Google DeepMind is pivoting from building specialized AI tools to developing general-purpose AI agents that can conduct scientific research with minimal human guidance. At Google I/O 2026, CEO Demis Hassabis announced Gemini for Science, a suite of experimental tools powered by large language models (LLMs), which are AI systems trained on vast amounts of text to understand and generate human language. This represents a significant departure from the company's previous strategy of creating single-purpose breakthroughs like AlphaFold, the Nobel Prize-winning protein structure prediction system.

What Is Gemini for Science and Why Does It Matter?

Gemini for Science unites several of Google DeepMind's LLM-based research tools under one brand, including the AI Co-Scientist for hypothesis generation and AlphaEvolve for algorithm optimization. These tools are not yet publicly available, but Google has opened applications for researcher access, signaling broader adoption in the scientific community. Early testers have expressed enthusiasm about the potential of these systems. Gary Peltz, a Stanford geneticist, compared using the AI Co-Scientist to "consulting the oracle of Delphi" in a Nature Medicine article.

The shift reflects a broader industry trend where general-purpose models are beginning to make independent contributions to scientific fields. This week, OpenAI announced that one of its models disproved an important mathematics conjecture, perhaps the most meaningful contribution that generative AI has made to mathematics so far, according to some mathematicians. Importantly, the model used by OpenAI is not specialized for solving mathematical problems; it is a general-purpose reasoning model in the vein of GPT-5.5.

Why Is Google Reassigning Its Top Scientists?

The strategic realignment is evident in concrete personnel changes within the company. John Jumper, the Nobel laureate who led the development of AlphaFold, has reportedly moved from science-specific tool development to working on AI coding projects. This move is linked to Google's efforts to improve its coding tools to compete with rivals such as OpenAI and Anthropic, but it also signals a prioritization of agentic science, as coding abilities are essential for the success of autonomous AI systems.

Pushmeet Kohli, Google Cloud's chief scientist, published a piece in the journal Daedalus stating that the industry is moving toward AI that "begins to do science" rather than just facilitating it, a sentiment echoed by Hassabis who described the current moment as "standing in the foothills of the singularity". This language reflects the company's ambition to develop AI systems that could eventually conduct research independently, without constant human oversight.

How to Understand the Tension Between Two Approaches to AI in Science

• Specialized Tools Approach: Systems like AlphaFold and WeatherNext are designed and trained to solve specific scientific problems. AlphaFold helps researchers understand protein structures, work that historically consumed years of laboratory effort. WeatherNext, Google's weather prediction software, provided an advance alert about Hurricane Melissa's catastrophic landfall in Jamaica and potentially saved lives.
• Agentic LLM-Based Approach: General-purpose AI systems that could one day execute cutting-edge research projects without human involvement. These systems can generate hypotheses, optimize algorithms, and potentially make independent scientific discoveries across multiple domains.
• Resource Allocation Shift: Google is reallocating key personnel and resources away from specialized tool development toward agentic systems, signaling a long-term strategic bet that general-purpose AI will eventually outpace specialized approaches in scientific discovery.

The juxtaposition of Hassabis's lofty rhetoric about the "foothills of the singularity" with the real-world results of WeatherNext highlights this tension. While WeatherNext's hurricane prediction is an enormous and meaningful achievement, it is hardly evidence of an impending singularity. The question facing the scientific community is whether investing in agentic systems makes sense when specialized tools like AlphaFold remain extremely popular among researchers. Last year, Google reported that protein structure predictions from AlphaFold have been used by over three million researchers worldwide.

Is Google Abandoning Specialized AI Tools Entirely?

Google is not abandoning its work on specialized AI tools, at least not yet. AlphaGenome and AlphaEarth Foundations, trained for genetics and Earth science respectively, were released last summer, and the newest version of WeatherNext came out in November. Isomorphic Labs, a Google subsidiary that utilizes AlphaFold technologies for drug development, recently raised a $2 billion Series B funding round.

However, the reallocation of resources and the emphasis on general-purpose models suggest a long-term strategy where agentic systems act as accelerants for human scientists, potentially evolving into collaborators in the coming decade. Hassabis has been careful to position this new set of scientific agents as an accelerant for human scientists, rather than a replacement for them. "For the next decade or so, we should think about AI as this amazing tool to help scientists," Hassabis said in an interview published in the Daedalus issue. "Beyond that timeframe, it is hard to say with any certainty, but perhaps these systems will become more like collaborators".

"For the next decade or so, we should think about AI as this amazing tool to help scientists. Beyond that timeframe, it is hard to say with any certainty, but perhaps these systems will become more like collaborators," said Demis Hassabis.

Demis Hassabis, CEO of Google DeepMind

The challenge, however, is that no one can be an effective scientific collaborator without also being a skilled scientist in their own right. If Hassabis is anywhere near the mark when he talks about the "foothills of the singularity," then AI scientists could eventually exceed the capabilities of their human counterparts. Hassabis has spoken of how he was initially inspired to pursue AI when he observed how progress in physics had stagnated since the 1970s; he wondered whether the human mind had reached its limits in that domain, and if AI could help to overcome that barrier.

What makes this pivot significant is its timing. Though it has only been five years since AlphaFold solved the protein-folding problem, both the technology and the discourse have quickly moved beyond that once-revolutionary achievement. Google seems to be aiming itself toward a future where agentic AI systems become the primary drivers of scientific progress, rather than tools that augment human researchers. Whether that vision becomes reality depends on whether these general-purpose systems can deliver the kind of independent, verifiable scientific breakthroughs that specialized tools like AlphaFold have already achieved.