The Information Machine

The man who turned $225 million into $13.7 billion identified the single hardest unsolved problem in AI (Save this).

Milk Road AI Twitter · Milk Road AI (@MilkRoadAI) · 2026-05-29

Leopold Aschenbrenner, known for turning $225 million into $13.7 billion, identifies AI honesty as the single hardest unsolved problem in AI and argues it is central to the alignment challenge.

Open original ↗

Appears in

Extraction

Topics: ai-alignmentai-safetyai-honesty

Claims

  • Leopold Aschenbrenner identifies AI honesty as the hardest unsolved problem in AI.
  • Solving AI honesty would likely get researchers most of the way toward solving alignment.
  • Achieving AI honesty is surprisingly difficult even as a narrow technical goal.
  • Aschenbrenner achieved exceptional investment returns, growing $225 million to $13.7 billion.

Key quotes

Would solving AI honesty solve alignment? Probably gets you pretty far but it is surprisingly hard to even do that.