Changes

Jump to: navigation, search

Timeline of AI safety

4,350 bytes added, 11:36, 19 April 2020
no edit summary
| 2017 || This is a great year for cryptocurrency prices, causing a number of donations to MIRI from people who got rich through cryptocurrency. The AI safety funding and support landscape changes somewhat with the launch of the Berkeley Existential Risk Initiative (BERI) (and funding of its grants program by Jaan Tallinn) and the Effective Altruism Funds, specifically the Long-Term Future Fund. Open Phil makes several grants in AI safety, including a $30 million grant to OpenAI and a $3.75 million grant to MIRI. AI safety attracts dismissive commentary from Mark Zuckerberg, while Elon Musk continues to highlight its importance. The year begins with the Asilomar Conference and the Asilomar AI Principles, and initiatives such as AI Watch and the AI Alignment Prize begin toward the end of the year.
|-
| 2018 || Activity in the field of AI safety becomes more steady, in terms of both ongoing discussion (with the launch of the AI Alignment Newsletter , AI Alignment Podcast, and the Alignment Forum) and funding (with structural changes to the Long-Term Future Fund to make it grant more regularly, the introduction of the annual Open Phil AI Fellowship grants, and more grantmaking by BERI). Near the end of the year, MIRI announces its nondisclosure-by-default policy.
|-
| 2019 || The Center for Security and Emerging Technology (CSET), that is focused on AI safety and other security risks, launches with a 5-year $55 million grant from Open Phil. Grantmaking from the Long-Term Future Fund picks up pace; BERI hands off its grantmaking of Jaan Tallinn's money to the Survival and Flourishing Fund (SFF). Open Phil begins using the Committee for Effective Altruism Support to decide grant amounts for some of its AI safety grants, including grants to MIRI. OpenAI unveils its GPT-2 model but does not release the full model initially; this sparks discussion on disclosure norms.
|}
| 1950 || || Publication || {{w|Alan Turing}} publishes ''Computing Machinery and Intelligence'' in the philosophy journal ''Mind''. The paper introduces the concept of the {{w|Turing test}}, the simple idea of which is that a machine would be said to have achieved human-level intelligence if it can convince a human that it is human. The Turing test would become a key part of discussions around benchmarking artificial intelligence, and also enter popular culture over the next few decades.<ref>{{cite web|url = http://www.mind.ilstu.edu/curriculum/turing_machines/turing_test_and_machine_intelligence.php|title = Turing Test and Machine Intelligence|last = Bradley|first = Peter|publisher = Consortium on Computer Science Instruction}}</ref><ref>{{cite web|url = https://intelligence.org/2013/08/11/what-is-agi/|title = What is AGI?|last = Muehlhauser|first = Luke|date = August 11, 2013|accessdate = September 8, 2019|publisher = Machine Intelligence Research Institute}}</ref>
|-
| 1960 || {{dts|May 6}} || Publication || {{w|Norbert Wiener}}'s article ''Some Moral and Technical Consequences of Automation'' is published.<ref>{{cite web|url = https://www.jstor.org/stable/1705998|title = Some Moral and Technical Consequences of Automation|last = Wiener|first = Norbert|date = May 6, 1960|accessdate = August 18, 2019}}</ref> In 2013, Jonah Sinick would note the similarities between the points raised in this article and the thinking of AI safety leader Eliezer Yudkowsky.<ref>{{cite web|url = https://www.lesswrong.com/posts/2rWfmahhqASnFcYLr/norbert-wiener-s-paper-some-moral-and-technical-consequences|title = Norbert Wiener's paper "Some Moral and Technical Consequences of Automation"|date = July 20, 2013|accessdate = August 18, 2019|publisher = LessWrong|last = Sinick|first = Jonah}}</ref> Wiener's work would also be cited by Allan Dafoe and Stuart Russell in 2016: "Rather, the risk arises from the unpredictability and potential irreversibility of deploying an optimization process more intelligent than the humans who specified its objectives. This problem was stated clearly by Norbert Wiener in 1960, and we still have not solved it."<ref name=dafoe-russell-2016>{{cite web|url = https://www.technologyreview.com/2016/11/02/156285/yes-we-are-worried-about-the-existential-risk-of-artificial-intelligence/|title = Yes, We Are Worried About the Existential Risk of Artificial Intelligence. A defense of the warnings about AI in philosopher Nick Bostrom’s book Superintelligence.|last = Dafoe|first = Allan|last2 = Russell|first2 = Stuart|publisher = Technology Review}}</ref>
|-
| 1965 || || Publication || {{w|I. J. Good}} [[w:Existential risk from artificial general intelligence#History|originates]] the concept of intelligence explosion in "Speculations Concerning the First Ultraintelligent Machine".
|-
| 2015 || {{dts|August}} || Publication || The Open Philanthropy Project publishes its cause report on potential risks from advanced artificial intelligence.<ref>{{cite web |url=http://www.openphilanthropy.org/research/cause-reports/ai-risk |title=Potential Risks from Advanced Artificial Intelligence |publisher=Open Philanthropy Project |accessdate=July 27, 2017}}</ref>
|-
| 2015 || {{dts|August 29}} || || The ControlProblem subreddit (full title: "The Artificial General Intelligence Control Problem") is created.<ref>{{cite web|url = https://www.reddit.com/r/ControlProblem/|title = The Artificial General Intelligence Control Problem|accessdate = April 18, 2020}}</ref>
|-
| 2015 || {{dts|October}} || Publication || The Open Philanthropy Project first publishes its page on AI timelines.<ref>{{cite web |url=http://www.openphilanthropy.org/focus/global-catastrophic-risks/potential-risks-advanced-artificial-intelligence/ai-timelines |title=What Do We Know about AI Timelines? |publisher=Open Philanthropy Project |accessdate=July 25, 2017}}</ref>
|-
| 2017 || {{dts|October}} || Project || The first commit for AI Watch, a repository of organizations, people, and products in AI safety, is made on October 23.<ref>{{cite web|url = https://github.com/riceissa/aiwatch/commit/563c48d9dcfd126b475f9a982e8d7af6c743411c|title = Initial commit: AI Watch|last = Rice|first = Issa|date = October 23, 2017|accessdate = April 19, 2020}}</ref> Work on the web portal at aiwatch.issarice.com would begin the next day.<ref>{{cite web|url = https://github.com/riceissa/aiwatch/commit/5329e8c34e599ca11f349cbe29427c1f4b73a20f#diff-b9e00bb3999fd5777328f867e74bbc9e|title = start on portal: AI Watch|date = October 24, 2017|accessdate = April 19, 2020|last = Rice|first = Issa}}</ref>
|-
| 2017 || {{dts|October}} {{snd}} December || Project || FHI launches its Governance of AI Program (later to be called the Centre for the Governance of AI, and shortened as GovAI), co-directed by Nick Bostrom and Allan Dafoe.<ref name="newsletter-winter-2017">{{cite web |url=https://www.fhi.ox.ac.uk/quarterly-update-winter-2017/ |author=Future of Humanity Institute - FHI |title=Quarterly Update Winter 2017 - Future of Humanity Institute |publisher=Future of Humanity Institute |date=January 19, 2018 |accessdate=March 14, 2018}}</ref>
|-
| 2017 || {{dts|November 3}} || Project || Zvi Mowshowitz and Vladimir Slepnev announce the AI Alignment Prize, a $5,000 prize funded by Paul Christiano for publicly posted work advancing AI alignment.<ref>{{cite web|url = https://www.lesswrong.com/posts/YDLGLnzJTKMEtti7Z/announcing-the-ai-alignment-prize|title = Announcing the AI Alignment Prize|date = November 3, 2017|accessdate = April 19, 2020|last = Slepnev|first = Vladimir|publisher = LessWrong}}</ref> The prize would be discontinued after the fourth round (ending December 31, 2018) due to reduced participation.<ref>{{cite web|url = https://www.alignmentforum.org/posts/nDHbgjdddG5EN6ocg/announcement-ai-alignment-prize-round-4-winners|title = Announcement: AI alignment prize round 4 winners|last = Slepnev|first = Vladimir|date = Janaury 20, 2019|accessdate = April 19, 2020|publisher = Alignment Forum}}</ref>
|-
| 2018 || {{dts|February 28}} || Publication || 80,000 Hours publishes a blog post ''A new recommended career path for effective altruists: China specialist'' suggesting specialization in China as a career path for people in the effective altruist movement. China's likely leading role in the development of artificial intelligence is highlighted as particularly relevant to AI safety efforts.<ref>{{cite web|url = https://80000hours.org/articles/china-careers/|title = 'A new recommended career path for effective altruists: China specialist|last = Todd|first = Benjamin|last2 = Tse|first2 = Brian|date = February 28, 2018|accessdate = September 8, 2019|publisher = 80,000 Hours}}</ref>
|-
| 2018 || {{dts|April 4}} || Podcast || The first AI Alignment Podcast from the Future of Life Institute is on this day. The AI Alignment Podcast would happen at an approximate frequency of one per month from this point onward, but with no set cadence.<ref>{{cite web|url = https://futureoflife.org/ai-alignment-podcast/|title = AI Alignment Podcast|publisher = Future of Life Institute|accessdate = April 4, 2020}}</ref>
|-
| 2018 || {{dts|April 5}} || Documentary || The documentary ''{{w|Do You Trust This Computer?}}'', directed by {{w|Chris Paine}}, is released. It covers issues related to AI safety and includes interviews with prominent individuals relevant to AI, such as {{w|Ray Kurzweil}}, {{w|Elon Musk}} and {{w|Jonathan Nolan}}.
|-
| 2019 || {{dts|February}} || Funding || The Open Philanthropy Project grants $2,112,500 to the Machine Intelligence Research Institute (MIRI) over two years. This is part of the first batch of grants decided by the Committee for Effective Altruism Support, which will set "grant sizes for a number of our largest grantees in the effective altruism community, including those who work on long-termist causes."<ref name="donations-portal-open-phil-ai-safety" /> Around the same time, BERI grants $600,000 to MIRI.<ref name="donations-portal-beri-ai-safety">{{cite web|url = https://donations.vipulnaik.com/donor.php?donor=Berkeley+Existential+Risk+Initiative&cause_area_filter=AI+safety|title = Berkeley Existential Risk Initiative donations made (filtered to cause areas matching AI safety)|accessdate = August 18, 2019}}</ref>
|-
| 2019 || {{dts|February 14}} || Disclosure norms || OpenAI unveils its language-generating system called GPT-2, a system able to write the news, answer reading comprehension problems, and is beginning to show promise at tasks like translation.<ref>{{cite web |title=An AI helped us write this article |url=https://www.vox.com/future-perfect/2019/2/14/18222270/artificial-intelligence-open-ai-natural-language-processing |website=vox.com |accessdate=28 June 2019}}</ref> However, the data or the parameters of the model are not released, under expressed concerns about potential abuse. This would lead to a lot of discussion and controversy.<ref>{{cite web |last1=Lowe |first1=Ryan |title=OpenAI’s GPT-2: the model, the hype, and the controversy |url=https://towardsdatascience.com/OpenAIs-gpt-2-the-model-the-hype-and-the-controversy-1109f4bfd5e8 |website=towardsdatascience.com |accessdate=10 July 2019}}</ref>
|-
| 2019 || {{dts|April 7}} || Funding || The Long-Term Future Fund, one of the Effective Altruism Funds, announces a set of 23 grants totaling $923,150. About half the grant money is to organizations or projects directly working in AI safety. Recipients include the Machine Intelligence Research Institute (MIRI), AI Safety Camp, Ought, and a number of individuals working on AI safety projects, including three in deconfusion research.<ref name="donations-portal-ea-funds-ai-safety" />
| 2019 || {{dts|June 7}} || Fictional portrayal || The movie ''{{w|I Am Mother}}'' is released on Netflix. According to a comment on Slate Star Codex: "you can use it to illustrate everything from paperclip maximization to deontological kill switches".<ref>{{cite web|url = https://slatestarcodex.com/2019/06/05/open-thread-129-25/|title = OPEN THREAD 129.25|date = June 8, 2019|accessdate = August 18, 2019}}</ref>
|-
| 2019 || {{dts|June 12}} || Disclosure norms || In a blog post titled "The Hacker Learns to Trust" Connor Leahy explains why he decided against releasing his own replication of OpenAI's GPT-2 model. Leahy says that helpful discussions with OpenAI team members (Jack Clark, Alec Radford and Jeff Wu), and a convincing argument from Buck Shlegeris of MIRI, led him to change his mind. Highlighted text, based on the conversation with Shlegeris: "Because this isn’t just about GPT2. What matters is that at some point in the future, someone will create something truly dangerous and there need to be commonly accepted safety norms before that happens."<ref>{{cite web|url = https://medium.com/@NPCollapse/the-hacker-learns-to-trust-62f3c1490f51|title = The Hacker Learns to Trust|date = June 12, 2019|accessdate = April 19, 2020|last = Leahy|first = Connor}}</ref> The post is linkposted to LessWrong, where it attracts more comments.<ref>{{cite web|url = https://www.lesswrong.com/posts/36fxiKdEqswkedHyG/the-hacker-learns-to-trust|title = The Hacker Learns to Trust|last = Pace|first = Ben|date = June 21, 2019|accessdate = April 19, 2020|publisher = LessWrong}}</ref>|-| 2019 || {{dts|August 25}} || Funding || Grantmaking done by Berkeley Existential Risk Initiative (BERI) funded by {{w|Jaan Tallinn}} moves to the newly created Survival and Flourishing Fund (SFF).<ref>{{cite web|url = http://existence.org/tallinn-grants-future/|title = The Future of Grant-making Funded by Jaan Tallinn at BERI|date = August 25, 2019|accessdate = April 18, 2020|publisher = Berkeley Existential Risk Initiative}}</ref> BERI's grantmaking in this space had previously included AI safety organizations. However, as of April 2020, SFF's grants have not included any grants to organizations exclusively focused on AI safety, but rather been to organizations working on broader global catastrophic risks and other adjacent topics.
|-
| 2019 || {{dts|August 30}} || Funding || The Long-Term Future Fund, one of the Effective Altruism Funds, announces a set 13 grants totaling $415,697 USD to organizations and individuals. About half the grant money is to organizations or projects working in AI safety and related AI strategy, governance, and policy issues. With the exception of a grant to AI Safety Camp, all the other grants related to AI safety are to individuals.<ref>{{cite web|url = https://app.effectivealtruism.org/funds/far-future/payouts/4UBI3Q0TBGbWcIZWCh4EQV|title = August 2019: Long-Term Future Fund Grants and Recommendations|date = August 30, 2019|accessdate = April 18, 2020|publisher = Effective Altruism Funds}}</ref>
| 2019 || {{dts|December 18}} || Publication || The "2019 AI Alignment Literature Review and Charity Comparison" is published on the Effective Altruism Forum. It surveys AI safety work in 2019. It continues an annual tradition of similar blog posts in 2016, 2017, and 2018. One feature new to the document this year is the author's effort to make it easier for readers to jump to and focus on parts of the document most relevant to them, rather than read it beginning to end. To make this easier, the author ends each paragraph with a hashtag, and lists the hashtags at the beginning of the document.<ref name=larks-2019>{{cite web|url = https://forum.effectivealtruism.org/posts/dpBB24QsnsRnkq5JT/2019-ai-alignment-literature-review-and-charity-comparison|title = 2019 AI Alignment Literature Review and Charity Comparison|author = Larks|publisher = Effective Altruism Forum|date = December 18, 2019|accessdate = April 18, 2020}}</ref>
|-
| 2020 || {{dts|January|| Publication || Rohin Shah (the person who started the weekly AI Alignment Newsletter) publishes a blog post on LessWrong (cross-posted to the Alignment Forum) titled "AI Alignment 2018-19 Review" that he describes as "a review post of public work in AI alignment over 2019, with some inclusions from 2018."<ref>{{cite web|url = https://www.lesswrong.com/posts/dKxX76SCfCvceJXHv/ai-alignment-2018-19-review|title = AI Alignment 2018-19 Review|last = Shah|first = Rohin|date = January 27, 2020|accessdate = April 18, 2020|publisher = LessWrong}}</ref>
|-
| 2020 || {{dts|February}} || Funding || The Open Philanthropy Project makes grants to AI safety organizations Ought ($1.59 million) and Machine Intelligence Research Institute ($7.7 million) with the money amount determined by the Committee for Effective Altruism Support (CEAS). Other organizations receiving money based on CEAS recommendations at around the same time are the Centre for Effective Altruism and 80,000 Hours, neither of which is primarily focused on AI safety.<ref name="donations-portal-open-phil-ai-safety" />
* [http://inst.eecs.berkeley.edu/~cs294-149/fa18/ CS 294-149: Safety and Control for Artificial General Intelligence (Fall 2018)], taught by Andrew Critch and Stuart Russell
* Median Group (insofar as some of their research is about AI)
* Norms set by OpenAI's decision not to release the full GPT2 model, and some of the discussion this sparked
More detail would help for these rows:
* [[Timeline of artificial intelligence]]
* [[Timeline of robotics]]
 
===Other timelines related to topics with potential applications to AI safety===
 
* [[Timeline of decision theory]]
===Other timelines about cause areas prioritized in effective altruism===
2,422
edits

Navigation menu