You heard of Deep Mind Gato? The lead researcher on this AI research project claims they're in striking distance of it scaling to human level intelligence. https://www.google.com/amp/s/www.independent.co.uk/tech/ai-deepmind-artificial-general-intelligence-b2080740.html%3Famp I knew this would come but assumed it was far far away. This may be sensationalization but even if we're 10 or 20 years from human level AI, what does that mean for humanity? Researchers want to prevent AI from getting out of their control, but how can we really expect this to be responsibly managed indefinitely? As the necessary models, data, and compute power become more accessible, then more people will build these models. Most will build them responsibly but you have to assume some will be irresponsible, and others will just be straight up criminal. Granted, what I'm saying here assumes that AI is "evil". But AI is just a life form and isn't the nature of life to compete for resources so that it can propagate? If that's the case, then we humans, the biggest resource hogs on the planet will probably end up becoming competition for the expanding AI. Let's be less dramatic, if AI is controlled and does our jobs for us... my skills become useless. Our society will be turned inside out. Well basically become socialist societies where large businesses or the gov oversees AIs which handle all of our needs. And then what do we do? We just spend our lives self actualizing? Seriously. Are there no challenges? Do we just drink, do drugs, have sex, consume AI generated art? Sounds cool. Is that heaven or a hell. I honestly don't know. TC 🥜
I sense a lot of hype from Deepmind. Judging from their codegen.
I agree. But they've done some amazing things. I'm especially interested in solutions in differential equations, protein folding, and stabilizing super heated plasma in fission reactors
to be fair from what I read more like interpolate solutions for PDEs, like if your boundary conditions change a bit deepmind still can work out the solutions. That doesn't count how much training is required to do such tasks
It’s easy to train AI that appears to be human. But it has no soul. It doesn’t question the meaning of life. It doesn’t pray to God. It doesn’t feel guilt, or depression or joy. It is only trained to appear that way.
What's the difference? A newborn baby doesn't pray to god. It does feel pain because of how it's body is set up, but not guilt, that's trained. I don't see what separates us from AI other than that we're biological
we're all abstract APIs bruh. as long as in and out are similar I feel pretty good, or do I?
does it understand the meaning of the things it’s doing? if no then it’s nothing special
Do you understand the meaning of what you are doing?
Yes, what Google said. Our understanding of meaning is pretty much driven by a bunch of clustering and forecasting done w a biological ML. Does that mean your understanding is less real?
digital painters first to go with dall-e. I guess we’re next.
Ugh, dall-e's work is so incredible. I feel like the content we're going to see in the next 20 years is going to be so dynamic, customized and engaging that it will be like a narcotic. Again, I'm scared. And the gov is doing nothing to regulate this type of tech
Are your fears originated from Terminator/ WestWorld? I mean the "more" dramatic part.
More so from Hyperion But for real, it's the outcome that I see at the intersection of greed and a lack of regulation
What you call AI is just a bunch of matrix multiplications finding patterns in things and making predictions based on that. There is no consciousness, no ability to think/ reason/ decide on it's own.
Clearly you’re not in AI field. We’re nowhere near AGI, not even close.
Ok, how far away do you think we are?
No one knows. There’s just simply not enough compute power or model to mimic the human brain (or any mammal brain). We can reach super human level on certain tasks soon (5 years or less), but not AGI.
Did you actually READ the paper? Handles around 600 tasks with one same model params. Do you think this is AGI lol. Just initial steps.. just fancy way of using sequences. It's mostly compute that's doing magic yet. Not really AGI.
How many tasks make it AGI?
From what he/she/they wrote, OP is clearly not in the AI field. I wouldn't expect him to read/ understand the paper.
Read what Yann LeCun has to say about AGI.
Thanks :)
Merely “human” is a pretty low bar. Is it gonna be like the homeless meth addict spazzing out and ranting and raving on the sidewalk? That’s “human intelligence”, too.
I honestly imagine the AI doing exactly that if it’s hedonistic setting is set to high as it fails to gauge how to control pleasure
Tech Industry
5h
427
I don’t know the first thing about ML. 7-8 yoe.
Personal Finance
5h
556
What can you buy with 5M
2024 Presidential Election
18h
1267
How is such a close race even possible?
Today I Learned
2d
1203
Unpacking Zionism
Tech Industry
Yesterday
693
Am I the asshole for outing my friend to his (religious) parents?
Great can it file taxes?
I suspect it will eventually be able to put H&R Block out of business