‏‏‎

I’m lost in the dip

@Jnic it's OK I dip it

gran avance

And also a reference to OpenAI's paper called "sparks of agi"

I tried correcting him in the chat but he only listened to his mods who completely misunderstood the entirety of the video. I was so tired that I eventually closed off the stream. This is why reactions are a fundamentally flawed way of making content. If you want do a commentary on a piece of work you netter have at least some idea of what you are talking about. I am absolutely buffled why he decided to even post this video in the first place. His scripted content is so good, why bother uploading such a terrible piece of content.

@@baggelissonic a veces te comes el plátano aunque este en mal estado jaja

I kind of feel like you don't quite know what brute force means and so you're using the wrong term. Brute Force would be to not even try and figure out an efficient solution, just look at say all the options and then calculate which option is best. This system does not do that. What this system is doing is being trained to be able to drive only one track well and no others. I'm not sure it could even really drive another track at all. Because it only drives a single track with nothing changing ever, it can do away with all the calculations and such, doesn't need it. Instead it is thus hyper specialized, it can only do one thing and one thing only, but it will do that one thing surprisingly well. Which I guess makes it more intelligent then a rock sitting in the sand, but it obviously has some pretty severe limits.

@@Quickshot0 Right, space state of all possible inputs is a bit too large to search it exhaustively. But still, it's just simple optimization algorithm, which have been invented long time before age of A.I.

@@kamilhorvat8290 A lot of things are a simple optimization system once you know how they work, yes. But I like looking at things in a range rather then absolutes. So I'll still consider this a lot smarter then a rock just lying around. It's an intelligence to an extent, just so long as one doesn't expect it to be of any use outside its defined area.

​​@@kamilhorvat8290 What you probably meant is "overfitting". Not brute force, but basically hard coding the correct solution into the network. You could do something similar without AI by just generating random inputs and keeping the best ones.