Former

Several of Google's most influential AI researchers are leaving the company to launch Discovery Loop, a startup that will build AI systems designed to accelerate scientific discovery. Led by former Google chief scientist Jeff Dean, the company's work could have implications far beyond Silicon Valley, particularly for drug discovery, where AI is already reshaping how researchers identify targets and design molecules. The founding team includes Sanjay Ghemawat, Oriol Vinyals and Quoc Le, all former Google researchers. The company has not released technical details, but its focus reflects growing confidence that foundation models can change how scientists identify new drug targets, design molecules and interpret complex biological data. Discovery Loop enters a competitive market where technology companies and pharmaceutical firms are investing heavily in AI to shorten drug development timelines and improve research productivity.

Why this matters for drug discovery

AI is already embedded across many stages of drug discovery, from predicting protein structures and identifying disease targets to generating novel compounds and prioritising candidates for laboratory testing. Many researchers believe today's tools are fragmented, with models often built for individual tasks rather than supporting the entire scientific discovery process. Discovery Loop is pursuing a broader idea: AI capable of reasoning across disciplines and assisting scientists throughout the research cycle. If successful, these systems could help researchers generate hypotheses, interpret experimental results and identify new directions more quickly than conventional approaches. This follows the direction of the pharmaceutical industry toward AI platforms that act as research collaborators rather than standalone prediction tools.

An evolving field

The launch comes as investment in AI-driven drug discovery continues to grow. Pharmaceutical companies have formed new partnerships with AI developers, while specialist biotech firms are using large language models and multimodal AI to integrate genomic, clinical and chemical datasets. Former Google researchers have played a key role in many of the technologies underpinning this shift, including machine learning architectures and protein structure prediction. Their decision to establish an independent company suggests some researchers see greater opportunities outside large technology firms to develop AI tailored for scientific applications. For scientists tracking these developments, the AI for Science & Research topic page covers how AI is being applied across research disciplines.

What researchers should watch

Discovery Loop has yet to announce its first products or partnerships, making it too early to judge how its technology will compare with existing AI platforms used in life sciences. The company's success will ultimately depend on whether it can produce tools that improve experimental research rather than simply automate existing workflows. Researchers will also be watching how closely the company works with pharmaceutical organisations, academic institutions and biomedical datasets. Access to high-quality scientific data is still one of the biggest barriers to developing AI systems capable of delivering meaningful advances in drug discovery.

Why this matters for research scientists

If Discovery Loop succeeds in building AI that can reason across biology, chemistry and clinical science, it could help create more autonomous research systems that support scientists from target identification through to therapeutic development. That would shift the daily work of researchers who currently spend significant time on data preparation, literature review and hypothesis testing. For scientists who want to work effectively with AI, understanding model capabilities and limitations is becoming a core skill. Researchers who need to build these skills can work through the AI Learning Path for Research Scientists, which is designed for scientists applying AI in their own work.