While the headline is correct, it speaks more about this particular human's expertise than the AI's.
GPT-5-Codex is a better AI researcher than me
31–39 of 39 posts
Re: GPT-5-Codex is a better AI researcher than me
#32The title makes it sound like we have reached the singularity. The real insight here is that amateurs may have a difficult time competing with AI
I agree but if this year the AI can keep up with amateurs, next year who knows?
If you got married this month you will have 12 wives in a year, who knows?
Re: GPT-5-Codex is a better AI researcher than me
#33Earlier quoted context omitted.
I agree but if this year the AI can keep up with amateurs, next year who knows?
If you got married this month you will have 12 wives in a year, who knows?
Re: GPT-5-Codex is a better AI researcher than me
#34[dead]
Re: GPT-5-Codex is a better AI researcher than me
#35This is written by someone who's not an AI researcher, working with tiny models on toy datasets. It's at the level of a motivated undergraduate student in their first NLP course, but not much more.
If one can easily reach parity with a motivated undergrad by leveraging LLMs I will still consider it impressive.
While the 5-minutes model will never be useful in itself it lays the groundwork for amateurs and small groups to getting into developing small models. There's at the moment another HN headline hyping up a tiny model that scores impressively at the arc-agi benchmarks so it's clearly not a dead end to explore what is "household-affordable" models.
Though an approach that doesn't lean on the authors $200/month OAI sub would've been more interesting to follow.
Re: GPT-5-Codex is a better AI researcher than me
#36This is written by someone who's not an AI researcher, working with tiny models on toy datasets. It's at the level of a motivated undergraduate student in their first NLP course, but not much more.
If one can easily reach parity with a motivated undergrad by leveraging LLMs I will still consider it impressive. While the 5-minutes model will never be useful in itself it lays the groundwork for amateurs and small groups to getting into developing small models. There's at the moment another HN headline hyping up a tiny model that scores impressively at the arc-agi benchmarks so it's clearly not a dead end to explo…
You can also reach research parity by downloading a Github repository. Is that impressive too?
Re: GPT-5-Codex is a better AI researcher than me
#37I love using AI to set up projects in 5 minutes but I hate to develop these projects using AI because inevitably it runs into a wall and I need to guide it and fix its code. I suppose in this case it picked up an existing project and DIDN’T walk off a cliff? Were the mutations really small?
The worst to me is trying to use someone else's AI code. This guy, for example, has thousands of lines of undocumented (there are a few docstrings but not much more) code. I'm not really motivated to go through that, and you can't trust AI to go through and document it correctly necessarily either. At least if I generated it, I was probably around for enough of the process to know how most of it works.
That's weird. I thought LLMs loved over-explaining their code?
Re: GPT-5-Codex is a better AI researcher than me
#38Earlier quoted context omitted.
If one can easily reach parity with a motivated undergrad by leveraging LLMs I will still consider it impressive. While the 5-minutes model will never be useful in itself it lays the groundwork for amateurs and small groups to getting into developing small models. There's at the moment another HN headline hyping up a tiny model that scores impressively at the arc-agi benchmarks so it's clearly not a dead end to explo…
You can also reach research parity by downloading a Github repository. Is that impressive too?
Downloading a file is not equivalent to having high level abstractified control over running software.
And if it is then I'm a farmer because I bought potatoes from the store.
Re: GPT-5-Codex is a better AI researcher than me
#39Right now AI is lifting up the floor, so if you don't know programming, mandarin or any other topic it will sure do better than you. (Vibe coding goes here) Same for tasks you know how to do but AI does them faster, there is also value there. (Claude Code used by a senior goes here) The interesting thing is when AI is lifting up the ceiling everywhere, but maybe then is when we are almost on AGI territory.
We already have a lifted ceiling.
We have several synthetic datasets and automated evaluation options for such things that were close to impossible to do before LLMs.