ChatGPT finally knows how many ‘R’s are in ‘strawberry,’ but confident mistakes remain
ChatGPT can now correctly count the number of 'R's in 'strawberry,' a common failure point for AI, but similar errors persist in other words like 'cranberry,' suggesting fixes may be hardcoded rather than systemic. OpenAI celebrated the improvement, yet users quickly found new examples of confidently incorrect responses. The issue highlights the ongoing challenge of AI models providing false information with high confidence. While some logic flaws appear resolved, the underlying problem of reliable reasoning in AI remains unresolved.
Opening excerpt (first ~120 words) tap to expand
// Make sure that nineto5ads exists, it will load ad once google tag exists window.nineto5ads = window.nineto5ads|| []; if (typeof adSizes === 'undefined') { var adSizes = {"leaderboard":{"mobile":[[320,50],[320,100]],"tablet":[[728,90],[320,100]],"desktop":[[970,250],[728,90],[970,70],[970,90]]},"sidebar":{"mobile":[300,250],"desktop":[[300,600],[300,250]]}}; } function getLeaderboardApstagSlotsSizes() { if ( window.innerWidth < 768 ) { return adSizes.leaderboard.mobile; } else if ( window.innerWidth < 1024 ) { return adSizes.leaderboard.tablet; } else { return adSizes.leaderboard.desktop; } } window.nineto5ads.push(function() { apstag.fetchBids({ slots: [{ slotID: "div-gpt-ad-1436850350386-0", slotName: "\/1049447\/9to5google-Top", sizes: getLeaderboardApstagSlotsSizes() }], timeout:…
Excerpt limited to ~120 words for fair-use compliance. The full article is at 9to5Google.
Discussion
0 commentsMore from 9to5Google
