# ?? Generative AI's ?? Enormous Promise for Health Care ??

Summary and Review of https://www2.deloitte.com/us/en/blog/health-care-blog/2023/generative-ai-holds-enormous-promise-for-health-care.html

**7?? Questions for Dr. John Halamka ??**

**By Brian Doty ?? ??, Deloitte Consulting LLP, and Jay Bhatt, D.O., ?? Health Equity Institute & Deloitte Center for Health Solutions, Deloitte Services, LP**

Generative AI ?? potentially holds a big promise for health care ??, opening the door ?? to a new era of tools ???. But, it's still growing ??, accuracy ?? isn't reliable, and there's a lack of rules ?? or regulations ??.

In April, the Coalition for Health AI (CHAI) ??? shared its Blueprint for Trustworthy AI ?? and is finalizing an AI Maturity Evaluation tool ???. This could help health organizations check their readiness to use AI ??.

Dr. John Halamka ??, who cofounded CHAI, works with people from federal agencies, the White House ???, tech companies, and top academic health systems. The goal is to provide guidelines for health AI tools, ensuring information quality ?? and increasing trust ??.

Dr. Halamka is also president of Mayo Clinic Platform and author of books ?? on tech in health care.

Recently, we talked ??? with Dr. Halamka about generative AI and its potential for health care.

**Here's a bit from our conversation:**

**Brian:** We've been talking a lot about generative AI ?? and its role in health care ??. Can you explain predictive AI versus generative AI?

**Dr. Halamka:** Predictive AI is math ??, not magic ?? ♂?. Generative AI is different - it can make a picture or text ????. Large language models (LLM) can create human-like communication ???. The AI predicts the next word in a sentence based on billions of previous human communications ??.

**Jay:** What do you see as the big challenges for health system executives, and how might generative AI help?

**Dr. Halamka:** Hospital CEOs usually face 3?? challenges: negative margins ??, hiring and keeping staff ?? ???? ??, and staff burnout ??. What if generative AI could help with administrative work, pre-op workflows, and appealing claim denials?

**Brian:** You wrote a blog ?? about a cardiology patient whose diagnosis was made by generative AI. It was compelling but inaccurate. What should happen to improve accuracy and trust?

**Dr. Halamka:** I'm hopeful that good health data could be found and used to retrain LLM models, considering data nature and interoperability and the ability to de-identify information.

**Brian:** Can you explain prompt engineering?

**Dr. Halamka:** Depending on how a question is asked, the AI could provide useful info or a hallucination. That's what prompt engineering is about. [An AI hallucination or confabulation happens when the model fills knowledge gaps with plausible sounding information that might not be accurate.]

**Jay:** You've served in the Obama, Bush, and Biden administrations and teach at Harvard Medical School. What role should government play in regulating AI?

**Dr. Halamka:** The U.S. government should set standards for training data, testing data, and prompt engineering. It should also make things clear for users

. We have suggested a data card ?? that explains how an AI was trained and how it should be used.

**Jay:** What do you think the future of generative AI looks like for health care?

**Dr. Halamka:** Generative AI will not be a cure-all ?? ♂?, but it can be a really important tool ??? if used in the right way. We should be ready to adjust and adapt. We don’t want to find ourselves stuck with a tool that doesn’t do what we need it to do.

---

## ?? Summary

In a recent blog post, Dr. John Halamka discusses the potential of generative AI in healthcare, noting its potential benefits and the challenges it faces. Dr. Halamka, who cofounded the Coalition for Health AI (CHAI), sees promise in generative AI’s potential to alleviate administrative burdens and streamline hospital workflows. However, he also cautions against potential misuse of the technology, such as in the context of insurance claim disputes. Furthermore, he advocates for greater transparency and regulation in AI’s deployment within healthcare.

## ?? Analysis

The blog emphasizes the necessity of a delicate balance between technological innovation and regulation when it comes to the application of generative AI in healthcare. The technology shows great promise in its potential applications, but it is also plagued by a lack of comprehensive regulatory safeguards and risks of inaccuracy.?

One of the key points raised by Dr. Halamka is the need for generative AI models to be trained specifically on healthcare data. Current AI models are trained using diverse data sources, and not specifically on healthcare data, which could potentially be a large untapped resource for improving the accuracy of these models.

## ?? Commentary

As healthcare technology continues to evolve, generative AI could potentially play a major role in transforming the sector. However, the challenges identified in the blog highlight the need for careful and considered deployment of this technology. Ultimately, the goal should be to augment human capabilities and not to replace them. Hence, the deployment of AI in healthcare must be approached with caution and a commitment to its ethical and responsible use.