4 ms·

Elon Musk: OpenAI is lying when it says it is not using copyrighted data

5 points by hhs 3y ago

noneoftheabove 3y ago
Ok another source validating what we already know. If open ai were half as transparent as they claim We would know on what sources they trained
- 3cats-in-a-coat 3y ago
  Everyone is using copyrighted data. Google is also using copyrighted data, or we'd have no search, and therefore we'd have no web, which thrives on search. Elon Musk's xAI Grok model is also using copyrighted data. So what, exactly, is supposed to be the "gotcha" or news here? What is there for OpenAI to reveal? Yes they scraped the web. That's blindingly obvious. But the idea they can license the world's content like you do when you buy a book for yourself, that's as silly as saying Google should buy a license for the world's websites before it can index them. We should instead be steering the conversation towards how can we use AI itself to identify sources to award when people pay for AI's services. Using the legal system or human judgement for this is laughable. The legal system acts on the scale of years. Humans act on the scale of days and months. AI acts on the scale is milliseconds.
renegade-otter 3y ago
At this point we know that Musk improvizes on the fly. He is like ChatGPT- good at presenting made up information in a credible way. Not to say that he is wrong, but the guy is following pizzagate conspiracies on Twitter in his spare time, which he seems to have tons of. And by the way, the same interview where his sounds like he is either high, severely sleep-deprived, or both.
dragonwriter 3y ago
In what context did OpenAI say they aren’t using copyrighted data in the first place? OpenAI has publicly said on multiple occasions [0] that it does use cooyrighted data for training, and that that use is fair use under the law. They may have said that they don’t use copyrighted data in some narrower context, but I am unfamiliar with the claim. I also wouldn’t take Elon at his word about anything related to this, given his history of both publicly lying about things he is in a position to know about when it suits him and his history of making up strong statements where he isn't in a position to know, and the fact that he has a commercial interest as someone trying to stand up an OpenAI competitor to spread lies about both what OpenAI has claimed and what OpenAI is factually doing. [0] e.g., https://www.uspto.gov/sites/default/files/documents/OpenAI_RFC-84-FR-58141.pdf https://www.uspto.gov/sites/default/files/documents/OpenAI_R...

5 points by hhs 3y ago

Ok another source validating what we already know. If open ai were half as transparent as they claim We would know on what sources they trained

3cats-in-a-coat 3y ago
Everyone is using copyrighted data. Google is also using copyrighted data, or we'd have no search, and therefore we'd have no web, which thrives on search. Elon Musk's xAI Grok model is also using copyrighted data. So what, exactly, is supposed to be the "gotcha" or news here? What is there for OpenAI to reveal? Yes they scraped the web. That's blindingly obvious. But the idea they can license the world's content like you do when you buy a book for yourself, that's as silly as saying Google should buy a license for the world's websites before it can index them. We should instead be steering the conversation towards how can we use AI itself to identify sources to award when people pay for AI's services. Using the legal system or human judgement for this is laughable. The legal system acts on the scale of years. Humans act on the scale of days and months. AI acts on the scale is milliseconds.

renegade-otter 3y ago

At this point we know that Musk improvizes on the fly. He is like ChatGPT- good at presenting made up information in a credible way. Not to say that he is wrong, but the guy is following pizzagate conspiracies on Twitter in his spare time, which he seems to have tons of. And by the way, the same interview where his sounds like he is either high, severely sleep-deprived, or both.

dragonwriter 3y ago

In what context did OpenAI say they aren’t using copyrighted data in the first place? OpenAI has publicly said on multiple occasions [0] that it does use cooyrighted data for training, and that that use is fair use under the law. They may have said that they don’t use copyrighted data in some narrower context, but I am unfamiliar with the claim. I also wouldn’t take Elon at his word about anything related to this, given his history of both publicly lying about things he is in a position to know about when it suits him and his history of making up strong statements where he isn't in a position to know, and the fact that he has a commercial interest as someone trying to stand up an OpenAI competitor to spread lies about both what OpenAI has claimed and what OpenAI is factually doing. [0] e.g., https://www.uspto.gov/sites/default/files/documents/OpenAI_RFC-84-FR-58141.pdf https://www.uspto.gov/sites/default/files/documents/OpenAI_R...