Google sees potential in leveraging generative AI models to assist healthcare workers or potentially delegate more healthcare responsibilities to these models.
The company has introduced MedLM, a suite of models specifically tailored for the medical field. Building upon Med-PaLM 2, a Google-developed model exhibiting high expertise across various medical exam questions, MedLM is accessible to select Google Cloud customers in the U.S. (with a preview available in specific markets) who have been whitelisted through Vertex AI, Google’s fully managed AI development platform.
Presently, there are two MedLM models offered: a larger model intended for handling complex tasks and a smaller, more adaptable model suited for a broader range of tasks.
“As we pilot our tools, we aim to integrate generative AI models into healthcare processes, potentially enhancing efficiency and aiding medical professionals in their responsibilities.”
“In a preview of today’s announcement, Yossi Matias, VP of engineering and research at Google, shared insights in a blog post with TechCrunch, stating, ‘Through our tool piloting across different organizations, we’ve discovered that the most efficient model for a specific task varies based on the use case.’ Matias elaborated, ‘For instance, summarizing conversations might be optimally managed by one model, while navigating through medications might be more effectively handled by another.’
According to Google, one of the early adopters of MedLM is HCA Healthcare, a for-profit facility operator. They’ve been conducting trials with physicians using these models to assist in drafting patient notes within emergency department hospital sites. Another participant, BenchSci, has integrated MedLM into its ‘evidence engine,’ employing it to identify, classify, and rank new biomarkers.
Matias emphasized the collaborative efforts with healthcare practitioners, researchers, and organizations within the health and life sciences sector. ‘We’re closely partnering with the pioneers of healthcare, working hand in hand with practitioners and researchers in this field,’ he added.
Google, alongside its primary competitors Microsoft and Amazon, is fervently competing to establish a foothold in the healthcare AI market, projected to potentially reach tens of billions of dollars by 2032. Recently, Amazon introduced AWS HealthScribe, utilizing generative AI to transcribe, summarize, and analyze patient-doctor conversation notes. Microsoft, on the other hand, is piloting various AI-powered healthcare products, including medical ‘assistant’ applications powered by extensive language models.”
There are valid reasons to approach such technology with caution, particularly in the healthcare domain where AI has a history of mixed outcomes.
Babylon Health, an AI startup supported by the U.K.’s National Health Service, has faced criticism for claiming superiority of its disease-diagnosing technology over doctors. Similarly, IBM had to divest its AI-focused Watson Health division at a financial loss due to technical issues leading to deteriorating customer relations.
While some argue that newer generative models like Google’s MedLM family are more sophisticated, studies reveal their limitations in accurately addressing healthcare queries, even basic ones. For instance, ophthalmologists co-authored a study testing ChatGPT and Google’s Bard chatbot on eye conditions and found inaccuracies in the majority of responses. Notably, ChatGPT has generated erroneous cancer treatment plans, while both ChatGPT and Bard have provided incorrect, debunked medical information regarding kidney function, lung capacity, and skin issues.
The World Health Organization (WHO) cautioned against the risks associated with using generative AI in healthcare, citing potential generation of harmful responses, spreading health-related misinformation, and possible disclosure of sensitive health data. The WHO expressed concerns about the inconsistent exercise of caution in adopting generative AI, fearing rushed implementation could lead to errors by healthcare workers, harm patients, erode trust in AI, and delay the technology’s potential long-term benefits worldwide.
Google has consistently emphasized its cautious approach in deploying generative AI healthcare tools, aiming for safe and responsible utilization. Yossi Matias reiterated Google’s commitment to advancing healthcare responsibly, ensuring accessibility of benefits for all involved.
Normally I do not read article on blogs however I would like to say that this writeup very forced me to try and do so Your writing style has been amazed me Thanks quite great post.