News for hackers. Stuff that matters.

A comment on Google Invests Almost $400M in ChatGPT Rival Anthropic. Parent comment

Re:Google Invests Almost $400M in ChatGPT Rival Anthropic by 0XYZxyzXYZ0 on Sunday February 05, 2023 @08:02AM (#34664593)
This doesn't make sense. They did invest in PaLM, which is ~2.5 larger than GPT-3, but no where near 100x better. I do agree that the limiting factor is the data, not scale though. What data has Google that OpenAI (or any other LLM company) can't scrape from the web as well? Using Gmail data without consent would be definitely a move towards a mountain of law suits.
[ Parent | on HN ]
Re:Google Invests Almost $400M in ChatGPT Rival Anthropic by CuriouslyC on Sunday February 05, 2023 @09:21AM (#34665093)
Google has a lot of scanned/ocr'd books that aren't available in public data sets.

I wouldn't be surprised if they used anonymized gmail data though. Even if they removed all proper nouns it'd probably provide a lot of extra tokens for training.

[ Parent | on HN ]
Re:Google Invests Almost $400M in ChatGPT Rival Anthropic by flangola7 on Sunday February 05, 2023 @09:23AM (#34665107)
PaLM isn't public, there's no way to know if it is better or if so how much better it is.
[ Parent | on HN ]