Jeff Dean, one of Google's most senior and influential artificial intelligence executives, is stepping down from the search and cloud giant to found a new AI startup. The move marks the end of a career at Google that spanned more than two and a half decades and helped shape much of the company's technical foundation. Dean will be joined by several prominent Google researchers who are leaving with him to serve as co-founders.

The new company, called Discovery Loop, is structured as a public benefit corporation and aims to apply large-scale AI systems to scientific research. Dean is expected to serve as chief executive officer. The startup's stated mission is to automate the experimental process that underlies scientific discovery, using machine learning to design, run, and learn from vast numbers of experiments at once.

Key facts at a glance

• Jeff Dean is leaving Google after 26 years to become CEO of a new AI startup called Discovery Loop.
• Co-founders include Sanjay Ghemawat, Quoc Le, and Oriol Vinyals.
• Discovery Loop is a public benefit corporation focused on using AI to accelerate scientific discovery.
• The startup plans to run thousands of simultaneous experiments and explore recursive self-improvement.
• Funding comes from Alphabet, Radical Ventures, Khosla Ventures, Kleiner Perkins, Lightspeed, and Doerr Capital.

What is Discovery Loop?

Discovery Loop intends to build advanced AI systems that can automate complete experimental loops. Instead of relying on scientists to manually set up and evaluate each test, the platform would use high-throughput algorithms to initiate and iterate experiments in parallel. The goal is to overcome what the founding team describes as a bottleneck created by slow, sequential human iteration.

In a press release, the company said science and engineering have advanced society tremendously over past centuries, but progress has traditionally depended on human iteration, which is slow and sequential. Discovery Loop says its AI systems will leverage massive computational scale to transform the speed and efficiency of innovation. The company's focus includes both experimental automation and the longer-term idea of using AI to improve AI itself, a process often called recursive self-improvement.

The concept of using AI to accelerate scientific discovery is not new. Researchers have been exploring it for years, particularly in fields such as drug discovery, materials science, and genomics. However, the field has remained largely experimental, with limited commercial applications until recently. Discovery Loop is among a new wave of ventures trying to move these techniques from academic prototypes to practical tools.

How the startup plans to work

The company's approach is based on the idea that modern AI systems can handle far more experiments than human researchers can. Traditional laboratory work and engineering discovery often move step by step: form a hypothesis, design an experiment, run it, analyze the results, and then adjust. Discovery Loop wants to compress that cycle by letting algorithms oversee many loops at the same time.

At the core is the concept of an 'experimental loop', in which an AI system proposes a test, performs it or simulates it, observes the outcome, and updates its understanding for the next round. By automating this loop, a single research program could explore enormous spaces of possibilities in a fraction of the time it would take a human team. The company says this will lead to both a higher quantity and a higher quality of experiments.

Dean expanded on this idea in an interview with The New York Times. 'We think there is opportunity for AI to more fully automate what has traditionally been a very human-intensive experimental loop,' he said. 'You will get both a higher quantity and a higher quality of experiments, and that will lead to scientific breakthroughs and advances.'

Funding and investors

Discovery Loop has already attracted significant financial backing. The initial funding round is being co-led by Radical Ventures and Khosla Ventures, two firms with deep ties to both AI and emerging technology. Other participants include Kleiner Perkins, Lightspeed, and Doerr Capital. Notably, Alphabet, Google's parent company, is also providing financial support, even as it loses a long-time executive.

The involvement of Alphabet is significant. It suggests that Google sees strategic value in staying connected to the new venture, even if it could eventually compete in areas that overlap with Google's own AI research. Discovery Loop's broad mission, using AI to speed up scientific discovery, could lead to valuable patents, tools, and discoveries that investors will be watching closely.

In a joint statement, the founding team said: 'The next great frontier for AI is to go beyond answering questions and to begin making discoveries. By fundamentally accelerating how engineering and scientific discovery are conducted, we can deliver the benefits of transformative technologies to the world far sooner.'

Jeff Dean's Google career

Jeff Dean joined Google in 1999 as the company's 30th employee. Over the following decades, he became one of the most consequential engineers in Silicon Valley. His early work helped build the core infrastructure for Google Search, including the crawling and indexing system and the query-serving system that processed billions of searches. Those systems became the backbone of Google's rise from a small startup to one of the world's largest technology companies.

Dean also played a leading role in Google's early AI research. He contributed to the development of systems that would eventually become part of Google Gemini, the company's family of multimodal models. His work helped bridge the gap between traditional software systems and modern machine learning, and he was known for his ability to lead ambitious research teams while remaining deeply involved in technical details.

In recent years, Dean held the role of chief scientist at Google DeepMind, following the merger of Google Brain and DeepMind. His departure marks a major shift for Google, which has relied on Dean as both a technical leader and a public face for its AI research. He is widely respected across the industry, and his decision to start a new company is likely to draw attention from researchers, investors, and potential partners.

The broader trend of AI researchers leaving Big Tech

Dean's departure is part of a broader pattern. Over the past few years, a number of prominent AI researchers have left large technology companies to launch their own startups. The combination of abundant venture capital and a wave of advances in generative AI has created a favorable environment for new ventures. Many founders believe they can move faster outside the corporate research environment and build products that have a more direct impact.

This trend has forced companies like Google to think carefully about how to retain talent. Google has responded by creating new divisions, adjusting compensation, and investing in research initiatives. But the lure of founding a company remains strong. For top researchers, the chance to set their own priorities and build a team from scratch is a powerful draw.

Discovery Loop is also notable because of its focus on the public benefit corporation structure. A PBC is a for-profit company that is legally required to consider the impact of its decisions on society, not just shareholders. That structure may appeal to founders and researchers who want to pursue ambitious goals while maintaining a commitment to positive social outcomes.

The startup's roadmap is still taking shape, but the founding team has been clear about its ambition. The group wants to push AI beyond the role of a digital assistant or content generator and into the realm of active scientific collaborator. By combining massive compute, advanced algorithms, and automated experimentation, Discovery Loop hopes to usher in an era where breakthroughs happen at machine speed rather than human speed.

For Google, the loss of Dean is both a symbolic and practical change. He has been a fixture at the company for more than a quarter of a century, and his influence can be found in many of the technologies that define the modern web. For the broader AI community, the launch of Discovery Loop could become a test case for whether AI can truly transform the pace of scientific discovery. The funding, the team, and the timing all position the startup to be a major player in the next phase of the AI industry.

Source:TechCrunch News