This article also highlights how quickly analogies can simplify the point you're trying to make down to uselessness. "Asking ChatGPT 50 questions consumes the equivalent of 500ml of water"...
This article also highlights how quickly analogies can simplify the point you're trying to make down to uselessness. "Asking ChatGPT 50 questions consumes the equivalent of 500ml of water" displaces the context so much that the comparison stops making sense. If someone was at a press conference, they'd definitely have drank at least half a liter of water by the 50th question. And, for all the shortcomings the human brain might have as a computing unit, everyone agrees it's pretty damn energy efficient compared to actual silicon based computers. How is that supposed to give you a proper idea of the amount of energy consumed by LLMs? I believe there are massive issues regarding the impact of the energy consumption caused by this GenAI arms race (especially when companies start floating around the idea of getting their hands on nuclear power plants for the sole purpose of powering their server farms) but as the article rightly points out, this kind of analogy hinders more than helps in conveying this concern.
Also doesn't consider that the AI can help replace a human worker who consumes significant water (to grow food to feed them or the animals that get fed to them, to raise them for their first 22+...
Also doesn't consider that the AI can help replace a human worker who consumes significant water (to grow food to feed them or the animals that get fed to them, to raise them for their first 22+ non-productive years, etc.).
AI (combined with future robots) can enable humanity to degrow its population without economic / quality of life sacrifice by plugging labor shortages in an inverted population pyramid.
Currently our economic paradigm demands population growth for economic growth: we need an upright population pyramid where there are more (young) workers than retirees to both grow and provide for...
Currently our economic paradigm demands population growth for economic growth: we need an upright population pyramid where there are more (young) workers than retirees to both grow and provide for the non-working population.
Aging population = economic stagnation like in Japan.
Better AI and robotics can free us from this. We can have fewer workers than retirees (inverted aging population pyramid) and not end up with economic stagnation.
Then a shrinking population is no longer an existential issue for countries. We don't have to have a world where there needs to be poorer, less developed countries to produce babies as cheap labor for wealthier, more developed countries. A high standard of living can become available for everyone, at least decoupled from the need for an underclass of laborers to clean offices, cook meals, harvest produce, or build houses.
You don't need population growth. You'd just at least want replacement levels. The issue is more that a birthrate below replacement causes EXPONENTIAL DECAY, which is quite bad. Even just a...
You don't need population growth. You'd just at least want replacement levels. The issue is more that a birthrate below replacement causes EXPONENTIAL DECAY, which is quite bad. Even just a linearly declining population can be worked around.
Obviously having some buffer above replacement level is desirable.
Population decay isn't a concern. We are nowhere near needing to worry about population collapse to the point of engagerment of our species. If anything a reduction would likely help our species...
Population decay isn't a concern. We are nowhere near needing to worry about population collapse to the point of engagerment of our species. If anything a reduction would likely help our species sustain.
The reason why governments freak out about population decline is because our economic systems are built on the idea of eternal growth. If there's no growth then company values are reconsidered and there's a massive market shrink. It's not a real problem in the sense of survival but it's a problem for the wealthy as it threatens their wealth which is typically speculative (Musk does not actually have billions in cash he can deploy).
There's also the racists who fear replacement of their race if they can't dominate the others by population but their arguments have no intellectual value.
I read that to mean that right now we run into problems when generations shrink because the amount of elderly that aren't working and need care puts too big of a burden on the smaller next...
I read that to mean that right now we run into problems when generations shrink because the amount of elderly that aren't working and need care puts too big of a burden on the smaller next generations.
I don't really buy that gen AI will help with this even slightly though. These systems are strained by not having enough carers, medical professionals, food producers, and so on. The actual physical needs of humans that need physical things to be moved around in the real world. The best argument I could make on that front is that eliminating the roles AI can do frees up more people to do care jobs they probably don't want.
I also don't really think people are making decisions on whether or not to have kids based on the labor market. I can't imagine there's really anyone out there that wants kids but would think, "eh, we don't need their labor anymore so why bother?"
Example off the top of my head is how gen AI can help reduce busy paperwork for doctors, so they can spend more time providing care than filling out forms. Gen AI is an important piece of the...
I don't really buy that gen AI will help with this even slightly though. These systems are strained by not having enough carers, medical professionals, food producers, and so on. The actual physical needs of humans that need physical things to be moved around in the real world. The best argument I could make on that front is that eliminating the roles AI can do frees up more people to do care jobs they probably don't want.
Gen AI is an important piece of the puzzle to creating robot laborers. There are efforts to use gen AI to help robots engage in complex reasoning about situations.
A simple humanoid task like washing dishes can be quite complex.
If a robot picks up an object, is it meant to be washed?
Is it even dirty?
Should it be placed up, down, or sideways in a dishwasher?
Which rack should it be placed on?
I just asked ChatGPT-4o:
I am a robot. I am looking at what I've identified as a mug.
Can it be washed?
It looks dirty, should it be washed?
Can it be placed in a dishwasher?
Should it be placed facing up, down, or sideways in the dishwasher?
Which rack should I place it on?
4o says
Can it be washed?
• Yes, if it’s a mug, it can generally be washed.
2. It looks dirty, should it be washed?
• Yes, if it’s dirty, you should wash it.
3. Can it be placed in a dishwasher?
• Most mugs can go in the dishwasher, but check if it has any special coatings (e.g., delicate decorations, metallic prints, or hand-painted designs) that might be damaged.
4. Should it be placed facing up, down, or sideways in the dishwasher?
• Place it facing down so water can properly rinse out the inside.
5. Which rack should I place it on?
• Place it on the top rack, as this is where mugs, cups, and glasses typically go to avoid excessive heat exposure.
It won't let me read that article, but from the headline that is terrifying. One of the last things I want in a doctor or lawyer is for them to just be like, "it's too much paperwork, I'll just...
It won't let me read that article, but from the headline that is terrifying. One of the last things I want in a doctor or lawyer is for them to just be like, "it's too much paperwork, I'll just throw some info into the lie machine and hope for the best." Maybe I've just been too close to the tech, but I don't trust it at all for anything important.
On your robot example, unless you're counting image classification as AI (which I don't because it's been around far longer than this buzz word hype) then AI is probably actively detrimental. It encourages people to imagine and hope for incredibly impractical human-like robots powered by black boxes, rather than practical purpose-made machines.
I largely just don't buy into the hype. The promises will be left unfilled because it is marketed that it'll do stuff it'll never do. Just like how Tesla has claimed FSD to be a year away for like 9 years now these hard problems are going to remain unsolved for a very long time, probably until someone realizes there are just better other routes. Like which is really more likely between a complex AI robot that includes loading a dishwasher correctly into it, or a smart dishwasher that can just detect the shape of a dish to determine which side water would pool in? The latter seems far more likely to me and doesn't require any complex AI, but is instead not done probably because of the complex moving parts, parts the AI robot still needs, that are required to reorient dishes.
This article also highlights how quickly analogies can simplify the point you're trying to make down to uselessness. "Asking ChatGPT 50 questions consumes the equivalent of 500ml of water" displaces the context so much that the comparison stops making sense. If someone was at a press conference, they'd definitely have drank at least half a liter of water by the 50th question. And, for all the shortcomings the human brain might have as a computing unit, everyone agrees it's pretty damn energy efficient compared to actual silicon based computers. How is that supposed to give you a proper idea of the amount of energy consumed by LLMs? I believe there are massive issues regarding the impact of the energy consumption caused by this GenAI arms race (especially when companies start floating around the idea of getting their hands on nuclear power plants for the sole purpose of powering their server farms) but as the article rightly points out, this kind of analogy hinders more than helps in conveying this concern.
Also doesn't consider that the AI can help replace a human worker who consumes significant water (to grow food to feed them or the animals that get fed to them, to raise them for their first 22+ non-productive years, etc.).
AI (combined with future robots) can enable humanity to degrow its population without economic / quality of life sacrifice by plugging labor shortages in an inverted population pyramid.
What do you mean by "enable humanity to degrow its population"?
Currently our economic paradigm demands population growth for economic growth: we need an upright population pyramid where there are more (young) workers than retirees to both grow and provide for the non-working population.
Aging population = economic stagnation like in Japan.
Better AI and robotics can free us from this. We can have fewer workers than retirees (inverted aging population pyramid) and not end up with economic stagnation.
Then a shrinking population is no longer an existential issue for countries. We don't have to have a world where there needs to be poorer, less developed countries to produce babies as cheap labor for wealthier, more developed countries. A high standard of living can become available for everyone, at least decoupled from the need for an underclass of laborers to clean offices, cook meals, harvest produce, or build houses.
You don't need population growth. You'd just at least want replacement levels. The issue is more that a birthrate below replacement causes EXPONENTIAL DECAY, which is quite bad. Even just a linearly declining population can be worked around.
Obviously having some buffer above replacement level is desirable.
Population decay isn't a concern. We are nowhere near needing to worry about population collapse to the point of engagerment of our species. If anything a reduction would likely help our species sustain.
The reason why governments freak out about population decline is because our economic systems are built on the idea of eternal growth. If there's no growth then company values are reconsidered and there's a massive market shrink. It's not a real problem in the sense of survival but it's a problem for the wealthy as it threatens their wealth which is typically speculative (Musk does not actually have billions in cash he can deploy).
There's also the racists who fear replacement of their race if they can't dominate the others by population but their arguments have no intellectual value.
If the exponent is close enough to one, it doesn't seem so bad. (Consider a 1% a year decline.) It wouldn't take much immigration to make up for this.
I read that to mean that right now we run into problems when generations shrink because the amount of elderly that aren't working and need care puts too big of a burden on the smaller next generations.
I don't really buy that gen AI will help with this even slightly though. These systems are strained by not having enough carers, medical professionals, food producers, and so on. The actual physical needs of humans that need physical things to be moved around in the real world. The best argument I could make on that front is that eliminating the roles AI can do frees up more people to do care jobs they probably don't want.
I also don't really think people are making decisions on whether or not to have kids based on the labor market. I can't imagine there's really anyone out there that wants kids but would think, "eh, we don't need their labor anymore so why bother?"
Example off the top of my head is how gen AI can help reduce busy paperwork for doctors, so they can spend more time providing care than filling out forms.
Gen AI is an important piece of the puzzle to creating robot laborers. There are efforts to use gen AI to help robots engage in complex reasoning about situations.
A simple humanoid task like washing dishes can be quite complex.
I just asked ChatGPT-4o:
4o says
It won't let me read that article, but from the headline that is terrifying. One of the last things I want in a doctor or lawyer is for them to just be like, "it's too much paperwork, I'll just throw some info into the lie machine and hope for the best." Maybe I've just been too close to the tech, but I don't trust it at all for anything important.
On your robot example, unless you're counting image classification as AI (which I don't because it's been around far longer than this buzz word hype) then AI is probably actively detrimental. It encourages people to imagine and hope for incredibly impractical human-like robots powered by black boxes, rather than practical purpose-made machines.
I largely just don't buy into the hype. The promises will be left unfilled because it is marketed that it'll do stuff it'll never do. Just like how Tesla has claimed FSD to be a year away for like 9 years now these hard problems are going to remain unsolved for a very long time, probably until someone realizes there are just better other routes. Like which is really more likely between a complex AI robot that includes loading a dishwasher correctly into it, or a smart dishwasher that can just detect the shape of a dish to determine which side water would pool in? The latter seems far more likely to me and doesn't require any complex AI, but is instead not done probably because of the complex moving parts, parts the AI robot still needs, that are required to reorient dishes.