He is a fields medal winner who is looking to see how AI can further mathematics research. I’m unsure of what your specific issue with that is.

This is for mathematicians and not software engineers. You look for ease of use and familiarity with tooling.

If I was a mathematician looking for a simple solution, I would go for GitHub too. Software engineering concerns don’t really apply here.

A product is what people (or now agents) can use to carry out any activity.

There is a lot of hand wringing in the software space about ai code being a mess - but from a product perspective code quality doesn’t matter. It’s the reliability of what people (or agents) use that matters.

Code quality used to further that goal but only because it was really hard to rewrite or fix code. If that barrier is removed does anyone really care what the code quality under the hood is?

Not true. Novel mathematical methods precede their application by at least a decade and widespread use by about a century.

Calculus was invented in 1670, it was about 1680-1700 till it started actually being used in astronomy. The uptake was probably faster because at that time a lot of mathematicians were Astronomers as well.

There is a lot of mathematics created but we don’t yet know how to use it. My hope is that AI can bridge the search gap to accelerate this.

> Novel mathematical methods precede their application by at least a decade and widespread use by about a century.

This isn't really a good argument. The assumption here is that the "applications" were possible because of the math itself, but it leaves out the possibility if the math didn't exist somehow it will be discovered/invented because the applications demand so.

> There is a lot of mathematics created but we don’t yet know how to use it.

The vast majority of mathematical work is complete useless. Only a small percentage finds use in the real world (even if you consider the maths from centuries back).

Quantitative scientists are also mathematicians. I'm not attacking mathematics, just probably-useless subfields. Is there any good quantitative evidence that actually estimates what percent of math work today will be useful? Because to me it seems like <1% and I feel like we could easily make that number a lot higher. Particularly I want to see massive improvements in quantitative social science; physics already gets a lot of attention so it wouldn't be able to see as much improvement to getting more resources but it could probably still get more.

Here are a couple of Fields medalists' work that had direct practical applications less than 5 years after publication:

Terence Tao's work (with Emmanuel Candès and Justin Romberg) on compressed sensing. Published in 2004-05. By 2007 that was being used for in-vivo MRI reconstruction with substantially undersampled data

June Huh's work on combinatorial Hodge theory was used within 3 years to improve sampling for random spanning forests.

Note that this is only Fields medalists (not all pure mathematics). There have been huge improvements in zero knowledge proofs and homomorphic encryption over the last ~5 years that are also directly applicable to pure mathematics, but no one has a Fields medal for it.

> Published in 2004-05. By 2007 that was being used for in-vivo MRI reconstruction with substantially undersampled data

I have argued exact claim before and from what I recall, Tao maybe contributed 1% to MRI improvements in the last 20 to 30 years. You are way over exaggerating the importance of what he did.

I think in social sciences it maybe more of an inability/reluctance of the domain group to use the mathematics as opposed to the mathematics being absent.

Take category theory for example. The initial mathematics appeared in 1942. The application to social sciences started in about 1970 and I’m not sure of the level of uptake at the current time but a quick AI search says applications have accelerated in the past decade (needs verification).

What open problems in quantitative social science do you have in mind where better math could achieve massive improvements? I'd expect the bottleneck to be data availability nearly always.

About time. We are a couple of decades late to the apocalypse.

A macabre counterpoint to "It's xxx year, where's my flying car?"

First pass through. After the machines invent time travel they end up accelerating the creation of skynet.

This is a really weird comparison. The Chan Zuckerberg foundation is nowhere involved with any war. All they’re doing is making money available for research. In which universe is $280m not enough?

When the people running it have a net worth of over $150 billion.

It can be made liquid. And in Zuck's case, it really, really should. It's not good to have publicly-traded companies where one person has total voting control over shares.

I would support that, for sure

I think an FDE is a post sales thing. It’s basically the sale has been done and as a company you don’t have enough expertise to scale what you bought.

I used to do that when my children were little. However, in another 3 years your child will not require you as much anymore and your downtime duration will improve.

My problem has been that once I started doing the audiobooks/podcasts, it has been really hard to reclaim my focus to read. I used to be able to power through books. Now, there always seems to be a distraction at hand.

Does meta have the research talent to create a SOTA frontier model? Yann LeCun has left Meta and I don’t think either alexandr wang or zuck have enough credibility to attract talent to create one.

it's possible Yann LeCun wasn't the right guy either. He seemed to be more focused at finding the next model architecture rather than iterating on the current LLM architecture to build a competitive frontier model.