AI predictions that missed their deadlines · every statement counted, not one per person

Counted statement by statement, the missed-deadline record grows to 44 claims from 30 people — and the loudest forecasters missed again and again

44 distinct prediction events by 30 named researchers, founders and commentators, stated between 1950 and 2025, every deadline judged elapsed by August 28, 2026, each with a cited later assessment. Where a person made several separate public commitments — Elon Musk made six, Eliezer Yudkowsky three, and seven others two each — every one is kept as its own row, because each had its own wording or deadline; some overlap or revise an earlier timeline. Twenty-one statements fall in 2016–2025 and lead the page; the 23 older events sit below as historical context. Repeated near-term misses cluster in autonomous driving and robotaxis, AGI and superintelligence, coding automation, AI agents, and job displacement. A miss grades the stated deadline and scope only — never motives — and a later arrival would not make the original deadline correct.

filters dim the timelines and hide table rows; era toggle governs the historical panel and table

Last 10 years · 21 statements by 13 people, grouped by person

statement made deadline passed unmet diamond — too pessimistic, the direction exception dashed span — low confidence: goal, hedge or debatable grading ×n beside a name — that person’s total statements in the record

Colour by topic family: autonomous vehicles & robotaxis jobs, coding & society AGI, singularity & superintelligence AI agents AI progress, language & chess

Repeat forecasters · share of the 44 statements

Elon Musk 6 Eliezer Yudkowsky 3 Dario Amodei 2 Herbert A. Simon 2 John Zimmer 2 Mark Zuckerberg 2 Marvin Minsky 2 Ray Kurzweil 2 Sam Altman 2 everyone else 1 each

Historical context, 1950–2015 · 23 statements by 18 people, muted for scale

What the timeline cannot say alone

Elon Musk is the record’s most frequent voice with six tracked statements: a coast-to-coast autonomous drive for 2017, a million robotaxis for 2020, no-hands full self-driving by 2020, “vastly smarter” AI within five years of 2020, smarter-than-any-human AI for 2025, and an unsupervised Austin robotaxi launch that used safety drivers instead.forbes.com

Coding deadlines shrank to months and still missed: Dario Amodei gave three to six months in 2025 for AI writing 90% of code, and twelve months for essentially all of it, while later public estimates at major companies ran around 25–30% — a figure that itself varies by organisation and measurement.sloppish.com

Multiple statements from one person can overlap or revise a previous timeline: John Zimmer’s 2016 pair set 2021 for mostly-autonomous Lyft rides and 2025 for the near-end of private car ownership; each is retained because it was a distinct public commitment with its own wording and deadline.newsletter.autogpt.net

Only two of the 44 statements erred pessimistic — Rodney Brooks on the deep-learning era ending by 2020, and Hubert Dreyfus on chess in the 1960s; every other miss was optimism about a deadline, not about direction.forecastfools.com

All 44 statements, recent rows first

Sort:

Person	Claim and outcome	Made → deadline	Topic	Confidence	Source

Confidence and caveats

“All” means all prediction events found and retained in this evidence-screened research pass — not every statement anyone ever made. Inclusion requires a named notable person, a falsifiable AI or automation claim, a deadline judged elapsed by August 28, 2026, and a cited later assessment.
Several rows are interpretive or low confidence. Shane Legg’s forecast was explicitly probabilistic — the row grades the lower tail he flagged as plausible. David Levy won his original 1978 wager; his inclusion reflects later extensions he lost, and should be treated cautiously.
Andrew Ng’s “near future,” Sam Altman’s “might,” Mark Zuckerberg’s stated goals, and broad AGI definitions are not as cleanly falsifiable as a dated robotaxi launch. Marc Benioff’s billion-agent item was a corporate goal, graded on the stated number and date only. Code-generation percentages vary by organisation and by how “AI-written” is measured.
Eliezer Yudkowsky’s 1999 item concerns molecular nanotechnology and is adjacent to AI and superintelligence rather than purely an AI claim; it is labelled as such in the timeline and table. Some claims’ source wording is paraphrased rather than verbatim.
Nothing here claims that a later arrival vindicates an earlier deadline, and nothing alleges dishonesty, incompetence or motives — only the stated deadline and scope are graded.

Statement-level record of 44 missed AI predictions by 30 named people, statements 1950–2025, deadlines through 2026, all judged elapsed by the 2026-08-28 cutoff with a cited later assessment; 21 statements were made in 2016–2025. Each row is one distinct public commitment with its own wording or deadline; a span measures statement year to stated deadline year. Repeat totals come from the dataset’s per-person counts. Timeline labels are shortened for space; full claim and outcome wording is in the table. An evidence-screened research pass, not an exhaustive census of mistaken forecasts.