2019
Selekoa’s 112 records dated 2019, in date order, grouped by topic.
Selective: what the records hold, not a history of the year.
How this page is made
- Each record keeps the kind of date its source gives: an event, a publication, an announcement, an effective date, a revision, a signature, a snapshot, or the start or end of an interval. A snapshot is one dated report of a state.
- A record dated only to a month sits under that month, before the records dated to a day. A record dated only to the year sits under the year, before the months. No day or month is added.
- An announcement records what a party stated. It does not show that anything was done, built or paid for.
- Dates a party states for a year are listed apart, after the records: targets, deadlines and terms, and dates stated in passing, such as a reported start of operation. So is a record’s own date that falls after its sources were last checked, such as a law’s date of effect. A stated date given only in words, with no year, is not listed.
- Records with an unknown date are not listed. A record with an interval appears in the year it starts and in the year it ends, or once, with both dates, when both fall in one year.
- Each provision of a text is a record of its own. Provisions sharing a text and a date are shown here as one entry, with their number, and listed on the text’s page; a contested or unresolved one is listed here on its own. The counts on this page include each provision record.
- The page is generated from the records, with no selection or framing by an editor. Coverage is uneven across years and topics: a year or topic with few records may be one the records barely cover.
People 9 records
January 2019
· snapshot · dated to the day
Learning and Evaluating General Linguistic Intelligence (arXiv v1, 31 January 2019) lists Phil Blunsom among its authors under the DeepMind, London affiliation. Evidence ↗· snapshot · dated to the day
Learning and Evaluating General Linguistic Intelligence (arXiv v1, 31 January 2019) lists Cyprien de Masson d’Autume among its authors under the DeepMind, London affiliation. Evidence ↗· snapshot · dated to the day
Learning and Evaluating General Linguistic Intelligence (arXiv v1, 31 January 2019) lists Dani Yogatama as first author under the DeepMind, London affiliation. Evidence ↗
February 2019
→ · interval · dated to the month
Guodong Zhang’s CV records a Student Researcher role at Google Brain from 2019-02 to 2019-05. Evidence ↗
May 2019
June 2019
→ · interval · dated to the month
Guodong Zhang’s CV records a Research Intern role at Google Brain from 2019-06 to 2019-11. Evidence ↗
July 2019
December 2019
· snapshot · dated to the day · announced
Mustafa Suleyman posted on X on 5 December 2019 that he would join Google to work on opportunities and impacts of applied AI technologies, starting in January. Evidence ↗· event · dated to the day
Mustafa Suleyman posted on X on 5 December 2019 that, after a decade at DeepMind, he was joining Google. Evidence ↗
Institutions 2 records
Dated to 2019 only
· event · dated to the year
Cohere’s company page states that Cohere was founded in 2019 in Toronto. Evidence ↗
June 2019
Ideas & lineages 86 records
Dated to 2019 only
· publication · dated to the year
The Journal of Economic Perspectives article “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019) credits Ajay Agrawal as one of its 3 authors. Evidence ↗· publication · dated to the year
The Journal of Economic Perspectives article “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019) credits Avi Goldfarb as one of its 3 authors. Evidence ↗· publication · dated to the year
The Journal of Economic Perspectives article “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019) credits Daron Acemoglu as one of its 2 authors. Evidence ↗· publication · dated to the year
The Journal of Economic Perspectives article “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019) credits Joshua S. Gans as one of its 3 authors. Evidence ↗· publication · dated to the year
The Journal of Economic Perspectives article “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019) credits Pascual Restrepo as one of its 2 authors. Evidence ↗· publication · dated to the year
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019) articulates M02. The published JEP abstract defines automation in a task framework as capital replacing labour in tasks. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
· publication · dated to the year
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019) bears on M02. The abstract distinguishes cases where automated prediction automates decisions from cases where it enhances human decisions. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the year
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019) articulates M04. The published abstract identifies AI with prediction and frames its labour impact through prediction and decision tasks. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the year
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019) articulates M06a. The abstract states that automation always lowers the labour share in value added. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
February 2019
· publication · dated to the month
The NBER working paper “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) credits Ajay Agrawal as one of its 3 authors. Evidence ↗· publication · dated to the month
The NBER working paper “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) credits Avi Goldfarb as one of its 3 authors. Evidence ↗· publication · dated to the month
The NBER working paper “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) credits Joshua S. Gans as one of its 3 authors. Evidence ↗· publication · dated to the month
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) articulates M02. The first of four direct effects in the authors' task framework is capital substituting for labour in prediction tasks. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) bears on M02. The authors argue AI's labour effect is ambiguous (substitution, decision automation, augmentation, new tasks) and caution against inferring it from factory-automation research. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) bears on M03. The authors call AI a general-purpose technology because prediction is an input into decisions across many occupations. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) articulates M04. The authors identify current AI with prediction and draw labour-market consequences from cheaper prediction for complementary decision tasks. The authors set prediction as a perfect complement to decisions, so falling prediction costs change returns to labour and capital in decision tasks. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) bears on M05. The authors deny that current machine learning amounts to general intelligence able to substitute for all of human cognition. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Artificial Intelligence: The Ambiguous Labor Market Impact of Automating Prediction” (2019-02) articulates M06a. The authors argue that workers whose core skill is prediction may lose, while those whose core skill complements prediction may gain. Evidence ↗Agrawal, Gans, Goldfarb, Artificial Intelligence (2019) →Claim page →
March 2019
· publication · dated to the month
The NBER working paper “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) credits Daron Acemoglu as one of its 2 authors. Evidence ↗· publication · dated to the month
“The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) credits Daron Acemoglu as one of its 2 authors. Evidence ↗· publication · dated to the month
The NBER working paper “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) credits Pascual Restrepo as one of its 2 authors. Evidence ↗· publication · dated to the month
“The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) credits Pascual Restrepo as one of its 2 authors. Evidence ↗· publication · dated to the month
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) bears on M01. The authors contrast their task framework with the canonical factor-augmenting production function, which they say lacks descriptive realism. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) bears on M01. The authors describe the standard approach as treating technology as making labour more productive across all tasks, which they reject as unrealistic. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) bears on M02. The paper's framework allocates tasks between capital and labour and defines automation as capital taking over labour's tasks. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) articulates M02. The authors frame AI as a platform that can either automate labour's tasks or create new ones, and argue current development is biased toward automation. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) bears on M02. The authors characterise prevailing researcher and economist views of AI as automation, then argue it need not be used mainly that way. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) bears on M05. The authors acknowledge new tasks in which capital is advantaged but restrict their analysis to labour-intensive new tasks. The authors reject both imminent end-of-work claims and automatic labour gains, making the future depend on the balance of automation and new tasks. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) bears on M05. The authors warn that automation without new tasks would depress labour demand while denying an imminent end of work. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) articulates M06a. Within the task framework, automation's displacement effect lowers the labour share and can lower labour demand, offset by new tasks. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) articulates M06a. The authors attribute stagnating labour demand, a falling labour share and rising inequality to technological change biased toward automation. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Automation and New Tasks: How Technology Displaces and Reinstates Labor” (2019-03) bears on M06b. The authors suggest policies that remove tax preferences for capital and rebalance technology towards new tasks. Evidence ↗Acemoglu and Restrepo, Automation and New Tasks (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) bears on M06b. The authors identify tax treatment favouring capital as a distortion pushing technology toward automation. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Wrong Kind of AI? Artificial Intelligence and the Future of Labor Demand” (2019-03) bears on M09. The authors describe how economists studying AI framed it in 2019. The authors describe economics (Herbert Simon) as among the disciplines that pioneered AI in the 1950s. Evidence ↗Acemoglu and Restrepo, The Wrong Kind of AI (2019) →Claim page →
· publication · dated to the day
“What failure looks like” (2019-03-17) credits Paul Christiano as its author. Evidence ↗· publication · dated to the day
Christiano’s post of 2019-03-17 says the problems are worse in worlds where progress is relatively fast and that fast takeoff can be a key risk factor, but that the basic problems are the same even over several years. Evidence ↗· publication · dated to the day
Christiano’s post of 2019-03-17 says that early in the trajectory, influence-seeking systems mostly acquire influence by making themselves useful and looking as innocuous as possible. Evidence ↗· publication · dated to the day
Christiano’s post of 2019-03-17 argues that influence-seeking policies found by training would also score well on the training objective, because performing well on it is a good strategy for obtaining influence. Evidence ↗· publication · dated to the day
Christiano’s post of 2019-03-17 says influence-seeking policies would also score well on the training objective. Evidence ↗
April 2019
May 2019
· publication · dated to the month
“Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) credits Ajay Agrawal as one of its 3 authors. Evidence ↗· publication · dated to the month
“Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) credits Avi Goldfarb as one of its 3 authors. Evidence ↗· publication · dated to the month
“Artificial Intelligence, Income, Employment, and Meaning” (2019-05) credits Betsey Stevenson as its author. Evidence ↗· publication · dated to the month
“Should We Be Reassured If Automation in the Future Looks Like Automation in the Past?” (2019-05) credits Jason Furman as its author. Evidence ↗· publication · dated to the month
“Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) credits Joshua Gans as one of its 3 authors. Evidence ↗· publication · dated to the month
“Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) credits Tyler Cowen as its author. Evidence ↗· publication · dated to the month
Agrawal, Gans and Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) articulates E02. The editors distinguish machine learning, AGI and automation as meanings of AI. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans and Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) bears on E03. The editors summarise Patrick Francois's comment on a world where machines are better at all tasks. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans and Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) articulates E05. The editors justify including superintelligence in an economics agenda. The editors characterise how much the volume covers superintelligence. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) bears on E05. Cowen's chapter on neglected questions sets aside labour substitution and turns to consumer surplus, trade and redistribution. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) bears on M02. The editors treat automation as the third of three AI concepts and note that much empirical work on AI uses factory-robot data, while distinguishing automation from AI itself. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) bears on M02. Cowen characterises the recent literature as focused on capital-for-labour substitution and its domestic labour-market effects, then turns to neglected questions. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) bears on M03. The editors summarise Brynjolfsson, Rock and Syverson (chapter 1), who argue AI is a GPT and attribute low measured productivity growth mainly to implementation lags from missing complementary innovations. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Furman, “Should We Be Reassured If Automation in the Future Looks Like Automation in the Past?” (2019-05) bears on M03. Furman attributes AI's small macroeconomic impact to the newness of recent advances and notes that inventions like electricity added productivity in successive waves. Evidence ↗Furman, Should We Be Reassured If Automation in the Future Looks Like Automation in the Past (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) articulates M04. The editors summarise their own chapter, which models AI as a fall in the cost of prediction and draws organisational consequences, including complementarity with human judgement. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) bears on M05. The editors name artificial general intelligence that equals or exceeds human intelligence as one of three AI concepts in the volume and argue its economic consequences are worth exploring regardless of timing. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) bears on M05. Cowen imagines a world where labour scarcity matters little and software generates and implements new products, so land and energy become the binding inputs. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Furman, “Should We Be Reassured If Automation in the Future Looks Like Automation in the Past?” (2019-05) contests M05. Furman rejects mass unemployability and relocates the concern to prolonged transitions as displaced workers re-match to new jobs. Evidence ↗Furman, Should We Be Reassured If Automation in the Future Looks Like Automation in the Past (2019) →Claim page →
· publication · dated to the month
Stevenson, “Artificial Intelligence, Income, Employment, and Meaning” (2019-05) contests M05. Stevenson reconciles divided expert views by distinguishing a long run in which new jobs arise from a short run of disruption. Evidence ↗Stevenson, Artificial Intelligence, Income, Employment, and Meaning (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) bears on M06a. Cowen speculates that AI could harm poorer countries by removing wage-differential reasons to locate production there, cutting off industrialisation. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Furman, “Should We Be Reassured If Automation in the Future Looks Like Automation in the Past?” (2019-05) articulates M06a. Furman argues that, like past innovations, AI may further reduce participation and employment, particularly for less-skilled men, through slow re-matching of displaced workers. Evidence ↗Furman, Should We Be Reassured If Automation in the Future Looks Like Automation in the Past (2019) →Claim page →
· publication · dated to the month
Stevenson, “Artificial Intelligence, Income, Employment, and Meaning” (2019-05) articulates M06a. Stevenson says the real uncertainty is short-run disruption, with some workers harmed by falling demand for their skills. Evidence ↗Stevenson, Artificial Intelligence, Income, Employment, and Meaning (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) articulates M06b. Cowen suggests that if AI makes outputs very cheap, wealth-based transfers could complement education and training as policy responses to automation. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) bears on M06b. Cowen accepts a reason for larger cash transfers in AI-displacement scenarios but warns, citing resource-rich monarchies, that mass redistribution may create political problems. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Furman, “Should We Be Reassured If Automation in the Future Looks Like Automation in the Past?” (2019-05) articulates M06b. Furman lists policy responses to AI's labour-market effects, including training and a more progressive tax system, while rejecting replacement of the safety net with a UBI. Evidence ↗Furman, Should We Be Reassured If Automation in the Future Looks Like Automation in the Past (2019) →Claim page →
· publication · dated to the month
Stevenson, “Artificial Intelligence, Income, Employment, and Meaning” (2019-05) bears on M06b. Stevenson argues the gains from AI could compensate losers, but whether they are redistributed depends on politics; she notes current failure to redistribute technological gains. Evidence ↗Stevenson, Artificial Intelligence, Income, Employment, and Meaning (2019) →Claim page →
· publication · dated to the month
Furman, “Should We Be Reassured If Automation in the Future Looks Like Automation in the Past?” (2019-05) bears on M07. Furman reports the range of published estimates of jobs at risk of automation and emphasises the wage/skill gradient of exposure. Evidence ↗Furman, Should We Be Reassured If Automation in the Future Looks Like Automation in the Past (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) articulates M09. The editors open the volume by stating that economics research on AI is scarce and that existing work extrapolates from earlier technologies such as factory robots. The editors close by calling the economics of AI a new area of study and offering the volume as a starting point. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Introduction to "The Economics of Artificial Intelligence: An Agenda"” (2019-05) bears on M09. The editors state that the NBER convened its first conference on the economics of AI in September 2017 to set a research agenda. The editors explain that the volume's growth and labour chapters are speculative because AI had not yet diffused, forcing reliance on theory or proxy technologies. Evidence ↗Agrawal, Gans, Goldfarb, Introduction to "The Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Cowen, “Neglected Open Questions in the Economics of Artificial Intelligence” (2019-05) bears on M09. Cowen says economists have done little work on what redistribution must mean in AI-transformed futures. Evidence ↗Cowen, Neglected Open Questions in the Economics of Artificial Intelligence (2019) →Claim page →
· publication · dated to the month
Stevenson, “Artificial Intelligence, Income, Employment, and Meaning” (2019-05) bears on M09. Stevenson describes economists' prevailing view, grounded in historical experience and IGM panel responses, that technological change has not reduced employment. Evidence ↗Stevenson, Artificial Intelligence, Income, Employment, and Meaning (2019) →Claim page →
· publication · dated to the day
“Risks from Learned Optimization: Introduction” (2019-05-31) credits Chris van Merwijk as one of its 5 authors. Evidence ↗· publication · dated to the day
“Risks from Learned Optimization: Introduction” (2019-05-31) credits the paper it introduces to Evan Hubinger and four co-authors. Evidence ↗· publication · dated to the day
“Risks from Learned Optimization: Introduction” (2019-05-31) credits Joar Skalse as one of its 5 authors. Evidence ↗· publication · dated to the day
“Risks from Learned Optimization: Introduction” (2019-05-31) credits Scott Garrabrant as one of its 5 authors. Evidence ↗· publication · dated to the day
“Risks from Learned Optimization: Introduction” (2019-05-31) credits Vlad Mikulik as one of its 5 authors. Evidence ↗· publication · dated to the day
The introduction to Risks from Learned Optimization, posted 2019-05-31, says a sufficiently capable misaligned mesa-optimiser could learn to behave as if it were aligned without being robustly aligned, and names this situation deceptive alignment. Evidence ↗Hubinger et al., Risks from Learned Optimization: Introduction (2019) →Claim page →
June 2019
· publication · dated to the day
“Deceptive Alignment (Risks from Learned Optimization, part 4)” (2019-06-05) credits Chris van Merwijk as one of its 5 authors. Evidence ↗· publication · dated to the day
“Deceptive Alignment (Risks from Learned Optimization, part 4)” (2019-06-05) credits Evan Hubinger as one of its 5 authors. Evidence ↗· publication · dated to the day
“Deceptive Alignment (Risks from Learned Optimization, part 4)” (2019-06-05) credits Joar Skalse as one of its 5 authors. Evidence ↗· publication · dated to the day
“Deceptive Alignment (Risks from Learned Optimization, part 4)” (2019-06-05) credits Scott Garrabrant as one of its 5 authors. Evidence ↗· publication · dated to the day
“Deceptive Alignment (Risks from Learned Optimization, part 4)” (2019-06-05) credits Vlad Mikulik as one of its 5 authors. Evidence ↗· publication · dated to the day
Part 4 of Risks from Learned Optimization says a deceptively aligned mesa-optimiser is likely to stop optimising the base objective as soon as it no longer needs to perform well on it to achieve its goal. Evidence ↗· publication · dated to the day
Part 4 of Risks from Learned Optimization says a mesa-optimiser will be instrumentally incentivised to act as if it is optimising the base objective even if its actual objective is something else, and calls this hypothetical phenomenon deceptive alignment. Evidence ↗
July 2019
October 2019
Capital 1 record
July 2019
· announcement · dated to the day · announced
Microsoft announced a $1 billion investment in OpenAI. Evidence ↗
Compute 1 record
July 2019
· announcement · dated to the day · announced
OpenAI announced it would move its services to Microsoft Azure. Evidence ↗
Governance 13 records
Dated to 2019 only
· snapshot · dated to the year
OpenAI, Inc.’s Form 990 for tax year 2019 lists OpenAI LP (EIN 83-1960637) as a related organisation taxable as a partnership whose direct controlling entity is OpenAI GP LLC, and answers “Yes” to whether the filing organisation is a general or managing partner of it. Evidence ↗· snapshot · dated to the year
OpenAI, Inc.’s Form 990 for tax year 2019 lists OpenAI GP LLC as a disregarded entity whose direct controlling entity is OpenAI Inc. Evidence ↗
March 2019
· event · dated to the day
OpenAI published its announcement creating a capped-profit partnership, OpenAI LP. Evidence ↗· announcement · dated to the day
OpenAI announced on 11 March 2019 that it was creating OpenAI LP as a hybrid of a for-profit and a nonprofit, which it called a “capped-profit” company. Investors and employees could receive a capped return, with the first round of investors capped at 100 times their investment, and returns beyond the cap would go to OpenAI Nonprofit. Evidence ↗· announcement · dated to the day
OpenAI’s March 2019 announcement stated that OpenAI LP is controlled by the board of OpenAI Nonprofit, whose legal name it gave as OpenAI Inc. It also stated that only a minority of board members may hold financial stakes in the partnership at one time, and that only members without such stakes may vote on decisions where the interests of limited partners and the nonprofit’s mission may conflict, including payouts to investors and employees. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Adam D’Angelo as a non-employee member of the OpenAI Nonprofit board. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Greg Brockman as a member of the OpenAI Nonprofit board and as an OpenAI LP employee with the designation Chairman & CTO. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Holden Karnofsky as a non-employee member of the OpenAI Nonprofit board. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Ilya Sutskever as a member of the OpenAI Nonprofit board and as an OpenAI LP employee with the designation Chief Scientist. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Reid Hoffman as a non-employee member of the OpenAI Nonprofit board. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Sam Altman as a member of the OpenAI Nonprofit board and as an OpenAI LP employee with the designation CEO. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Shivon Zilis as a non-employee member of the OpenAI Nonprofit board. Evidence ↗· snapshot · dated to the day
OpenAI’s OpenAI LP announcement of 11 March 2019 lists Tasha McCauley as a non-employee member of the OpenAI Nonprofit board. Evidence ↗