2018
Selekoa’s 70 records dated 2018, 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 4 records
Dated to 2018 only
· interval ends · dated to the year
Durk Kingma’s biography dates his OpenAI Research Scientist role to 2015–2018. Evidence ↗
February 2018
· event · dated to the day
OpenAI announced Elon Musk’s departure from its board in February 2018. Evidence ↗
April 2018
· snapshot · dated to the day · announced
Demis Hassabis announced on 11 April 2018 that Lila Ibrahim was joining DeepMind as its first Chief Operating Officer. Evidence ↗· snapshot · dated to the day
A Johns Hopkins University seminar announcement for Jeff Dean’s lecture of 26 April 2018 states that he leads the Google Brain team. Evidence ↗
Institutions 2 records
January 2018
November 2018
Ideas & lineages 64 records
January 2018
· publication · dated to the month
“Prediction, Judgment and Complexity: A Theory of Decision Making and Artificial Intelligence” (2018-01) credits Ajay K. Agrawal as one of its 3 authors. Evidence ↗· publication · dated to the month
“Prediction, Judgment and Complexity: A Theory of Decision Making and Artificial Intelligence” (2018-01) credits Avi Goldfarb as one of its 3 authors. Evidence ↗· publication · dated to the month
“Artificial Intelligence, Automation and Work” (2018-01) credits Daron Acemoglu as one of its 2 authors. Evidence ↗· publication · dated to the month
“Prediction, Judgment and Complexity: A Theory of Decision Making and Artificial Intelligence” (2018-01) credits Joshua S. Gans as one of its 3 authors. Evidence ↗· publication · dated to the month
“AI as the Next GPT: A Political-Economy Perspective” (2018-01) credits Manuel Trajtenberg as its author. Evidence ↗· publication · dated to the month
“Artificial Intelligence, Automation and Work” (2018-01) credits Pascual Restrepo as one of its 2 authors. Evidence ↗· publication · dated to the month
Acemoglu and Restrepo, “Artificial Intelligence, Automation and Work” (2018-01) bears on M01. Argues that the canonical factor-augmenting production function cannot represent automation's displacement of labour. Evidence ↗Acemoglu and Restrepo, Artificial Intelligence, Automation and Work (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Artificial Intelligence, Automation and Work” (2018-01) articulates M02. Applies the task framework explicitly to AI and robotics, emphasising a displacement effect absent from factor-augmenting models. Evidence ↗Acemoglu and Restrepo, Artificial Intelligence, Automation and Work (2018) →Claim page →
· publication · dated to the month
Trajtenberg, “AI as the Next GPT: A Political-Economy Perspective” (2018-01) bears on M03. Frames AI as the coming GPT and turns to policies to soften its displacement effects. Evidence ↗· publication · dated to the month
Agrawal, Gans, Goldfarb, “Prediction, Judgment and Complexity: A Theory of Decision Making and Artificial Intelligence” (2018-01) articulates M04. Treats AI as prediction technology and models how better prediction changes the value of judgment, the process of determining payoffs. Makes the fall in the cost of prediction the premise of the analysis. Draws the consequence of better prediction for a complement, human judgment. Evidence ↗Agrawal, Gans, Goldfarb, Prediction, Judgment and Complexity (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Artificial Intelligence, Automation and Work” (2018-01) bears on M05. Identifies new labour-intensive tasks as the main force offsetting automation and keeping growth balanced. Evidence ↗Acemoglu and Restrepo, Artificial Intelligence, Automation and Work (2018) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Prediction, Judgment and Complexity: A Theory of Decision Making and Artificial Intelligence” (2018-01) bears on M05. Notes that existing automation models (here Acemoglu and Restrepo 2017) rely on tasks that only labour can perform, and offers a prediction-based approach instead. Evidence ↗Agrawal, Gans, Goldfarb, Prediction, Judgment and Complexity (2018) →Claim page →
· publication · dated to the month
Trajtenberg, “AI as the Next GPT: A Political-Economy Perspective” (2018-01) bears on M05. Summarises a prevailing view that AI will perform most new tasks too, so net displacement will exceed past GPTs. Evidence ↗· publication · dated to the month
Trajtenberg, “AI as the Next GPT: A Political-Economy Perspective” (2018-01) contests M05. Concludes, from history, that human obsolescence is not to be presumed, while expecting many occupations to vanish. Evidence ↗· publication · dated to the month
Acemoglu and Restrepo, “Artificial Intelligence, Automation and Work” (2018-01) articulates M06a. Even with offsetting forces, automation lowers the labour share. Evidence ↗Acemoglu and Restrepo, Artificial Intelligence, Automation and Work (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Artificial Intelligence, Automation and Work” (2018-01) bears on M06b. Attributes part of the productivity slowdown to automation pushed beyond the social optimum by the tax treatment of capital and labour-market rents. Evidence ↗Acemoglu and Restrepo, Artificial Intelligence, Automation and Work (2018) →Claim page →
· publication · dated to the month
Trajtenberg, “AI as the Next GPT: A Political-Economy Perspective” (2018-01) articulates M06b. Calls for policies (education, professionalising care work, steering technical change) to ameliorate AI's detrimental effects on displaced workers. Argues that mass displacement and ageing may require governments to manage transitions, not merely compensate losers. Evidence ↗· publication · dated to the month
Acemoglu and Restrepo, “Artificial Intelligence, Automation and Work” (2018-01) bears on M09. Opens the conclusion by saying the debate on automation and the future of work lacks a satisfactory framework, which the authors blame for a false dichotomy. The authors describe two sides of the debate, alarmist predictions of the end of work and economists' reassurance from history, and reject both as a false dichotomy. Evidence ↗Acemoglu and Restrepo, Artificial Intelligence, Automation and Work (2018) →Claim page →
· publication · dated to the month
Trajtenberg, “AI as the Next GPT: A Political-Economy Perspective” (2018-01) bears on M09. Self-critique of economists' tendency to treat GDP growth as welfare and give only lip service to distribution. Says economics has traditionally neglected the direction of technical change, motivating the human-enhancing versus human-replacing distinction. Evidence ↗
February 2018
· publication · dated to the month
“Self-Regulating Artificial General Intelligence” (2018-02) credits Joshua S. Gans as its author. Evidence ↗· publication · dated to the month
Gans, “Self-Regulating Artificial General Intelligence” (2018-02) bears on E01. After quoting Good and citing Yudkowsky (2007) and Omohundro (2007) on "recursive self-improvement", Gans recasts it in economic terms. Gans's main result: a paperclip AI refrains from activating power-accumulating successor AIs. Evidence ↗Gans, Self-Regulating Artificial General Intelligence (2018) →Claim page →
· publication · dated to the month
Gans, “Self-Regulating Artificial General Intelligence” (2018-02) articulates E03. Gans takes Bostrom's control problem and "paperclip apocalypse" into economic theory. Gans states the scope of his self-regulation result. Evidence ↗Gans, Self-Regulating Artificial General Intelligence (2018) →Claim page →
· publication · dated to the month
Gans, “Self-Regulating Artificial General Intelligence” (2018-02) bears on E04. Gans draws a governance implication from his model. Evidence ↗Gans, Self-Regulating Artificial General Intelligence (2018) →Claim page →
· publication · dated to the month
Gans, “Self-Regulating Artificial General Intelligence” (2018-02) articulates E05. Gans explains why he takes a different modelling route. Evidence ↗Gans, Self-Regulating Artificial General Intelligence (2018) →Claim page →
· publication · dated to the day
Christiano’s post says it is not about whether an intelligence explosion will occur, which seems very likely to its author, and that the author does not disagree with the central thesis of Intelligence Explosion Microeconomics. Evidence ↗· publication · dated to the day
Christiano’s post says that before AI is great at self-improvement it will be mediocre at it, and that powerful AI can be used to develop better AI, leading to runaway growth. Evidence ↗· publication · dated to the day
Christiano’s post operationalises slow takeoff as a complete four-year interval in which world output doubles before the first one-year doubling, argues that it is more likely, and holds that recursive self-improvement, though it will lead to runaway growth, is not on its own an argument for a discontinuity. Evidence ↗· publication · dated to the day
Christiano’s post says powerful AI can be used to develop better AI, among other things. Evidence ↗· publication · dated to the day
Christiano’s post contrasts the fast-takeoff view that whoever builds AGI has a decisive strategic advantage with a slow takeoff, in which being twelve months ahead might give such an advantage but is about as hard as building aeroplanes thirty years before anyone else. Evidence ↗
May 2018
· publication · dated to the month
“Modeling Automation” (2018-05) credits Daron Acemoglu as one of its 2 authors. Evidence ↗· publication · dated to the month
“Modeling Automation” (2018-05) credits Pascual Restrepo as one of its 2 authors. Evidence ↗· publication · dated to the month
Acemoglu and Restrepo, “Modeling Automation” (2018-05) bears on M01. Opens by rejecting the factor-augmenting treatment of automation, implying it was an available modelling convention that the authors replace with a task-based one. Evidence ↗Acemoglu and Restrepo, Modeling Automation (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Modeling Automation” (2018-05) bears on M02. Argues for modelling automation as machines taking over labour-performed tasks rather than as factor-augmenting technical change. Evidence ↗Acemoglu and Restrepo, Modeling Automation (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “Modeling Automation” (2018-05) articulates M06a. Contrasts task automation with factor-augmenting change: the former always lowers the labour share. Evidence ↗Acemoglu and Restrepo, Modeling Automation (2018) →Claim page →
· publication · dated to the day
AI safety via debate credits Paul Christiano as a co-author. Evidence ↗· publication · dated to the day
“Open question: are minimal circuits daemon-free?” (2018-05-05) credits the username “paulfchristiano” as its author. Evidence ↗· publication · dated to the day
Christiano’s post of 2018-05-05 describes a predictor found by search that is a consequentialist motivated to make good predictions, perhaps because it has realised that only good predictors survive, and that may stop predicting well under different conditions. Evidence ↗Christiano, Are minimal circuits daemon-free? (2018) →Claim page →
· publication · dated to the day
Christiano’s post of 2018-05-05 says a daemon found by search may predict well only instrumentally and may stop doing so under different conditions. Evidence ↗Christiano, Are minimal circuits daemon-free? (2018) →Claim page →
· publication · dated to the day
“Challenges to Christiano’s capability amplification proposal” (2018-05-19) credits Eliezer Yudkowsky as its author. Evidence ↗· publication · dated to the day
The post “Challenges to Christiano’s capability amplification proposal” (2018-05-19) contests Y21. Discussing imitation-based alignment, he argues that inexact imitation of humans does not carry over properties such as being a good person, and exact imitation is harder than surpassing human capability. Evidence ↗Yudkowsky, Challenges to Christiano’s capability amplification proposal (2018) →Claim page →
June 2018
· publication · dated to the month
“Economic Policy for Artificial Intelligence” (2018-06) credits Ajay K. Agrawal as one of its 3 authors. Evidence ↗· publication · dated to the month
“Economic Policy for Artificial Intelligence” (2018-06) credits Avi Goldfarb as one of its 3 authors. Evidence ↗· publication · dated to the month
“The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment” (2018-06) credits Daron Acemoglu as one of its 2 authors. Evidence ↗· publication · dated to the month
“AI and the Economy” (2018-06) credits Jason Furman as one of its 2 authors. Evidence ↗· publication · dated to the month
“Economic Policy for Artificial Intelligence” (2018-06) credits Joshua S. Gans as one of its 3 authors. Evidence ↗· publication · dated to the month
“The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment” (2018-06) credits Pascual Restrepo as one of its 2 authors. Evidence ↗· publication · dated to the month
“AI and the Economy” (2018-06) credits Robert Seamans as one of its 2 authors. Evidence ↗· publication · dated to the month
Acemoglu and Restrepo, “The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment” (2018-06) articulates M02. The abstract sets the paper in a framework in which labour-performed tasks are automated while new labour-advantaged tasks are created, to examine concerns that labour will become redundant. Evidence ↗Acemoglu and Restrepo, The Race between Man and Machine (2018) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Economic Policy for Artificial Intelligence” (2018-06) articulates M03. Following the preceding paragraph's point that GPT co-invention "is costly and slow", infers that AI's productivity impact will lag, citing Brynjolfsson, Rock and Syverson. Evidence ↗Agrawal, Gans, Goldfarb, Economic Policy for Artificial Intelligence (2018) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Economic Policy for Artificial Intelligence” (2018-06) articulates M04. Recasts machine learning as a fall in the price of prediction and asks how policy shapes its diffusion and consequences. Defines prediction and summarises the authors' programme on consequences of cheaper prediction. Evidence ↗Agrawal, Gans, Goldfarb, Economic Policy for Artificial Intelligence (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment” (2018-06) articulates M05. A conditional model result: under a low relative cost of capital, all tasks (including those created later) end up automated; the abstract presents the alternative stable balanced-growth case as the other outcome. Evidence ↗Acemoglu and Restrepo, The Race between Man and Machine (2018) →Claim page →
· publication · dated to the month
Acemoglu and Restrepo, “The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment” (2018-06) articulates M06a. In the static model automation lowers employment and the labour share and may reduce wages, while new tasks push the other way. Evidence ↗Acemoglu and Restrepo, The Race between Man and Machine (2018) →Claim page →
· publication · dated to the month
Agrawal, Gans, Goldfarb, “Economic Policy for Artificial Intelligence” (2018-06) articulates M06a. Gives a rising capital share as one of two reasons AI might increase inequality, with skill bias the other. Evidence ↗Agrawal, Gans, Goldfarb, Economic Policy for Artificial Intelligence (2018) →Claim page →
· publication · dated to the month
Furman and Seamans, “AI and the Economy” (2018-06) articulates M06a. Opens the labour-market policy section by noting that AI may worsen declining participation and rising inequality. Evidence ↗· publication · dated to the month
Agrawal, Gans, Goldfarb, “Economic Policy for Artificial Intelligence” (2018-06) articulates M06b. Concludes that AI's consequences for jobs, inequality and competition call for education policy, the safety net and antitrust. Evidence ↗Agrawal, Gans, Goldfarb, Economic Policy for Artificial Intelligence (2018) →Claim page →
· publication · dated to the month
Furman and Seamans, “AI and the Economy” (2018-06) articulates M06b. Weighs UBI against EITC expansion and wage subsidies as responses to AI-related labour disruption, favouring the latter. Evidence ↗· publication · dated to the month
Agrawal, Gans, Goldfarb, “Economic Policy for Artificial Intelligence” (2018-06) bears on M09. Summarises economists' long-run optimism on jobs while stressing short-run adjustment costs. Evidence ↗Agrawal, Gans, Goldfarb, Economic Policy for Artificial Intelligence (2018) →Claim page →
· publication · dated to the month
Furman and Seamans, “AI and the Economy” (2018-06) bears on M09. Concludes that economic research on AI's productivity and labour effects has only begun and calls for more empirical work. Explains why the NBER volume's chapters rely on theoretical models and national statistics rather than firm-level data. Evidence ↗
October 2018
· publication · dated to the day
Supervising strong learners by amplifying weak experts cites AI safety via debate. Evidence ↗Supervising strong learners by amplifying weak experts →Claim page →
· publication · dated to the day
Supervising strong learners by amplifying weak experts credits Paul Christiano as a co-author. Evidence ↗
November 2018
· publication · dated to the day
Scalable agent alignment via reward modeling credits Jan Leike as a co-author. Evidence ↗· publication · dated to the day
Scalable agent alignment via reward modeling: a research direction cites Supervising strong learners by amplifying weak experts. Evidence ↗Scalable agent alignment via reward modeling: a research direction →Claim page →
· publication · dated to the day
Scalable agent alignment via reward modeling: a research direction cites Alignment for Advanced Machine Learning Systems. Evidence ↗Scalable agent alignment via reward modeling: a research direction →Claim page →
December 2018
· publication · dated to the day
An Empirical Model of Large-Batch Training names Sam McCandlish as a co-author. Evidence ↗