Physicists Find Formula to Predict When AI Chatbots Fail
Researchers at George Washington University say a mathematical formula can pinpoint when AI models start giving bad output.
Advertising disclosure: we may earn a commission when you join an operator via links on this page, at no cost to you. This never dictates our ratings. How we make money
Key takeaways
- George Washington University physicists developed a formula to predict AI chatbot degradation.
- The model identifies when an AI shifts from accurate answers to incorrect or hallucinated data.
- Early testing on smaller AI models successfully backed up the team's theoretical formula.
Physicists at George Washington University say they've built a mathematical formula that calculates precisely when an AI chatbot will start feeding you bad answers.
Instead of sitting around waiting for an artificial intelligence system to hallucinate or break down, the researchers used mathematical modeling to predict that breaking point ahead of time. Early lab tests on smaller AI models confirmed the math, proving that model failure follows predictable mathematical patterns.
Predicting the flip
AI models don't just fail out of nowhere. As context windows grow or query complexity spikes, neural networks hit a mathematical threshold where response quality drops off sharply. By isolating these key metrics, the team's formula catches the shift from reliable output to bad data before you ever see it on your screen.
Why it matters
For crypto traders and Web3 users relying on autonomous AI agents to parse smart contracts, manage liquidity pools, or execute trades, bad AI data translates directly to lost capital. An algorithm that flags when an AI tool is about to produce junk gives developers a way to shut down or re-route automated agents before they execute bad transactions on-chain.
Source: Decrypt
Top Crypto Casinos Right Now
🇫🇷 Showing sites that accept players from France Change country
Advertising disclosure: we may earn a commission when you join an operator via links on this page, at no cost to you. This never dictates our ratings. How we make money
More info Less info
Why we chose it: Phemex is a derivatives-first venue for Europe, Asia and Latin America. Good engine, mid-size liquidity, and a blocklist that grew in 2025.
- Online since
- 2019
- Licence
- Seychelles
- KYC
- Required
Restricted countries: United Arab Emirates, Afghanistan, Australia, Belgium, Bermuda, Belarus, Ontario (Canada), China and 18 more.
More info Less info
Why we chose it: A US-oriented casino that turns away only New Jersey. The cashier takes crypto, but the site publishes fewer specifics than any operator we list.
- KYC
- Required
Restricted countries: New Jersey (US).
Sofia Marek
Sofia reviews exchanges and crypto casinos, focused on fees, safety and what actually reaches the player.
Related News
OpenAI and Anthropic Rehearse for Major AI Cyberattacks
AI 'Vibe Coders' Are Blending Elden Ring, Minecraft, and Call of Duty
AI Could Crack Crypto Encryption Before Quantum Computers, Drake Says
