He’s probably wrong on medicine.
Ray Kurzweil writes:
After working in the field [of AI] for 61 years—longer than anyone else alive—I am gratified to see ai at the heart of global conversation. Yet most commentary misses how large language models like Chatgpt and Gemini fit into an even larger story. ai is about to make the leap from revolutionising just the digital world to transforming the physical world as well. This will bring countless benefits, but three areas have especially profound implications: energy, manufacturing and medicine.
This is from, “Ray Kurzweil on how AI will transform the physical world,” The Economist, June 17, 2024. (gated)
Kurzweil makes his case well.
Another excerpt:
By contrast, AI can rapidly sift through billions of chemistries in simulation, and is already driving innovations in both photovoltaics and batteries. This is poised to accelerate dramatically. In all of history until November 2023, humans had discovered about 20,000 stable inorganic compounds for use across all technologies. Then, Google’s gnome AI discovered far more, increasing that figure overnight to 421,000. Yet this barely scratches the surface of materials-science applications. Once vastly smarter AGI [artificial general intelligence] finds fully optimal materials, photovoltaic megaprojects will become viable and solar energy can be so abundant as to be almost free.
Energy abundance enables another revolution: in manufacturing. The costs of almost all goods— from food and clothing to electronics and cars— come largely from a few common factors such as energy, labour (including cognitive labour like r&d and design) and raw materials. AI is on course to vastly lower all these costs.
Where he falls short is on medicine. It’s not that he doesn’t make a good case that in a relatively unregulated market, AI could easily have huge positive effects on the kinds of drugs that we put in our bodies. It’s that he seems unaware of the immense power that the Food and Drug Administration has over what drugs we will be allowed to have.
He writes:
Much more laboratory research is needed to populate larger simulations accurately, but the roadmap is clear. Next, AI will simulate protein complexes, then organelles, cells, tissues, organs and—eventually—the whole body.
This will ultimately replace today’s clinical trials, which are expensive, risky, slow and statistically underpowered. Even in a phase-3 trial, there’s probably not one single subject who matches you on every relevant factor of genetics, lifestyle, comorbidities, drug interactions and disease variation.
Digital trials will let us tailor medicines to each individual patient. The potential is breathtaking: to cure not just diseases like cancer and Alzheimer’s, but the harmful effects of ageing itself.
This will happen only if the FDA backs off in a substantial way. Let’s hope, but don’t let hope overrule painful learning from experience.
READER COMMENTS
roundtree
Jun 27 2024 at 8:36am
Michael Sandifer
Jun 27 2024 at 9:08am
Yes, the FDA’s approach to the vaccines during the pandemic, while representing an improvement, may have prevented many thousands, if not hundreds of thousands from saving their lives. The new vaccines should have been on the open market for use on a voluntary basis, immediately, with as good of an associated data collection effort as possible.
I had been an advocate of reducing the FDA’s role to merely labeling drugs as having research basis for safety and efficacy or not, but now I think the agency does more harm than good, even in that role. We saw during the pandemic that there were people who trusted the FDA’s standard process too much, and then were skeptical of emergency authorization.
I think AI will help medicine and research publishing in general though, in that publishing via chatbot will become standard. Though somewhat suboptimal in ways now, publishing via chatbot will allow researchers to write informal papers, without even have to necessarily fully reference them, upload them into chatbots. Then the chatbots will produce formal papers/books on an overview, and/or any related topic, subtopic, or combination thereof at the reader’s request, and at levels of explanation the readers can understand.
Dylan
Jun 27 2024 at 9:37am
I think he is wrong on medicine as well, but for different reasons. I’m not a biologist, but I’ve worked with many over the years. The one thing I’ve learned from them is just how incomplete our model of any particular disease or of the whole human body. AI can do a lot, but I think it still chokes (at least LLMs) on that bit. And until we get those fundamental breakthroughs, we’re still going to be going slow, since we still have no idea if a new drug is going to help you or hurt you without trying it out. The promise of personalized medicine is great… It just always seems to be a few years or more away.
Richard W Fulmer
Jun 27 2024 at 10:56am
Could AI-identified medicines swap the FDA’s ability to review new drugs to the point that public demand drives real reform?
steve
Jun 27 2024 at 5:41pm
I would second Dylan. If and when we have perfect understanding of biology and the human body in particular then maybe AI can do what the author claims. Based upon current levels of understanding we cant predict the outcomes for drugs across a population let alone at the individual level. As a very long time, avid reader of sci-fi I would like to think that we eventually get to the level where we we have a tricorder that tells us exactly what is wrong with someone and what molecules we need to treat, then have the ability to manufacture those molecules on the spot, but that is a long time away. That said AI will help with a lot of other things in medicine well before we have tricorders.
Just a reminder about how the FDA really works. Almost everything it approves has to be looked at by their own people and then goes out to people in academia or the private sector for review. If you want that done faster you will need to pay more people so they have time to do the reviews. At present approval is in about 10 months, priority approval 6 months. Truly lifesaving drugs can get faster, conditional approval.
Query- Having been a practicing doc for a long time and chair of a department for over 10 years, people seem to want to skip phase 2 and 3 testing. First, I am not sure drug companies actually want that since they wouldn’t have a leg to stand on when they get sued. Second, how would I as a doc know how to dose the drug, if it was effective and how does it compare to other drugs? Why would I order such a drug not knowing those things?
Steve
Matthias
Jun 30 2024 at 7:29am
Doctors are already allowed to prescribe medication off-label, and that is common enough practice.
For those not in the know: off-label means for a purpose other than what was officially tested and approved.
Eg think of aspirin for heart attack prevention, and not as a mild painkiller. (Though aspirin is perhaps a bad example, because that drug predates modern approval processes anyway.)
So all the issues your mention in your query are already being navigated by doctors today.
Comments are closed.