VC firm Andressen Horowitz recently published an article on how the latest AI technologies will have a big impact on healthcare. Is it finally time to say good-bye to the trusted fax machine? I read with guarded optimism, given AI’s lack of success in healthcare over the last decade and during the COVID-19 pandemic. While I see great potential, I do not believe AI to have as much impact in a field where human connection and privacy are vital.
Less (cumbersome) software, more productivity
I love new technology. I love it because it makes me more productive. As I was booking holiday travel recently, I realized how easy it has become to do things on my own - on my own time, in my own way. No calling a travel agent to arrange a trip. No taking out a paper map to get to the hotel. I don’t think you’ll find too many people who will complain about the convenience this provides.
We use new software all the time these days. Everytime a new app comes out, we can easily download it and check it out. It integrates seamlessly into our everyday lives; it makes things a bit more efficient.
Unfortunately, healthcare seems to have gotten it backwards - every new healthcare software solution promises more productivity, but instead does the opposite. EHRs require more clicks for physicians. For patients, my personal favorite is when I am asked to fill out insurance information in the EHR before my next appointment, even though I provided it at the previous appointment. So, why bother filling out anything online?
That’s not productivity.
The article states that instead of healthtech companies developing new software, they’ll develop “AI ‘people’ who are cheap, fast, cheerful and empathic.” However, if I’m a patient, I want a human being to help me feel better. The thought of a machine having empathy discounts any reason to see a doctor at all.
A good piece of software is one that will handle the simple things like document capture, reconciliation, etc. I agree with the article that the first revolution will be non-clinical use cases. If AI can help us get past that hurdle, then I think we will have achieved productivity.
Marginal improvements are worth the effort
Having worked in high-tech technical roles over the last decade, I have seen the evolution of more advanced computing capabilities. Working in big data and cloud platforms have shown the ability to create powerful solutions with seemingly limitless compute power. I may even argue that next-gen AI solutions and the infrastructure that powers them will produce 10x the value that will displace outdated technologies.
Hence, replace the humble fax machine.
Yet, I am wary about moving at that pace in healthcare. On the one hand, cloud computing can process billions of data points for genomic sequencing or population health management. The article is correct in that AI will require more compute resources due to algorithm complexity and compute resources required. There will be more insights, more personalization, more productivity.
But there could also be a cloud outage, a malicious entity that wants to conduct a cybersecurity attack on a hospital, a patient record that ends up in the wrong hands.
There’s nothing wrong with marginal improvements. Is it better to misdiagnose 1 patient or 1000 patients? A small error allows time to investigate and course correct.
A world-class AI doctor in your pocket: Do I really want one?
Good technology works seamlessly and increases productivity. Much like today’s travel apps, it would be amazing to have a doctor that knows everything about me in the future. There are some models already out there that attempt this - Livongo by Teladoc, as an example. However, even if it is covered by insurance, I haven’t really used it unless out of necessity.
An AI doctor might sit as one of those unused apps on my phone. Most people can find a satisfactory answer to a medical question via an internet search, coupled with an in-person visit to the doctor’s office. Can an AI doctor change things for the better? Maybe. However, a virtual doctor will never replace a conversation with a real human being.
However, I like the idea of AI enabling earlier diagnosis - this is one use case I believe will see more rapid adoption.
Conclusion
I appreciated this insightful and hopeful article for the future of healthcare. Perhaps with this new wave of AI, there is a renewed commitment by people and processes to enable a better healthcare experience for everyone. Perhaps we can put that fax machine to rest.