3 Health Care Takeaways from Google I/O 2025

It was no surprise that artificial intelligence (AI) dominated news coming out of last month’s Google I/O 2025, the tech giant’s annual developer conference.

And as in past conferences, excitement abounded about where AI is headed and how fast some of the grandest visions for the technology’s promise could be realized. Here’s a rundown of what grabbed our attention.

Could AGI Really Be Here in Five Years?

Google's co-founder Sergey Brin and Google DeepMind CEO Demis Hassabis shared slightly different projections about whether artificial general intelligence (AGI) would arrive by 2030 or slightly later.

AGI generally has been understood to mean AI that matches or surpasses most human capabilities, which could have potentially huge implications for health care and other fields.

The trouble is, AGI remains somewhat of a unicorn. Plenty of tech gurus and AI developers can describe AGI, but no consensus exists on what it would look like or how it could change lives once it gets here, notes a recent Axios report.

Hassabis predicts developers likely will need a couple more big breakthroughs to get to AGI. One of those breakthroughs already may have been partially achieved via reasoning approaches that Google, OpenAI and others unveiled recently, he said.

Takeaway

AGI developments bear watching in health care, regardless at what pace they occur. And make no mistake. Experts believe AGI has a future in all fields — whether that’s two, five or 10 years from now. AGI's self-learning can transform health care, improving diagnoses and personalized treatments. However, its integration presents regulatory, ethical and public perception challenges.

MedGemma Could Speed Development of Health AI Apps

The launch of MedGemma, an open model for multimodal medical text and image comprehension, has the potential to accelerate development of new health applications.

It is designed to be a starting point for developers building such applications as analyzing radiology images or summarizing clinical data, and its small size makes it efficient for fine-tuning specific needs. When evaluated on the MedQA benchmark, its baseline performance on clinical knowledge and reasoning tasks is similar to that of larger models, Google states.

MedGemma follows on the heels of Google’s Articulate Medical Intelligence Explorer (AMIE) launch last year. AMIE is a research AI system based on a large language model that is optimized for diagnostic reasoning and conversations. Google says it trained and evaluated AMIE along many dimensions that reflect quality in real-world clinical conversations from the perspective of both clinicians and patients.

To scale AMIE across a multitude of disease conditions, specialties and scenarios, Google developed a self-play-based simulated diagnostic dialog environment with automated feedback mechanisms to enrich and accelerate its learning process.

Takeaway

Speeding development of medical AI programs is important, but experts note that users must remain diligent in testing programs for the possibility of errors, privacy breaches, biases in decision-making and potential replacement of human judgment.

An AI operating system to enhance efficiency

Google outlined how its Search with AI Mode powered by Gemini 2.5 (an experimental tool at this point) could help health care organizations improve operational efficiency, improve the patient experience and support better clinical outcomes. A new report shows one health care AI expert’s experiments with Gemini 2.5.

The AI Mode rollout to all U.S. users employs Gemini 2.5’s advanced reasoning, multimodal capabilities and contextual understanding to handle complex, longer queries (two to three times longer than traditional searches) and ask follow-up questions. It provides conversational answers rather than traditional link-based results.

Google’s Deep Search, meanwhile, breaks down queries, performs web exploration, iteratively browses and synthesizes information into structured responses or reports. AI Mode incorporates Deep Search, which analyzes hundreds of sources in real time to generate comprehensive research reports, enhancing its utility for in-depth queries.

Takeaway

Potential use cases include image classification, clinical reports, summarization, triage and medical Q&A. The models are for research, not clinical use, requiring developers to validate and adapt them before deployment. Google states that health care users can use Deep Search to assist physicians in summarizing a patient’s medical history across multiple providers, analyzing lab results and other tests, while generating a comprehensive overview to support decision-making.