Google unveiled a groundbreaking addition to its Vertex AI platform at HLTH 2023, introducing powerful generative AI capabilities tailored for health and life sciences enterprises. This innovative feature empowers organizations to delve into patient data, including valuable clinical sources such as FHIR data and clinical notes, all of which can be seamlessly integrated with Google's medical large language model, Med-PaLM 2.
Google Vertex AI, the cornerstone of this development, serves as a versatile search engine that enables businesses to construct their generative AI-driven search engines, delivering superior customer experiences. Meanwhile, this model harnesses the prowess of Google's Language Models (LLMs) to address a myriad of medical inquiries.
This transformative Vertex AI Search feature, when combined with this PaLM 2 model, equips healthcare providers with the ability to uncover precise medical insights from patients' medical records and provides responses to general medical questions. These capabilities seamlessly extend the existing repertoire of Google Vertex AI, renowned for facilitating conversational search applications and comprehensive question-answering capabilities.
Crucially, this feature aligns with the stringent standards of HIPAA compliance and seamlessly integrates with Google Cloud's Healthcare API and Healthcare Data Engine. Furthermore, it collaborates with Google Health's pioneering product, Care Studio, and is now available for trial usage by healthcare and life sciences organizations.
Google's vision for this offering is multifaceted. It seeks to alleviate the mounting workforce shortages within the healthcare sector, reduce administrative burdens on medical professionals, and mitigate physician burnout. Ultimately, it aims to empower customers with the ability to retrieve precise clinical data, thereby enhancing their decision-making capabilities.
Burak Gokturk, VP, and General Manager of Cloud AI and Industry Solutions for Google Cloud, underscored the significance of this development, stating, "Bringing Google-quality, gen AI capabilities across an organization's entire ecosystem, including EHRs, has the potential to dramatically improve efficiencies, provide clinical decision support, and increase the quality of care clinicians can give patients. Making Vertex AI Google Search more useful for healthcare and life science organizations is a priority for us because we know that having the right information and insights at the right time can make all the difference in health."
In a notable achievement, Med-PaLM 2 demonstrated its prowess by achieving an "expert" test-taker level with an accuracy rate exceeding 85% on U.S. Medical Licensing Examination-style questions during tests conducted in March. The model also successfully attained a passing score on the MedMCQA dataset, specially designed to address real-world medical entrance exam questions.
One month following this milestone, Google extended access to Med-PaLM 2 for a select group of Google Cloud customers, enabling them to explore diverse use cases and contribute valuable feedback. This collaborative effort culminated in July with a study conducted by Google researchers, as reported in Nature. The study revealed that Med-PaLM consistently delivered long-form answers that aligned with the scientific consensus on 92.6% of questions, mirroring clinician-generated answers at a remarkable 92.9% accuracy rate.