The role of generative artificial intelligence in healthcare was front and center at the annual Healthy Longevity Global Innovation Summit, where experts discussed its potential to help people live longer and healthier lives. Microsoft's Corporate Vice President of Research and Incubation, Peter Lee, who delivered the conference's keynote, had a personal connection to the topic of AI and medicine.
Lee said ChatGPT-4 helped his family monitor their elderly father's health remotely, adding that the technology is more than just explaining complex medical issues. He recalled that he and his siblings were trying to find time to talk with doctors and caregivers about the information they were being told. The stress of the situation caused a strain on the siblings' relationship.
“Some of that was the fuel, and the fire became our lack of understanding of these lab tests and reports,” Lee said, giving his father's medical information and encouraging three very important questions: when to ask Chatjipit. They talk to the doctor.
“Being able to give us GPT guidance kept the temperature down and kept the family together,” Lee said.
Regarding longevity and AI, Lee expressed his hope that there are many areas where GPT-4 can contribute to new scientific discoveries, drugs, treatments and diagnostics.
“What I've always been interested in is empowering people themselves, as we live longer and the people who care for us,” Lee said. “For this, I really have high hopes.”
Although the concept of artificial intelligence is almost a century old, Generative AI has developed rapidly in a short period of time. OpenAI's ChatGPT was released to the public in November last year and is already the standard for AI chatbot development. Generative AI refers to AI tools that respond to user queries and can generate text, images and videos based on those queries.
While AI tools like ChatGPT are colloquially referred to as chatbots, Lee emphasized how much better GPT-4 is than its predecessors GPT-3 and GPT-3.5. However, despite the GPT-4's impressive capabilities, Lee emphasized that it is still an unconscious machine, citing its limitations, which can be described as human-like.
“Of course GPT-4 is not human,” he said. “If you think of a computer as a perfect memory, perfect computing machine, then GPT-4 is not a computer either – it's a new tool that we're all trying to figure out how to use.”
Lee also found that AI excels at identifying flaws in medical studies—but when asked to write its own reports, it introduces those same flaws.
“GPT-4 was able to outperform human reviewers in identifying biased, harmful, and non-inclusive language in submitted manuscripts in a way that human reviewers could not,” he said. “Even if GPT-4 itself were asked to write the same paper, it could be guilty of those same biases.”
Lee thanked the conference organizer for taking an active role in leading the development and use of AI in healthcare.
“I'm very happy that the National Academy of Medicine is taking the mantle on this, especially because I understand well enough that this will be a net positive benefit for us,” he said.
Major corporations including Microsoft, Google, Amazon and Meta have invested heavily in artificial intelligence in the past year. Microsoft has deepened its relationship with OpenAI by investing $10 billion in the company and integrating ChatGPT with the Bing search engine and Office suite of tools to create a new digital assistant, dubbed Copilot, that will replace its previous digital assistant, Cortana. Microsoft's investment in OpenAI has brought its AI developer image generator Dall-E 3 to Microsoft's Bing.
The collaboration between Microsoft and OpenAI has not been without controversy. In July, OpenAI disabled access to its “Bing by Bing” feature for its subscription-based ChatGPT Plus service after discovering that users were using it to get around website paywalls and privacy settings. In September, Bing reinstated OpenAI Browse, with the service no longer limited to data before September 2021.
As generative AI becomes more entrenched in everyday life, these tools are used for small-scale tasks like creating large documents and requiring human hours for users to query the document. In July, British startup Twin Health unveiled an AI platform designed to analyze MRI scans and detect diseases at an early stage.
