These are interesting articles from a few weeks ago, AI is getting better at solving hard problems:
https://fortune.com/2025/02/12/openai-deepresearch-humanity-last-exam/
https://www.techradar.com/computing/artificial-intelligence/could-you-pass-humanitys-last-exam-probably-not-but-neither-can-ai
3 comments:
These news stories are not that accurate.
9:06 -- I agree some of this is hype, but they are correct that there have been rapid improvements.
LANL plans to be the "AI" lab! The problem is they are way behind ORNL, ANL, LBNL and way way behind industry, like way behind. The other issue is that LANLs publications are falling while these other labs and industry the publications are increasing. If you really want to make an impact on AI you need to have top scientists with freedom and the resources to to publish.
LANL seems to think that AI research is just about having computers and people have nothing to do with it. Most LANL staff are now saying amongst themselves that LANLs AI efforts are poorly planed and have little hope of ever having impact.
LANL will never be the "AI" lab and there is little to no chance it will ever have any impact in this field. In fact I would venture to say that LANLs scientific impact is going to continue to decline.
I would guess if DOE really wants a AI lab they will need to start a new one, or it will be at ANL or ORNL. There was recent APS meeting which was the largest physics meeting of all time and I know for a fact that people from ex-LANL people at ORNL where telling everyone how much better it was and to get out of LANL.
Post a Comment