AGI Zeitgeist

Panel evidence

A13 panel checks: group d1 (core tier)

A working document. It is one of the documents the site is built from, published as it is written for anyone checking the work, and it uses the project's own shorthand. The plain account of the same material is on the home page.

A13 panel checks: group d1 (core tier)

Checked 2026-09-30. "Fetched" = I read the page myself. "Seen in search result" = the item showed up in a search result's title, URL or summary but I could not open the page. A few domains (nytimes.com, theguardian.com, wired.com, newscientist.com, theatlantic.com, bbc) can't be searched or fetched with these tools, which drags down the national-outlet counts (C3) for everyone. Treat those counts as "at least", not as a full survey.

Steven Pinker

Tier checked: core C1 earliest: 2008-06-01 | possibility (rejects the singularity) | "There is not the slightest reason to believe in a coming singularity." | https://spectrum.ieee.org/tech-luminaries-address-singularity | primary (his quoted answer in IEEE Spectrum's "Tech Luminaries Address Singularity"), fetched C2 latest: 2024-06-18 | no verbatim quote found | Episode "Steven Pinker: Why Smart People Believe Stupid Things", Honestly with Bari Weiss (host Michael Moynihan). Air date seen in search result. A critic's write-up dated 2024-06-21 (https://lironshapira.substack.com/p/ai-doom-debate-steven-pinker-vs-liron, fetched) describes Pinker on the episode arguing that "AI doom worries are overblown" but quotes none of his words | secondhand. Most recent verbatim statement found: 2023-12-20, Joseph Walker podcast with David Deutsch (https://josephnoelwalker.com/153-deutsch-pinker-dialogue/, fetched): "We'd be best off building AIs as tools rather than as agents or rivals". That predates 2024. Warning: a search-engine summary claimed Pinker gave 2026 testimony to Canada's House of Commons calling self-improving AI dangerous. That quote is unverified, I believe it is misattributed, and it should not be used. C3 pieces: 1. 2008-06-01 | IEEE Spectrum | "Tech Luminaries Address Singularity" | https://spectrum.ieee.org/tech-luminaries-address-singularity | fetched. (Borderline: IEEE Spectrum is a professional society's magazine. I count it as a national science and technology magazine, but John should rule on it.) 2. None found. Checked and ruled out: Popular Science 2018-02-14 by Pinker himself (https://www.popsci.com/?p=147391, too late and written by him); ArtsJournal 2018-04-27 (aggregator, too late); Fortune 2018-03-03 on Musk vs Pinker (seen in search result, too late); Time 2011-02-10 "2045: The Year Man Becomes Immortal" and Time 1996-03-25 "Can Machines Think?" (both fetched, neither names Pinker). The 2015 Edge essays are not a national outlet. 3. None found. Quota tags: skeptic: before 2017 he dismissed the singularity outright, "not the slightest reason to believe" (2008-06-01, https://spectrum.ieee.org/tech-luminaries-address-singularity) | philosopher: cognitive scientist by appointment. He taught in MIT's Department of Brain and Cognitive Sciences 1982–2003 and co-directed its Center for Cognitive Science 1985–94; PhD in experimental psychology (https://en.wikipedia.org/wiki/Steven_Pinker, fetched 2026-09-30) | non_us_uk: does not apply. Canadian-born, but at MIT and Harvard (US) through his pre-2017 statements | frontier_lab: does not apply Verdict: fails on C3 as checked (1 borderline piece of the 3 needed); C2 is unclear (a 2024 appearance on the record, but no verbatim 2024+ quote fetched). C1 passes. He is a strong skeptic candidate, so a second pass with access to NYT and Guardian archives is worth doing before dropping him. Pinker is often quoted in the press, so this failure may come from the tools rather than the record. Queries: (1) Steven Pinker artificial general intelligence 2024 interview AGI; (2) Pinker "AGI" 2025 superintelligence skeptic interview; (3) Steven Pinker artificial intelligence 2024 OR 2025 "Pinker" said AI "general intelligence" incoherent [extended]; (4) "Pinker" AI doom superintelligence podcast 2024; (5) Steven Pinker Honestly Bari Weiss Moynihan AI 2024; (6) Steven Pinker artificial intelligence robots takeover fears 2015 OR 2016 OR 2014 newspaper; (7) Pinker artificial intelligence singularity future [restricted to national-outlet domains; two earlier attempts were rejected for blocked domains and never ran]; (8) "Steven Pinker" artificial intelligence threat robots 2015 Hawking Musk fears overblown. 8 of 8 used.

David Chalmers

Tier checked: core C1 earliest: 2010 (Journal of Consciousness Studies 17:7-65) | timing | "I would be surprised if there were human-level AI within the next three decades. Nevertheless, my credence that there will be human-level AI before 2100 is somewhere over one-half." (37 words, so over the 30-word cap. If one clause is needed, use "my credence that there will be human-level AI before 2100 is somewhere over one-half.") | https://consc.net/papers/singularity.pdf | primary, fetched (text pulled from the PDF). The paper says it grew out of his talk at the 2009 Singularity Summit in New York, so the first public statement may be October 2009, but the talk's content wasn't verified. C2 latest: 2026-07-10 (lecture given May 2026 at UC Berkeley; podcast page dated July 10) | "I think they're beginning to approach the level of faculty of philosophy, professors in philosophy." | https://news.berkeley.edu/2026/07/10/berkeley-talks-when-we-talk-to-ai-what-are-we-talking-to/ | primary (transcript), fetched. Thin: this is about what language models can do now, not an explicit statement about human-level AI or its timing. No explicit 2024+ statement on human-level AI or its timing was found. His 2023 statement that language models "could become serious candidates for consciousness within a decade" (Wikipedia, fetched) predates 2024 and is about consciousness. C3 pieces: 1. 2009-10-02 | Forbes | "Calling All Transhumanists" (headline from search result) | https://www.forbes.com/2009/10/02/machine-intelligence-technology-breakthroughs-singularity.html | fetched ("Australian philosopher David Chalmers, whose idea inspired the Matrix film series") 2. 2009-10-12 | Forbes (Courtney Boyd Myers) | headline not legible in the archived copy; file named "forbes the transhumanists", about the 2009 Singularity Summit | https://synthneuro.org/wp-content/uploads/2021/08/news-091012-forbesthetranshumanists.pdf | fetched (text pulled from the PDF copy) 3. 2010-12-22 | Scientific American (Carl Zimmer) | "Can You Live Forever? Maybe Not--But You Can Have Fun Trying" | https://www.scientificamerican.com/article/e-zimmer-can-you-live-forever | fetched Also: Slate, April 2016, "Your Artificial Intelligence Cheat Sheet" (names him only in a reading list) | https://www.slate.com/articles/technology/future_tense/2016/04/killer_a_i_101_a_cheat_sheet_to_the_terminology_the_ethical_debates_the.html | fetched. And Time, 1996-03-25, "Can Machines Think?" (Robert Wright) | https://time.com/archive/6728762/can-machines-think/ | fetched. Quota tags: philosopher: philosopher by appointment (NYU philosophy professor; PhD at Indiana University under Douglas Hofstadter) (https://en.wikipedia.org/wiki/David_Chalmers, fetched 2026-09-30) | non_us_uk: Australian National University, Australia. Professor at ANU from 2004. Part-time at NYU from 2009 and full-time only from 2014 (Wikipedia, above). Forbes calls him "Australian philosopher" (2009-10-02, URL above). The 2010 paper thanks audiences at ANU and the University of New South Wales. He held US and Australian posts at the same time in 2009–14, so this is a judgment call, though a defensible one | skeptic: near miss. In 2010 he "would be surprised" by human-level AI within three decades (doubts it within decades), but put it above even odds before 2100. Under A13's examples ("no sooner than 2070" counts, "2029" does not) he sits in between; John should rule | frontier_lab: does not apply Verdict: passes. C1 and C3 are solid. C2 passes only on a statement about what current models can do (2026); there is no explicit 2024+ timing statement, so flag it. Queries: (1) David Chalmers AGI 2024 interview human-level AI; (2) David Chalmers 2025 AGI timeline "Chalmers" superintelligence LLM consciousness interview; (3) "David Chalmers" 2024 "human-level" OR "AGI" said within decade [extended]; (4) Chalmers philosopher AGI "next decade" OR "by 2030" 2025 podcast language models superintelligence; (5) "Chalmers" singularity artificial intelligence philosopher superintelligent machines [national-outlet domains; one earlier attempt was rejected for blocked domains and never ran]; (6) "David Chalmers" Singularity Summit 2009 New York Times OR Guardian OR "New Scientist" machines smarter than humans. 6 of 8 used.

Daniel Dennett

Tier checked: core Alive in 2024: yes, until 19 April 2024. He died that day, aged 82, of interstitial lung disease, at Maine Medical Center in Portland, Maine (https://en.wikipedia.org/wiki/Daniel_Dennett, fetched 2026-09-30). No statements after that date are possible, so his within-person record ends in April 2024. C1 earliest: 1985 | possibility | "Yes, I do think that it's possible to program self-consciousness into a computer." | "Can Machines Think?", first published in M. Shafto (ed.), How We Know (Harper & Row, 1985); reprinted in Brainchildren (MIT Press, 1998). Text read at https://vdoc.pub/documents/brainchildren-essays-on-designing-minds-7e3p21ej3pp0, an unofficial copy of the book | primary, fetched. Caveat: the quote is from the "Discussion" section following the essay, and I couldn't confirm that section ran in the 1985 printing rather than being added in 1998. The 1998 note introducing the essay says it shows "how huge the task posed by the Turing test is, and hence how unlikely it is that any computer will ever pass it" (dated 1998, primary, same URL). C2 latest: given before 2024-04-19, published 2024-04-25 (spectator.com) or 2024-04-27 (Australian edition) | "some thinkers have looked ahead and seen the possibility of strong artificial general intelligence becoming conscious, becoming alive and taking over the world and enslaving us. Yawn!" | https://spectator.com/?p=485505 (Gus Carter, "Daniel Dennett's last interview"); also https://spectator.com.au/2024/04/daniel-dennetts-last-interview-ai-could-signal-the-end-of-human-civilisation | primary (interview), fetched. C2 passes, narrowly: he made the statement in 2024 and it is about artificial general intelligence. But it dismisses a takeover scenario as not worth worrying about rather than saying whether or when such AI arrives, and it was published after his death. Other 2024 items, seen in search results only: David Pakman Show interview, 2024-03-06; Harvard Kennedy School Student Policy Review interview, posted 2024-04-20 (https://studentreview.hks.harvard.edu/interview-with-daniel-dennett-ai-free-will-and-the-future-of-humankind/, fetched, but it carries no Dennett quotes). C3 pieces: 1. 1996-03-25 | Time (Robert Wright) | "Can Machines Think?" | https://time.com/archive/6728762/can-machines-think/ | fetched (names Dennett and Chalmers; its subject is whether machines can think, after Deep Blue) 2. 2013-06-17 | Time (Joel Stein) | "Man vs. Machine" | https://time.com/archive/6643630/man-vs-machine-2/ | fetched ("Philosopher Daniel Dennett ... told me that my idea of a computer as a mere logic machine was impoverished.") 3. 2017-06-16 | NPR (13.7: Cosmos and Culture) | "Is The Concern Artificial Intelligence — Or Autonomy?" | https://www.npr.org/sections/13.7/2017/06/16/533196904/is-the-concern-artificial-intelligence-or-autonomy | seen in search result (two fetch attempts timed out). The search summary describes it as built around a Dennett interview. Caveat: 13.7 is NPR's guest-columnist blog, so it may not count as a national-outlet news piece. Also: Guardian interview 2017-02-12, "The whole singularity stuff, that's preposterous." Seen only via https://nautil.us/10-brilliant-insights-from-daniel-dennett-567922/ (fetched, secondhand); the Guardian itself is blocked to these tools. It is a general profile, not a piece about AI futures. Quota tags: skeptic: his pre-2017 statements doubt it. In 1998 he called it "unlikely ... that any computer will ever pass" the Turing test (1998, vdoc URL above). In 2015 he called the singularity possibly "an urban legend" and said the real danger is "clueless machines being ceded authority far beyond their competence" (2015, https://www.edge.org/response-detail/26035, fetched). In 2017 (From Bacteria to Bach and Back) he put superintelligence "at least 50 years away" (Wikipedia, fetched). The clearest timing statement is from 2017, not before 2017, but the earlier ones support the tag | philosopher: philosopher by appointment (Tufts University) (Wikipedia, fetched 2026-09-30) | non_us_uk: does not apply (Tufts, US) | frontier_lab: does not apply (he advised MIT's Cog humanoid-robot project, an academic lab, not a company) Verdict: unclear, leaning pass. C1 passes. C2 passes narrowly (2024, spoken before his death, about AGI-as-threat rather than its timing). C3 has two solid Time pieces plus one NPR blog item seen only in search. Before freezing the panel, John should rule on whether someone who died in April 2024 belongs in a series meant to trace people to the present. Queries: (1) Daniel Dennett 2024 artificial intelligence interview before death; (2) Daniel Dennett 1994 OR 1985 OR 1998 conscious robot "strong AI" possible in principle humanoid Cog essay; (3) Dennett "Can Machines Think?" 1985 text Turing test essay pdf; (4) "Daniel Dennett" artificial intelligence future machines singularity article 2015 OR 2016 OR 2013 OR 2009 [national-outlet domains]; (5) Dennett robots "artificial intelligence" will machines become conscious philosopher interview 2016 Guardian OR BBC OR "New Scientist" OR Wired OR Atlantic (the tool ran follow-up searches inside this one call); (6) Dennett philosopher artificial intelligence robots future consciousness [national-outlet domains]. 6 of 8 used.

Erik Brynjolfsson

Tier checked: core C1 earliest: 2017-11-30 | possibility/timing (distance) | "But it's also clear we are a long way from artificial general intelligence." | https://www.technologyreview.com/2017/11/30/147363/progress-in-ai-isnt-as-impressive-as-you-might-think/ | primary quote reported in MIT Technology Review (Will Knight, on the AI Index report), fetched. This meets the "2017 or earlier" rule by one month. No earlier statement on AGI was found. His 2011 and 2014 books are about automation and jobs, not human-level AI. C2 latest: 2025-10 | "AGI scores (e.g., GPT-4 at 27%, GPT-5 at 57%) concretely quantify both rapid progress and the substantial gap remaining before AGI." | https://www.agidefinition.ai/ ("A Definition of AGI", Hendrycks, Song, Brynjolfsson and others) | primary but group-authored, fetched. Also 2025-09: "As we approach Transformative Artificial Intelligence (TAI), there is an urgent need to advance our understanding..." | https://www.nber.org/papers/w34256 (with Korinek and Agrawal) | primary, fetched. C3 pieces: 1. 2013-06-12 | MIT Technology Review (David Rotman) | "How Technology Is Destroying Jobs" | https://www.technologyreview.com/2013/06/12/178008/how-technology-is-destroying-jobs/ | fetched 2. 2014-05-04 | Slate (Niall Firth) | "Robots Will Take Our Jobs" | https://slate.com/technology/2014/05/robots-taking-jobs-technology-will-replace-driving-routine-physical-labor-information-processing.html | fetched (names him as co-author of The Second Machine Age; the interview is with McAfee) 3. 2017-11-30 | MIT Technology Review (Will Knight) | "Progress in AI Isn't as Impressive as You Might Think" | URL as in C1 | fetched Also: 2015-09-28 | MIT Technology Review | "A Closer Look at the Future of Work" (a reading guide, which is thin) | https://www.technologyreview.com/s/541566/a-closer-look-at-the-future-of-work | fetched Caveat: pieces 1 and 2 are about automation and jobs. That counts as "where AI is going" only on a broad reading. Quota tags: skeptic: near miss / does not apply. "A long way from" AGI is from late 2017, not before 2017, and gives no number of decades | frontier_lab: does not apply (he co-founded Workhelix, Flexplay and Foundation Technologies, none of which build frontier AI models) | non_us_uk: does not apply (MIT, US) | philosopher: does not apply (economist) Verdict: passes, with flags. C1 rests on one late-2017 quote. Two of the three C3 pieces are about jobs, not AI's direction. C2 is a co-authored paper rather than a personal statement. Queries: (1) Brynjolfsson "artificial general intelligence" OR "human-level" AI decades away 2014 OR 2015 OR 2016 OR 2017 quote; (2) Brynjolfsson "Second Machine Age" singularity Kurzweil skeptical "artificial general intelligence" 2014; (3) Erik Brynjolfsson AGI 2025 interview "AGI" timeline Stanford Digital Economy Lab; (4) Brynjolfsson McAfee machines artificial intelligence future jobs 2011–2016 [national-outlet domains]; (5) "Brynjolfsson" "artificial general intelligence" 2024 OR 2025 OR 2026 said [extended]. 5 of 8 used.

Jürgen Schmidhuber

Tier checked: core C1 earliest: 2006-06 (page updated 2012) | timing | "It seems that history itself is about to converge around 2040 in an Omega Point or Historic Singularity." | https://people.idsia.ch/~juergen/history.html | primary (his own site), fetched. His homepage says his goal since about age 15 has been a self-improving AI smarter than himself, but that is undated in the source. C2 latest: 2024-10 (TEDAI Vienna) | "why 2042 might just be the year AI reaches superhuman levels" | https://www.ted.com/talks/juergen_schmidhuber_why_2042_will_be_a_big_year_for_ai | fetched, but this line is TED's own description of the talk, not Schmidhuber speaking. The date is primary; the quote is secondhand. A January 2025 Machine Learning Street Talk interview was seen in a search result only. C3 pieces: 1. 2016 | New York Times | "When A.I. Matures, It May Call Jürgen Schmidhuber 'Dad'" (John Markoff) | headline seen in search result; nytimes.com is blocked. A syndicated copy was fetched at https://www.qatar-tribune.com/article/36838/CHILLOUT/When-AI-matures-it-may-call-Jurgen-Schmidhuber-39dad-39, but the date it gives (2016-08-12) conflicts with my recollection of 2016-11-27 for the NYT original, so treat the exact date as unverified. 2. 2017-11-20 | PAP (Polish Press Agency, via Science in Poland) | "Scientist: We're only a few decades away from artificial intelligence with the capabilities of the human brain" | https://scienceinpoland.pl/en/node/27249 | fetched ("I would be very surprised if artificial intelligence does not become more intelligent than we are within a few decades.") 3. 2017 | The Guardian | interview: AI will "march out to where the resources are" and colonise the Milky Way | seen in search result; no date or URL verified Also: 2017-12-15 | Axios | "Computers to be smarter than man by 2050, says AI expert" | https://www.axios.com/2017/12/15/computers-to-be-smarter-than-man-by-2050-says-ai-expert-1513301690 | seen in search result (403 when fetched; that the "AI expert" is Schmidhuber is inferred from the search, not confirmed) Quota tags: non_us_uk: Swiss AI lab IDSIA (Dalle Molle Institute for Artificial Intelligence Research), Lugano, Switzerland. Scientific director since 1995, so through all his pre-2017 statements (https://en.wikipedia.org/wiki/J%C3%BCrgen_Schmidhuber, fetched 2026-09-30). Since 2021 he has also been at KAUST, Saudi Arabia | skeptic: does not apply (2040 convergence point, 2006) | frontier_lab: near miss. He co-founded NNAISENSE in 2014 (Wikipedia, above), but it was not building the most capable models of its time. Several early DeepMind researchers came from his lab, but he did not | philosopher: does not apply Verdict: passes, with flags. C1 is solid and primary. C2's 2024 date is solid but the quote is TED's paraphrase; a verbatim line from the talk or the 2025 interview would firm it up. C3 needs at least one of the NYT or Guardian items confirmed by a fetch; only the Polish Press Agency piece is fully verified. Queries: (1) Jürgen Schmidhuber AGI superintelligence 2025 OR 2024 interview prediction; (2) Schmidhuber "When A.I. Matures" Markoff New York Times 2016 OR Guardian 2017 robot future ants; (3) Schmidhuber artificial intelligence smarter than humans 2014 OR 2015 OR 2016 interview newspaper Swiss lab LSTM. 3 of 8 used.

Kai-Fu Lee

Tier checked: core C1 earliest: 2017-06-24 | timing | general AI and the singularity "may not arise for hundreds of years, if ever"; "no known path from our best A.I. tools" to general A.I. | New York Times op-ed, "The Real Threat of Artificial Intelligence". nytimes.com is blocked; text read via a 2017-07-10 mailing-list repost at https://server-nexa.polito.it/hyperkitty/list/nexa@server-nexa.polito.it/thread/GTHAMI4ZB7TXLHXYR45CFQNXWLUX75ZC/ (fetched) | primary (his own op-ed), read via the repost. Meets the "2017 or earlier" rule. Nothing earlier was found. C2 latest: 2025-10-02 (Digital Bridge 2025 panel) | "The moment when AI will be better than 90% of people at 90% of tasks. I think this will happen in about five years." | https://qazinform.com/amp/ai-will-surpass-90-of-people-in-five-years-2c7ca2/ (Kazinform, Kazakhstan's national news agency) | his words quoted by the agency, fetched C3 pieces: 1. 2017-04-27 | CNBC | "Kai-Fu Lee: Robots will replace half of all jobs" | https://www.cnbc.com/2017/04/27/kai-fu-lee-robots-will-replace-half-of-all-jobs.html | seen in search result (403 when fetched) 2. 2017-06-27 | MIT Technology Review (Will Knight) | "China's Tech Moguls Warn of AI's Troubling Trajectory" | https://www.technologyreview.com/2017/06/27/150705/chinas-tech-moguls-warn-of-ais-troubling-trajectory/ | fetched 3. 2017-11-13 | CNBC | "AI will obliterate half of all jobs, starting with white collar, says ex-Google China president" | https://www.cnbc.com/2017/11/13/ex-google-china-president-a-i-to-obliterate-white-collar-jobs-first.html | seen in search result Also: 2017-11-05 Forbes (Gil Press, a contributor post, so weaker) "6 Reasons Why China Will Lead In AI", seen in search result. All qualifying pieces are from 2017, and his own NYT op-ed of 2017-06-24 is not counted because he wrote it. Quota tags: non_us_uk: Beijing, China. He co-founded Sinovation Ventures (originally Innovation Works) in 2009 and was its chief executive when he made his 2017 statements; before that he founded Microsoft Research Asia in Beijing (1998) and was president of Google China (2005–09) (https://en.wikipedia.org/wiki/Kai-Fu_Lee, fetched 2026-09-30) | skeptic: near miss on date. "Hundreds of years, if ever" (2017-06-24, NYT, URL above) is a clear doubt that AGI arrives within decades, but it is from 2017, and A13 says "pre-2017". It qualifies only if John reads that as "2017 or earlier", the wording the core eligibility rule uses. His view has since flipped (about five years, 2025) | frontier_lab: judgment call. He founded 01.AI in March 2023 and runs it; it released the Yi-34B large language model (November 2023) (Wikipedia, above). It was a leading Chinese lab but arguably not building "the most capable AI models of its time" worldwide. Google China (2005–09) was a business role, not research. I lean toward not applying the tag, but flag it | philosopher: does not apply Verdict: passes, with flags. C1 sits right at the 2017 boundary. C3 rests entirely on 2017 coverage, two of the three items seen only in search results. He is useful for the panel's balance because his view flipped from "hundreds of years" (2017) to "about five years" (2025). Queries: (1) Kai-Fu Lee "The Real Threat of Artificial Intelligence" New York Times 2017 superintelligence general intelligence; (2) Kai-Fu Lee AGI 2025 "AGI" years away prediction interview 01.AI; (3) Kai-Fu Lee artificial intelligence future jobs China 2016 OR 2017 OR 2015 OR 2013 article [national-outlet domains]. 3 of 8 used.

Group summary

Rendered from docs/panel-evidence-2026-10/a13-core-new-d1.md at version a7bd17a.