OpenAI CTO SLIPS UP on TRAINING DATA? #ai #openai #technology #allinpodcast #investing #chatgpt
00:00:00.000 | - Did OpenAI just get caught with their hand
00:00:01.480 | in the cookie jar or the training data cookie jar?
00:00:04.120 | OpenAI's CTO was interviewed
00:00:06.440 | by the Wall Street Journal this week.
00:00:07.760 | - What data was used to train Sora?
00:00:10.040 | - We used publicly available data and licensed data.
00:00:14.740 | - So videos on YouTube?
00:00:16.820 | - I'm actually not sure about that.
00:00:21.180 | - You got to think the CTO of an organization
00:00:24.080 | whose job it is to build models based on training data
00:00:26.760 | knows where the training data is from.
00:00:28.840 | My interpretation of that answer is that she is hesitating
00:00:33.360 | 'cause she doesn't want to make a statement against interest.
00:00:35.520 | I don't know if I would say that she's lying per se,
00:00:38.560 | but I think she's probably concerned
00:00:40.960 | that if she comes right out and answers the question
00:00:44.000 | as directly as she could, that it could be a problem
00:00:46.900 | for them in all these lawsuits that they're now facing.