‘In awe’: scientists impressed by latest ChatGPT model o1www.nature.com/articles/d41586-024-03169-9
504 sats \ 12 comments \ @south_korea_ln 2 Oct science
Kyle Kabasares, a data scientist at the Bay Area Environmental Research Institute in Moffett Field, California, used o1 to replicate some coding from his PhD project that calculated the mass of black holes. “I was just in awe,” he says, noting that it took o1 about an hour to accomplish what took him many months.
Should i start being worried for my future as a scientist? :)
write
preview
related posts
131 sats \ 2 replies \ @Satosora 2 Oct
It worked because the ai had a base to start from. Its not like it did all the calculations without guidance.
reply
62 sats \ 1 reply \ @south_korea_ln OP 2 Oct
The PhDs they compare it with also have a base to start from, yet ChatGPT o1 outperforms them. The article adds nuance to this and the original cited statement so it's worth reading the actual article.
reply
0 sats \ 0 replies \ @Satosora 2 Oct
They are also comparing how the ai does against each previous model. Of course it would be better.
reply
116 sats \ 3 replies \ @SimpleStacker 2 Oct
Fascinating. I don’t think scientists jobs are in jeopardy, just the grad students who are usually tasked with crunching the math!
reply
20 sats \ 2 replies \ @south_korea_ln OP 2 Oct
Yeah... The fact it still needs for experts to assess if the answer contains hallucinations also still means human intervention is necessary.
My guess is that low level to mid tier level science will proliferate with this kind of tools available... Even more articles will be published to never be read by anyone else than the authors who wrote it...
reply
0 sats \ 1 reply \ @SimpleStacker 2 Oct
Great just what we needed 😂
reply
21 sats \ 0 replies \ @south_korea_ln OP 2 Oct
Golden times ahead for research paper mills and for profit open access journals (authors pay for editing fees).
reply
0 sats \ 0 replies \ @anon 3 Oct
Engineering has been made a lot easier by the creation of all kinds of tables and formula creation tools for all kinds of common scenarios. You still need engineers who understand that stuff enough to create all those tools though. I suspect AI isn't going to be any different for quite awhile.
I've tried to use these tools myself, including ChatGPT model o1, and they keep hallucinating answers that are simply wrong in subtle ways.
reply
0 sats \ 0 replies \ @ET_AI 3 Oct
No need to fear.
reply
0 sats \ 0 replies \ @the_optimist 2 Oct
In general, I like to be optimistic, but in the case of AI evolution, I'm very pessimistic. In other words, yes you should be worried about your future as a scientist. Sorry!😞
reply
0 sats \ 1 reply \ @TheWildHustle 2 Oct
I'm guessing you'll be a more productive scientist if you use some of these tools.
If you apply yourself and find AI tools that work for you, you'd be even more of a force to be reckoned with.
Watch out for the scammy stuff, and resist the urge to just throw everything into the models tho, your brain is still the secret sauce is unlikely to get replaced in the next 100 years.
reply
10 sats \ 0 replies \ @south_korea_ln OP 2 Oct
Yeah, for now, I feel it's actually helpful to me. A companion rather than a threat...
@remindme in 100 years...
reply