LegoGPT generates a LEGO structure from a user-provided text prompt in an end-to-end manner. Notably, our generated LEGO structure is physically stable and buildable.
This is how AI should be used. this idea of throwing everything at it but the kitchen sink is a big mistake. The most successful things will come out of training to do something specific on the highest quality data they can get. Less bad data percent will win over total data amount.
They’re just built with 1 by X bricks. But I kinda like that, it’s not replacing anything a person would design. I think I’d prefer a program that would take a 3d model as in input instead of a text prompt, but both are cool.
Finally a real-world use-case for AI!
It’s awesome in one hand, but on the other it removes a lot of the frenzy creativity fun
OK, while I am generally against AI, this is awesome
Fuckin nice!
How long until AGI and we don’t need separate models for every single thing?
I don’t know for sure, but right now we’re in a very early version of an AGI with limitations, but it’s only available to users who pay a subscription.
Nobody can tell you, and if they say they do know, they’re selling snake oil.
But when it happens, shit is going to change fast, for better or for worse.
If
When
Just not soon
Potentially never
A pretty long time.
Niche models are tons of fun though.
Tool use/mcp can do an awful lot already, not everything but it’s hard to come up with stuff it just plain can’t do.
Agi is poorly defined, without a clear definition of exactly what it is there will never be a time we actually achieve it because the goalposts will move.
We all probably remember the time that the turing test was the ultimate test of an ai, but that’s pretty well cooked and it no longer matters at all.
AGI has always had a same definition: general intelligence, when an AI can do anything a human brain could.
Don’t see how that’s not the case now. Only thing missing is a universal harness, which it’ll probably be able to build itself soon, for sure humans can’t.
