Hacker News new | ask | show | jobs
by yalogin 2002 days ago
The phrase AI always bothered me. What we have is a generic way to do “curve fitting” on a large amount of data. Nothing more. The one difference is the “curve” is a black box but it still strictly adheres to the input used.
2 comments

While true, this statement is completely vacuous. The curves being fitted are extremely complicated and if you can easily construct curve fitting problems (meaning: regression) that would be breakthroughs in many scientific fields. This says more about the generality of regression tasks than it says about ML.
My comment was focused on the phrase “artificial intelligence “ and not the tech itself. Obviously the techniques are awesome but the naming is unfortunate
Came here to post the same. I’m no expert, but when I’ve played with AI frameworks it just seems like curve fitting.

The few corporate deployments that make it to production barely outperform a simple regression model and are therefore over engineered.

The fact that many self contained vision and language tasks can be solved with curve fitting is in itself an interesting finding. Certainly I did not expect this 10 years ago.