3 ms·

Terence Tao: creative strategies, this aspect of LLM tools is still weak

77 points by ColinWright 2y ago

perihelions 2y ago
https://news.ycombinator.com/item?id=41540902 https://news.ycombinator.com/item?id=41540902 ("Terence Tao on O1 (mathstodon.xyz)", 432 comments)
- dang 2y ago
  Comments moved thither. Thanks!
  - ColinWright 2y ago
    It's not a copy, it's not a dupe, it's a genuine update. I think it's an error to think of this as a dupe, and I think it's an error to move the comments there. But too late now.
    - dang 2y ago
      It's not too late. But the criterion here isn't whether the articles are technically different, it's whether the discussions are substantially the same or not. In this case the answer was yes: the discussion in the current thread was generic and therefore indistinguishable from the discussion in the previous thread. You can see this for yourself—here are the subthreads I moved: https://news.ycombinator.com/item?id=41549800 https://news.ycombinator.com/item?id=41549800 https://news.ycombinator.com/item?id=41549690 https://news.ycombinator.com/item?id=41549690 https://news.ycombinator.com/item?id=41549645 https://news.ycombinator.com/item?id=41549645 https://news.ycombinator.com/item?id=41549625 https://news.ycombinator.com/item?id=41549625 https://news.ycombinator.com/item?id=41549535 https://news.ycombinator.com/item?id=41549535 https://news.ycombinator.com/item?id=41549483 https://news.ycombinator.com/item?id=41549483 https://news.ycombinator.com/item?id=41549342 https://news.ycombinator.com/item?id=41549342 https://news.ycombinator.com/item?id=41549200 https://news.ycombinator.com/item?id=41549200
ChrisArchitect 2y ago
Related: Terence Tao on O1 https://news.ycombinator.com/item?id=41540902 https://news.ycombinator.com/item?id=41540902
ChaseRensberger 2y ago
Terence Tao has a remarkable ability (among many others obviously) to convey information that I think a lot of us may intuitively understand but struggle to explain clearly.
grutetc 2y ago
> Here the result was mildly disappointing: Essentially the model proposed the same strategy that was already identified in the most recent work on the problem (and which I restated in the blog post This was my concern when the first post hit HN. Tao seemed excited that “this time around” the “new” model proposed Cramer’s Theorem: > This time around, Cramer's theorem was identified and a perfectly satisfactory answer was given. All I could think was, well of course it did now, it has likely since been trained on Tao’s blog posts talking about its failure to suggest Cramers Theorem. I’m unsure why Tao is allowing himself to be an advertisement for this company and its paid product.

77 points by ColinWright 2y ago

https://news.ycombinator.com/item?id=41540902 https://news.ycombinator.com/item?id=41540902 ("Terence Tao on O1 (mathstodon.xyz)", 432 comments)

dang 2y ago
Comments moved thither. Thanks!
- ColinWright 2y ago
  It's not a copy, it's not a dupe, it's a genuine update. I think it's an error to think of this as a dupe, and I think it's an error to move the comments there. But too late now.
  - dang 2y ago
    It's not too late. But the criterion here isn't whether the articles are technically different, it's whether the discussions are substantially the same or not. In this case the answer was yes: the discussion in the current thread was generic and therefore indistinguishable from the discussion in the previous thread. You can see this for yourself—here are the subthreads I moved: https://news.ycombinator.com/item?id=41549800 https://news.ycombinator.com/item?id=41549800 https://news.ycombinator.com/item?id=41549690 https://news.ycombinator.com/item?id=41549690 https://news.ycombinator.com/item?id=41549645 https://news.ycombinator.com/item?id=41549645 https://news.ycombinator.com/item?id=41549625 https://news.ycombinator.com/item?id=41549625 https://news.ycombinator.com/item?id=41549535 https://news.ycombinator.com/item?id=41549535 https://news.ycombinator.com/item?id=41549483 https://news.ycombinator.com/item?id=41549483 https://news.ycombinator.com/item?id=41549342 https://news.ycombinator.com/item?id=41549342 https://news.ycombinator.com/item?id=41549200 https://news.ycombinator.com/item?id=41549200

ChrisArchitect 2y ago

Related: Terence Tao on O1 https://news.ycombinator.com/item?id=41540902 https://news.ycombinator.com/item?id=41540902

ChaseRensberger 2y ago

Terence Tao has a remarkable ability (among many others obviously) to convey information that I think a lot of us may intuitively understand but struggle to explain clearly.

grutetc 2y ago

> Here the result was mildly disappointing: Essentially the model proposed the same strategy that was already identified in the most recent work on the problem (and which I restated in the blog post This was my concern when the first post hit HN. Tao seemed excited that “this time around” the “new” model proposed Cramer’s Theorem: > This time around, Cramer's theorem was identified and a perfectly satisfactory answer was given. All I could think was, well of course it did now, it has likely since been trained on Tao’s blog posts talking about its failure to suggest Cramers Theorem. I’m unsure why Tao is allowing himself to be an advertisement for this company and its paid product.