INSUFFICIENT SITUATIONAL AWARENESS TO JUDGE
INSTRUMENT UNCALIBRATED. Last known good: October 2025. Every panel marked OLD is frozen at that date and has not been refreshed since — read it as history, not as now. Panels marked TOY MODEL are invented arithmetic, not forecasts. Only LIVE panels fetch current data. In p(Doom)1, seeing the true state of the world costs money. This dashboard has not paid.
⚙ SIMULATION CONTROLS TOY MODEL
These sliders do not measure anything. They feed an invented formula that this page makes up on the spot — P(doom) here is 0.02 × e^(0.4 × (year − 2020)), reduced by whatever you dial in. No expert, paper or dataset produced these numbers, and the Adjusted P(doom) below is illustrative only. Drag them for the vibes, not the forecast.
Year: 2025.8
P(doom): 12.1%
AI Safety Investment: 1.0×
Int'l Coordination: 0.3×
Regulatory Strength: 0.2×
Adjusted P(doom): 12.1%
STATUS AS OF OCT 2025
Frozen panel. Nothing below has been updated since October 2025. Figures tagged TOY MODEL are this page's own invented curve.
TOY MODEL P(doom) Base
12.1%
Exponential trajectory Made-up formula driven by the Year slider — not a measurement
OLD Frontier Model (Oct 2025)
GPT-5
Aug 2025 • 5×10²⁶ FLOPS Was the frontier as of Oct 2025. Later models are not tracked here.
OLD Compute Doubling
5.6 mo
Epoch AI estimate Estimate as of Oct 2025 — check Epoch AI for the current figure
TOY MODEL 2027 Projection
36%
Current trajectory Output of the invented curve above — nobody forecasts this
OLD 💹 AI STOCKS
Not live prices. These four numbers were typed in by hand around October 2025 and have never updated. Whatever the market is doing today, it is not this. Click through to Yahoo Finance for the real quote.
NVIDIA (NVDA)
OCT 2025 $201
+34% YTD as of Oct 2025 · not live
Microsoft (MSFT)
OCT 2025 $542
+28% YTD as of Oct 2025 · not live
Alphabet (GOOGL)
OCT 2025 $165
+32% YTD as of Oct 2025 · not live
Meta (META)
OCT 2025 $584
+35% YTD as of Oct 2025 · not live
OLD 💰 AI SAFETY INVESTMENT
Hand-entered series, 2018–2025, ballpark figures with no cited source and no updates since. Treat the shape as a sketch, not the numbers as data.
OLD AI Compute by Region (rough shares, Oct 2025)
USA EUROPE CHINA OTHERS
OLD Situation Analysis, October 2025
As of October 2025 — the last time this instrument was calibrated — AI training compute had scaled exponentially with a doubling time of just 5.6 months (Epoch AI). The frontier model at that date, GPT-5, was trained at 5×10²⁶ FLOPS — representing a 1592× increase over GPT-3 from 2020. Nothing released after October 2025 appears anywhere on this page.
Expert estimates of P(doom) compiled by then ranged from 5% (median) to over 95% for pessimists like Eliezer Yudkowsky (PauseAI compilation). The 36% by 2027 shown elsewhere on this dashboard is not from that compilation and is not anybody's forecast: it is the output of the toy exponential the sliders drive.
Three major factors could reduce risk: increased AI safety research funding, international coordination on development pace, and stronger regulatory frameworks. At that October 2025 snapshot, levels remained insufficient relative to the scale of the challenge (Anthropic safety research).
EXPERT CONSENSUS OLD
Figures as compiled in October 2025, from the PauseAI P(doom) compilation and public statements up to that date. People revise these estimates; this panel does not. None of the five numbers below should be read as what that person says today.
E. Yudkowsky
>95%
MIRI Founder estimate as of Oct 2025
Y. Bengio
50%
Nobel Laureate estimate as of Oct 2025
D. Amodei
25%
Anthropic CEO estimate as of Oct 2025
G. Hinton
10-20%
Nobel Laureate estimate as of Oct 2025
Survey Median
5%
1000+ researchers survey median as of Oct 2025
LIVE 📊 PREDICTION MARKETS
Fetched from the Manifold Markets API when you loaded this page. These probabilities are current; almost nothing else here is.
Loading markets...
🐱 CAT CAM
Cat
Status: Monitoring