Doctor says AI could change health care for the better

When folks hear the term artificial intelligence (AI), they get a bit anxious because they remember the movie that depicted robots taking over the planet. There has been concern for human well-being for some time. Now, it is making its way into the health care world, but an expert on the subject says that may be a good thing.

University of New Mexico's Health Sciences Dr. David L Perkins (M.D., Ph.D., professor of medicine, director of Center for Personalized Health and co-director of M.D./Ph.D. program) is spearheading AI use in the UNM hospitals, eventually including Sandoval Regional Medical Center, and has been working with local politicians on regulating it.

"It has potential to change everything from our education to how we manage clinic appointments to how we manage patient-doctor interaction," Perkins said.

An example Perkins says people would see the most impact would be in patient-doctor interaction.

"If you have been to the doctor lately, you may have seen your doctor sitting there typing and sort of ignoring you, but not intentionally, and they have to get that data down or doctors are going home at night writing notes, which can lead to burnout. So, AI has the potential with voice recognition to record the whole interaction, to write the note, to recommend action, etc. Also, with the chatbots and stuff, they can answer simple queries from the patient afterward," he explained.

The point of this, he says, is to free up the interaction between the doctor and patient so the human aspect of the appointment returns. The doctor can focus on the patient rather than dividing attention so they can get data down.

"So, I can actually look at you, talk to you, right? I think it's a win-win for the patient and the doctor," he added.

This is not the first time Perkins has worked on introducing AI to the health care world. At the University Of Chicago, he and other faculty members developed training for medical students that would expose them to AI.

"The important thing here though is AI is still new. I would tell the students that this is five to 10 years out. However, I was wrong there because Chat GPT exploded and only took five to 10 months," he added.

The current form of ChatGPT was introduced in November 2022. The dialogue format makes it possible for ChatGPT to answer follow-up questions, admit its mistakes, challenge incorrect premises and reject inappropriate requests. ChatGPT is a sibling model to InstructGPT, which is trained to follow an instruction in a prompt and provide a detailed response.

Typically, it is used for correcting computer code but in a way that individuals can ask for help, like an IT service without a human.

The problem with AI is ensuring true responses. In other words, ChatGPT is capable of answering questions but is not able to understand truth from lies. But the AI is trained to decline questions it doesn't understand. The creators of ChatGPT also say the program often overuses some terms and can repeat itself.

"Ideally, the model would ask clarifying questions when the user provided an ambiguous query. Instead, our current models usually guess what the user intended," the website reads.

It also reads that while they made efforts to make the model refuse inappropriate requests, it will sometimes respond to harmful instructions or exhibit biased behavior.

"We’re using the Moderation API to warn or block certain types of unsafe content, but we expect it to have some false negatives and positives for now. We’re eager to collect user feedback to aid our ongoing work to improve this system," it reads.

Perkins says while AI would be useful to doctors and patients, it comes with its faults. Like humans, AI is capable of prejudice and bias because it is created by humans with those same prejudices and biases.

"For example, AI was used to predict the 2020 election winner. The data they used showed the AI all the previous presidents. Then they showed the AI photos of (Donald) Trump and Hilary (Clinton) next to each other," he explained.

Because almost all previous presidents have been white and male, the AI assumed the next one would be too, he added.

He went on to say that because AI is still in its new stages, fear of the unknown is common.

"But whenever something new comes along, people are always skeptical. Think back to the industrial revolution," Perkins said.

Part of the issue is regulating AI. Perkins has discussed with many New Mexico representatives, including state Rep. Alan Martinez, how to go about it.

"I think the work Dr. Perkins is doing here is very important," Martinez said.

While he is excited at some of the prospects AI could have in health care, he is also scared.

"But I think any time there is a new thing we are trying it is a little bit scary. AI is just another tool and with the right safeguards, it could be great," he added.

Martinez says he is interested to see how AI can help with preventative care, especially in detecting cancer in its early stages.

According to the National Health Institute, AI had 97% accuracy in predicting two types of lung cancer last year in a research study.

Martinez's concerns are with the funding of it and how to properly regulate it.

"If we do it right, the state can make a real difference," he said.

AI's young start is making it difficult to properly regulate it. There are still some loose ends to tie up, according to Perkins.

Right now, Perkins and other doctors just want to spread awareness about it. So far, feedback is mixed.

"There are a few of those who are skeptical, and there are also some doctors that have that same concern that AI is going to exterminate us or something horrible," he said.

On the flip side of that, Perkins says people have positively attached to AI robot use as well.

"Some universities now use robots that deliver food to the dorms, they do everything. They cross the street, take the food and sometimes when they come back, they have love notes on stuff on them. They have these robots with AI in them that essentially live with the elderly. They're living alone at home but need assistance. So, they do everything from remind them to take their medicines and can have basic conversations with them. A lot of these people get very, very attached to robots," he added.

He also said a chatbot in California got positive feedback with patients.

"They said it was more compassionate than the doctor at times," he said.

Perkins' main worry is that people, namely medical students, won't be educated in AI before it is implemented in the health care setting. The first generation of students to learn AI at UNM will get a chance to do so in a pilot course offered this summer.

"Because right now, there's no AI education in medical school. But we also will be teaching faculty because some of this is happening so fast that the faculty aren't ready to teach it," he added.

He also says people needn't worry about their jobs because "there will always be another job."

"It's a matter of adapting and learning what AI is about," he said.

As far as concern about imminent danger from AI, Perkins is not worried: "I am more worried about nuclear weapons and global warming than AI attacking us."