2 ms·
There's a network effect in that they are used more so they've generated more feedback from users, which is then used to improve GPT.
by BryanLegend 3y ago
There's a network effect in that they are used more so they've generated more feedback from users, which is then used to improve GPT.
- vitorgrs 3y agoThat doesn't make sense because GPT4 finished training in August 2022 - before ChatGPT 3.5 release in November 30. You could say they got data to train RLHF after the training, but that seems unlikely. Bing was launched in February 7 with GPT4 - that's just 2 months after ChatGPT launch.