Catch up on stories from the past week (and beyond) at the Slashdot story archive

 


Forgot your password?
Close
typodupeerror
×
Google AI Technology

Google Denies Bard Was Trained With ChatGPT Data (theverge.com) 17

Posted by msmash from the closer-look dept.
An anonymous reader shares a report: Google's Bard hasn't exactly had an impressive debut -- and The Information is reporting that the company is so interested in changing the fortunes of its AI chatbots, it's forcing its DeepMind division to help the Google Brain team beat OpenAI with a new initiative called Gemini. The Information's report also contains the potentially staggering thirdhand allegation that Google stooped so low as to train Bard using data from OpenAI's ChatGPT, scraped from a website called ShareGPT. A former Google AI researcher reportedly spoke out against using that data, according to the publication. But Google is firmly and clearly denying the data was used: "Bard is not trained on any data from ShareGPT or ChatGPT," spokesperson Chris Pappas tells The Verge.
This discussion has been archived. No new comments can be posted.

Google Denies Bard Was Trained With ChatGPT Data More

Google Denies Bard Was Trained With ChatGPT Data

Comments Filter:

  • Trash article (Score:4, Insightful)

    by timeOday ( 582209 ) on Thursday March 30, 2023 @09:52AM (#63411588)
    Wow Verge, thanks for the rumor-mongering and hyperbole. "it's forcing its DeepMind division to help the Google Brain team beat OpenAI." Oh the humanity, Alphabet's AI research division is reduced to conducting AI research to support its product line!

  • Right. (Score:4, Interesting)

    by cirby ( 2599 ) on Thursday March 30, 2023 @10:01AM (#63411606)

    Like Google needs help finding a large amount of data to train their AIs.

    "We already have all of the data. And by that, we mean all of it, even the stuff you don't think we know about."

    • Weren't they scanning massive numbers of old books just a few short years ago? I can't imagine how much text that adds up to. And that would only be part of the training data.

      • They bought Dejanews and have Usenet from like the 80s to present. They have emails from Gmail. Youtube transcriptions. I think they even back up most sites as we see with their AMP system. They have to crawl the web to generate Pagerank rating to rank search results, may as well save it (like archive.org).
        The idea they have to steal someone's data is laughable. I can't imagine any entity other than maybe the NSA that has more stored data. The only thing I can imagine is one of their people on their own tri

  • This is just embarrassing now. He has accomplished nothing and has got you playing catch up.
  • How would that even be lower than OpenAI? They got all their data from things their engineers didn't write. The intellect of chatgpt is literally the stolen intellect of masses of morons talking about random shit. Until OpenAI starts paying people for their work which constituted training data at a price set by those people that content was produced by it's at best an argument of who's worse: a thief, or the guy who stole from a thief? OpenAI is the least deserving of sympathy no matter how you cut it (

    • Re: (Score:3)

      by Bahbus ( 1180627 )

      Stop applying "open source" to anything and everything with open in the name. OpenAI's name was not ever a reference to open source to begin with, and only morons think that narrowly to begin with. And if you think that anyone who used it, should be compensated for that "data" if it was used, then you have a very over-inflated sense of worth for yourself and other users. Even if OpenAI did pay out to the "masses of morons", as you put it, their data would be worth pennies each, at best. The data is worth ne

      • It literally was. They raised all that money as a non-profit under the guise of being 100% open with AI research. Once they saw the money they could make from it they transitioned to a for-profit corp and kept the money. Their entire existence is predicated on fleecing investors under the notion of open AI research.
      • And it's not about "their data would be worth pennies each" (it would actually be in the multi-dollar range) it's about the cost to OpenAI in paying those pennies a piece. Everything they have was stolen, from other people's data, their thought processes (yes, that's what LLMs are: it's literally how people think distilled down into a model,) to the donations initially aimed at "open AI research." The simple fact is: they should not exist as an organization, they are thieves, and the last thing we need ar

  • Obviously (Score:3)

    by RJFerret ( 1279530 ) on Thursday March 30, 2023 @11:07AM (#63411778)

    Anyone who has used both knows Bard has access to recent data ChatGPT doesn't, but ChatGPT provides dramatically more useful different results that match the intent of the query better.

    You can't get ChatGPT style responses from Bard, and similarly if you want info post 2021, you need to ask Bard.

    Were the conjecture the case, the results would be a subset, heck, Bard even will outright tell you it can access websearches although it doesn't do great with social media posts from today.

  • Jacob Devlin left Google to immediately join its rival OpenAI after attempting to warn Google not to use that ChatGPT data because it would violate OpenAI’s terms of service

    Lol, since when AI tech bros are worried about copyrights issues? I guess this guy is not yet aware of the 2$/hour kenyan workers used by his new employer and will resign as soon he is informed.

  • If Google really did train Bard on ChatGPT, why would it be so much worse than even GPT 3.5? It seems to me that they are doing their own thing, and at some point their tool will get better.

  • Everybody knows that.

Slashdot Top Deals

Most public domain software is free, at least at first glance.

Close