Researchers Debate AI Spend on Millennium Prize Problems
Susan Zhang flags a bet against quick Millennium Prize solutions while Leo Boytsov raises cost estimates and broader goals.
Susan Zhang posted that $1T of infrastructure buildout targets $1M prize math problems and linked to a wager that no Millennium Prize problem will be solved by 2027 even with AI assistance. Leo Boytsov replied that the figure is closer to 5T and that the spend aims to create continuous capabilities across coding, math, and science problems. He asked what would happen if the effort produced thousands of discoveries that raise US GDP by 10 percent. Both posts appear in the visible conversation.
Combined views
20.8K
4 posts, first seen 11d ago
Reactions
4 postsSZ 
Susan Zhang@suchenzang$1T of infra buildout to solve $1M prize math problems
- likes: 162
- replies: 10
- bookmarks: 10
- reposts: 4
LB Leo Boytsov@srchvrsIt's not 1T, it's more like 5T, but it's clearly not solve just the Millennium Prize problem, it is to provide capabilities to solve many coding, math, and science problems (in a continuous fashion). What if this helps making 1000s of new discoveries that will increase US GDP by…
- likes: 0
- replies: 0
- bookmarks: 0
- reposts: 0
HE heiner@HeinrichKuttler@suchenzang There's 6 remaining ones. That's already one OOM. We will find a few others.
- likes: 3
- replies: 0
- bookmarks: 0
- reposts: 0
CP Chris Paxton@chris_j_paxton@suchenzang Math problems? You mean the problems so easy even a computer can solve them. Why even bother
- likes: 1
- replies: 0
- bookmarks: 0
- reposts: 0