Jeff Dean, the chief scientist at Google, along with three leading artificial intelligence researchers, is departing the tech giant to launch their own start-up. This significant shift highlights the evolving dynamics within Google and underscores the robust interest from venture capitalists in innovative A.I. concepts.
As Google’s 30th employee, Dr. Dean has been a pivotal figure in the company’s history. He was instrumental in establishing the global network that powers Google’s search capabilities and has been a frontrunner in A.I. research.
Despite a decline in his visibility within Google as leaders like Sergey Brin took charge of A.I. strategies, Dr. Dean has remained a key player in developing the company’s most critical technologies.
His new venture, named Discovery Loop, joins an expanding list of notable start-ups that aim to harness A.I. capable of self-improvement with minimal human intervention.
The concept of recursive self-improvement is viewed by Dr. Dean and his team as a transformative approach to rapidly advancing A.I. They believe this technique can aid in designing advanced hardware, discovering new pharmaceuticals, creating innovative materials, and expediting scientific breakthroughs.
“We foresee an opportunity for A.I. to automate traditionally human-intensive experimental processes,” Dr. Dean stated during an interview in his Silicon Valley home. “This will result in a larger volume and improved quality of experiments, leading to major scientific advancements.”
As Google navigates competition with smaller A.I. players like OpenAI and Anthropic, many distinguished researchers are being drawn away by lucrative venture funding and the ability to lead their own initiatives.
Over the last year, several top-tier researchers, including Peter Norvig—who headed Google’s research for 25 years—have transitioned to A.I. start-ups, including Recursive Superintelligence, another company devoted to recursive self-improvement.
“The environment at a start-up is vastly different—it’s more agile and focused,” remarked Tim Rocktäschel, who recently co-founded Recursive Superintelligence after leaving Google. “With a smaller team, there’s much less bureaucracy and political jargon.”
Dr. Dean is poised to step into the role of chief executive at his new start-up in the next few days, accompanied by long-time collaborators Sanjay Ghemawat, Quoc Le, and Oriol Vinyals, who have also made significant strides in developing Google’s chatbot technology since 2023.
The start-up is receiving initial funding from Radical Ventures, Khosla Ventures, and various investors from Silicon Valley, including Google’s parent company. While the financial specifics remain undisclosed, it seems the researchers are parting ways with Google amicably, with Alphabet agreeing to support their computing needs for their A.I. development for the next year.
“For over 27 years, Jeff and Sanjay have been at the forefront of pivotal technological changes, from the initial search infrastructure to the neural networks that shaped the current A.I. landscape,” commented Google’s CEO, Sundar Pichai. “Personally, it has been an honor to work alongside Jeff and Sanjay.”
Google is expected to collaborate with the new start-up, according to Pichai.
Dr. Vinyals emphasized the importance of having autonomy in shaping their own direction. “Extreme focus has immense power,” he remarked in an interview. “The ultimate form of extreme focus occurs when it aligns perfectly with the company’s mission.”
Dr. Dean joined Google in 1999, where he, along with Dr. Ghemawat and his team, developed complex software systems that enabled Google’s search engine to handle billions of queries efficiently and reliably.
Their innovations, including MapReduce, BigTable, and Spanner, have become benchmarks for the industry, guiding other tech giants like Amazon and Meta in managing global traffic. Notably, Dr. Dean and Dr. Ghemawat are the only Google employees honored as “senior fellows,” the company’s highest technical accolade.
In 2011, Dr. Dean spearheaded the establishment of Google Brain, an A.I. research lab that catalyzed a major industry movement towards A.I. technologies. Early members included Dr. Le and Dr. Vinyals, who were instrumental in creating the groundbreaking Google Translate technology.
However, after catalyzing technologies that are now foundational to contemporary chatbots, Google was slow to capitalize on its potential. After the launch of ChatGPT by OpenAI in late 2022 sparked an A.I. revolution, Google merged Google Brain with DeepMind, its London-based lab.
The new entity, named Google DeepMind, is led by DeepMind founder Demis Hassabis, with Dr. Dean and Dr. Vinyals focusing on designing Gemini, Google’s response to ChatGPT. Dr. Noam Shazeer, another leading researcher, joined this initiative but left for OpenAI less than two months ago.
Following Mr. Brin’s return to Google, the company ramped up its A.I. efforts, although by some measures, its technologies still lag behind offerings from U.S. start-ups such as Anthropic and OpenAI, as well as Chinese firms like Moonshot and Z.ai.
As Google cultivates a closer relationship with the U.S. government, Dr. Dean has been one of the few voices urging caution. Earlier this year, he publicly condemned the actions of federal agents in Minneapolis, labeling the incident “absolutely shameful.” He also participated in a legal brief supporting a lawsuit filed by Anthropic against the Defense Department, which banned government agencies from utilizing the start-up’s technologies.
Dr. Dean mentioned that unlike Google, his new venture will function as a public benefit corporation (P.B.C.), aiming to prioritize public good alongside profit. This structure is also adopted by other A.I. companies like Anthropic and OpenAI.
“We have the potential to effect substantial positive change globally by broadly distributing scientific discoveries,” he asserted. “Our decisions may not always align with pure financial interests.”
Source: www.nytimes.com
