[anonylizard]

If one compares O3-mini's coding abilities to the original GPT-4. It is as large as GPT-3 to GPT-4 gap

GPT-3: Useful as autocomplete. Still error prone, but vastly better than any pre-AI autocomplete

GPT-4: Already capable of independently coding up simple functions based on natural language.

O3-mini: Can code in say top 5% of codeforces.

There's a 2 years gap between each of them.

More over, intelligence has a superexponential return, 90IQ->100IQ < 100IQ->110IQ in terms of returns.

[ben_w]

> More over, intelligence has a superexponential return, 90IQ->100IQ < 100IQ->110IQ in terms of returns

That's the second time I've seen the claim that linear increases in intelligence have "superexponential" results, first time was Altman's blog.

But I've not seen any justification for this.

(As you specifically say IQ, note that an IQ is defined as a mapping of standard deviations rather than a mapping of absolute skill, the normal mapping is 15 points being 1σ).