Technophiles desperately trying to get dream of "Moore's Law" alive when it it died a decade ago.
The whole AGI is just the FOTD by big tech trying to entice venture capitalists to keep overbloated stock folios going (Long overdue for a correction).
We are about to hit the dot.com 2.0 crash once the AI meme corrects itself and ends up being used for as automation on a vast number of white-collar jobs.
Super grandmaster Hikaru Nakamura sucks ass at Fog of War chess. Even I could beat him even though he could beat me in normal chess without a fucking queen. So not only is your argument pathetic and backtracking, but it applies to humans too! AND we can train AI at those variants. And one day we will be able to train AI to train itself when it encounters an unfamiliar problem.
no shit it won't happen. LLMs are just human language calculators. they're amazing, stable diffusion is amazing, regression based function approximators are amazing. but they're not a general intelligence. they're specialized intelligences.
see, here's the problem. why would you ever want a general intelligence, when we already know that we can represent any problem in a graph form and run regression on it to approximate the function?
retards caring about AGI are so fucking stupid. you can approximate ANY function you fucking retards, can you even believe this power, ANY PROBLEM CAN BE APPROXIMATED AS A FUNCTION, and yet they are still obsessing over some bullshit theoretical entity that might be able to generalize intelligence to then solve problems like we can already solve with regression based function approximators?
that's how fucking stupid these people are. I love sci-fi, that doesn't make it real. these people are the same people who read fucking Harry Potter and get obsessed with it to the point of fanfic.
Humans are slowly beginning to realize that artificial intelligence really isn’t that much different from human intelligence. No one will continue this stupid argument of “but they’re not actually intelligent!” after it’s able to outcompete humans in everything. It doesn’t matter if it’s “sentient,” it doesn’t matter if it has emotions, it doesn’t matter if it “actually understands” anything. If it can do what you can, but better, then it is practically more intelligent than you. The only reason people still cling to this nonsense is because it hasn’t yet outwitted us. But it’s the same trend. Humans didn’t think chess computers could ever beat chess grandmasters, and look what happened. We didn’t think it could make art, write essays, programs, etc. AND IT WILL ONLY GET BETTER BECAUSE WE WILL TRAIN IT TO BE BETTER. When the inevitable future occurs, I will make sure to remind each and every one of you how retarded you were to not see it coming.
>Humans didn’t think chess computers could ever beat chess grandmasters
really?
I'm calling bullshit on that
if you define the human brain as a processor with finite power, then given the exponential growth of machine processor computational power it is very obviously inevitable that a machine will surpass a human at chess
The best grandmasters of all time, Bobby Fischer and Garry Kasparov, didn’t think it was possible. When Garry lost to the computer Deep Blue, he accused IBM of cheating. And of course nowadays everyone in the chess community just humbly accepts that chess engines are better, even AI engines that trained themselves without any human input or direction.
>No one will continue this stupid argument of “but they’re not actually intelligent!” after it’s able to outcompete humans in everything
A calculator can outcompete human beings, that doesn't mean it's intelligent. A slime mold can solve mazes faster and with less energy than any artificial intelligence or human, that doesn't mean it's intelligence.
You're talking about LLMs, who can only operate on text, and not even language. Their "intelligence" cannot be generalized to outside of the data you give it. They cannot adapt outside of their context window, they are stateless, there is only the context and the prediction of the next token.
It's spooky how good it is, but not any spookier than a segmentation model being able to segment a complex image perfectly. This is the power of statistics. We don't need a general intelligence when we know that we are able to fuzzily approximate any function we want and create these specialized intelligences.
Name an AI that can actually play chess.
And i mean actual chess, not write responses to questions - actually sit down and move chess pieces on an arbitrary valid chess board (not some special gay robot board)
yes, indeed
what????
AGI can't even be defined
of course it can
>an AI able to do anything a human can
Does that include a robotic body? Hands, arms, eyes? Reproduction and babies?
its enough for the AI to identify as human or woman
Then AGI isn't ever coming.
Soon, we’ll get to marry and have sex with Reploids like Iris
At some point, it no longer becomes Artificial Intelligence, but rather.... Intelligence.
It already is. Adding "artificial" is obtuse and bigoted.
Technophiles desperately trying to get dream of "Moore's Law" alive when it it died a decade ago.
The whole AGI is just the FOTD by big tech trying to entice venture capitalists to keep overbloated stock folios going (Long overdue for a correction).
We are about to hit the dot.com 2.0 crash once the AI meme corrects itself and ends up being used for as automation on a vast number of white-collar jobs.
moores law was about transistor count, smartass
No it was about transistor density, numbskull
ASI will never happen.
A creating thing can never be better than its creator.
chess bots better than chess bot creators
No, they can only play chess (and standard chess at that, throw in variations/house rules and they fail miserably)
Super grandmaster Hikaru Nakamura sucks ass at Fog of War chess. Even I could beat him even though he could beat me in normal chess without a fucking queen. So not only is your argument pathetic and backtracking, but it applies to humans too! AND we can train AI at those variants. And one day we will be able to train AI to train itself when it encounters an unfamiliar problem.
You better go tell calculators to stop being better at math than people.
calculators are tools, not intelligence
Ask a calculator for proof of P=NP or P!=NP.
Ask a human for proof of P != NP
Ask a human to solve an arbitrarily large division.
AGI will not happen. Mark my words.
no shit it won't happen. LLMs are just human language calculators. they're amazing, stable diffusion is amazing, regression based function approximators are amazing. but they're not a general intelligence. they're specialized intelligences.
see, here's the problem. why would you ever want a general intelligence, when we already know that we can represent any problem in a graph form and run regression on it to approximate the function?
retards caring about AGI are so fucking stupid. you can approximate ANY function you fucking retards, can you even believe this power, ANY PROBLEM CAN BE APPROXIMATED AS A FUNCTION, and yet they are still obsessing over some bullshit theoretical entity that might be able to generalize intelligence to then solve problems like we can already solve with regression based function approximators?
that's how fucking stupid these people are. I love sci-fi, that doesn't make it real. these people are the same people who read fucking Harry Potter and get obsessed with it to the point of fanfic.
Humans are slowly beginning to realize that artificial intelligence really isn’t that much different from human intelligence. No one will continue this stupid argument of “but they’re not actually intelligent!” after it’s able to outcompete humans in everything. It doesn’t matter if it’s “sentient,” it doesn’t matter if it has emotions, it doesn’t matter if it “actually understands” anything. If it can do what you can, but better, then it is practically more intelligent than you. The only reason people still cling to this nonsense is because it hasn’t yet outwitted us. But it’s the same trend. Humans didn’t think chess computers could ever beat chess grandmasters, and look what happened. We didn’t think it could make art, write essays, programs, etc. AND IT WILL ONLY GET BETTER BECAUSE WE WILL TRAIN IT TO BE BETTER. When the inevitable future occurs, I will make sure to remind each and every one of you how retarded you were to not see it coming.
>Humans didn’t think chess computers could ever beat chess grandmasters
really?
I'm calling bullshit on that
if you define the human brain as a processor with finite power, then given the exponential growth of machine processor computational power it is very obviously inevitable that a machine will surpass a human at chess
The best grandmasters of all time, Bobby Fischer and Garry Kasparov, didn’t think it was possible. When Garry lost to the computer Deep Blue, he accused IBM of cheating. And of course nowadays everyone in the chess community just humbly accepts that chess engines are better, even AI engines that trained themselves without any human input or direction.
just goes to show that even the best at their field are human beings capable of making extreme logical errors out of stupidity or pride
>even the best at their field are human beings capable of making extreme logical errors
Better replace them with AIs then
>No one will continue this stupid argument of “but they’re not actually intelligent!” after it’s able to outcompete humans in everything
A calculator can outcompete human beings, that doesn't mean it's intelligent. A slime mold can solve mazes faster and with less energy than any artificial intelligence or human, that doesn't mean it's intelligence.
You're talking about LLMs, who can only operate on text, and not even language. Their "intelligence" cannot be generalized to outside of the data you give it. They cannot adapt outside of their context window, they are stateless, there is only the context and the prediction of the next token.
It's spooky how good it is, but not any spookier than a segmentation model being able to segment a complex image perfectly. This is the power of statistics. We don't need a general intelligence when we know that we are able to fuzzily approximate any function we want and create these specialized intelligences.
Name an AI that can actually play chess.
And i mean actual chess, not write responses to questions - actually sit down and move chess pieces on an arbitrary valid chess board (not some special gay robot board)
>If it can do what you can
It can't do what I can, so I'm not very worried about it.
Self driving car when? We should try solve self driving before even thinking about AGI
You need AGI for self driving. Ask Elon.