normative
Hallucination Makes AI Unfit Where Errors Are Unforgivable
The hallucination problem is extremely damaging for applications where making a mistake is an unforgivable error; AI's greatest near-term opportunities are in domains without a single correct answer, and merely upgrading to higher-priced model versions does not solve hallucination—developers must change the penalty function, not just add data.
normativepending
Speaker
Austan GoolsbeeEvidence Quote
“If you’re going to steer me to o1 Pro, all we’re going to identify is that I’m too cheap to shell out the money for the higher and higher levels... I don’t accept that... if you just upgrade to the higher-priced one, then those hallucination problems would stop.”
Created: 6/12/2026, 4:48:31 PM
Your notes
Loading notes...