On: Google's Gemini model is delayed

[riku_iki]

> so it makes absolutely no sense to me that they can't manage to throw 200 data annotators at their model and make it best-in-class.

these foundation models are trained on trillions of tokens to achieve quality. 200 data annotators with unknown quality issues may be not enough..

[azinman2 → riku_iki]

Not to mention that you don’t think OpenAI is paying probably many, many more annotators?

[kristianp → azinman2]

Not to mention the free feedback they get from the 100 million chatgpt users.

[riku_iki → kristianp]

google has their own piles of data