Home > Article > Technology peripherals > When generative AI meets healthcare
Although it contains huge energy, we still need to carefully analyze the best application scenarios to find an ideal platform for it. This is especially true in healthcare – a field known for being slow to change, where any hasty deployment of emerging technologies can pose huge risks. You may still remember that IBM Watson, which attracted much attention in the past few years, once claimed to be able to diagnose complex cancers, but this is not actually the case. Ultimately, Big Blue sold it for a low price last year.
So in terms of healthcare, we might as well use a simple five-step method to evaluate the contributions that generative AI can make:
1. Start with the problems that technology can help solve Start with the problem and figure out what generative AI is good at.
2. Search the overall area where these problems exist.
3. Understand the motivations and barriers to using generative AI in core use cases, including what people need to abandon old approaches before embracing new ones.
4. Make priority assessment based on business dynamics.
5. Broadly understand the elements necessary to build a complete solution, including technology, workflow consulting, patient education, etc.
To apply this approach to healthcare, we first need to be clear about what we are evaluating—not deep learning that can interpret medical imaging or population health data sets. These efforts are already underway. In addition, we are not considering simple applications of AI in specific scenarios, such as diagnostic appointments. There is only one focus here, generative AI and emerging healthcare services.
First of all, what problems can generative AI help solve? There are many answers, but to keep it simple, we can focus on four of them: 1) Interpret unstructured data; 2) Interpret data in a coherent way; 3) Engage people in conversation; 4) Generate new ideas.
Second, what overall areas do these questions correspond to? Starting from the above four points, we can correspond to the following examples:
1) Interpret unstructured data: summarize the key facts expressed in the diagnostic instructions in the electronic medical record, require the medical insurance company to provide pre-authorization, and provide pre-authorization in the clinical Extract patterns from trial data, such as finding commonalities among patient-reported outcomes or treatment failure.
2) Interpret data in a coherent manner: Provide customer service, diagnose and develop treatment plans for health insurance companies.
3) Involve people in the conversation: Capture screening data (e.g., does the patient feel safe at home?) and offer talk therapy for less sensitive health issues.
4) Generate new ideas: Use proteomics and genomics data sets to discover new active ingredients and some new therapeutic effects on existing therapies.
Third, what are the motivations and obstacles for adopting new technologies. This issue is likely to directly determine whether certain use cases can actually be implemented. For example, until generative AI is approved by the FDA as a medical device, no company may use AI to provide a clear diagnosis or treatment plan for U.S. patients. However, the market outlook may change in the future. Considering that many clinicians are already overwhelmed by patient demand, perhaps appropriately relaxing regulatory requirements is the way forward for sustainable development. In addition, this part of the analysis can also help identify areas suitable for rapid innovation (areas with low dependencies, high demand, and low risk/switching costs). For example, talk therapy that was originally self-funded is now expected to be done by AI.
Fourth, determine the priority of implementation in different markets based on business dynamics. This issue is too complex to be discussed in depth in this article. But in general, we can make judgments based on factors such as individual/scale economies, market channels, sales processes, and competition intensity.
Finally, take a broad look at the complete solution. Few new technologies have the potential to revolutionize long-term work practices like generative AI. For example, widespread adoption of generative AI may require customer training and the establishment of an ecosystem of complementary products. In addition, generative AI can also help differentiate products in the market when competitors imitate certain underlying medical technologies.
In short, if you are in the healthcare or life sciences industry, you might as well try various ways to explore the value of generative AI. And instead of starting with technology alone, we might as well focus on the overall challenges brought about by this and think macroscopically about what solutions we really need. Then we will study the implementation methods based on this to see if, in addition to generative AI, there are any mature solutions with lower thresholds that can also bring similar effects.
In the field of healthcare, enterprise-level generative AI is facing a vast blue ocean. The five-step methodology mentioned earlier is enough to show the rich opportunities contained therein. So even in a traditionally conservative industry like healthcare, disruptive changes will occur quickly.
The above is the detailed content of When generative AI meets healthcare. For more information, please follow other related articles on the PHP Chinese website!