What do AI predictions and promises mean for your work? Pick an industry or role to see the claims tracked on SiPoly that bear on it: which came true, which are still open and when they are due, and which did not happen. Every claim links to its verbatim quote, source and verdict.
What changes for Research & science?
Claims about AI doing or speeding up research: mathematics, the sciences, and AI research itself.
On SiPoly: 6 claims came true (fully or partly), 16 are open, 4 did not come true.
Came true (6)
- Kept An automated "intern-level research assistant" by September 2026
- Correct Mid-2025: specialized coding and research agents begin to transform their professions
- Correct Global AI investment could approach $200 billion by 2025
- Kept AI Action Plan within 180 days (EO 14179)
- Correct 4% that AI solves the hardest IMO problem by 2025
- Kept EO 14110: launch a National AI Research Resource pilot within 90 days
Still open, with deadline (16)
- Pending EO 14434: proposed legal definition of "Super Intelligence" due to the President within 60 days
- Pending First clinical trials of its AI-designed drugs by end of 2026
- Open In 2026 AI will be able to make "very small discoveries"
- Open AI "world-class mathematicians" within a year, world-class programmers within one or two (Jul 2025)
- Open An autonomous AI team beats human-led model research on equal time and compute, setting its agenda (within ~12 months)
- Open "Strikingly plausible" that models do the work of an AI researcher/engineer by 2027
- Open About a 50% chance that AI companies automate the whole AI research process by end of 2028
- Open 1.6% of recently approved drug sales in 2027 will come from AI-discovered drugs (5% by 2030)
- Open "Pretty confident" of AI that makes more significant discoveries in 2028 and beyond
- Pending A fully automated "legitimate AI researcher" by 2028
- Open Bet: AI produces Annals-quality number theory papers for under $100k each within 5 years
- Open By 2030 AI will autonomously fix issues, implement features and solve hard scientific programming problems
- Open By 2030 many scientific fields will have AI assistants comparable to today's coding assistants
- Pending UK to expand its AI Research Resource compute "by at least 20x by 2030"
- Open By 2030 AI will outcompete most professional mathematicians at research
- Open ~30% of physics, materials science and medicine papers will engage AI by 2030 (from 3% in 2022)
Did not come true (4)
“Came true” and “did not” are our verdicts on each claim as worded (a forecast with a stated probability is graded on its probability). Claims are matched to sectors by keyword rules and editors’ review, so a claim can appear under several sectors. These are other people’s claims; SiPoly makes no forecast of its own. Methodology