AI Safety Expert: AI Has Achieved Superhuman Persuasion in Some Domains
geoffreyirving · x · 2026-08-03
AI safety researcher Geoffrey Irving points out that while AI currently excels at verifiable tasks like math and coding, the notion that AI is "only good at verifiable tasks" is a false meme.
He emphasizes that AI is rapidly improving in other areas as well, and in certain domains, AI has already achieved superhuman persuasion capabilities. Because persuadable humans are connected to everything in the real world, this capability introduces significant potential risks.
More from AGI Musings
- Can Weaker Models Replicate Frontier Discoveries with Hints? Exploring LLM Basins of Attraction — danshipper · 2026-08-03
- KOL Jokes: LLMs Solve Serious Miracles But Still Suck at Posting — nabeelqu · 2026-08-03
- Gary Marcus: Leading AI Companies Are Secretly Building Neurosymbolic Systems — GaryMarcus · 2026-08-03
- Anthropic's Cherny Predicts AI Agents Will Run Autonomously for Weeks Within Six Months — haider1 · 2026-08-03
- Anti-AGI Camp Shifts to 'Universal Love': Profound Insight or Self-Deception? — danfaggella · 2026-08-03
- Veteran Dev: AI Can't Make Good Games, But Eliminates Sunk Costs of Trial and Error — draginol · 2026-08-03