PhD mathematics; scientific software architect; mountain biker / runner. Opinions are my own and do not reflect any associated entity.
Colorado
Joined July 2022
- Tweets4.3K
- Following510
- Followers1.4K
- Likes5.5K
Beautiful perspective
increasingly concerned that ai is revealing the mathematical world as negatively curved. instead of everyone wandering about in idea-space and clustering around the same famous conjectures (local positive curvature), ai is pushing out the boundaries rapidly and revealing that the space of mathematical ideas is ~hyperbolic (negative curvature). though this means there's lots left to discover, it also means that small separations between people/agents will lead to wandering alone.
we may be emerging into an idea-space where there's so much to see and do but everyone is alone in a private heaven, unable to share their subtle joys with others...
Glad theorem proving with AI is taking off. Helpful for both mathematics and software. Kinda sad I missed chances to join in on this development
We have to remove certain things from the models’ training:
- how long things take (e.g., “this is a day of work”)
- certain things are challenging (e.g., setting up App Store deployments)
- not finishing a task or over-chunking it (e.g., asked to do something e2e, stops halfway)
Claude knows more than I do and can do more than I can.
(Doesn’t mean I’m useless and for sure doesn’t mean I agree with Sabine about other things)
AI has seemingly increased drama across the board. Stocks up or down, problem solved or not, permanent underclass or machines of loving grace. Extreme hyperbole.
The only thing I’m sure of is this technology has changed my life and it’s only just begun.
Been using Opus/Sonnet 5.5 for a good while now. Noticed no nerf in subscription or bedrock.
I’m honestly not convinced nerfs are that real. I think we just fixate on the bad experiences and lose track of how good we have it.
Echoing: The Anthropic 5.5 model line is really incredible. Been using Opus at work, and Sonnet/Haiku at home with a $20 plan after having the $100 OpenAI plan. Sonnet on $20 is far better than what I was getting with Codex.
“But what are you building?”
At work, it’s HPC/simulations predominantly, but also some full-stack work because… you just can.
At home: managing Proxmox and building my own health app since none of the others do what I want.
People reading this seem to believe I think professional mathematicians couldn’t learn the technical aspects of AI/ML or quantitative finance.
No, most of them could. When someone says “any” and neglects the fact that these institutions have their pick of the litter, it’s a no.
I also want to claim that there are many mathematicians that would not put up with the working requirements of these jobs at all. It’s not a knock on them, people are different. You can’t just get a math degree and waltz onto Wall Street or into Anthropic.
> Any one of them could work as a quant or “AI Scientist” …
No, no they could not.
