In my point of view the copyright issue is as follows: If you obtain (legally) a book, you can used it for what ever you want , which includes the training of an AI or your brain, but you are not allowed to publish a copy of the learned text by yourself, that’s copyright infringement.
I disagree. They should be allowed to train on it, but we shouldn't be allowed to generate copyrighted works. I can read all of an author's novels and write a novel based on what I've learned from reading their copyrighted work, but I can't just write a novel that is not materially different from one of their's and sell it. I don't see the moral disconnect. Are you saying it's immoral for me to for example, study copyrighted works and use it as a "base model" for producing my own works using the same techniques, structures, etc?
I agree with this comment. I'm a songwriter, and I have no problem having my music used as training because it would be hypocritical as I have trained on and influenced by artists prior to me.
They’re mixing up the concepts of being a tool or a service provider with the products created through it. I get that’s different from what you say but I thought it’d be nice if I piled on a bit
You are a human, not a massive scale AI. There is no comparison. 20th century copyright laws that couldn’t have envisioned an ai are not fit for purpose now. Not saying the horse hasn’t bolted but I find it strange.no one clocks this
@@simonw7580 oh good so you agree that AI is not violating any current copyright laws, just the ones that you imagine should be in place. Thankfully we only have to follow the actual law.
GPT could simply be overly familiar with common tests like "Create a snake game". I enjoyed this, but maybe it would be worth creating another test that the models couldn't possibly be pre trained for. Also, you can toggle off the usage of your data to train the model in the settings in the bottom left, this has stayed off for me for months.
Correct me if I’m wrong but isn’t Gemini based on a larger router? I’ve heard that Gemini will route requests to different models depending on the tasks, like a simple “hello” messages get Gemini pro, while maybe a science questions would get Gemini advanced if you pay for it.
ive been reading ian m banks books recently that posits a Culture society that is post scarcity and AGI. people just hung out doing what they liked in a "hedonistic" society. i really wonder what will happen?
There's no way that Sam said GPT-4 is the dumbest model we'll ever have to use again (by a lot) and the difference between the two models are so minimal, if that's truly GPT-5. I would lose all confidence in OpenAI if that were the case and if "i'm also a good blah" is gpt5, we can toss any chance of AGI any time soon in the trash.
When you're talking about model spec what you're asking for is an AGI it needs to be intelligent enough to make these decisions for itself and understand the scenario and ramifications. The technology just isnt there yet.
In my point of view the copyright issue is as follows:
If you obtain (legally) a book, you can used it for what ever you want , which includes the training of an AI or your brain, but you are not allowed to publish a copy of the learned text by yourself, that’s copyright infringement.
I disagree. They should be allowed to train on it, but we shouldn't be allowed to generate copyrighted works. I can read all of an author's novels and write a novel based on what I've learned from reading their copyrighted work, but I can't just write a novel that is not materially different from one of their's and sell it. I don't see the moral disconnect. Are you saying it's immoral for me to for example, study copyrighted works and use it as a "base model" for producing my own works using the same techniques, structures, etc?
I agree with this comment. I'm a songwriter, and I have no problem having my music used as training because it would be hypocritical as I have trained on and influenced by artists prior to me.
They’re mixing up the concepts of being a tool or a service provider with the products created through it. I get that’s different from what you say but I thought it’d be nice if I piled on a bit
You are a human, not a massive scale AI. There is no comparison. 20th century copyright laws that couldn’t have envisioned an ai are not fit for purpose now. Not saying the horse hasn’t bolted but I find it strange.no one clocks this
@@simonw7580 oh good so you agree that AI is not violating any current copyright laws, just the ones that you imagine should be in place. Thankfully we only have to follow the actual law.
GPT could simply be overly familiar with common tests like "Create a snake game". I enjoyed this, but maybe it would be worth creating another test that the models couldn't possibly be pre trained for.
Also, you can toggle off the usage of your data to train the model in the settings in the bottom left, this has stayed off for me for months.
love the content keep it up!
great work! excellent updates
Correct me if I’m wrong but isn’t Gemini based on a larger router? I’ve heard that Gemini will route requests to different models depending on the tasks, like a simple “hello” messages get Gemini pro, while maybe a science questions would get Gemini advanced if you pay for it.
If you do the tie thing in the arena are you providing bad data to the rankings? If you say the worst model ties with gpt4t?
ive been reading ian m banks books recently that posits a Culture society that is post scarcity and AGI. people just hung out doing what they liked in a "hedonistic" society. i really wonder what will happen?
There's no way that Sam said GPT-4 is the dumbest model we'll ever have to use again (by a lot) and the difference between the two models are so minimal, if that's truly GPT-5. I would lose all confidence in OpenAI if that were the case and if "i'm also a good blah" is gpt5, we can toss any chance of AGI any time soon in the trash.
People who are against AI are hindering progress, should be recognized as alarmists and punished by a tribunal!
Sound?
Have you tried refreshing or changing the quality? It definitely has sound, sometimes early on there's issues.
No sound here
Have you tried refreshing or changing the quality? It definitely has sound, sometimes early on there's issues.
When you're talking about model spec what you're asking for is an AGI it needs to be intelligent enough to make these decisions for itself and understand the scenario and ramifications. The technology just isnt there yet.
English is the new Computer Science! Do a joint degree in English and CS
Interesting how people are (quite rightly) turning against OpenAI's policies. Maybe Mr Altman shouldn't have returned.
Jensen Huang gives me weird vibes.
yeah - education is ripe for destruction, but won't happen because of union power. shame.