i mean you say that, but every time i see an AI or algorithm try to work in one of my fields of expertise it’s laughable how bad it is. maybe it’s not obvious to the layman, but try to ask it about something you’re an expert in. like it gives explicitly wrong information. people think it only hallucinates 1/5 of the time because the other 4/5 they’re asking about shit they don’t know
The current state of LLMs is not the end of development. Right now, they’re sloppy and they consume a lot of resources to function, but future iterations will be better at it, will compute faster, and use fewer resources. Hopefully, we will have escaped the disease that is capitalism before we get there, and the gains in productivity can be shared alike with everyone, rather than a select few.
An industrial revolution expert on Wired pointed out the quality of products improved with power looms and assembly lines and the countless other inventions that rose with mass-produced steel. AI has had a few successes in science by making discoveries that humans hadn’t yet, but other than that, we’re seeing AI slop, vibecoding, etc. produces worse results than when human beings do it all by hand.
The horror of this moment is not that AI will someday out perform humans in complex and creative labor, and do it affordably. It’s that corporations and billionaires believe they can force that moment to the present if they throw enough money and resources at it. And they’re doing this specifically and brazenly with the intention of shutting out the majority of the human race as if this were an Ayn Rand fantasy.
And they’re blind to the consequences of trying, which can break the global economy (and the global ecology).
Yup. Like I also said. It might not be LLMs. The same was once true of steam engines, which were great at long distances perhaps but were huge, heavy, expensive and complicated and required a team of people to operate. Horses became even more valuable for their last mile stuff. It took time before the internal combustion engine and 4 stroke compression suddenly made engine power smaller, cheaper, more efficient and easy to operate by a single person. In 1918, the US for example had 27 million head of horse, by 1960 it was 3 million. All I’m saying is that it will happen. When? I have no idea.
so like, i’m with you, only it’s not just artificial intelligence it’s also the power requirements. usable fusion is always 5 years away. like there’s even been articles about it. scaling up green energy just to meet our demands (and i’m not an expert, i’m just really good at estimating) could provide the power to a society, but the first few AI will need many times that. as i understand my Machine Learning colleagues from college (and it’s been a while since i’ve spoken to these friends, you know the type you see them every few years and no time has passed) we’re really good at throwing spaghetti at the wall and hoping it sticks. getting the spaghetti to cook itself, throw itself at the wall, and stick the first try? well we can make it look like that happens
For sure man. Haha yeah we are. It’s true. This is the critical aspect: teaching the spaghetti to harvest itself, make itself and cook itself. I think we are a ways away from it, but who knows. Energy requirements is also a huge issue.
i mean you say that, but every time i see an AI or algorithm try to work in one of my fields of expertise it’s laughable how bad it is. maybe it’s not obvious to the layman, but try to ask it about something you’re an expert in. like it gives explicitly wrong information. people think it only hallucinates 1/5 of the time because the other 4/5 they’re asking about shit they don’t know
The current state of LLMs is not the end of development. Right now, they’re sloppy and they consume a lot of resources to function, but future iterations will be better at it, will compute faster, and use fewer resources. Hopefully, we will have escaped the disease that is capitalism before we get there, and the gains in productivity can be shared alike with everyone, rather than a select few.
An industrial revolution expert on Wired pointed out the quality of products improved with power looms and assembly lines and the countless other inventions that rose with mass-produced steel. AI has had a few successes in science by making discoveries that humans hadn’t yet, but other than that, we’re seeing AI slop, vibecoding, etc. produces worse results than when human beings do it all by hand.
The horror of this moment is not that AI will someday out perform humans in complex and creative labor, and do it affordably. It’s that corporations and billionaires believe they can force that moment to the present if they throw enough money and resources at it. And they’re doing this specifically and brazenly with the intention of shutting out the majority of the human race as if this were an Ayn Rand fantasy.
And they’re blind to the consequences of trying, which can break the global economy (and the global ecology).
Yup. Like I also said. It might not be LLMs. The same was once true of steam engines, which were great at long distances perhaps but were huge, heavy, expensive and complicated and required a team of people to operate. Horses became even more valuable for their last mile stuff. It took time before the internal combustion engine and 4 stroke compression suddenly made engine power smaller, cheaper, more efficient and easy to operate by a single person. In 1918, the US for example had 27 million head of horse, by 1960 it was 3 million. All I’m saying is that it will happen. When? I have no idea.
so like, i’m with you, only it’s not just artificial intelligence it’s also the power requirements. usable fusion is always 5 years away. like there’s even been articles about it. scaling up green energy just to meet our demands (and i’m not an expert, i’m just really good at estimating) could provide the power to a society, but the first few AI will need many times that. as i understand my Machine Learning colleagues from college (and it’s been a while since i’ve spoken to these friends, you know the type you see them every few years and no time has passed) we’re really good at throwing spaghetti at the wall and hoping it sticks. getting the spaghetti to cook itself, throw itself at the wall, and stick the first try? well we can make it look like that happens
For sure man. Haha yeah we are. It’s true. This is the critical aspect: teaching the spaghetti to harvest itself, make itself and cook itself. I think we are a ways away from it, but who knows. Energy requirements is also a huge issue.