Even after dissolving Google Health, an attempt at creating a health record repository that directly connects doctors, hospitals and pharmacies, Google has continued working on healthcare technology solutions. With one AI tool already in place that reads, stores, and labels medical imaging and another established tool that helps insurers speed up prior authorization, Google has now turned to developing a healthcare-specific large language model.

The large language model, made viral by OpenAI and its chatbot ChatGPT, is known for summarizing and organizing data. When applied to healthcare, one use case would involve the language model solving a simple, yet time-consuming process: patient handoff at shift change. The AI would ingest patient data from the last 12 hours (medications, events, lab results, etc.), summarize it, and include suggestions for what the oncoming nurse should be considering for the upcoming shift. Nashville-based HCA Healthcare’s UCF Lake Nona hospital is already piloting the technology. The latest version is being dubbed “Med-PaLM 2,” and Google plans to release the technology to more customers this fall.

Of course, there are obstacles. Large language models are not constantly learning, and it’s not realistic to expect them to have a comprehensive knowledge of medicine. Consent, data privacy, and security are also concerns, as hospitals would be asking AI to analyze millions of medical records. Tackling these barriers, however, is an ongoing effort from many tech giants like Google that see clear benefits to having AI as an addition to care teams.

To read more, click here.