
AGI Predictions and Why Not Yet

AI is Smart Until It's Dumb: Why LLM Will Fail When You Least Expect It - Emmanuel Maggiori
Topic 17 of 19
Published Nov 10, 2025 · Clip 1:02:04–1:07:23 (5:20)
Emmanuel argues AGI predictions sell—people like Ray Kurzweil built careers on the singularity—and a prominent book agent told him a balanced AI book wouldn't sell because you must claim AI will either change or destroy everything. He says AGI is not coming anytime soon because current machine learning methodology has flaws (like hallucinations) that are unsolvable within it; only a new, unknown methodology could change that. He cites GPT-5's flop as unsurprising since no secret ingredient was discovered, compares it to nuclear fusion power perpetually 'around the corner,' and notes philosophical AGI questions are fascinating but not the reality of current machine learning.