If a human copies someone else’s art, that is copyright infringement. You say that AI works exactly the same way. Somehow you think that people should not be angry when AI is used to copy their work. If I ask you what is 2+2, will you tell me 5?
The only way AI training is even remotely like how humans learn is if you have a very limited understanding of both how humans learn and how AI model training works.
Also, AI image generation is largely built off datasets of images that are classified and made useful for training by what is effectively slave labor, or at least considerably underpaid third world people.
Please stop anthropomorphizing code. We have generations of study into how the brain forms connections, various aspects of how memories are formed and what keeps some salient and others not, how different people process information and learn differently, how people build skills over time and applied effort… the list goes on. And all of AI is built off absurdly complex math in application, but not as much so in concept.
What I’m getting at is that actual scholarly information sources are out there about how humans learn and develop skills, and there are likewise scholarly sources on how AI and the underlying algorithms work (albeit harder to find unbiased papers with the current hype bubble around AI at the moment).
It doesn’t take much effort to expose yourself to enough to understand that comparing AI training to human learning is an incredibly lossy analogy that doesn’t hold up under barely any scrutiny.
Bro, literally all I said was that the training dataset was the same.
I’m well aware of how generative AI works. I’m an industry professional who attended the very first AI/ML training course by a major cloud provider back in 2023 and was instrumental in changing the way they operated the AI side of the house.
But the parallels are there. Neural networks are designed based on the human brain, and vector driven databases aren’t super dissimilar to how neurons interact. A ton of human memory and processing is based on referential data, like gen AI.
Yes, it might not be able to approximate human intelligence and the actual workings of the brain yet, but it’s on its way.
People tend to forget how computers used to take up entire floors of office buildings and a simple hard drive was the size of an industrial washing machine. Gen AI is in its infancy, and while its current state is highly inefficient and flawed, that’s all the reason for it to improve.
Is current gen AI a solution for anything? Absolutely not. Is it a stepping stone towards true artificial intelligence? Absolutely.
I swear, Lemmy users would have been the people complaining that Excel and Word would be the downfall of society. There are absolutely legitimate complaints about the technology, but to completely dismiss it without looking at the nuances of the situation is assanine.
The truth is that it’s entirely possible for a business to ethically use an LLM. I know this for a fact because I was intimately involved in the implementation of one. The entire thing was trained on our proprietary dataset that had been built over 40 years of industry experience, on our servers which were powered by green energy.
All artists copy.
“Imitation is the sincerest form of flattery”
It’s also how you learn to do anything.
That’s why I think it’s funny that people hate on AI generated art for using copyrighted artwork for training.
It’s the exact same thing a human would do to learn how to do art.
If a human copies someone else’s art, that is copyright infringement. You say that AI works exactly the same way. Somehow you think that people should not be angry when AI is used to copy their work. If I ask you what is 2+2, will you tell me 5?
The only way AI training is even remotely like how humans learn is if you have a very limited understanding of both how humans learn and how AI model training works.
Also, AI image generation is largely built off datasets of images that are classified and made useful for training by what is effectively slave labor, or at least considerably underpaid third world people.
Please stop anthropomorphizing code. We have generations of study into how the brain forms connections, various aspects of how memories are formed and what keeps some salient and others not, how different people process information and learn differently, how people build skills over time and applied effort… the list goes on. And all of AI is built off absurdly complex math in application, but not as much so in concept.
What I’m getting at is that actual scholarly information sources are out there about how humans learn and develop skills, and there are likewise scholarly sources on how AI and the underlying algorithms work (albeit harder to find unbiased papers with the current hype bubble around AI at the moment).
It doesn’t take much effort to expose yourself to enough to understand that comparing AI training to human learning is an incredibly lossy analogy that doesn’t hold up under barely any scrutiny.
Bro, literally all I said was that the training dataset was the same.
I’m well aware of how generative AI works. I’m an industry professional who attended the very first AI/ML training course by a major cloud provider back in 2023 and was instrumental in changing the way they operated the AI side of the house.
But the parallels are there. Neural networks are designed based on the human brain, and vector driven databases aren’t super dissimilar to how neurons interact. A ton of human memory and processing is based on referential data, like gen AI.
Yes, it might not be able to approximate human intelligence and the actual workings of the brain yet, but it’s on its way.
People tend to forget how computers used to take up entire floors of office buildings and a simple hard drive was the size of an industrial washing machine. Gen AI is in its infancy, and while its current state is highly inefficient and flawed, that’s all the reason for it to improve.
Is current gen AI a solution for anything? Absolutely not. Is it a stepping stone towards true artificial intelligence? Absolutely.
I swear, Lemmy users would have been the people complaining that Excel and Word would be the downfall of society. There are absolutely legitimate complaints about the technology, but to completely dismiss it without looking at the nuances of the situation is assanine.
The truth is that it’s entirely possible for a business to ethically use an LLM. I know this for a fact because I was intimately involved in the implementation of one. The entire thing was trained on our proprietary dataset that had been built over 40 years of industry experience, on our servers which were powered by green energy.
Not exactly the same the human would have to pay to view the art first
Not necessarily but also irrelevant whether they did
What are you talking about? Art is available on the Internet without paying to view it. The same art that the AI models train on.