The debate about the use of artificial intelligence and platforms like ChatGPT across sectors is heating up as enthusiasts tout recent breakthroughs, while critics point out AI’s potential pitfalls and ways the technology could be misused.
And while AI is already being deployed in some areas of the healthcare sector, there are major hurdles facing the industry as it seeks to bring the latest technological advances into the field.
Geeta Nayyar, former chief medical officer at Salesforce, likened these challenges to those facing self-driving cars — not everyone is raising their hand to get into a car driven entirely by AI.
“Now, the idea of a self-driving car that is a complement to someone actually sitting in the driver’s seat, that becomes interesting. That’s what we need in health care,” Nayyar told Yahoo Finance in an interview.
“Everyone is fixated on this, ‘We won’t need the doctor anymore…AI is here to save the day.’ That’s not so true,” she added.
The topic was a focal point at a major healthcare conference this month, where executives cautioned that there is still more to understand about machine learning and its impact on the sector.
One of the biggest uptake problems facing the healthcare sector is one unique to the industry — most healthcare data is largely protected and hidden. And this privacy complicates efforts to build a data pool robust enough to deploy across the sector.
The US government has already lodged its concerns about AI’s role in healthcare and determined data quality was a key area that needed improvement.
“Developing or expanding access to high-quality datasets could help facilitate training and testing ML (machine learning) technologies across diverse and representative conditions,” according to the report from the Government Accountability Office. “This could improve the technologies’ performance and generalizability, help developers understand their performance and areas for improvement, and help to build trust and adoption in these technologies.”
Currently, more than 97% of health data being put into systems does not get used — and the volume of that data has grown in just ten years from 15% of all health information being digital to almost 98% today, GE Healthcare (GE) chief medical officer Taha Kass-Hout told Yahoo Finance.
But this data isn’t yet reliable, in Kass-Hout’s view, because it lacks structure.”[A]nd without that you can’t query and analyze it efficiently or make any decisions,” Kass-Hout said.
‘Much more efficient’ tools
Kass-Hout is still optimistic about the use of AI in health, noting that it is already being used in radiology, for example, as well as with creating new therapies.
But he said that still having a human involved in these processes is key to AI’s success. Nayyar agreed, saying AI could also be used as a tool to help patients stick to their medications. She explained, drawing again on the self-driving car example.
“It doesn’t necessarily have to drive the car for you, but can it get you in the right car and tell you to wear your seatbelt,” Nayyar said. “There are many different things that AI can do to help complement that patient journey,”
AI also has potential to help the drug industry be more efficient in its research and development. Moderna (MRNA), for example, used AWS (AMZN) to help identify which vaccine formula to pursue in its efforts to protect against COVID-19.
“Moderna was able to reach (development) milestones so quickly due to the programmable nature of mRNA,” said Dave Johnson, vice president of informatics, data, science and AI at Moderna.
And the biotech giant just announced a new partnership with IBM (IBM) to further explore the use of AI in medicine.
Bill Gates also recently touted AI’s capacity not only to help drug companies, but the potential this technology holds for public health in developing countries.
“There is already software that can look at this data, infer what the pathways are, search for targets on pathogens, and design drugs accordingly.” the Microsoft (MSFT) founder said in a recent letter. Gates said the next generation of AI tools would be much more efficient, “and they’ll be able to predict side effects and figure out dosing levels.”
Some experts are hoping AI can help with paperwork as well, a significant burden for clinicians in the era of electronic medical records and claims, as well as on the insurer side in deciding about coverage of claims.
Bad data can ‘jeopardize progress’
Experts are also concerned about AI’s potential to produce results based on skewed data the machines learn from, a known risk with machine learning models in healthcare.
Addressing these problems would also help ease concerns of clinicians, the GAO’s report noted. And a familiar pathway — robust clinical trials — could be a solution.
“While there is excitement and demonstrated benefits to bringing (AI) tools into clinical practice, poor data quality and prevalent biases in health care can jeopardize progress towards achieving health equity and fuel ongoing uncertainties and hesitancies about adopting these tools,” the report stated.
That’s partly why the recent debate has focused on moving ahead too quickly with the technology, as highlighted in a recent letter signed by Tesla (TSLA) CEO Elon Musk and Apple (AAPL) CEO Tim Cook, among others. It doesn’t call for a pause on AI development in general, but “merely a stepping back from the dangerous race to ever-larger unpredictable black-box models with emergent capabilities” for at least six months. “If such a pause cannot be enacted quickly, governments should step in and institute a moratorium,” the letter stated.
But some, like Gates, have said calling for a pause is unrealistic. Instead, there should be responsible building, Kass-Hout said. Which is easier in the health world, thanks to the amount of regulation already in place. That makes it more likely that responsible use of AI will be top of mind, which means knowing how the models are behaving over time, how to measure outcomes, and ensuring that “there is a human in the loop that’s evaluating, correcting, editing,” Kass-Hout added.
And the data have to get to a point of being robust, or multi-modal. That includes storing a person’s medical notes, texts, medical images, records, and lab results all in one place to help with analysis and diagnosis.
All of which, in Kass-Hout’s view, means it will take time for this technology to have a large impact on the health care sector. “These models have never been trained on these kinds of data.”
Follow Anjalee on Twitter @AnjKhem
Click here for the latest stock market news and in-depth analysis, including events that move stocks
Read the latest financial and business news from Yahoo Finance
Add Comment