3 ms·
IMHO The biggest roadblock to use AI to rewrite code and blindly trust results is having good tests to validate results. This is mood for simulations and models
by rlupi 3y ago
IMHO The biggest roadblock to use AI to rewrite code and blindly trust results is having good tests to validate results. This is mood for simulations and models.
Also, neural ODEs are already taking traction. It won't be hard to replace lots of obscure spaghetti code that deal with edge cases in models with maybe inscrutable neural networks that deliver the same results. Better tools are already developing to deal with interpretability of NN, as higher-level blocks become standard abstractions.
- iudqnolq 3y agoDo you mean to say simulations and models have good tests? In my (very limited) experience with scientific computing that's absolutely not the case. There's a decent chance the software is impossible to test because it can only be run all together and is nondeterministic.