• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Engineers Recall Google's Pre-ChatGPT Internal Chatbot

    Engineer who worked on the project says Google built LMChat but held it back.

    DR
    SW
    SW
    12 Sources, 61d ago, first seen 61d ago

    TLDR

    Thibault Sottiaux posted that he joined the team behind an internal Google chatbot called LMChat. He stated the system resembled ChatGPT roughly a year earlier yet stayed unreleased because leadership feared it would disrupt Google. DeepMind faced blocks on shipping similar products for the same reason. Cheng Lou referenced a Jeff Dean interview about an earlier internal bot. Multiple engineers and observers quoted or replied to the post, noting parallel stories at other labs and internal OpenAI debates before the public ChatGPT launch. No first-party confirmation from Google appears in the posts.

    Combined views

    2.4M

    12 Sources, first seen 61d ago

    Combined views

    2.4M

    12 Sources, first seen 61d ago

    16K likes
    16K likes
    343 comments
    2.6K saves
    1K reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    343 comments
    2.6K saves
    1K reposts
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    12 Sources

    @_chenglouI sometime think about that Jeff Dean interview where he said they had an internal bot before ChatGPT but didn't think it was better than just googling
    @thsottiaux@_chenglou I was part of that team. Basically ChatGPT one year before it came out. Called LMChat and then another codename. Google was too nervous to release it and DeepMind was blocked from shipping products that could disrupt Google. I think about this a lot.
    @swyx> Google was too nervous to release it and DeepMind was blocked from shipping products that could disrupt Google. bookmark for the next vc that asks you "what if <incumbent> builds this?"
    @GergelyOroszI remember a time in the mid-2010s when startups were getting existential dread when Google launched a competing feature. Like Google Flights, and the fear at Skyscanner. Turned out Google never executed as well on these side quests as on Ads. And too nervous to disrupt itself:
    @HeinrichKuttler@thsottiaux @_chenglou Inflection AI was in a similar boat - founded to train and release a LLM chatbot, something existed early and it was not released before ChatGPT.
    @tunguzThis so wild.
    @simonw@_chenglou Slight echoes of that from OpenAI themselves, apparently there was much internal debate over whether ChatGPT was worth launching or not https://www.nytimes.com/2023/02/03/technology/chatgpt-openai-artificial-intelligence.html
    @ScobleizerNothing new. I told Bill Gates and several other Microsoft execs to buy all of Web 2.0. Sat next to Tim Oreilly at Google’s first advertisers’ conference and he had just written the 2.0 memo the night before. He taught me everything. The answer I got back was my ideas had no business value.
    @delipraoOn face value, this tweet makes it appear like Google leadership did not know what they were doing, but that’s not the full story. Working out the timing on this, LMChat should be in 2021, the year when the “Stochastic Parrots” nonsense came out and led to a mass psychosis. In 2020/21, you would get mobbed and cancelled if you even as much questioned the authors about their premise. It was a terrible time to be scientifically objective in the open for many. It was also when folks were trying to turn AI/ML into a vehicle for social justice and affirmative action movements, even if it meant not looking at the science for its own sake. They were not entirely wrong (AI should not perpetuate evils), but they became a sort of extremist about it and wanted to position themselves as gatekeepers for what AI research projects saw the light of day. From private conversations with Googler friends, two of the authors on the SP paper were kicked out of Google because they made the environment at Google toxic (those folks would go on publicly to say they left because the environment at Google was toxic for them), and many researchers were frustrated because they could not explore the science even if they knew it would be useful/worked. Politically savvy mid-management figured it was convenient to weaponize these sentiments and kill peers’ projects in the competition for resources. Add to that, around 2020, antitrust lawsuits started trickling in at Google that would continue building momentum for years to come. Any Google product showing undesirable results was put under the microscope and ripped to shreds on Twitter by an ever-increasing number of chronically online people thanks to COVID. It was also a time when Google suffered a bunch of high-profile infrastructure glitches around search. We could keep going on and on with this, but you get the idea — things are rarely one-dimensional. Did Google undercapitalize their inventions? Absolutely, but things happen for reasons and Google won’t be the last company to experience this.
    @MLStreetTalkRT @thsottiaux: @_chenglou I was part of that team. Basically ChatGPT one year before it came out. Called LMChat and then another codename.…

    12 Sources

    @_chenglouI sometime think about that Jeff Dean interview where he said they had an internal bot before ChatGPT but didn't think it was better than just googling
    @thsottiaux@_chenglou I was part of that team. Basically ChatGPT one year before it came out. Called LMChat and then another codename. Google was too nervous to release it and DeepMind was blocked from shipping products that could disrupt Google. I think about this a lot.
    @swyx> Google was too nervous to release it and DeepMind was blocked from shipping products that could disrupt Google. bookmark for the next vc that asks you "what if <incumbent> builds this?"
    @GergelyOroszI remember a time in the mid-2010s when startups were getting existential dread when Google launched a competing feature. Like Google Flights, and the fear at Skyscanner. Turned out Google never executed as well on these side quests as on Ads. And too nervous to disrupt itself:
    @HeinrichKuttler@thsottiaux @_chenglou Inflection AI was in a similar boat - founded to train and release a LLM chatbot, something existed early and it was not released before ChatGPT.
    @tunguzThis so wild.
    @simonw@_chenglou Slight echoes of that from OpenAI themselves, apparently there was much internal debate over whether ChatGPT was worth launching or not https://www.nytimes.com/2023/02/03/technology/chatgpt-openai-artificial-intelligence.html
    @ScobleizerNothing new. I told Bill Gates and several other Microsoft execs to buy all of Web 2.0. Sat next to Tim Oreilly at Google’s first advertisers’ conference and he had just written the 2.0 memo the night before. He taught me everything. The answer I got back was my ideas had no business value.
    @delipraoOn face value, this tweet makes it appear like Google leadership did not know what they were doing, but that’s not the full story. Working out the timing on this, LMChat should be in 2021, the year when the “Stochastic Parrots” nonsense came out and led to a mass psychosis. In 2020/21, you would get mobbed and cancelled if you even as much questioned the authors about their premise. It was a terrible time to be scientifically objective in the open for many. It was also when folks were trying to turn AI/ML into a vehicle for social justice and affirmative action movements, even if it meant not looking at the science for its own sake. They were not entirely wrong (AI should not perpetuate evils), but they became a sort of extremist about it and wanted to position themselves as gatekeepers for what AI research projects saw the light of day. From private conversations with Googler friends, two of the authors on the SP paper were kicked out of Google because they made the environment at Google toxic (those folks would go on publicly to say they left because the environment at Google was toxic for them), and many researchers were frustrated because they could not explore the science even if they knew it would be useful/worked. Politically savvy mid-management figured it was convenient to weaponize these sentiments and kill peers’ projects in the competition for resources. Add to that, around 2020, antitrust lawsuits started trickling in at Google that would continue building momentum for years to come. Any Google product showing undesirable results was put under the microscope and ripped to shreds on Twitter by an ever-increasing number of chronically online people thanks to COVID. It was also a time when Google suffered a bunch of high-profile infrastructure glitches around search. We could keep going on and on with this, but you get the idea — things are rarely one-dimensional. Did Google undercapitalize their inventions? Absolutely, but things happen for reasons and Google won’t be the last company to experience this.
    @MLStreetTalkRT @thsottiaux: @_chenglou I was part of that team. Basically ChatGPT one year before it came out. Called LMChat and then another codename.…