4 ms·
While I want to be very optimistic about open models, the fact that almost all top open models heavily use GPT-4 for synthetic data is discouraging. Even now it
by YetAnotherNick 3y ago
While I want to be very optimistic about open models, the fact that almost all top open models heavily use GPT-4 for synthetic data is discouraging. Even now it is against TOS and over time I think over time OpenAI would become better at detecting meaning the gap between open and closed models would increase.
- dimask 3y agoConsidering these models are released by "Microsoft AI", I doubt they do anything against the ToS of "Open"AI.
- dartos 3y agoI doubt it. They’re using gpt-4 data because it’s cheaper than getting real data, not because it’s better. If openai cares enough to stop people scraping responses, then we’ll just crowdsource them like open assistant did with their dataset.
- YetAnotherNick 3y agoThere is some news OpenAI spent 4 billion dollars for data sourcing.
- dartos 3y agoYeah but that’s what OpenAI might’ve spent
- 0x008 3y agoWhat can OpenAI realistically do about it except block the user account? I think the TOS should have no effect on the license and copyright of the resulting model, since outputs of AI Models cannot be copyrighted (at least in the US).
- YetAnotherNick 3y agoOpenAI would likely start using AI to detect the content generation and ban the users. It wouldn't even need to be GPT level. Just they could use some AI labelling and counting.