AI safety

From Wikipedia, the free encyclopedia

AI safety is an interdisciplinary field concerned with preventing accidents, misuse, or other harmful consequences that could result from artificial intelligence (AI) systems. It encompasses machine ethics and AI alignment (which aim to make AI systems moral and beneficial), monitoring AI systems for risks and making them highly reliable. Beyond AI research, it involves developing norms and policies that promote safety.


Scholars discuss current risks from critical systems failures,[1] bias,[2] and AI-enabled surveillance,[3] as well as emerging risks like technological unemployment, digital manipulation,[4] weaponization,[5] AI-enabled cyberattacks[6] and bioterrorism.[7] They also discuss speculative risks from losing control of future artificial general intelligence (AGI) agents,[8] or from AI enabling perpetually stable dictatorships.[9]

Existential safety[edit]

Some ways in which an advanced misaligned AI could try to gain more power.[10] Power-seeking behaviors may arise because power is useful to accomplish virtually any objective[11] (see instrumental convergence).

AI safety has a significant focus on existential risks and other large-scale risks. The part of AI safety aiming to address concerns of existential risk is sometimes called AI existential safety.[12]

Some have criticized concerns about AGI, such as Andrew Ng who compared them in 2015 to "worrying about overpopulation on Mars when we have not even set foot on the planet yet."[13] Stuart J. Russell on the other side urges caution, arguing that "it is better to anticipate human ingenuity than to underestimate it."[14]

AI researchers have widely differing opinions about the severity and primary sources of risk posed by AI technology[15][16][17] – though surveys suggest that experts take high consequence risks seriously. In two surveys of AI researchers, the median respondent was optimistic about AI overall, but placed a 5% probability on an "extremely bad (e.g. human extinction)" outcome of advanced AI.[15] In a 2022 survey of the natural language processing community, 37% agreed or weakly agreed that it is plausible that AI decisions could lead to a catastrophe that is "at least as bad as an all-out nuclear war."[18]


Risks from AI began to be seriously discussed at the start of the computer age:

Moreover, if we move in the direction of making machines which learn and whose behavior is modified by experience, we must face the fact that every degree of independence we give the machine is a degree of possible defiance of our wishes.

— Norbert Wiener (1949)[19]

From 2008 to 2009, the Association for the Advancement of Artificial Intelligence (AAAI) commissioned a study to explore and address potential long-term societal influences of AI research and development. The panel was generally skeptical of the radical views expressed by science-fiction authors but agreed that "additional research would be valuable on methods for understanding and verifying the range of behaviors of complex computational systems to minimize unexpected outcomes."[20]

In 2011, Roman Yampolskiy introduced the term "AI safety engineering"[21] at the Philosophy and Theory of Artificial Intelligence conference,[22] listing prior failures of AI systems and arguing that "the frequency and seriousness of such events will steadily increase as AIs become more capable."[23]

In 2014, philosopher Nick Bostrom published the book Superintelligence: Paths, Dangers, Strategies. He has the opinion that the rise of AGI has the potential to create various societal issues, ranging from the displacement of the workforce by AI, manipulation of political and military structures, to even the possibility of human extinction.[24] His argument that future advanced systems may pose a threat to human existence prompted Elon Musk,[25] Bill Gates,[26] and Stephen Hawking[27] to voice similar concerns.

In 2015, dozens of artificial intelligence experts signed an open letter on artificial intelligence calling for research on the societal impacts of AI and outlining concrete directions.[28] To date, the letter has been signed by over 8000 people including Yann LeCun, Shane Legg, Yoshua Bengio, and Stuart Russell.

In the same year, a group of academics led by professor Stuart Russell founded the Center for Human-Compatible AI at the University of California Berkeley and the Future of Life Institute awarded $6.5 million in grants for research aimed at "ensuring artificial intelligence (AI) remains safe, ethical and beneficial."[29]

In 2016, the White House Office of Science and Technology Policy and Carnegie Mellon University announced The Public Workshop on Safety and Control for Artificial Intelligence,[30] which was one of a sequence of four White House workshops aimed at investigating "the advantages and drawbacks" of AI.[31] In the same year, Concrete Problems in AI Safety – one of the first and most influential technical AI Safety agendas – was published.[32]

In 2017, the Future of Life Institute sponsored the Asilomar Conference on Beneficial AI, where more than 100 thought leaders formulated principles for beneficial AI including "Race Avoidance: Teams developing AI systems should actively cooperate to avoid corner-cutting on safety standards."[33]

In 2018, the DeepMind Safety team outlined AI safety problems in specification, robustness, and assurance.[34] The following year, researchers organized a workshop at ICLR that focused on these problem areas.[35]

In 2021, Unsolved Problems in ML Safety was published, outlining research directions in robustness, monitoring, alignment, and systemic safety.[36]

In 2023, Rishi Sunak said he wants the United Kingdom to be the "geographical home of global AI safety regulation" and to host the first global summit on AI safety.[37] The AI safety summit took place in November 2023, and focused on the risks of misuse and loss of control associated with frontier AI models.[38]

In 2024, The US and UK forged a new partnership on the science of AI safety. The MoU was signed on 1 April 2024 by US commerce secretary Gina Raimondo and UK technology secretary Michelle Donelan to jointly develop advanced AI model testing, following commitments announced at an AI Safety Summit in Bletchley Park in November.[39]

Research focus[edit]

AI safety research areas include robustness, monitoring, and alignment.[36][34]


Adversarial robustness[edit]

AI systems are often vulnerable to adversarial examples or "inputs to machine learning (ML) models that an attacker has intentionally designed to cause the model to make a mistake".[40] For example, in 2013, Szegedy et al. discovered that adding specific imperceptible perturbations to an image could cause it to be misclassified with high confidence.[41] This continues to be an issue with neural networks, though in recent work the perturbations are generally large enough to be perceptible.[42][43][44]

Carefully crafted noise can be added to an image to cause it to be misclassified with high confidence.

All of the images on the right are predicted to be an ostrich after the perturbation is applied. (Left) is a correctly predicted sample, (center) perturbation applied magnified by 10x, (right) adversarial example.[41]

Adversarial robustness is often associated with security.[45] Researchers demonstrated that an audio signal could be imperceptibly modified so that speech-to-text systems transcribe it to any message the attacker chooses.[46] Network intrusion[47] and malware[48] detection systems also must be adversarially robust since attackers may design their attacks to fool detectors.

Models that represent objectives (reward models) must also be adversarially robust. For example, a reward model might estimate how helpful a text response is and a language model might be trained to maximize this score.[49] Researchers have shown that if a language model is trained for long enough, it will leverage the vulnerabilities of the reward model to achieve a better score and perform worse on the intended task.[50] This issue can be addressed by improving the adversarial robustness of the reward model.[51] More generally, any AI system used to evaluate another AI system must be adversarially robust. This could include monitoring tools, since they could also potentially be tampered with to produce a higher reward.[52]


Estimating uncertainty[edit]

It is often important for human operators to gauge how much they should trust an AI system, especially in high-stakes settings such as medical diagnosis.[53] ML models generally express confidence by outputting probabilities; however, they are often overconfident,[54] especially in situations that differ from those that they were trained to handle.[55] Calibration research aims to make model probabilities correspond as closely as possible to the true proportion that the model is correct.

Similarly, anomaly detection or out-of-distribution (OOD) detection aims to identify when an AI system is in an unusual situation. For example, if a sensor on an autonomous vehicle is malfunctioning, or it encounters challenging terrain, it should alert the driver to take control or pull over.[56] Anomaly detection has been implemented by simply training a classifier to distinguish anomalous and non-anomalous inputs,[57] though a range of additional techniques are in use.[58][59]

Detecting malicious use[edit]

Scholars[5] and government agencies have expressed concerns that AI systems could be used to help malicious actors to build weapons,[60] manipulate public opinion,[61][62] or automate cyber attacks.[63] These worries are a practical concern for companies like OpenAI which host powerful AI tools online.[64] In order to prevent misuse, OpenAI has built detection systems that flag or restrict users based on their activity.[65]


Neural networks have often been described as black boxes,[66] meaning that it is difficult to understand why they make the decisions they do as a result of the massive number of computations they perform.[67] This makes it challenging to anticipate failures. In 2018, a self-driving car killed a pedestrian after failing to identify them. Due to the black box nature of the AI software, the reason for the failure remains unclear.[68] It also raises debates in healthcare over whether statistically efficient but opaque models should be used.[69]

One critical benefit of transparency is explainability.[70] It is sometimes a legal requirement to provide an explanation for why a decision was made in order to ensure fairness, for example for automatically filtering job applications or credit score assignment.[70]

Another benefit is to reveal the cause of failures.[66] At the beginning of the 2020 COVID-19 pandemic, researchers used transparency tools to show that medical image classifiers were 'paying attention' to irrelevant hospital labels.[71]

Transparency techniques can also be used to correct errors. For example, in the paper "Locating and Editing Factual Associations in GPT," the authors were able to identify model parameters that influenced how it answered questions about the location of the Eiffel tower. They were then able to 'edit' this knowledge to make the model respond to questions as if it believed the tower was in Rome instead of France.[72] Though in this case, the authors induced an error, these methods could potentially be used to efficiently fix them. Model editing techniques also exist in computer vision.[73]

Finally, some have argued that the opaqueness of AI systems is a significant source of risk and better understanding of how they function could prevent high-consequence failures in the future.[74] "Inner" interpretability research aims to make ML models less opaque. One goal of this research is to identify what the internal neuron activations represent.[75][76] For example, researchers identified a neuron in the CLIP artificial intelligence system that responds to images of people in spider man costumes, sketches of spiderman, and the word 'spider.'[77] It also involves explaining connections between these neurons or 'circuits'.[78][79] For example, researchers have identified pattern-matching mechanisms in transformer attention that may play a role in how language models learn from their context.[80] "Inner interpretability" has been compared to neuroscience. In both cases, the goal is to understand what is going on in an intricate system, though ML researchers have the benefit of being able to take perfect measurements and perform arbitrary ablations.[81]

Detecting trojans[edit]

ML models can potentially contain 'trojans' or 'backdoors': vulnerabilities that malicious actors maliciously build into an AI system. For example, a trojaned facial recognition system could grant access when a specific piece of jewelry is in view;[36] or a trojaned autonomous vehicle may function normally until a specific trigger is visible.[82] Note that an adversary must have access to the system's training data in order to plant a trojan. [citation needed] This might not be difficult to do with some large models like CLIP or GPT-3 as they are trained on publicly available internet data.[83] Researchers were able to plant a trojan in an image classifier by changing just 300 out of 3 million of the training images.[84] In addition to posing a security risk, researchers have argued that trojans provide a concrete setting for testing and developing better monitoring tools.[52]


In the field of artificial intelligence (AI), AI alignment research aims to steer AI systems toward a person's or group's intended goals, preferences, and ethical principles. An AI system is considered aligned if it advances its intended objectives. A misaligned AI system may pursue some objectives, but not the intended ones.[85]

It is often challenging for AI designers to align an AI system due to the difficulty of specifying the full range of desired and undesired behaviors. To aid them, they often use simpler proxy goals, such as gaining human approval. But that approach can create loopholes, overlook necessary constraints, or reward the AI system for merely appearing aligned.[85][86]

Misaligned AI systems can malfunction and cause harm. AI systems may find loopholes that allow them to accomplish their proxy goals efficiently but in unintended, sometimes harmful, ways (reward hacking).[85][87][88] They may also develop unwanted instrumental strategies, such as seeking power or survival because such strategies help them achieve their final given goals.[85][89][90] Furthermore, they may develop undesirable emergent goals that may be hard to detect before the system is deployed and encounters new situations and data distributions.[91][92]

Today, these problems affect existing commercial systems such as language models,[93][94][95] robots,[96] autonomous vehicles,[97] and social media recommendation engines.[93][90][98] Some AI researchers argue that more capable future systems will be more severely affected, since these problems partially result from the systems being highly capable.[99][87][86]

Many of the most-cited AI scientists,[100][101][102] including Geoffrey Hinton, Yoshua Bengio, and Stuart Russell, argue that AI is approaching human-like (AGI) and superhuman cognitive capabilities (ASI) and could endanger human civilization if misaligned.[103][90]

AI alignment is a subfield of AI safety, the study of how to build safe AI systems.[104] Other subfields of AI safety include robustness, monitoring, and capability control.[105] Research challenges in alignment include instilling complex values in AI, developing honest AI, scalable oversight, auditing and interpreting AI models, and preventing emergent AI behaviors like power-seeking.[105] Alignment research has connections to interpretability research,[106][107] (adversarial) robustness,[104] anomaly detection, calibrated uncertainty,[106] formal verification,[108] preference learning,[109][110][111] safety-critical engineering,[112] game theory,[113] algorithmic fairness,[104][114] and social sciences.[115]

Systemic safety and sociotechnical factors[edit]

It is common for AI risks (and technological risks more generally) to be categorized as misuse or accidents.[116] Some scholars have suggested that this framework falls short.[116] For example, the Cuban Missile Crisis was not clearly an accident or a misuse of technology.[116] Policy analysts Zwetsloot and Dafoe wrote, "The misuse and accident perspectives tend to focus only on the last step in a causal chain leading up to a harm: that is, the person who misused the technology, or the system that behaved in unintended ways… Often, though, the relevant causal chain is much longer." Risks often arise from 'structural' or 'systemic' factors such as competitive pressures, diffusion of harms, fast-paced development, high levels of uncertainty, and inadequate safety culture.[116] In the broader context of safety engineering, structural factors like 'organizational safety culture' play a central role in the popular STAMP risk analysis framework.[117]

Inspired by the structural perspective, some researchers have emphasized the importance of using machine learning to improve sociotechnical safety factors, for example, using ML for cyber defense, improving institutional decision-making, and facilitating cooperation.[36]

Cyber defense[edit]

Some scholars are concerned that AI will exacerbate the already imbalanced game between cyber attackers and cyber defenders.[118] This would increase 'first strike' incentives and could lead to more aggressive and destabilizing attacks. In order to mitigate this risk, some have advocated for an increased emphasis on cyber defense. In addition, software security is essential for preventing powerful AI models from being stolen and misused.[5]

Improving institutional decision-making[edit]

The advancement of AI in economic and military domains could precipitate unprecedented political challenges.[119] Some scholars have compared AI race dynamics to the cold war, where the careful judgment of a small number of decision-makers often spelled the difference between stability and catastrophe.[120] AI researchers have argued that AI technologies could also be used to assist decision-making.[36] For example, researchers are beginning to develop AI forecasting[121] and advisory systems.[122]

Facilitating cooperation[edit]

Many of the largest global threats (nuclear war,[123] climate change,[124] etc.) have been framed as cooperation challenges. As in the well-known prisoner's dilemma scenario, some dynamics may lead to poor results for all players, even when they are optimally acting in their self-interest. For example, no single actor has strong incentives to address climate change even though the consequences may be significant if no one intervenes.[124]

A salient AI cooperation challenge is avoiding a 'race to the bottom'.[125] In this scenario, countries or companies race to build more capable AI systems and neglect safety, leading to a catastrophic accident that harms everyone involved. Concerns about scenarios like these have inspired both political[126] and technical[127] efforts to facilitate cooperation between humans, and potentially also between AI systems. Most AI research focuses on designing individual agents to serve isolated functions (often in 'single-player' games).[128] Scholars have suggested that as AI systems become more autonomous, it may become essential to study and shape the way they interact.[128]

Challenges of Large Language Models[edit]

In recent years, the development of large language models (LMs) has raised unique concerns within the field of AI safety. Researchers Bender and Gebru et al.[129] have highlighted the environmental and financial costs associated with training these models, emphasizing that the energy consumption and carbon footprint of training procedures like those for Transformer models can be substantial. Moreover, these models often rely on massive, uncurated Internet-based datasets, which can encode hegemonic and biased viewpoints, further marginalizing underrepresented groups. The large-scale training data, while vast, does not guarantee diversity and often reflects the worldviews of privileged demographics, leading to models that perpetuate existing biases and stereotypes. This situation is exacerbated by the tendency of these models to produce seemingly coherent and fluent text, which can mislead users into attributing meaning and intent where none exists, a phenomenon described as 'stochastic parrots.' These models, therefore, pose risks of amplifying societal biases, spreading misinformation, and being used for malicious purposes, such as generating extremist propaganda or deepfakes. To address these challenges, researchers advocate for more careful planning in dataset creation and system development, emphasizing the need for research projects that contribute positively towards an equitable technological ecosystem.[130][131]

In governance[edit]

The AI Safety Summit of November 2023.[132]

AI governance is broadly concerned with creating norms, standards, and regulations to guide the use and development of AI systems.[120]


AI safety governance research ranges from foundational investigations into the potential impacts of AI to specific applications. On the foundational side, researchers have argued that AI could transform many aspects of society due to its broad applicability, comparing it to electricity and the steam engine.[133] Some work has focused on anticipating specific risks that may arise from these impacts – for example, risks from mass unemployment,[134] weaponization,[135] disinformation,[136] surveillance,[137] and the concentration of power.[138] Other work explores underlying risk factors such as the difficulty of monitoring the rapidly evolving AI industry,[139] the availability of AI models,[140] and 'race to the bottom' dynamics.[125][141] Allan Dafoe, the head of longterm governance and strategy at DeepMind has emphasized the dangers of racing and the potential need for cooperation: "it may be close to a necessary and sufficient condition for AI safety and alignment that there be a high degree of caution prior to deploying advanced powerful systems; however, if actors are competing in a domain with large returns to first-movers or relative advantage, then they will be pressured to choose a sub-optimal level of caution.".[126] A research stream focuses on developing approaches, frameworks, and methods to assess AI accountability, guiding and promoting audits of AI-based systems.[142][143][144]

Scaling Local AI Safety Measures to Global Solutions[edit]

In addressing the AI safety problem it is important to stress the distinction between local and global solutions. Local solutions focus on individual AI systems, ensuring they are safe and beneficial, while global solutions seek to implement safety measures for all AI systems across various jurisdictions. Some researchers[145] argue for the necessity of scaling local safety measures to a global level, proposing a classification for these global solutions. This approach underscores the importance of collaborative efforts in the international governance of AI safety, emphasizing that no single entity can effectively manage the risks associated with AI technologies. This perspective aligns with ongoing efforts in international policy-making and regulatory frameworks, which aim to address the complex challenges posed by advanced AI systems worldwide.[146][147]

Government action[edit]

Some experts have argued that it is too early to regulate AI, expressing concerns that regulations will hamper innovation and it would be foolish to "rush to regulate in ignorance."[148][149] Others, such as business magnate Elon Musk, call for pre-emptive action to mitigate catastrophic risks.[150]

Outside of formal legislation, government agencies have put forward ethical and safety recommendations. In March 2021, the US National Security Commission on Artificial Intelligence reported that advances in AI may make it increasingly important to "assure that systems are aligned with goals and values, including safety, robustness and trustworthiness."[151] Subsequently, the National Institute of Standards and Technology drafted a framework for managing AI Risk, which advises that when "catastrophic risks are present – development and deployment should cease in a safe manner until risks can be sufficiently managed."[152]

In September 2021, the People's Republic of China published ethical guidelines for the use of AI in China, emphasizing that AI decisions should remain under human control and calling for accountability mechanisms. In the same month, The United Kingdom published its 10-year National AI Strategy,[153] which states the British government "takes the long-term risk of non-aligned Artificial General Intelligence, and the unforeseeable changes that it would mean for ... the world, seriously."[154] The strategy describes actions to assess long-term AI risks, including catastrophic risks.[154] The British government held first major global summit on AI safety. This took place on the 1st and 2 November 2023 and was described as "an opportunity for policymakers and world leaders to consider the immediate and future risks of AI and how these risks can be mitigated via a globally coordinated approach."[155][156]

Government organizations, particularly in the United States, have also encouraged the development of technical AI safety research. The Intelligence Advanced Research Projects Activity initiated the TrojAI project to identify and protect against Trojan attacks on AI systems.[157] The DARPA engages in research on explainable artificial intelligence and improving robustness against adversarial attacks.[158][159] And the National Science Foundation supports the Center for Trustworthy Machine Learning, and is providing millions of dollars in funding for empirical AI safety research.[160]

In 2024, the United Nations General Assembly adopted the first global resolution on the promotion of “safe, secure and trustworthy” AI systems that emphasized the respect, protection and promotion of human rights in the design, development, deployment and the use of AI.[161]

Corporate self-regulation[edit]

AI labs and companies generally abide by safety practices and norms that fall outside of formal legislation.[162] One aim of governance researchers is to shape these norms. Examples of safety recommendations found in the literature include performing third-party auditing,[163] offering bounties for finding failures,[163] sharing AI incidents[163] (an AI incident database was created for this purpose),[164] following guidelines to determine whether to publish research or models,[140] and improving information and cyber security in AI labs.[165]

Companies have also made commitments. Cohere, OpenAI, and AI21 proposed and agreed on "best practices for deploying language models," focusing on mitigating misuse.[166] To avoid contributing to racing-dynamics, OpenAI has also stated in their charter that "if a value-aligned, safety-conscious project comes close to building AGI before we do, we commit to stop competing with and start assisting this project"[167] Also, industry leaders such as CEO of DeepMind Demis Hassabis, director of Facebook AI Yann LeCun have signed open letters such as the Asilomar Principles[33] and the Autonomous Weapons Open Letter.[168]

See also[edit]


  1. ^ De-Arteaga, Maria (2020-05-13). Machine Learning in High-Stakes Settings: Risks and Opportunities (PhD). Carnegie Mellon University.
  2. ^ Mehrabi, Ninareh; Morstatter, Fred; Saxena, Nripsuta; Lerman, Kristina; Galstyan, Aram (2021). "A Survey on Bias and Fairness in Machine Learning". ACM Computing Surveys. 54 (6): 1–35. arXiv:1908.09635. doi:10.1145/3457607. ISSN 0360-0300. S2CID 201666566. Archived from the original on 2022-11-23. Retrieved 2022-11-28.
  3. ^ Feldstein, Steven (2019). The Global Expansion of AI Surveillance (Report). Carnegie Endowment for International Peace.
  4. ^ Barnes, Beth (2021). "Risks from AI persuasion". Lesswrong. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  5. ^ a b c Brundage, Miles; Avin, Shahar; Clark, Jack; Toner, Helen; Eckersley, Peter; Garfinkel, Ben; Dafoe, Allan; Scharre, Paul; Zeitzoff, Thomas; Filar, Bobby; Anderson, Hyrum; Roff, Heather; Allen, Gregory C; Steinhardt, Jacob; Flynn, Carrick (2018-04-30). "The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigation". Apollo-University Of Cambridge Repository, Apollo-University Of Cambridge Repository. Apollo - University of Cambridge Repository. doi:10.17863/cam.22520. S2CID 3385567. Archived from the original on 2022-11-23. Retrieved 2022-11-28. {{cite journal}}: Cite journal requires |journal= (help)
  6. ^ Davies, Pascale (December 26, 2022). "How NATO is preparing for a new era of AI cyber attacks". euronews. Retrieved 2024-03-23.
  7. ^ Ahuja, Anjana (February 7, 2024). "AI's bioterrorism potential should not be ruled out". Financial Times. Retrieved 2024-03-23.
  8. ^ Carlsmith, Joseph (2022-06-16). "Is Power-Seeking AI an Existential Risk?". arXiv:2206.13353. {{cite journal}}: Cite journal requires |journal= (help)
  9. ^ Minardi, Di (16 October 2020). "The grim fate that could be 'worse than extinction'". BBC. Retrieved 2024-03-23.
  10. ^ Carlsmith, Joseph (2022-06-16). "Is Power-Seeking AI an Existential Risk?". arXiv:2206.13353 [cs.CY].
  11. ^ "'The Godfather of A.I.' warns of 'nightmare scenario' where artificial intelligence begins to seek power". Fortune. Retrieved 2023-06-10.
  12. ^ Sinclair, Sebastian (2021-09-24). "Future of Life Institute to Launch Vitalik Buterin Fellowships Centered on AI Safety Research". CoinDesk. Retrieved 2024-03-23.
  13. ^ "AGI Expert Peter Voss Says AI Alignment Problem is Bogus |". 2023-04-04. Retrieved 2023-07-23.
  14. ^ Dafoe, Allan (2016). "Yes, We Are Worried About the Existential Risk of Artificial Intelligence". MIT Technology Review. Archived from the original on 2022-11-28. Retrieved 2022-11-28.
  15. ^ a b Grace, Katja; Salvatier, John; Dafoe, Allan; Zhang, Baobao; Evans, Owain (2018-07-31). "Viewpoint: When Will AI Exceed Human Performance? Evidence from AI Experts". Journal of Artificial Intelligence Research. 62: 729–754. doi:10.1613/jair.1.11222. ISSN 1076-9757. S2CID 8746462. Archived from the original on 2023-02-10. Retrieved 2022-11-28.
  16. ^ Zhang, Baobao; Anderljung, Markus; Kahn, Lauren; Dreksler, Noemi; Horowitz, Michael C.; Dafoe, Allan (2021-05-05). "Ethics and Governance of Artificial Intelligence: Evidence from a Survey of Machine Learning Researchers". Journal of Artificial Intelligence Research. 71. arXiv:2105.02117. doi:10.1613/jair.1.12895.
  17. ^ Stein-Perlman, Zach; Weinstein-Raun, Benjamin; Grace (2022-08-04). "2022 Expert Survey on Progress in AI". AI Impacts. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  18. ^ Michael, Julian; Holtzman, Ari; Parrish, Alicia; Mueller, Aaron; Wang, Alex; Chen, Angelica; Madaan, Divyam; Nangia, Nikita; Pang, Richard Yuanzhe; Phang, Jason; Bowman, Samuel R. (2022-08-26). "What Do NLP Researchers Believe? Results of the NLP Community Metasurvey". Association for Computational Linguistics. arXiv:2208.12852.
  19. ^ Markoff, John (2013-05-20). "In 1949, He Imagined an Age of Robots". The New York Times. ISSN 0362-4331. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  20. ^ Association for the Advancement of Artificial Intelligence. "AAAI Presidential Panel on Long-Term AI Futures". Archived from the original on 2022-09-01. Retrieved 2022-11-23.
  21. ^ Yampolskiy, Roman V.; Spellchecker, M. S. (2016-10-25). "Artificial Intelligence Safety and Cybersecurity: a Timeline of AI Failures". arXiv:1610.07997. {{cite journal}}: Cite journal requires |journal= (help)
  22. ^ "PT-AI 2011 – Philosophy and Theory of Artificial Intelligence (PT-AI 2011)". Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  23. ^ Yampolskiy, Roman V. (2013), Müller, Vincent C. (ed.), "Artificial Intelligence Safety Engineering: Why Machine Ethics is a Wrong Approach", Philosophy and Theory of Artificial Intelligence, Studies in Applied Philosophy, Epistemology and Rational Ethics, vol. 5, Berlin; Heidelberg, Germany: Springer Berlin Heidelberg, pp. 389–396, doi:10.1007/978-3-642-31674-6_29, ISBN 978-3-642-31673-9, archived from the original on 2023-03-15, retrieved 2022-11-23
  24. ^ McLean, Scott; Read, Gemma J. M.; Thompson, Jason; Baber, Chris; Stanton, Neville A.; Salmon, Paul M. (2023-07-04). "The risks associated with Artificial General Intelligence: A systematic review". Journal of Experimental & Theoretical Artificial Intelligence. 35 (5): 649–663. Bibcode:2023JETAI..35..649M. doi:10.1080/0952813X.2021.1964003. hdl:11343/289595. ISSN 0952-813X. S2CID 238643957.
  25. ^ Wile, Rob (August 3, 2014). "Elon Musk: Artificial Intelligence Is 'Potentially More Dangerous Than Nukes'". Business Insider. Retrieved 2024-02-22.
  26. ^ Kuo, Kaiser (2015-03-31). Baidu CEO Robin Li interviews Bill Gates and Elon Musk at the Boao Forum, March 29, 2015. Event occurs at 55:49. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  27. ^ Cellan-Jones, Rory (2014-12-02). "Stephen Hawking warns artificial intelligence could end mankind". BBC News. Archived from the original on 2015-10-30. Retrieved 2022-11-23.
  28. ^ Future of Life Institute. "Research Priorities for Robust and Beneficial Artificial Intelligence: An Open Letter". Future of Life Institute. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  29. ^ Future of Life Institute (October 2016). "AI Research Grants Program". Future of Life Institute. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  30. ^ "SafArtInt 2016". Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  31. ^ Bach, Deborah (2016). "UW to host first of four White House public workshops on artificial intelligence". UW News. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  32. ^ Amodei, Dario; Olah, Chris; Steinhardt, Jacob; Christiano, Paul; Schulman, John; Mané, Dan (2016-07-25). "Concrete Problems in AI Safety". arXiv:1606.06565. {{cite journal}}: Cite journal requires |journal= (help)
  33. ^ a b Future of Life Institute. "AI Principles". Future of Life Institute. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  34. ^ a b Research, DeepMind Safety (2018-09-27). "Building safe artificial intelligence: specification, robustness, and assurance". Medium. Archived from the original on 2023-02-10. Retrieved 2022-11-23.
  35. ^ "SafeML ICLR 2019 Workshop". Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  36. ^ a b c d e Hendrycks, Dan; Carlini, Nicholas; Schulman, John; Steinhardt, Jacob (2022-06-16). "Unsolved Problems in ML Safety". arXiv:2109.13916. {{cite journal}}: Cite journal requires |journal= (help)
  37. ^ Browne, Ryan (2023-06-12). "British Prime Minister Rishi Sunak pitches UK as home of A.I. safety regulation as London bids to be next Silicon Valley". CNBC. Retrieved 2023-06-25.
  38. ^ Bertuzzi, Luca (October 18, 2023). "UK's AI safety summit set to highlight risk of losing human control over 'frontier' models". Euractiv. Retrieved March 2, 2024.
  39. ^ Shepardson, David (1 April 2024). "US, Britain announce partnership on AI safety, testing". Retrieved 2 April 2024.
  40. ^ Goodfellow, Ian; Papernot, Nicolas; Huang, Sandy; Duan, Rocky; Abbeel, Pieter; Clark, Jack (2017-02-24). "Attacking Machine Learning with Adversarial Examples". OpenAI. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  41. ^ a b Szegedy, Christian; Zaremba, Wojciech; Sutskever, Ilya; Bruna, Joan; Erhan, Dumitru; Goodfellow, Ian; Fergus, Rob (2014-02-19). "Intriguing properties of neural networks". ICLR. arXiv:1312.6199.
  42. ^ Kurakin, Alexey; Goodfellow, Ian; Bengio, Samy (2017-02-10). "Adversarial examples in the physical world". ICLR. arXiv:1607.02533.
  43. ^ Madry, Aleksander; Makelov, Aleksandar; Schmidt, Ludwig; Tsipras, Dimitris; Vladu, Adrian (2019-09-04). "Towards Deep Learning Models Resistant to Adversarial Attacks". ICLR. arXiv:1706.06083.
  44. ^ Kannan, Harini; Kurakin, Alexey; Goodfellow, Ian (2018-03-16). "Adversarial Logit Pairing". arXiv:1803.06373. {{cite journal}}: Cite journal requires |journal= (help)
  45. ^ Gilmer, Justin; Adams, Ryan P.; Goodfellow, Ian; Andersen, David; Dahl, George E. (2018-07-19). "Motivating the Rules of the Game for Adversarial Example Research". arXiv:1807.06732. {{cite journal}}: Cite journal requires |journal= (help)
  46. ^ Carlini, Nicholas; Wagner, David (2018-03-29). "Audio Adversarial Examples: Targeted Attacks on Speech-to-Text". IEEE Security and Privacy Workshops. arXiv:1801.01944.
  47. ^ Sheatsley, Ryan; Papernot, Nicolas; Weisman, Michael; Verma, Gunjan; McDaniel, Patrick (2022-09-09). "Adversarial Examples in Constrained Domains". arXiv:2011.01183. {{cite journal}}: Cite journal requires |journal= (help)
  48. ^ Suciu, Octavian; Coull, Scott E.; Johns, Jeffrey (2019-04-13). "Exploring Adversarial Examples in Malware Detection". IEEE Security and Privacy Workshops. arXiv:1810.08280.
  49. ^ Ouyang, Long; Wu, Jeff; Jiang, Xu; Almeida, Diogo; Wainwright, Carroll L.; Mishkin, Pamela; Zhang, Chong; Agarwal, Sandhini; Slama, Katarina; Ray, Alex; Schulman, John; Hilton, Jacob; Kelton, Fraser; Miller, Luke; Simens, Maddie (2022-03-04). "Training language models to follow instructions with human feedback". NeurIPS. arXiv:2203.02155.
  50. ^ Gao, Leo; Schulman, John; Hilton, Jacob (2022-10-19). "Scaling Laws for Reward Model Overoptimization". ICML. arXiv:2210.10760.
  51. ^ Yu, Sihyun; Ahn, Sungsoo; Song, Le; Shin, Jinwoo (2021-10-27). "RoMA: Robust Model Adaptation for Offline Model-based Optimization". NeurIPS. arXiv:2110.14188.
  52. ^ a b Hendrycks, Dan; Mazeika, Mantas (2022-09-20). "X-Risk Analysis for AI Research". arXiv:2206.05862. {{cite journal}}: Cite journal requires |journal= (help)
  53. ^ Tran, Khoa A.; Kondrashova, Olga; Bradley, Andrew; Williams, Elizabeth D.; Pearson, John V.; Waddell, Nicola (2021). "Deep learning in cancer diagnosis, prognosis and treatment selection". Genome Medicine. 13 (1): 152. doi:10.1186/s13073-021-00968-x. ISSN 1756-994X. PMC 8477474. PMID 34579788.
  54. ^ Guo, Chuan; Pleiss, Geoff; Sun, Yu; Weinberger, Kilian Q. (2017-08-06). "On calibration of modern neural networks". Proceedings of the 34th international conference on machine learning. Proceedings of machine learning research. Vol. 70. PMLR. pp. 1321–1330.
  55. ^ Ovadia, Yaniv; Fertig, Emily; Ren, Jie; Nado, Zachary; Sculley, D.; Nowozin, Sebastian; Dillon, Joshua V.; Lakshminarayanan, Balaji; Snoek, Jasper (2019-12-17). "Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shift". NeurIPS. arXiv:1906.02530.
  56. ^ Bogdoll, Daniel; Breitenstein, Jasmin; Heidecker, Florian; Bieshaar, Maarten; Sick, Bernhard; Fingscheidt, Tim; Zöllner, J. Marius (2021). "Description of Corner Cases in Automated Driving: Goals and Challenges". 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW). pp. 1023–1028. arXiv:2109.09607. doi:10.1109/ICCVW54120.2021.00119. ISBN 978-1-6654-0191-3. S2CID 237572375.
  57. ^ Hendrycks, Dan; Mazeika, Mantas; Dietterich, Thomas (2019-01-28). "Deep Anomaly Detection with Outlier Exposure". ICLR. arXiv:1812.04606.
  58. ^ Wang, Haoqi; Li, Zhizhong; Feng, Litong; Zhang, Wayne (2022-03-21). "ViM: Out-Of-Distribution with Virtual-logit Matching". CVPR. arXiv:2203.10807.
  59. ^ Hendrycks, Dan; Gimpel, Kevin (2018-10-03). "A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks". ICLR. arXiv:1610.02136.
  60. ^ Urbina, Fabio; Lentzos, Filippa; Invernizzi, Cédric; Ekins, Sean (2022). "Dual use of artificial-intelligence-powered drug discovery". Nature Machine Intelligence. 4 (3): 189–191. doi:10.1038/s42256-022-00465-9. ISSN 2522-5839. PMC 9544280. PMID 36211133.
  61. ^ Center for Security and Emerging Technology; Buchanan, Ben; Lohn, Andrew; Musser, Micah; Sedova, Katerina (2021). "Truth, Lies, and Automation: How Language Models Could Change Disinformation". doi:10.51593/2021ca003. S2CID 240522878. Archived from the original on 2022-11-24. Retrieved 2022-11-28. {{cite journal}}: Cite journal requires |journal= (help)
  62. ^ "Propaganda-as-a-service may be on the horizon if large language models are abused". VentureBeat. 2021-12-14. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  63. ^ Center for Security and Emerging Technology; Buchanan, Ben; Bansemer, John; Cary, Dakota; Lucas, Jack; Musser, Micah (2020). "Automating Cyber Attacks: Hype and Reality". Center for Security and Emerging Technology. doi:10.51593/2020ca002. S2CID 234623943. Archived from the original on 2022-11-24. Retrieved 2022-11-28.
  64. ^ "Lessons Learned on Language Model Safety and Misuse". OpenAI. 2022-03-03. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  65. ^ Markov, Todor; Zhang, Chong; Agarwal, Sandhini; Eloundou, Tyna; Lee, Teddy; Adler, Steven; Jiang, Angela; Weng, Lilian (2022-08-10). "New-and-Improved Content Moderation Tooling". OpenAI. Archived from the original on 2023-01-11. Retrieved 2022-11-24.
  66. ^ a b Savage, Neil (2022-03-29). "Breaking into the black box of artificial intelligence". Nature. doi:10.1038/d41586-022-00858-1. PMID 35352042. S2CID 247792459. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  67. ^ Center for Security and Emerging Technology; Rudner, Tim; Toner, Helen (2021). "Key Concepts in AI Safety: Interpretability in Machine Learning". doi:10.51593/20190042. S2CID 233775541. Archived from the original on 2022-11-24. Retrieved 2022-11-28. {{cite journal}}: Cite journal requires |journal= (help)
  68. ^ McFarland, Matt (2018-03-19). "Uber pulls self-driving cars after first fatal crash of autonomous vehicle". CNNMoney. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  69. ^ Felder, Ryan Marshall (July 2021). "Coming to Terms with the Black Box Problem: How to Justify AI Systems in Health Care". Hastings Center Report. 51 (4): 38–45. doi:10.1002/hast.1248. ISSN 0093-0334. PMID 33821471.
  70. ^ a b Doshi-Velez, Finale; Kortz, Mason; Budish, Ryan; Bavitz, Chris; Gershman, Sam; O'Brien, David; Scott, Kate; Schieber, Stuart; Waldo, James; Weinberger, David; Weller, Adrian; Wood, Alexandra (2019-12-20). "Accountability of AI Under the Law: The Role of Explanation". arXiv:1711.01134. {{cite journal}}: Cite journal requires |journal= (help)
  71. ^ Fong, Ruth; Vedaldi, Andrea (2017). "Interpretable Explanations of Black Boxes by Meaningful Perturbation". 2017 IEEE International Conference on Computer Vision (ICCV). pp. 3449–3457. arXiv:1704.03296. doi:10.1109/ICCV.2017.371. ISBN 978-1-5386-1032-9. S2CID 1633753.
  72. ^ Meng, Kevin; Bau, David; Andonian, Alex; Belinkov, Yonatan (2022). "Locating and editing factual associations in GPT". Advances in Neural Information Processing Systems. 35. arXiv:2202.05262.
  73. ^ Bau, David; Liu, Steven; Wang, Tongzhou; Zhu, Jun-Yan; Torralba, Antonio (2020-07-30). "Rewriting a Deep Generative Model". ECCV. arXiv:2007.15646.
  74. ^ Räuker, Tilman; Ho, Anson; Casper, Stephen; Hadfield-Menell, Dylan (2022-09-05). "Toward Transparent AI: A Survey on Interpreting the Inner Structures of Deep Neural Networks". IEEE SaTML. arXiv:2207.13243.
  75. ^ Bau, David; Zhou, Bolei; Khosla, Aditya; Oliva, Aude; Torralba, Antonio (2017-04-19). "Network Dissection: Quantifying Interpretability of Deep Visual Representations". CVPR. arXiv:1704.05796.
  76. ^ McGrath, Thomas; Kapishnikov, Andrei; Tomašev, Nenad; Pearce, Adam; Wattenberg, Martin; Hassabis, Demis; Kim, Been; Paquet, Ulrich; Kramnik, Vladimir (2022-11-22). "Acquisition of chess knowledge in AlphaZero". Proceedings of the National Academy of Sciences. 119 (47): e2206625119. arXiv:2111.09259. Bibcode:2022PNAS..11906625M. doi:10.1073/pnas.2206625119. ISSN 0027-8424. PMC 9704706. PMID 36375061.
  77. ^ Goh, Gabriel; Cammarata, Nick; Voss, Chelsea; Carter, Shan; Petrov, Michael; Schubert, Ludwig; Radford, Alec; Olah, Chris (2021). "Multimodal neurons in artificial neural networks". Distill. 6 (3). doi:10.23915/distill.00030. S2CID 233823418.
  78. ^ Olah, Chris; Cammarata, Nick; Schubert, Ludwig; Goh, Gabriel; Petrov, Michael; Carter, Shan (2020). "Zoom in: An introduction to circuits". Distill. 5 (3). doi:10.23915/distill.00024.001. S2CID 215930358.
  79. ^ Cammarata, Nick; Goh, Gabriel; Carter, Shan; Voss, Chelsea; Schubert, Ludwig; Olah, Chris (2021). "Curve circuits". Distill. 6 (1). doi:10.23915/distill.00024.006 (inactive 31 January 2024). Archived from the original on 5 December 2022. Retrieved 5 December 2022.{{cite journal}}: CS1 maint: DOI inactive as of January 2024 (link)
  80. ^ Olsson, Catherine; Elhage, Nelson; Nanda, Neel; Joseph, Nicholas; DasSarma, Nova; Henighan, Tom; Mann, Ben; Askell, Amanda; Bai, Yuntao; Chen, Anna; Conerly, Tom; Drain, Dawn; Ganguli, Deep; Hatfield-Dodds, Zac; Hernandez, Danny; Johnston, Scott; Jones, Andy; Kernion, Jackson; Lovitt, Liane; Ndousse, Kamal; Amodei, Dario; Brown, Tom; Clark, Jack; Kaplan, Jared; McCandlish, Sam; Olah, Chris (2022). "In-context learning and induction heads". Transformer Circuits Thread. arXiv:2209.11895.
  81. ^ Olah, Christopher. "Interpretability vs Neuroscience [rough note]". Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  82. ^ Gu, Tianyu; Dolan-Gavitt, Brendan; Garg, Siddharth (2019-03-11). "BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain". arXiv:1708.06733. {{cite journal}}: Cite journal requires |journal= (help)
  83. ^ Chen, Xinyun; Liu, Chang; Li, Bo; Lu, Kimberly; Song, Dawn (2017-12-14). "Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning". arXiv:1712.05526. {{cite journal}}: Cite journal requires |journal= (help)
  84. ^ Carlini, Nicholas; Terzis, Andreas (2022-03-28). "Poisoning and Backdooring Contrastive Learning". ICLR. arXiv:2106.09667.
  85. ^ a b c d Russell, Stuart J.; Norvig, Peter (2021). Artificial intelligence: A modern approach (4th ed.). Pearson. pp. 5, 1003. ISBN 9780134610993. Retrieved September 12, 2022.
  86. ^ a b Ngo, Richard; Chan, Lawrence; Mindermann, Sören (2022). "The Alignment Problem from a Deep Learning Perspective". International Conference on Learning Representations. arXiv:2209.00626.
  87. ^ a b Pan, Alexander; Bhatia, Kush; Steinhardt, Jacob (2022-02-14). The Effects of Reward Misspecification: Mapping and Mitigating Misaligned Models. International Conference on Learning Representations. Retrieved 2022-07-21.
  88. ^ Zhuang, Simon; Hadfield-Menell, Dylan (2020). "Consequences of Misaligned AI". Advances in Neural Information Processing Systems. Vol. 33. Curran Associates, Inc. pp. 15763–15773. Retrieved 2023-03-11.
  89. ^ Carlsmith, Joseph (2022-06-16). "Is Power-Seeking AI an Existential Risk?". arXiv:2206.13353 [cs.CY].
  90. ^ a b c Russell, Stuart J. (2020). Human compatible: Artificial intelligence and the problem of control. Penguin Random House. ISBN 9780525558637. OCLC 1113410915.
  91. ^ Christian, Brian (2020). The alignment problem: Machine learning and human values. W. W. Norton & Company. ISBN 978-0-393-86833-3. OCLC 1233266753. Archived from the original on February 10, 2023. Retrieved September 12, 2022.
  92. ^ Langosco, Lauro Langosco Di; Koch, Jack; Sharkey, Lee D.; Pfau, Jacob; Krueger, David (2022-06-28). "Goal Misgeneralization in Deep Reinforcement Learning". Proceedings of the 39th International Conference on Machine Learning. International Conference on Machine Learning. PMLR. pp. 12004–12019. Retrieved 2023-03-11.
  93. ^ a b Bommasani, Rishi; Hudson, Drew A.; Adeli, Ehsan; Altman, Russ; Arora, Simran; von Arx, Sydney; Bernstein, Michael S.; Bohg, Jeannette; Bosselut, Antoine; Brunskill, Emma; Brynjolfsson, Erik (2022-07-12). "On the Opportunities and Risks of Foundation Models". Stanford CRFM. arXiv:2108.07258.
  94. ^ Ouyang, Long; Wu, Jeff; Jiang, Xu; Almeida, Diogo; Wainwright, Carroll L.; Mishkin, Pamela; Zhang, Chong; Agarwal, Sandhini; Slama, Katarina; Ray, Alex; Schulman, J.; Hilton, Jacob; Kelton, Fraser; Miller, Luke E.; Simens, Maddie; Askell, Amanda; Welinder, P.; Christiano, P.; Leike, J.; Lowe, Ryan J. (2022). "Training language models to follow instructions with human feedback". arXiv:2203.02155 [cs.CL].
  95. ^ Zaremba, Wojciech; Brockman, Greg; OpenAI (2021-08-10). "OpenAI Codex". OpenAI. Archived from the original on February 3, 2023. Retrieved 2022-07-23.
  96. ^ Kober, Jens; Bagnell, J. Andrew; Peters, Jan (2013-09-01). "Reinforcement learning in robotics: A survey". The International Journal of Robotics Research. 32 (11): 1238–1274. doi:10.1177/0278364913495721. ISSN 0278-3649. S2CID 1932843. Archived from the original on October 15, 2022. Retrieved September 12, 2022.
  97. ^ Knox, W. Bradley; Allievi, Alessandro; Banzhaf, Holger; Schmitt, Felix; Stone, Peter (2023-03-01). "Reward (Mis)design for autonomous driving". Artificial Intelligence. 316: 103829. arXiv:2104.13906. doi:10.1016/j.artint.2022.103829. ISSN 0004-3702. S2CID 233423198.
  98. ^ Stray, Jonathan (2020). "Aligning AI Optimization to Community Well-Being". International Journal of Community Well-Being. 3 (4): 443–463. doi:10.1007/s42413-020-00086-3. ISSN 2524-5295. PMC 7610010. PMID 34723107. S2CID 226254676.
  99. ^ Russell, Stuart; Norvig, Peter (2009). Artificial Intelligence: A Modern Approach. Prentice Hall. p. 1003. ISBN 978-0-13-461099-3.
  100. ^ Bengio, Yoshua; Hinton, Geoffrey; Yao, Andrew; Song, Dawn; Abbeel, Pieter; Harari, Yuval Noah; Zhang, Ya-Qin; Xue, Lan; Shalev-Shwartz, Shai (2023-11-12), Managing AI Risks in an Era of Rapid Progress, arXiv:2310.17688
  101. ^ "Statement on AI Risk | CAIS". Retrieved 2024-02-11.
  102. ^ Grace, Katja; Stewart, Harlan; Sandkühler, Julia Fabienne; Thomas, Stephen; Weinstein-Raun, Ben; Brauner, Jan (2024-01-05), Thousands of AI Authors on the Future of AI, arXiv:2401.02843
  103. ^ Smith, Craig S. "Geoff Hinton, AI's Most Famous Researcher, Warns Of 'Existential Threat'". Forbes. Retrieved 2023-05-04.
  104. ^ a b c Amodei, Dario; Olah, Chris; Steinhardt, Jacob; Christiano, Paul; Schulman, John; Mané, Dan (2016-06-21). "Concrete Problems in AI Safety". arXiv:1606.06565 [cs.AI].
  105. ^ a b Ortega, Pedro A.; Maini, Vishal; DeepMind safety team (2018-09-27). "Building safe artificial intelligence: specification, robustness, and assurance". DeepMind Safety Research – Medium. Archived from the original on February 10, 2023. Retrieved 2022-07-18.
  106. ^ a b Rorvig, Mordechai (2022-04-14). "Researchers Gain New Understanding From Simple AI". Quanta Magazine. Archived from the original on February 10, 2023. Retrieved 2022-07-18.
  107. ^ Doshi-Velez, Finale; Kim, Been (2017-03-02). "Towards A Rigorous Science of Interpretable Machine Learning". arXiv:1702.08608 [stat.ML].
  108. ^ Russell, Stuart; Dewey, Daniel; Tegmark, Max (2015-12-31). "Research Priorities for Robust and Beneficial Artificial Intelligence". AI Magazine. 36 (4): 105–114. arXiv:1602.03506. doi:10.1609/aimag.v36i4.2577. hdl:1721.1/108478. ISSN 2371-9621. S2CID 8174496. Archived from the original on February 2, 2023. Retrieved September 12, 2022.
  109. ^ Wirth, Christian; Akrour, Riad; Neumann, Gerhard; Fürnkranz, Johannes (2017). "A survey of preference-based reinforcement learning methods". Journal of Machine Learning Research. 18 (136): 1–46.
  110. ^ Christiano, Paul F.; Leike, Jan; Brown, Tom B.; Martic, Miljan; Legg, Shane; Amodei, Dario (2017). "Deep reinforcement learning from human preferences". Proceedings of the 31st International Conference on Neural Information Processing Systems. NIPS'17. Red Hook, NY, USA: Curran Associates Inc. pp. 4302–4310. ISBN 978-1-5108-6096-4.
  111. ^ Heaven, Will Douglas (2022-01-27). "The new version of GPT-3 is much better behaved (and should be less toxic)". MIT Technology Review. Archived from the original on February 10, 2023. Retrieved 2022-07-18.
  112. ^ Mohseni, Sina; Wang, Haotao; Yu, Zhiding; Xiao, Chaowei; Wang, Zhangyang; Yadawa, Jay (2022-03-07). "Taxonomy of Machine Learning Safety: A Survey and Primer". arXiv:2106.04823 [cs.LG].
  113. ^ Clifton, Jesse (2020). "Cooperation, Conflict, and Transformative Artificial Intelligence: A Research Agenda". Center on Long-Term Risk. Archived from the original on January 1, 2023. Retrieved 2022-07-18.
  114. ^ Prunkl, Carina; Whittlestone, Jess (2020-02-07). "Beyond Near- and Long-Term". Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society. New York NY USA: ACM. pp. 138–143. doi:10.1145/3375627.3375803. ISBN 978-1-4503-7110-0. S2CID 210164673. Archived from the original on October 16, 2022. Retrieved September 12, 2022.
  115. ^ Irving, Geoffrey; Askell, Amanda (2019-02-19). "AI Safety Needs Social Scientists". Distill. 4 (2): 10.23915/distill.00014. doi:10.23915/distill.00014. ISSN 2476-0757. S2CID 159180422. Archived from the original on February 10, 2023. Retrieved September 12, 2022.
  116. ^ a b c d Zwetsloot, Remco; Dafoe, Allan (2019-02-11). "Thinking About Risks From AI: Accidents, Misuse and Structure". Lawfare. Archived from the original on 2023-08-19. Retrieved 2022-11-24.
  117. ^ Zhang, Yingyu; Dong, Chuntong; Guo, Weiqun; Dai, Jiabao; Zhao, Ziming (2022). "Systems theoretic accident model and process (STAMP): A literature review". Safety Science. 152: 105596. doi:10.1016/j.ssci.2021.105596. S2CID 244550153. Archived from the original on 2023-03-15. Retrieved 2022-11-28.
  118. ^ Center for Security and Emerging Technology; Hoffman, Wyatt (2021). "AI and the Future of Cyber Competition". CSET Issue Brief. doi:10.51593/2020ca007. S2CID 234245812. Archived from the original on 2022-11-24. Retrieved 2022-11-28.
  119. ^ Center for Security and Emerging Technology; Imbrie, Andrew; Kania, Elsa (2019). "AI Safety, Security, and Stability Among Great Powers: Options, Challenges, and Lessons Learned for Pragmatic Engagement". doi:10.51593/20190051. S2CID 240957952. Archived from the original on 2022-11-24. Retrieved 2022-11-28. {{cite journal}}: Cite journal requires |journal= (help)
  120. ^ a b Future of Life Institute (2019-03-27). AI Strategy, Policy, and Governance (Allan Dafoe). Event occurs at 22:05. Archived from the original on 2022-11-23. Retrieved 2022-11-23.
  121. ^ Zou, Andy; Xiao, Tristan; Jia, Ryan; Kwon, Joe; Mazeika, Mantas; Li, Richard; Song, Dawn; Steinhardt, Jacob; Evans, Owain; Hendrycks, Dan (2022-10-09). "Forecasting Future World Events with Neural Networks". NeurIPS. arXiv:2206.15474.
  122. ^ Gathani, Sneha; Hulsebos, Madelon; Gale, James; Haas, Peter J.; Demiralp, Çağatay (2022-02-08). "Augmenting Decision Making via Interactive What-If Analysis". Conference on Innovative Data Systems Research. arXiv:2109.06160.
  123. ^ Lindelauf, Roy (2021), Osinga, Frans; Sweijs, Tim (eds.), "Nuclear Deterrence in the Algorithmic Age: Game Theory Revisited", NL ARMS Netherlands Annual Review of Military Studies 2020, Nl Arms, The Hague: T.M.C. Asser Press, pp. 421–436, doi:10.1007/978-94-6265-419-8_22, ISBN 978-94-6265-418-1, S2CID 229449677
  124. ^ a b Newkirk II, Vann R. (2016-04-21). "Is Climate Change a Prisoner's Dilemma or a Stag Hunt?". The Atlantic. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  125. ^ a b Armstrong, Stuart; Bostrom, Nick; Shulman, Carl. Racing to the Precipice: a Model of Artificial Intelligence Development (Report). Future of Humanity Institute, Oxford University.
  126. ^ a b Dafoe, Allan. AI Governance: A Research Agenda (Report). Centre for the Governance of AI, Future of Humanity Institute, University of Oxford.
  127. ^ Dafoe, Allan; Hughes, Edward; Bachrach, Yoram; Collins, Tantum; McKee, Kevin R.; Leibo, Joel Z.; Larson, Kate; Graepel, Thore (2020-12-15). "Open Problems in Cooperative AI". NeurIPS. arXiv:2012.08630.
  128. ^ a b Dafoe, Allan; Bachrach, Yoram; Hadfield, Gillian; Horvitz, Eric; Larson, Kate; Graepel, Thore (2021). "Cooperative AI: machines must learn to find common ground". Nature. 593 (7857): 33–36. Bibcode:2021Natur.593...33D. doi:10.1038/d41586-021-01170-0. PMID 33947992. S2CID 233740521. Archived from the original on 2022-11-22. Retrieved 2022-11-24.
  129. ^ Bender, E.M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜. FAccT '21: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610-623.
  130. ^ Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and Policy Considerations for Deep Learning in NLP. arXiv preprint arXiv:1906.02243.
  131. ^ Schwartz, R., Dodge, J., Smith, N.A., & Etzioni, O. (2020). Green AI. Communications of the ACM, 63(12), 54-63.
  132. ^ Satariano, Adam; Specia, Megan (2023-11-01). "Global Leaders Warn A.I. Could Cause 'Catastrophic' Harm". The New York Times. ISSN 0362-4331. Retrieved 2024-04-20.
  133. ^ Crafts, Nicholas (2021-09-23). "Artificial intelligence as a general-purpose technology: an historical perspective". Oxford Review of Economic Policy. 37 (3): 521–536. doi:10.1093/oxrep/grab012. ISSN 0266-903X. Archived from the original on 2022-11-24. Retrieved 2022-11-28.
  134. ^ 葉俶禎; 黃子君; 張媁雯; 賴志樫 (2020-12-01). "Labor Displacement in Artificial Intelligence Era: A Systematic Literature Review". 臺灣東亞文明研究學刊. 17 (2). doi:10.6163/TJEAS.202012_17(2).0002. ISSN 1812-6243.
  135. ^ Johnson, James (2019-04-03). "Artificial intelligence & future warfare: implications for international security". Defense & Security Analysis. 35 (2): 147–169. doi:10.1080/14751798.2019.1600800. ISSN 1475-1798. S2CID 159321626. Archived from the original on 2022-11-24. Retrieved 2022-11-28.
  136. ^ Kertysova, Katarina (2018-12-12). "Artificial Intelligence and Disinformation: How AI Changes the Way Disinformation is Produced, Disseminated, and Can Be Countered". Security and Human Rights. 29 (1–4): 55–81. doi:10.1163/18750230-02901005. ISSN 1874-7337. S2CID 216896677. Archived from the original on 2022-11-24. Retrieved 2022-11-28.
  137. ^ Feldstein, Steven (2019). The Global Expansion of AI Surveillance. Carnegie Endowment for International Peace.
  138. ^ Agrawal, Ajay; Gans, Joshua; Goldfarb, Avi (2019). The economics of artificial intelligence: an agenda. Chicago, Illinois. ISBN 978-0-226-61347-5. OCLC 1099435014. Archived from the original on 2023-03-15. Retrieved 2022-11-28.{{cite book}}: CS1 maint: location missing publisher (link)
  139. ^ Whittlestone, Jess; Clark, Jack (2021-08-31). "Why and How Governments Should Monitor AI Development". arXiv:2108.12427. {{cite journal}}: Cite journal requires |journal= (help)
  140. ^ a b Shevlane, Toby (2022). "Sharing Powerful AI Models | GovAI Blog". Center for the Governance of AI. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  141. ^ Askell, Amanda; Brundage, Miles; Hadfield, Gillian (2019-07-10). "The Role of Cooperation in Responsible AI Development". arXiv:1907.04534. {{cite journal}}: Cite journal requires |journal= (help)
  142. ^ Gursoy, Furkan; Kakadiaris, Ioannis A. (2022-08-31), System Cards for AI-Based Decision-Making for Public Policy, arXiv:2203.04754
  143. ^ Cobbe, Jennifer; Lee, Michelle Seng Ah; Singh, Jatinder (2021-03-01). "Reviewable Automated Decision-Making: A Framework for Accountable Algorithmic Systems". Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency. FAccT '21. New York, NY, USA: Association for Computing Machinery. pp. 598–609. doi:10.1145/3442188.3445921. ISBN 978-1-4503-8309-7.
  144. ^ Raji, Inioluwa Deborah; Smart, Andrew; White, Rebecca N.; Mitchell, Margaret; Gebru, Timnit; Hutchinson, Ben; Smith-Loud, Jamila; Theron, Daniel; Barnes, Parker (2020-01-27). "Closing the AI accountability gap: Defining an end-to-end framework for internal algorithmic auditing". Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency. FAT* '20. New York, NY, USA: Association for Computing Machinery. pp. 33–44. doi:10.1145/3351095.3372873. ISBN 978-1-4503-6936-7.
  145. ^ Turchin, Alexey; Dench, David; Green, Brian Patrick (2019). "Global Solutions vs. Local Solutions for the AI Safety Problem". Big Data and Cognitive Computing. 3 (16): 1–25. doi:10.3390/bdcc3010016.
  146. ^ Ziegler, Bart (8 April 2022). "Is It Time to Regulate AI?". Wall Street Journal.
  147. ^ Smith, John (15 May 2022). "Global Governance of Artificial Intelligence: Opportunities and Challenges". The Guardian.
  148. ^ Ziegler, Bart (8 April 2022). "Is It Time to Regulate AI?". Wall Street Journal. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  149. ^ Reed, Chris (2018-09-13). "How should we regulate artificial intelligence?". Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 376 (2128): 20170360. Bibcode:2018RSPTA.37670360R. doi:10.1098/rsta.2017.0360. ISSN 1364-503X. PMC 6107539. PMID 30082306.
  150. ^ Belton, Keith B. (2019-03-07). "How Should AI Be Regulated?". IndustryWeek. Archived from the original on 2022-01-29. Retrieved 2022-11-24.
  151. ^ National Security Commission on Artificial Intelligence (2021), Final Report
  152. ^ National Institute of Standards and Technology (2021-07-12). "AI Risk Management Framework". NIST. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  153. ^ Richardson, Tim (2021). "Britain publishes 10-year National Artificial Intelligence Strategy". Archived from the original on 2023-02-10. Retrieved 2022-11-24.
  154. ^ a b "Guidance: National AI Strategy". GOV.UK. 2021. Archived from the original on 2023-02-10. Retrieved 2022-11-24.
  155. ^ Hardcastle, Kimberley (2023-08-23). "We're talking about AI a lot right now – and it's not a moment too soon". The Conversation. Retrieved 2023-10-31.
  156. ^ "Iconic Bletchley Park to host UK AI Safety Summit in early November". GOV.UK. Retrieved 2023-10-31.
  157. ^ Office of the Director of National Intelligence, Intelligence Advanced Research Projects Activity. "IARPA – TrojAI". Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  158. ^ Turek, Matt. "Explainable Artificial Intelligence". Archived from the original on 2021-02-19. Retrieved 2022-11-24.
  159. ^ Draper, Bruce. "Guaranteeing AI Robustness Against Deception". Defense Advanced Research Projects Agency. Archived from the original on 2023-01-09. Retrieved 2022-11-24.
  160. ^ National Science Foundation (23 February 2023). "Safe Learning-Enabled Systems". Archived from the original on 2023-02-26. Retrieved 2023-02-27.
  161. ^ "General Assembly adopts landmark resolution on artificial intelligence". UN News. 21 March 2024. Archived from the original on 20 April 2024. Retrieved 21 April 2024.
  162. ^ Mäntymäki, Matti; Minkkinen, Matti; Birkstedt, Teemu; Viljanen, Mika (2022). "Defining organizational AI governance". AI and Ethics. 2 (4): 603–609. doi:10.1007/s43681-022-00143-x. ISSN 2730-5953. S2CID 247119668.
  163. ^ a b c Brundage, Miles; Avin, Shahar; Wang, Jasmine; Belfield, Haydn; Krueger, Gretchen; Hadfield, Gillian; Khlaaf, Heidy; Yang, Jingying; Toner, Helen; Fong, Ruth; Maharaj, Tegan; Koh, Pang Wei; Hooker, Sara; Leung, Jade; Trask, Andrew (2020-04-20). "Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims". arXiv:2004.07213. {{cite journal}}: Cite journal requires |journal= (help)
  164. ^ "Welcome to the Artificial Intelligence Incident Database". Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  165. ^ Wiblin, Robert; Harris, Keiran (2022). "Nova DasSarma on why information security may be critical to the safe development of AI systems". 80,000 Hours. Archived from the original on 2022-11-24. Retrieved 2022-11-24.
  166. ^ OpenAI (2022-06-02). "Best Practices for Deploying Language Models". OpenAI. Archived from the original on 2023-03-15. Retrieved 2022-11-24.
  167. ^ OpenAI. "OpenAI Charter". OpenAI. Archived from the original on 2021-03-04. Retrieved 2022-11-24.
  168. ^ Future of Life Institute (2016). "Autonomous Weapons Open Letter: AI & Robotics Researchers". Future of Life Institute. Archived from the original on 2023-09-22. Retrieved 2022-11-24.

External links[edit]