OpenAI founding member backtracks: 'maybe the problem is fixable' after predicting email-only future
iamtrask · x · 2026-11-11
Andrew Trask (iamtrask), a founding member of OpenAI, previously argued that the bot/spam problem online is unfixable: future trusted communication will happen via email exchanges among people you know and their connections, with code-generated UIs built on top of inboxes enabling high-quality exchanges — with no inherent scale limit, since you can reach the whole world intermediated through email and the social graph. In this post he softens his stance: "Ok maybe the problem is fixable."
Related event: OpenAI Veteran on Bot Flood: Social Media May Retreat to Trusted Circles(3 posts)→
More from AGI Musings
- Is Anthropic's alleged 2x AI R&D speedup real? A thread stress-tests the 'benchmark-only' counterargument — dfrsrchtwts · 2026-11-11
- Top RL researcher: stay paranoid about AI until GPT-9's alignment is formally verified — AaronBergman18 · 2026-11-11
- Which benchmarks do humans still win against frontier AI? Reddit debates AGI criteria — ImproveOurWorld · 2026-11-11
- Stage-4 cancer patient gets free help from ~20 strangers on X, including AI tumor RNA analysis — prasanna_says · 2026-11-11
- Scott Alexander dissects the "outside criticism" Kafkatrap aimed at rationalists and EA — Astral Codex Ten · 2026-11-11
- Sunstein's 'Hayek for Our Time' Free for a Month: AI Calculation Debate Mirrors Socialist Calculation Debate — sebkrier · 2026-11-11