News for hackers. Stuff that matters.

A comment on Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking. Parent comment

Re:Quiet-STaR: Language Models Can Teach Themselves to Think... by biosed on Friday March 15, 2024 @07:46AM (#39714449)
Wasn't most of the claims in that book refuted, some even by the author. I really enjoyed it and found some great insights only to be later told by a friend in that sphere that the book was not correct and even the author had "retracted" some of the assertions.
[ Parent | on HN ]
Re:Quiet-STaR: Language Models Can Teach Themselves to Think... by jerpint on Friday March 15, 2024 @08:04AM (#39714608)
He won a Nobel prize for his works so not sure how much of it would be refuted
[ Parent | on HN ]
Re:Quiet-STaR: Language Models Can Teach Themselves to Think... by gryn on Friday March 15, 2024 @08:26AM (#39714758)
One quick google search and you can find multiple links for that, including some that were posted here. wasn't proven to be false but that the evidence used was not much of evidence either.

here the first one in my results:

https://retractionwatch.com/2017/02/20/placed-much-faith-und...

[ Parent | on HN ]
Re:Quiet-STaR: Language Models Can Teach Themselves to Think... by mistermann on Friday March 15, 2024 @08:42AM (#39714872)
As luck would have it, a System 1 vs System 2 scenario falls into our laps.
[ Parent | on HN ]
Re:Quiet-STaR: Language Models Can Teach Themselves to Think... by jerpint on Saturday March 16, 2024 @12:18AM (#39723352)
Cunningham's Law states "the best way to get the right answer on the internet is not to ask a question; it's to post the wrong answer."

https://meta.wikimedia.org/wiki/Cunningham%27s_Law

[ Parent | on HN ]
Re:Quiet-STaR: Language Models Can Teach Themselves to Think... by mannykannot on Friday March 15, 2024 @09:23AM (#39715187)
It might still be a useful concept in developing LLMs.
[ Parent | on HN ]