Despite creating AI chatbot LMChat before OpenAI’s ChatGPT, Jeff Dean says Google AI leaders failed to ‘appreciate…’


Legendary computer scientist and software engineer Jeff Dean made an astonishing revelation recently. According to 58-year-old former chief scientist of Google, Google had built a ChatGPT-style artificial intelligence (AI) chatbot much before OpenAI released its own. Jeff Dean on 5 August disclosed that leading search engine created its version of chatbot — LMChat — a year before the launch of OpenAI’s groundbreaking product in 2022.

He made this announcement in a podcast video recorded before he announced that he would leave the company after 27 years. During the days of Google’s invention of chatbot. Dean led Google’s AI division. He served this position from 2018 to 2023 before taking up the chief scientist role.

In the clip, he can be heard saying, “We actually had an internal chatbot system that Googlers could play with even before ChatGPT came out.” Recalling the pandemic days, he said, “A lot of Googlers would enjoy… everyone’s lockdown at home and so they’d enjoy spending time chatting with it during lunch because was a nice engaging lunch partner.”

He added, “And I think one of the things we were a little you know our view of things from a search perspective was like these models elucidate a lot and they don’t get things right correctly you know a lot of the time or some of the time and that means that they aren’t as useful as they could be…. we’d like to make that better and you know from a search perspective. You want to get the right answer 100% of the time. Ideally, we’re going to be very high on factuality and these models were not near that bar…. I think what we were a little unsure about is that they were incredibly useful.”

What is LMChat?

Notably, LMChat, developed by Google and DeepMind, was an internal conversational AI chatbot which offered a ChatGPT-like user experience, but Google kept it unreleased due to fears that it would reduce Google Search usage and concerns over inaccurate or problematic outputs. Google invented the Transformer architecture in 2017 that underpins modern Large Language Models (LLMs).

‘We didn’t quite appreciate’

Describing the use case of chatbots, he noted, “We didn’t quite appreciate how useful they could be for things you wouldn’t ask a search engine. Like, help me write a note or take this stack and give a quick summary of it. That’s the kind of thing we have seen people flock to in terms of using chatbots as amazing new capabilities rather than a search engine.”

Jeff Dean’s statement on chatbot invention comes at a time the cofounder of Google Brain, chief scientist of Google DeepMind and Google Research joined the list Google’s veteran departures, including Sanjay Ghemawat, Quoc Le and Oriol Vinyals.

Dean’s left Google to co-found the AI start-up Discovery Loop with Sanjay Ghemawat. Built as a public benefit corporation, Discovery Loop uses AI for automating scientific and engineering experiments.

Leave a Reply

Your email address will not be published. Required fields are marked *