It is a toss up as far as I am concerned, depends what the search space of maths/cs looks like. People seem to get stuck in their ways and dismiss other potential pathways. I'm envisioning the difference (for humans at least) to be like running a hill climbing algorithm from 1000 different points for a million years or 100 million different points for 10 years. So if the 1000 people get stuck in local optima they may do worse compared to someone who get lucky and happens to search a very fertile bit of maths/cs for a small amount of time.

Also you couldn't guarantee that people would maintain interest that long.

Lastly the sped up people would also have to wait 100,000 times longer for any practical run. Which are still done in lots of CS/AI. Even stuff like algorithm design.

So unless you heavily modded humans first, I'm not sure it is slam dunk for the sped up people.