WelcomeUser Guide
ToSPrivacyCanary
DonateBugsLicense

©2025 Poal.co

509

(post is archived)

[–] 1 pt (edited )

I'm one of the ones here with the best grasp of where machine learning and robotics is as a science. And when I say it's not there, it's not there.

Probably in ten years if we dont hit another AI winter, but realistically the cutting edge is slicing their gains thinner and thinner.

Whats lacking is generalization. There is still very little understanding conceptually (let alone algorithmic) about how even a child learns after only a few examples (well, that is, outside very specific domains). We're getting there, but going from "play any nintendo game we throw at you" to "perform any task in real life after only being shown it a few times", is a tall order. We had ML bots play warcraft, but thats mostly specialized problem solving, not generalization.

Also it doesnt cover "heres what we want to accomplish, we'll leave it up to you to figure out robot." It also doesn't cover issues of ethics. By "ethics" I don't mean woke shit, but questions like "if you have a choice between saving one person or ten people, which would you save?" The machines have no procedures for wondering or asking about if who the people are matters, or pondering what other parameters and questions are important to consider. Theres been some work with simple scenario generators (accessible online but fuck if I can find it now), that present various either/or questions to people involving hypothetical driving situations, to try and build up a dataset for training 'ethics' but really thats a whole can of worms we've only just begun to explore as a science.

So again, the science isn't there.

We'll still make plenty of conceptual progress between now and then supposing the world doesn't destroy itself because the medicis and their jewish middlemen cant stop strangling world economies and blaming it on scapegoat boogiemen like the rothchilds, but I digress. We'll see lots of development in unifying existing narrow intelligences, and even new algorithms for tasks previously unsolved, but until we start to understand generalization and self-modification we'll still be behind the curve. And then theres 'consciousness', which we cant even contemplate without some morons inventing idle philosophical masturbatory material like "the chinese room" as thinly-veiled self-promotion strategies.

[–] 0 pt

Pretty much agreed from my perspective. All the cool "AI" stuff right now almost doesn't even truly qualify as being AI, they're more like machine(-'learning')-optimized algorithms.

[–] 0 pt

"Artificial Intelligence" is a moving target. Todays AI is tomorrows ML algorithm.

Which is why we see stories about sentient robots being denied rights. It's true to life.

Doesn't matter how smart the machines are, there will be people that deny their capabilities.