Do you guys think this AI thing's gonna last?
I feel like the current thing everyone's pursuing, generative AI will hit a dead-end pretty soon. There's a gorillion startup being founded around these thing and zillions in funding. But this aren't even the most important problems in AI, in fact, they're the easiest to solve.
Everyone seems to be optimistic that AI will soon be able to do most, if not all jobs. There's even a segment with the huge optimism expecting AGI by 2029.
But, the most important problems and limitations of AI are still there. AI still can't do math well, can't do most jobs humans do that isn't related to writing or art, AI needs huge computation requirements etc.
I think the LLM, text to video/image AI hype train will come to a crashing halt and we'll face another AI winter. I can't huff anymore hopium and just need to face reality, I will never have an AI gf, much less AGI.
> I will never have an AI gf
You can have an AI gf right now if you know the tech
> I think the LLM, text to video/image AI hype train will come to a crashing halt
There is absolutely no reason to believe this. Also, even if technological development ceased tomorrow, the tech we have even right now would be very useful in a wide variety of applications
> we'll face another AI winter
The last two times this happened it was because AI was literally useless and could not do anything it was hyped up to do. This time, clearly, AI is quite capable of actually making things happen.
> But, the most important problems and limitations of AI are still there
How sure are you that these are insurmountable limitations of the paradigm instead of, you know, engineering issues?
> AI still can't do math well
So hook it up to a calculator.
> can't do most jobs humans do that isn't related to writing or art
Yet
It's been less than a year since the boom started with GPT-4. This is just the beginning.
Yeah, they aren't insurmountable issues but there seems to be much less attention given to them compared to generative AI. I haven't heard about any fundamental problems of AI solved, only a .02% improvement in yet another LLM, image generator.
What you think are the fundamental problems are far from it. A lot of people will dismiss LLMs by saying something to the tune of "Oh, it's just regurgitating", "Oh, it's a parrot", "Oh, it can't generalize" - it's a distilled version of the human condition. I think a lot of these problems aren't really fundamental, and a lot of companies are working on them now. You'll see a lot more "progress" being made than you think.
I just hope they come within the next few decades and not centuries later.
I'd welcome such a world tbh
Your post smells like curry.
i just watched a 20 minute long AI generated south park episode that actually has funny moments in it.
https://fablestudio.github.io/showrunner-agents/
I got that from another g thread
once again, thats generative AI. I'm not denying that generative AI is shit, I'm quite impressed by it. But, we aren't making progress in the parts that truly matter.
Saying AI will hit a dead end is like saying the internet was a phase. Its BS, there may come a point where many startups fail, but the tech is here to stay forever. Its just a matter of shaking out which companies are gonna make it and which are not.
>AI can't do math
Not specifically true. LLM'S suck at math but then again, they're language models. You could train an AI on math and it would probably be much better.
>You could train an AI on math and it would probably be much better.
Holy fuck bro. Doing math is a computer's primary function. You don't need to "train" an AI to do math.
then how come we still don't have an AI that can't solve a single millenium prize problem?
your brain consists of nothing but chemical reactions and yet you still need to spend years studying neuroscience
>Doing math is a computer's primary function
Explain why CPU's still need arithmetic logic units then?
The primary function of computers is to do math but its a very specific kind of math, binary math. Computers still suck at basic arithmetic, they make floating point precision errors all the time.
Health and coding assistant are already enough to say it is here to stay forever.
AI doesn't steal your secret sauce or get poached by the competition. It won't be long until most devs are at an arms length from the real work AI is doing in secret. The breakthrough of AI is reasoning in a bottle, not writing spam.
>Do you guys think this AI thing's gonna last?
i surely hope not, anon, i'm already quite tired of it.
you know, by delivering everything for a machine to do without needing proper instructions (it is intelligent, so with a few words it figures it out on its own) we more than ever lose control over the process.
if that becomes the norm for anything we do, the world will be quite a boring place.
longer than humans will last
Yes it's here to stay. But ([{~They~}]) might castrate - or shall we say.. circumcise(?) it too much, to ruin it.
I think all the big corps have reached a mutual agreement to sabotage their own LLMs.