Hacker News Reader

Comment by jokoon | Hacker News Reader

Machine learning isn't really science, since it's only statistical methods. It doesn't provide insight into what intelligence is. It's only techniques, so it's just engineering. It's brute force hacking at best, and when it sort of works, it's impossible to figure out why it does because it's black boxes all the way down.

So of course there are cool things like gpt, but it's not like it's scientific progress. It doesn't really to understand how brains work, and how to understand what general intelligence really is.

View on HN

It's possible you may have misunderstood the title of the post. It isn't about the science of ML, or GPT-3, or brains. Rather, it's about using ML as a tool to do actual science, like medicine or political science or chemistry or whatnot. The first sentence of the post explains this.

Machine learning is not about getting insight into what intelligence is (it might do so as a byproduct but very few people are using it with that goal in mind).

However, ML is useful to generalist science as long as you are be aware of its shortcomings and not just trying to replace something with ML without thinking about it.

To give you an example I worked on (to be published): I worked with some physicists that use an incredibly slow and expensive iterative solver to get information on particules. We introduced a machine learning algorithm that predicts the end result. It does not replace the solver (you could not trust its results, contrary to a physics based numerical algorithm) but, using its guess as a starting point for the iterative solver, you can make the overall solving process orders of magnitude faster.

> It does not replace the solver (you could not trust its results, contrary to a physics based numerical algorithm)

And I guess the outcome variable in the train set for the ML model was produced by the solver?

By an unmodified solver, yes (we did have a test/train split in case people are wondering after having read the above article).

Statistics are the backbone of many natural sciences.

It is also valid to make scientific progress just inside of a field and not in the grand scheme of things.

I feel that the deterministic computing theologists are going to be in for a rude awakening over time. Computing need not be perfect to work and the thing about recent advancements in ML is that they scale Extremely well.

Reply on news.ycombinator.com