improvements will become increasingly difficult to find. Despite the numerous speculated means for amplifying human intelligence, non-human artificial intelligence (specifically seed AI) is the most popular option for organizations trying to advance the singularity.
Whether or not an intelligence explosion occurs depends on three factors. The first, accelerating factor, is the new intelligence enhancements made possible by each previous improvement. Contrariwise, as the intelligences become more advanced, further advances will become more and more complicated, possibly overcoming the advantage of increased intelligence. Each improvement must be able to beget at least one more improvement, on average, for the singularity to continue. Finally the laws of physics will eventually prevent any further improvements.
There are two logically independent, but mutually reinforcing causes of intelligence improvements: increases in the speed of computation, and improvements to the algorithms used. The former is predicted by Moore’s Law and the forecast improvements in hardware, and is comparatively similar to previous technological advance. On the other hand, most AI researchers[who?] believe that software is more important than hardware.
A 2017 email survey of authors with publications at the 2015 NIPS and ICML machine learning conferences asked them about the chance of an intelligence explosion. 12% said it was "quite likely", 17% said it was "likely", 21% said it was "about even", 24% said it was "unlikely" and 26% said it was "quite unlikely".
Both for human and artificial intelligence, hardware improvements increase the rate of future hardware improvements. Oversimplified, Moore's Law suggests that if the first doubling of speed took 18 months, the second would take 18 subjective months; or 9 external months, whereafter, four months, two months, and so on towards a speed singularity. An upper limit on speed may eventually be reached, although it is unclear how high this would be. Hawkins (2008), responding to Good, argued that the upper limit is relatively low;
Belief in this idea is based on a naive understanding of what intelligence is. As an analogy, imagine we had a computer that could design new computers (chips, systems, and software) faster than itself. Would such a computer lead to infinitely fast computers or even computers that were faster than anything humans could ever build? No. It might accelerate the rate of improvements for a while, but in the end there are limits to how big and fast computers can be. We would end up in the same place; we'd just get there a bit faster. There would be no singularity.
Whereas if it were a lot higher than current human levels of intelligence, the effects of the singularity would be great enough as to be indistinguishable (to humans) from a singularity with an upper limit. For example, if the speed of thought could be increased a million-fold, a subjective year would pass in 30 physical seconds.
It is difficult to directly compare silicon-based hardware with neurons. But Berglas (2008) notes that computer speech recognition is approaching human capabilities, and that this capability seems to require 0.01% of the volume of the brain. This analogy suggests that modern computer hardware is within a few orders of magnitude of being as powerful as the human brain.
Some intelligence technologies, like "seed AI", may also have the potential to make themselves more efficient, not just faster, by modifying their source code. These improvements would make further improvements possible, which would make further improvements possible, and so on.
The mechanism for a recursively self-improving set of algorithms differs from an increase in raw computation speed in two ways. First, it does not require external influence: machines designing faster hardware would still require humans to create the improved hardware, or to program factories appropriately. An AI which was rewriting its own source code, however, could do so while contained in an AI box.
Second, as with Vernor Vinge’s conception of the singularity, it is much harder to predict the outcome. While speed increases seem to be only a quantitative difference from human intelligence, actual algorithm improvements would be qualitatively different. Eliezer Yudkowsky compares it to the changes that human intelligence brought: humans changed the world thousands of times more rapidly than evolution had done, and in totally different ways. Similarly, the evolution of life had been a massive departure and acceleration from the previous geological rates of change, and improved intelligence could cause change to be as different again.
There are substantial dangers associated with an intelligence explosion singularity originating from a recursively self-improving set of algorithms. First, the goal structure of the AI may not be invariant under self-improvement, potentially causing the AI to optimise for something other than was intended. Secondly, AIs could compete for the scarce resources mankind uses to survive.
While not actively malicious, there is no reason to think that AIs would actively promote human goals unless they could be programmed as such, and if not, might use the resources currently used to support mankind to promote its own goals, causing human extinction.
Carl Shulman and Anders Sandberg suggest that algorithm improvements may be the limiting factor for a singularity because whereas hardware efficiency tends to improve at a steady pace, software innovations are more unpredictable and may be bottlenecked by serial, cumulative research. They suggest that in the case of a software-limited singularity, intelligence explosion would actually become more likely than with a hardware-limited singularity, because in the software-limited case, once human-level AI was developed, it could run serially on very fast hardware, and the abundance of cheap hardware would make AI research less constrained. An abundance of accumulated hardware that can be unleashed once the software figures out how to use it has been called "computing overhang."
Dramatic changes in the rate of economic growth have occurred in the past because of some technological advancement. Based on population growth, the economy doubled every 250,000 years from the Paleolithic era until the Neolithic Revolution. The new agricultural economy doubled every 900 years, a remarkable increase. In the current era, beginning with the Industrial Revolution, the world’s economic output doubles every fifteen years, sixty times faster than during the agricultural era. If the rise of superhuman intelligence causes a similar revolution, argues Robin Hanson, one would expect the economy to double at least quarterly and possibly on a weekly basis.
A superintelligence, hyperintelligence, or superhuman intelligence is a hypothetical agent that possesses intelligence far surpassing that of the brightest and most gifted human minds. "Superintelligence" may also refer to the form or degree of intelligence possessed by such an agent.
Technology forecasters and researchers disagree about if or when human intelligence is likely to be surpassed. Some argue that advances in artificial intelligence (AI) will probably result in general reasoning systems that lack human cognitive limitations. Others believe that humans will evolve or directly modify their biology so as to achieve radically greater intelligence. A number of futures studies scenarios combine elements from both of these possibilities, suggesting that humans are likely to interface with computers, or upload their minds to computers, in a way that enables substantial intelligence amplification.
Berglas (2008) notes that there is no direct evolutionary motivation for an AI to be friendly to humans. Evolution has no inherent tendency to produce outcomes valued by humans, and there is little reason to expect an arbitrary optimisation process to promote an outcome desired by mankind, rather than inadvertently leading to an AI behaving in a way not intended by its creators (such as Nick Bostrom's whimsical example of an AI which was originally programmed with the goal of manufacturing paper clips, so that when it achieves superintelligence it decides to convert the entire planet into a paper clip manufacturing facility). Anders Sandberg has also elaborated on this scenario, addressing various common counter-arguments. AI researcher Hugo de Garis suggests that artificial intelligences may simply eliminate the human race for access to scarce resources, and humans would be powerless to stop them. Alternatively, AIs developed under evolutionary pressure to promote their own survival could outcompete humanity.
Bostrom (2002) discusses human extinction scenarios, and lists superintelligence as a possible cause:
When we create the first superintelligent entity, we might make a mistake and give it goals that lead it to annihilate humankind, assuming its enormous intellectual advantage gives it the power to do so. For example, we could mistakenly elevate a subgoal to the status of a supergoal. We tell it to solve a mathematical problem, and it complies by turning all the matter in the solar system into a giant calculating device, in the process killing the person who asked the question.
A significant problem is that unfriendly artificial intelligence is likely to be much easier to create than friendly AI. While both require large advances in recursive optimisation process design, friendly AI also requires the ability to make goal structures invariant under self-improvement (or the AI could transform itself into something unfriendly) and a goal structure that aligns with human values and does not automatically destroy the human race. An unfriendly AI, on the other hand, can optimize for an arbitrary goal structure, which does not need to be invariant under self-modification.
Eliezer Yudkowsky proposed that research be undertaken to produce friendly artificial intelligence in order to address the dangers. He noted that the first real AI would have a head start on self-improvement and, if friendly, could prevent unfriendly AIs from developing, as well as providing enormous benefits to mankind.
Bill Hibbard (2014) proposes an AI design that avoids several dangers including self-delusion, unintended instrumental actions, and corruption of the reward generator. He also discusses social impacts of AI and testing AI. His 2001 book Super-Intelligent Machines advocates the need for public education about AI and public control over AI. It also proposed a simple design that was vulnerable to corruption of the reward generator.
One hypothetical approach towards attempting to control an artificial intelligence is an AI box, where the artificial intelligence is kept constrained inside a simulated world and not allowed to affect the external world. However, a sufficiently intelligent AI may simply be able to escape by outsmarting its less intelligent human captors.
Stephen Hawking said in 2014 that "Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last, unless we learn how to avoid the risks." Hawking believes that in the coming decades, AI could offer "incalculable benefits and risks" such as "technology outsmarting financial markets, out-inventing human researchers, out-manipulating human leaders, and developing weapons we cannot even understand." Hawking believes more should be done to prepare for the singularity:
So, facing possible futures of incalculable benefits and risks, the experts are surely doing everything possible to ensure the best outcome, right? Wrong. If a superior alien civilisation sent us a message saying, "We'll arrive in a few decades," would we just reply, "OK, call us when you get here – we'll leave the lights on"? Probably not – but this is more or less what is happening with AI.
Hard vs. soft takeoff
In a hard takeoff scenario, an AGI rapidly self-improves, "taking control" of the world (perhaps in a matter of hours), too quickly for significant human-initiated error correction or for a gradual tuning of the AGI's goals. In a soft takeoff scenario, AGI still becomes far more powerful than humanity, but at a human-like pace (perhaps on the order of decades), on a timescale where ongoing human interaction and correction can effectively steer the AGI's development.
Ramez Naam argues against a hard takeoff by pointing out that we already see recursive self-improvement by superintelligences, such as corporations. For instance, Intel has "the collective brainpower of tens of thousands of humans and probably millions of CPU cores to.. design better CPUs!" However, this has not led to a hard takeoff; rather, it has led to a soft takeoff in the form of Moore's law. Naam further points out that the computational complexity of higher intelligence may be much greater than linear, such that "creating a mind of intelligence 2 is probably more than twice as hard as creating a mind of intelligence 1."
J. Storrs Hall believes that "many of the more commonly seen scenarios for overnight hard takeoff are circular – they seem to assume hyperhuman capabilities at the starting point of the self-improvement process" in order for an AI to be able to make the dramatic, domain-general improvements required for takeoff. Hall suggests that rather than recursively self-improving its hardware, software, and infrastructure all on its own, a fledgling AI would be better off specializing in one area where it was most effective and then buying the remaining components on the marketplace, because the quality of products on the marketplace continually improves, and the AI would have a hard time keeping up with the cutting-edge technology used by the rest of the world. Ben Goertzel agrees with Hall's suggestion that a new human-level AI would do well to use its intelligence to accumulate wealth. The AI's talents might inspire companies and governments to disperse its software throughout society. Goertzel is skeptical of a very hard, 5-minute takeoff but thinks a takeoff from human to superhuman level on the order of 5 years is reasonable. He calls this a "semihard takeoff".
Max More disagrees, arguing that if there were only a few superfast human-level AIs, they wouldn't radically change the world, because they would still depend on other people to get things done and would still have human cognitive constraints. Even if all superfast AIs worked on intelligence augmentation, it's not clear why they would do better in a discontinuous way than existing human cognitive scientists at producing super-human intelligence, although the rate of progress would increase. More also argues that a superintelligence would not transform the world overnight, because a superintelligence would need to engage with existing, slow human systems to accomplish physical impacts on the world. "The need for collaboration, for organization, and for putting ideas into physical changes will ensure that all the old rules are not thrown out overnight or even within years."
- Accelerating change
- Artificial consciousness
- Flynn effect
- Human intelligence § Improving intelligence
- Outline of transhumanism
- Postbiological evolution
- Robot learning
- Ehrlich, Paul. The Dominant Animal: Human Evolution and the Environment
- Superbrains born of silicon will change everything. Archived August 1, 2010, at the Wayback Machine.
- Yampolskiy, Roman V. "Analysis of types of self-improving software." Artificial General Intelligence. Springer International Publishing, 2015. 384-393.
- Eliezer Yudkowsky. General Intelligence and Seed AI-Creating Complete Minds Capable of Open-Ended Self-Improvement, 2001
- Vinge, Vernor. "The Coming Technological Singularity: How to Survive in the Post-Human Era", in Vision-21: Interdisciplinary Science and Engineering in the Era of Cyberspace, G. A. Landis, ed., NASA Publication CP-10129, pp. 11–22, 1993.
- "What is the Singularity? | Singularity Institute for Artificial Intelligence". Singinst.org. Archived from the original on 2011-09-08. Retrieved 2011-09-09.
- David Chalmers John Locke Lecture, 10 May, Exam Schools, Oxford, Presenting a philosophical analysis of the possibility of a technological singularity or "intelligence explosion" resulting from recursively self-improving AI Archived 2013-01-15 at the Wayback Machine..
- The Singularity: A Philosophical Analysis, David J. Chalmers
- "ITRS" (PDF). Archived from the original (PDF) on 2011-09-29. Retrieved 2011-09-09.
- Grace, Katja; Salvatier, John; Dafoe, Allan; Zhang, Baobao; Evans, Owain (24 May 2017). "When Will AI Exceed Human Performance? Evidence from AI Experts". arXiv: [cs.AI].
- Siracusa, John (2009-08-31). "Mac OS X 10.6 Snow Leopard: the Ars Technica review". Arstechnica.com. Retrieved 2011-09-09.
- Eliezer Yudkowsky, 1996 "Staring into the Singularity"
- Eliezer S. Yudkowsky. "Power of Intelligence". Yudkowsky. Retrieved 2011-09-09.
- Omohundro, Stephen M., "The Basic AI Drives." Artificial General Intelligence, 2008 proceedings of the First AGI Conference, eds. Pei Wang, Ben Goertzel, and Stan Franklin. Vol. 171. Amsterdam: IOS, 2008
- "Artificial General Intelligence: Now Is the Time". KurzweilAI. Retrieved 2011-09-09.
- Omohundro, Stephen M., "The Nature of Self-Improving Artificial Intelligence." Self-Aware Systems. 21 Jan. 2008. Web. 07 Jan. 2010.
- Barrat, James (2013). "6, "Four Basic Drives"". Our Final Invention (First ed.). New York: St. Martin's Press. pp. 78–98. ISBN 978-0312622374.
- "Max More and Ray Kurzweil on the Singularity". KurzweilAI. Retrieved 2011-09-09.
- "Concise Summary | Singularity Institute for Artificial Intelligence". Singinst.org. Retrieved 2011-09-09.
- Bostrom, Nick, The Future of Human Evolution, Death and Anti-Death: Two Hundred Years After Kant, Fifty Years After Turing, ed. Charles Tandy, pp. 339–371, 2004, Ria University Press.
- Shulman, Carl; Anders Sandberg (2010). Mainzer, Klaus, ed. "Implications of a Software-Limited Singularity" (PDF). ECAP10: VIII European Conference on Computing and Philosophy. Retrieved 17 May 2014.
- Muehlhauser, Luke; Anna Salamon (2012). "Intelligence Explosion: Evidence and Import". In Amnon Eden; Johnny Søraker; James H. Moor; Eric Steinhart. Singularity Hypotheses: A Scientific and Philosophical Assessment (PDF). Springer.
- Robin Hanson, "Economics Of The Singularity", IEEE Spectrum Special Report: The Singularity, retrieved 2008-09-11 & Long-Term Growth As A Sequence of Exponential Modes
- Nick Bostrom, "Ethical Issues in Advanced Artificial Intelligence", in Cognitive, Emotive and Ethical Aspects of Decision Making in Humans and in Artificial Intelligence, Vol. 2, ed. I. Smit et al., Int. Institute of Advanced Studies in Systems Research and Cybernetics, 2003, pp. 12–17
- Eliezer Yudkowsky: Artificial Intelligence as a Positive and Negative Factor in Global Risk Archived 2012-06-11 at the Wayback Machine.. Draft for a publication in Global Catastrophic Risk from August 31, 2006, retrieved July 18, 2011 (PDF file)
- The Stamp Collecting Device, Nick Hay
- 'Why we should fear the Paperclipper', 2011-02-14 entry of Sandberg's blog 'Andart'
- Omohundro, Stephen M., "The Basic AI Drives." Artificial General Intelligence, 2008 proceedings of the First AGI Conference, eds. Pei Wang, Ben Goertzel, and Stan Franklin. Vol. 171. Amsterdam: IOS, 2008.
- de Garis, Hugo. "The Coming Artilect War", Forbes.com, 22 June 2009.
- Coherent Extrapolated Volition, Eliezer S. Yudkowsky, May 2004 Archived 2010-08-15 at the Wayback Machine.
- Hibbard, Bill (2012), "Model-Based Utility Functions", Journal of Artificial General Intelligence, 3: 1, arXiv: , Bibcode:2012JAGI....3....1H, doi:10.2478/v10229-011-0013-5.
- Avoiding Unintended AI Behaviors. Bill Hibbard. 2012 proceedings of the Fifth Conference on Artificial General Intelligence, eds. Joscha Bach, Ben Goertzel and Matthew Ikle. This paper won the Machine Intelligence Research Institute's 2012 Turing Prize for the Best AGI Safety Paper.
- Hibbard, Bill (2008), "The Technology of Mind and a New Social Contract", Journal of Evolution and Technology, 17.
- Decision Support for Safe AI Design|. Bill Hibbard. 2012 proceedings of the Fifth Conference on Artificial General Intelligence, eds. Joscha Bach, Ben Goertzel and Matthew Ikle.
- Yudkowsky, Eliezer (2008), Bostrom, Nick; Cirkovic, Milan, eds., "Artificial Intelligence as a Positive and Negative Factor in Global Risk" (PDF), Global Catastrophic Risks, Oxford University Press: 303, Bibcode:2008gcr..book..303Y, ISBN 978-0-19-857050-9, archived from the original (PDF) on 2008-08-07
- Artificial Intelligence Will Kill Our Grandchildren (Singularity), Dr Anthony Berglas
- The Singularity: A Philosophical Analysis David J. Chalmers
- Stephen Hawking (1 May 2014). "Stephen Hawking: 'Transcendence looks at the implications of artificial intelligence - but are we taking AI seriously enough?'". The Independent. Retrieved May 5, 2014.
- Eliezer Yudkowsky. "Artificial intelligence as a positive and negative factor in global risk." Global catastrophic risks (2008).
- Bugaj, Stephan Vladimir, and Ben Goertzel. "Five ethical imperatives and their implications for human-AGI interaction." Dynamical Psychology (2007).
- Sotala, Kaj, and Roman V. Yampolskiy. "Responses to catastrophic AGI risk: a survey." Physica Scripta 90.1 (2014): 018001.
- Naam, Ramez (2014). "The Singularity Is Further Than It Appears". Retrieved 16 May 2014.
- Naam, Ramez (2014). "Why AIs Won't Ascend in the Blink of an Eye - Some Math". Retrieved 16 May 2014.
- Hall, J. Storrs (2008). "Engineering Utopia" (PDF). Artificial General Intelligence, 2008: Proceedings of the First AGI Conference: 460–467. Retrieved 16 May 2014.
- Goertzel, Ben (26 Sep 2014). "Superintelligence — Semi-hard Takeoff Scenarios". h+ Magazine. Retrieved 25 October 2014.
- More, Max. "Singularity Meets Economy". Retrieved 10 November 2014.
- Good, I. J. (1965), Franz L. Alt; Morris Rubinoff, eds., "Speculations Concerning the First Ultraintelligent Machine", Advances in Computers, Advances in Computers, Academic Press, 6: 31–88, doi:10.1016/S0065-2458(08)60418-0, ISBN 9780120121069, archived from the original on 2001-05-27, retrieved 2007-08-07
- Hanson, Robin (1998), Some Skepticism, Robin Hanson, archived from the original on 2009-08-28, retrieved 2009-06-19
- Berglas, Anthony (2008), Artificial Intelligence will Kill our Grandchildren, retrieved 2008-06-13
- Bostrom, Nick (2002), "Existential Risks", Journal of Evolution and Technology, 9, retrieved 2007-08-07
- Hibbard, Bill (5 November 2014). "Ethical Artificial Intelligence". arXiv: [cs.AI].