Just a few years ago, the prevailing wisdom was that Google was so far ahead in the AI race, that they would be unstoppable. The early leader IBM – with it’s Jeopardy winning Watson juggernaut – had all but disappeared from the scene. Then Microsoft was still busy with a major course correction under Satya Nadella. Amazon was tending to business while wallowing in the profits of lockdown demand. Finally Apple was too busy playing with alleged self-driving cars and VR headsets to really care.

Googles lead in AI was all-but-insurmountable after Google acquired DeepMind in 2014. DeepMind had numerous algos that pushed deep learning into new territory. DeepMind stunned the computing and gaming world by beating the worlds lead Go champion.

Googles AI lead was so laughable, that Google had the audacity to name their first working AI, Google Brain. Google used it as if to say to the entire internet, challenge us at your own peril – we got this. Google Brain was presented as the heart-and-soul of the entire Google empire itself.

It has left many in the tech industry gasping to realize that Google’s AI efforts have been well behind everyone else. We all assumed Google was several years ahead. Then came ChatGPT and the famous code red meeting that had all our jaws on keyboards wondering, “Hey wait a minute, weren’t they set to lap the field with AI“? Daily after ChatGPT 3.5 released and went viral, story after story was printed about ChatGPT being a Google killer. Google actually did more than call a code red, they called in the founders Larry and Sergey.

Googles positioning of Google Brain now seems more like an over eager PR department than anything else, but when you are the 800lb gorilla that Google is – you can get away with such bluster. Given all the PR issues Google has had, it is more clear than ever, that Google has been dealing with significant internal issues that held up their AI developments.

Attention is All You Need

Published in mid-2017, it has become the most important paper written by Googlers since founders Larry Page and Sergey Brins original Page Rank paper that laid out the famous Google search engine algorithm. “Attention Is All You Need” (AIAYN) broke the AI processing log jam and enabled the new era of AI. The paper introduced a simple network architecture hack (a Transformer) that reduced the laborious and time consuming training overhead. It enabled smaller infrastructure kits to do the work of entire data centers. It has become one of the most cited papers in the history of AI with over 70k references on Google Scholar alone. The importance of this paper to AI can not be over stated.

The Gutting of Google Brain

There are eight credited authors of the AIAYN paper. All but one of those authors has left Google for positions elsewhere. Looking in from the outside, one can only guess what goes on inside some companies, but it is clear that Google has had issues retaining key personnel in the last five years. It would generously best be described as a mass exodus of AI engineers that have left Google for other companies or startups. One can only speculate that it has something to do with what a former employee called “a complete dumpster fire“.

PR Problems

In the five plus years since Attention is All You Need was published, Google AI programs have suffered one harsh public relations debacle after another. Here are some of the highlights:

Brain Drain – Attention is All You Need Authors

Lets take a look at where the authors of “Attention is all you Need” are at now.

Ashish Vaswani started Adept. Vaswani was a research scientist at Google Brain. Adept recently raised $350m to build LLM generative AI tools for software. Strangely, he has left Adept (a company he found) for some sort of unknown stealth startup. LinkedIn Adept.ai CrunchBase (Google Scholar Citations)

Noam Shazeer moved to Character.ai a chatbot or “dialog agent“. Shazeer was heavily involved in Googles famed ‘TenserFlow’ project that trained Googles LLM LaMDA systems. (LinkedIn)

Niki Parmar was a key component of Google’s “Brain” AI team. She left for Adept and now left there with Vaswamni for a stealthy backroom startup. (LinkedIn) (Google Scholar Citations)

Jakob Uszkoreit was also on the Google Brain team at Google for over a decade. He started Inceptive.life. Inceptive CrunchBase
(GoogleScholar Citations). (LinkedIn)

Aidan Gomez a cofounder at Cohere – a platform that gives developers and businesses access to NLP, powered by the latest generation of large language models. (CruchBase. (Personal Site), (LinkedIn), (Google Scholar Citations).

Lukasz Kaiser was an integral part of the Google Brain for eight years. He currently is an investor Pathway that adds machine learning to live event streams (CrunchBase Pathway). CrunchBase Kaiser, LinkedIn Google Scholar Citations)

Illia Polosukhin was at Google for three and a half years working on TensorFlow projects and managing a team of deep learning researchers. He CoFounded Near and is also listed as investor in nine startups (CrunchBase). Illia on Twitter and (LinkedIn).

Other players to have left the plex in search of greener pastures:

  • Mustafa Suleyman CoFounder DeepMind. He is now the CEO and co-founder of Inflection AI and a venture partner at Greylock Partners, (LinkedIn)

  • Ilya Sutskever is probably the best known of all the former Googlers. He is the Cofounder of OpenAI. He also was cofounder of DNNResearch with Geoffrey Hinton and Alex Krizhevsky that was acquired by Google and laid ground work for everything AI at Google afterwards.

  • Colin Raffel – Another Google Brain contributor. Now at Hugging Face.

  • Daniel De Freitas – core research scientist for seven years in the Google Brain project. Now also cofounder with Noam Shazeer at Character.ai.

  • Azalia Mirhoseini Worked on Google’s Tensor Processor chip design. He left to join Anthropic in 2022.
  • Fred Bertsch is now Co-Founder at Adept. He was a 12 year vetran at Google working on Google Brain.
  • Anna Goldie Also worked on Google’s Tensor Processor chip design. She left to join Anthropic in 2022.

  • Aravind Srinivas, was an intern at Google now Co-Founder at AI search engine startup Perplexity.ai. (CrunchBase), (LinkedIn)

  • Romal Thoppilan is at Chracter.ai as well after seven years at Google working on LaMDA. (LinkedIn)

  • Alicia Jin now at Character.ai worked for four years on Google Brain.

  • Jacob Devlin Now with OpenAi. Most noted for reporting that Google Bard was being trained on OpenAI data. Seems to be most noted for working on the Google BERT project. (LinkedIn)

  • Maithra Raghu and Fabio Petroni Co-founded Samaya AI. Petroni worked as an AI researcher at Meta (Faceboo) while Raghu spent five years working on Google Brain at Google. What makes their stealth startup interesting is that it was funded by Googles Head of Ai Jeff Dean, Facebooks head AI lead Yann LeCun, Mark Cuban of SharkTank/Dallas Mavericks, and Anne Wojcicki CEO of 23AndMe.


    Googlers that left for Adept.ai: Ashish Vaswani (also a staff research scientist at Google Brain), Niki Parmer (formerlystaff research scientist at Google Brain), Erich Elsen, Kelsey Szot, David Luan (previously a director at Google Research and VP of engineering at OpenAI), Fred Bertsch, and Anmol Gulaiti


    Googlers that left for Character.ai: Romal Thoppilan, Irwan Bello, and Daniel De Freitas.


    Googlers at Mustafa Suleyman Inflection.ai include: Rewon Child (former Google Brain and OpenAI researcher), Karen Simonyan, Maarten Bosma (research engineer at Googl), and Joe Fenton (Senior Product Manager at Google).


    Aidan Gomez, Ivan Zhang, Nick Frosst. Ed Grefenstette, and Phil Blunsom.


    Dario Amodei, Jack Clark, and ex OpenAI employees Tom Brown, Sam McCandlish and Amodei’s sister Daniela Amodei. Notable investers include Google, Skype co-founder Jaan Tallinn, and former Google CEO Eric Schmidt.

Related Stories