OpenPrediction
"Real AI" (beyond today's chatbots) arrives in about 5 years
“No. We’re not there yet. But the real AI is coming. I think we’ve got about 5 years.”
Risk & safety
12 claims
OpenPrediction
"10% to 20%" chance AI leads to human extinction within three decades
“said there was a “10% to 20%” chance that AI would lead to human extinction within the next three decades.”
PendingPromise
Binding regulation on the most powerful AI model developers
“Labour will ensure the safe development and use of AI models by introducing binding regulation on the handful of companies developing the most powerful AI models”
KeptPromise
Seoul commitment: publish a safety framework before the France summit
“to demonstrate how they have achieved this by publishing a safety framework focused on severe risks by the upcoming AI Summit in France.”
KeptPromise
Seoul commitment: publish a safety framework before the France summit
“to demonstrate how they have achieved this by publishing a safety framework focused on severe risks by the upcoming AI Summit in France.”
KeptPromise
Frontier Safety Framework "fully implemented by early 2025"
“We aim to have this initial framework fully implemented by early 2025.”
KeptPromise
EO 14110: NIST generative-AI companion to the AI Risk Management Framework within 270 days
“Within 270 days of the date of this order, to help ensure the development of safe, secure, and trustworthy AI systems, the Secretary of Commerce, acting through the Director of the…”
KeptPromise
EO 14110: AI risk assessments for critical infrastructure within 90 days
“Within 90 days of the date of this order, and at least annually thereafter, the head of each agency with relevant regulatory authority over critical infrastructure”
KeptPromise
EO 14110: frontier-model developers must report to the government within 90 days
“Within 90 days of the date of this order, to ensure and verify the continuous availability of safe, reliable, and effective AI in accordance with the Defense Production Act”
BrokenPromise
20% of secured compute to superalignment over four years
“We are dedicating 20% of the compute we’ve secured to date over the next four years to solving the problem of superintelligence alignment.”
CorrectPrediction
50%: AI risk will feel more widely accepted as a field by 2023 (40% same, 10% less)
“7. AI risk as a field subjectively feels more/same/less widely accepted than today: 50%/40%/10%”
CorrectPrediction
80%: MIRI still exists in 2023
“6. MIRI still exists in 2023: 80%”
See the leaderboards by topic.