Human beings are in cycles.
The term has become shorthand for the industry, making people nervous about artificial intelligence's growing role in health care. There are always people paying attention to what artificial intelligence is doing. We are there to make sure nothing goes wrong. But the thing is, sometimes we can get in the way. So now, humanity faces an interesting dilemma. To unleash the power of AI, sometimes we need to make room. But it can also be dangerous – what if something goes wrong?
Despite this Catch-22 atmosphere, healthcare companies like Merck are still taking risks, employing generative AI and agent-like AI tools (where software has the agency/power to take action, rather than simply displaying information about human action) to combat disease. In a recent interview, Asheesh Chhabra, vice president and chief data officer at Merck, cast a new light on the role of humans.
“Let's take humans out of the unnecessary middle but put them in the right rooms,” Chhabra declared. “The human never leaves. How do we put him in the right room?”
He shared an example of what is unnecessary in-between: publishing clinical trial results based on all the data collected. The process can take up to six months. But now with generative AI capabilities, the output takes days, and Merck employees with the appropriate expertise can simply oversee the final output and approve it.
Of course, clinical trial reform through artificial intelligence is something that all pharmaceutical companies are exploring, as is artificial intelligence drug and target discovery. Chhabra mentioned how TEDDY—a large language model developed by Merck and external collaborators—is being used in the field of biologics macromolecules, with a focus on oncology. There is another type of artificial intelligence language model that can also be used for small molecules. He added that Merck has released data for both models.
“So we're not really limiting them to one therapeutic area,” he said. “But we'll start with identification of a set of targets that our subject matter experts are working on and prove it…”
In the area of agent AI, Merck is looking at several different areas where they can be deployed. One involves communicating with health care providers around drug market access. This will be on the business side of the business. Another area is improving productivity and efficiency in engineering and manufacturing. Finally, agent AI can be used for digital twins. Digital twin technology enables pharmaceutical companies to create digital avatars of real humans, which may one day reduce the need to recruit real humans into clinical trials in the distant future.
Enrollment is an issue, but keeping people in clinical trials long-term is also challenging. This is where Chhabra sees the opportunity in digital twins.
“I think we see a huge opportunity for digital twins to sustain what I call the behavioral aspects of patients,” Chhabra said.
In other words, after initial enrollment, and even if people drop out during the clinical trial, enough information can be collected that the digital twin technology will be able to predict how the “person” will behave after completing the clinical trial.
The success of AI or humans in this new, more complex role, for that matter, will depend heavily on the strength of the underlying data.
“Data is really the fuel for AI,” Chhabra said. “I can't create a simple UI mockup using ChatGPT. I now need access to powerful data.”
