I have not watched the video yet, however I think it is fairly safe to say that "Yes, [AI Company X] is training AI on [any or all publicly accessible information]."
I have not watched the video yet, however I think it is fairly safe to say that "Yes, [AI Company X] is training AI on [any or all publicly accessible information]."
Watched the video the other day before it was posted to Tildes. There's a part in the video where he clips an interview with a YouTube exec. They were point blank asked if Google uses YouTube...
Watched the video the other day before it was posted to Tildes. There's a part in the video where he clips an interview with a YouTube exec. They were point blank asked if Google uses YouTube videos to train Gemini. The exec dances around the question with "Google uses YouTube videos within it's terms of service". Which is basically "yes but I don't want to say yes".
Hank goes into a lot more details and in depth thoughts about the whole situation though and it's well worth the watch.
He explains Google AI's position on open web access. The video is more about whether Google considers the YouTube TOS as specifically allowing scraping for LLM learning, and his mixed feelings...
He explains Google AI's position on open web access.
The video is more about whether Google considers the YouTube TOS as specifically allowing scraping for LLM learning, and his mixed feelings about the implications of creating online content under corporate TOS that can be reinterpreted however it suits the parent company.
He eventually settles on asking for an opt-out option, and encourages other creators to chime in in the comments. It's interesting looking at the comments and seeing a veritable who's who of educational YouTubers.
A thought-provoking vlogbrothers entry by Hank Green (now with chemo curls) making an argument that Google is probably training their LLMs on YouTube videos, and why he (might) have a problem with...
A thought-provoking vlogbrothers entry by Hank Green (now with chemo curls) making an argument that Google is probably training their LLMs on YouTube videos, and why he (might) have a problem with that.
I have not watched the video yet, however I think it is fairly safe to say that "Yes, [AI Company X] is training AI on [any or all publicly accessible information]."
Watched the video the other day before it was posted to Tildes. There's a part in the video where he clips an interview with a YouTube exec. They were point blank asked if Google uses YouTube videos to train Gemini. The exec dances around the question with "Google uses YouTube videos within it's terms of service". Which is basically "yes but I don't want to say yes".
Hank goes into a lot more details and in depth thoughts about the whole situation though and it's well worth the watch.
He explains Google AI's position on open web access.
The video is more about whether Google considers the YouTube TOS as specifically allowing scraping for LLM learning, and his mixed feelings about the implications of creating online content under corporate TOS that can be reinterpreted however it suits the parent company.
He eventually settles on asking for an opt-out option, and encourages other creators to chime in in the comments. It's interesting looking at the comments and seeing a veritable who's who of educational YouTubers.
A thought-provoking vlogbrothers entry by Hank Green (now with chemo curls) making an argument that Google is probably training their LLMs on YouTube videos, and why he (might) have a problem with that.