The studio behind the Postal franchise just announced that it will be publishing a brand new Postal game, to be developed by Goonswarm Games. Unfortunately, it even more recently announced it would be cancelling said game, due to accusations that Goonswarm had used generative AI in its development.
That is correct, but you completely miss the scales, not to mention the lack of learning when the AI makes a mistake.
When a human makes a mistake, they can learn from it, understand what went wrong and improve, when an AI makes a mistake it is just being told “bad”, it is really difficult to define why you rejected a specific response.
So not only does AI use more energy, far more is wasted
Getting off topic but these comments have me wondering, how much fresh water do we actually use to survive? Obviously we consume water to drink. And use it in our kitchens and bathrooms. But we know that’s just a small percentage of water use. The largest use is agriculture. I’ll count that as our consumption as well, because we eat the food produced. But I wonder what percentage of overall use goes to things that do not directly contribute to our survival. Like landscaping e.g. watering lawns. Especially in desert areas. Golf courses, that sort of thing.
My postulate was that an AI will not die, and eventually just learn more and more (or get better tuned). It can just function forever in theory.
We mere humans first spend many years learning to read and write, and so on, from 6 to 26 or more. Then we get a job and learn all the time, the welder gets better at welding, the doctor better at diagnosing and so on.
But then we humans grow old gets dementia and dies and society loses all that usefulness. If the society needs a new welder, we have to train someone to read, write, and eventually weld. They will get better at it until they die too. And so on.
An “AI welder” (or doctor or whatever) that is bad but “learns” over the years and softwate updates doesn’t go away suddenly, and can learn indefinitely, becoming extremely good at what it does, because it’s not resetted every 70 years or so.
The main issue with AI regarding learning is that while AI may be very logical, AI is not resonable, it can’t detect if something is resonable, leading to stupid conclusions and bad learnings
We are talking about to vastly different entities, to even consider that they would be equal in terms of energy use for a similar task is dumb. That would be like expecting the energy use for transporting box by truck or by ship to be equal. That just doesn’t work.
Just to play devils advocate, how much energy goes into growing our food, processing our food, transporting our food. Our education, and entertainment. All of which go into the output we produce. I wondering if at that point the numbers get closer.
When talking about these data centers, they often use the term enough to power a “small city.” So how much output could a “small city” of people produce? You’d have to break the result down into required man hours, and spread that across the population, as obviously the AI will do things faster than individual humans.
What a bunch of Bullshit no we don’t learn the same way as we don’t have perfect recall. And we do not consume the amount of energy AI data centers do. They consume a cities worth or more of energy to make a select few rich. Whatever we are done as a species it was a good run. All this fucking consumerism and must have infinite growth on a finite planet. We are to stupid to realize we are dead.
Ironically AI doesn’t have perfectly recall either, and that’s kind of one of the main problems with it and hallucinations. It can easily get poisoned by a handful of data points in it’s training set. But even then, it can only really blend 2 data points together, it’s got no ability to extrapolate and think outside the box.
People learn in the same way
People consume that too. Even while not working.
Human specialists also let other people stop being specialists. When did specialisation become the norm? In the early paleolith? Earlier?
That’s not a waste, though. I’d want them around even they weren’t working.
That is correct, but you completely miss the scales, not to mention the lack of learning when the AI makes a mistake.
When a human makes a mistake, they can learn from it, understand what went wrong and improve, when an AI makes a mistake it is just being told “bad”, it is really difficult to define why you rejected a specific response.
So not only does AI use more energy, far more is wasted
Getting off topic but these comments have me wondering, how much fresh water do we actually use to survive? Obviously we consume water to drink. And use it in our kitchens and bathrooms. But we know that’s just a small percentage of water use. The largest use is agriculture. I’ll count that as our consumption as well, because we eat the food produced. But I wonder what percentage of overall use goes to things that do not directly contribute to our survival. Like landscaping e.g. watering lawns. Especially in desert areas. Golf courses, that sort of thing.
On the other hand, the human will grow old and the hard learned information will die with him.
Computers? The software will be there forever.
Now, that IP should be FOSS of course.
Humm, have you ever heard of the concept of teaching or books?
Both are methods of knowledge transfer.
You still have to teach and learn.
When you die, that knowledge has to be re-learned by someone else, taking maybe tens of years.
What are you talking about?
Knowledge can and is in fact far easier to transfer when you are alive.
Yes. I didn’t say you should learn while dead 😋.
My postulate was that an AI will not die, and eventually just learn more and more (or get better tuned). It can just function forever in theory.
We mere humans first spend many years learning to read and write, and so on, from 6 to 26 or more. Then we get a job and learn all the time, the welder gets better at welding, the doctor better at diagnosing and so on.
But then we humans grow old gets dementia and dies and society loses all that usefulness. If the society needs a new welder, we have to train someone to read, write, and eventually weld. They will get better at it until they die too. And so on.
An “AI welder” (or doctor or whatever) that is bad but “learns” over the years and softwate updates doesn’t go away suddenly, and can learn indefinitely, becoming extremely good at what it does, because it’s not resetted every 70 years or so.
The main issue with AI regarding learning is that while AI may be very logical, AI is not resonable, it can’t detect if something is resonable, leading to stupid conclusions and bad learnings
Sure, but that will evolve. We’ll have better AI five ten years from now.
I don’t. I see no reason to believe that machines use more energy than humans while achieving comparable results.
Just to play devils advocate, how much energy goes into growing our food, processing our food, transporting our food. Our education, and entertainment. All of which go into the output we produce. I wondering if at that point the numbers get closer.
When talking about these data centers, they often use the term enough to power a “small city.” So how much output could a “small city” of people produce? You’d have to break the result down into required man hours, and spread that across the population, as obviously the AI will do things faster than individual humans.
What a bunch of Bullshit no we don’t learn the same way as we don’t have perfect recall. And we do not consume the amount of energy AI data centers do. They consume a cities worth or more of energy to make a select few rich. Whatever we are done as a species it was a good run. All this fucking consumerism and must have infinite growth on a finite planet. We are to stupid to realize we are dead.
Ironically AI doesn’t have perfectly recall either, and that’s kind of one of the main problems with it and hallucinations. It can easily get poisoned by a handful of data points in it’s training set. But even then, it can only really blend 2 data points together, it’s got no ability to extrapolate and think outside the box.