MD/PhD student at University of Pittsburgh, interested in computational approaches to modeling cancer and identifying optimal therapies
tyleryasaka
Karma: 18
Agreed, this part was oversimplified. Sicker, more complex patients will have multiple interacting conditions, and choosing the right course of action is often not trivial (this is why we have medical school and residency). I think if you talk to physicians in different specialties, they will also argue that their disease of interest is as complicated as cancer. Also, if you were to look at UpToDate (the clinical guidelines used by physicians), you would be hard pressed to find any diagnosis with a single treatment for all patients. Nothing in medicine is simple.
I’m a graduate student pursuing a career working on precisely the problem you’re describing here: using computation to tackle the tremendous complexity of cancer. Overall I think your post is spot on. I think I mostly would just challenge you on your conclusion:
but all of them at once and more, fused into a single representation, and presented on a platter to an impossibly large statistical model for it to gorge itself on.
I create large models and feed them data, so I respect what machine learning can do these days. And yet, I have found success with combining some of these tools with a more transparent, interpretable approach. I really think the biggest breakthroughs will come from figuring out how to model the most important features of cancer, rather than from giant black box neural networks. I also think that clinicians and humanity as a whole will be much better off understanding to some degree how we are able to decode cancer, not just blindly accepting a treatment protocol that is regurgitated by an algorithm. These days, we have very sophisticated protocols to manage treatment for cancer and other diseases, and yet clinicians still learn the meaning of every individual lab value, every reported symptom. This allows physicians to optimize care, and allows patients to participate in informed, shared decision making. I think this should be the gold standard for our computational biomarkers as well.
I’m new to the lesswrong community, but I’ll share my 2 cents. First, my sense of meaning has always come from an appreciation of beauty in the world. This includes a fascination with the natural world, mathematics, music, sunsets, food, etc. (Occasionally even people!) Second, my life became much more interesting (read: meaningful) when I left silicon valley. I think there is something soul crushing about building over-engineered solutions to first-world problems of dubious relevance to anyone. I quit my job working for a Blockchain startup in 2018 and switched careers—now I’m still coding a lot, but I do it in the context of medicine and cancer biology, and my skillset is now much broader than code. I don’t think this is the only way to increase a sense of meaning, but I think it’s about finding ideas that genuinely excite you, and choosing what is interesting over what is easy (my life sucks in a lot of ways, but it is more meaningful now). I don’t know how much this is helpful or answers your question, but this is what I have found so far.
I’m curious whether you think this epistemic disruption could be analogous at all to the Internet, social media, and smartphones? These technologies were greeted with great optimism not unlike the AI hype we are seeing today, and undoubtedly we all agree that these technologies have been beneficial in many ways. And yet we have seen the rise of vaccine hesitancy, loss of trust in our institutions, and adolescent anxiety that has led to a push to ban smartphone use in schools. I genuinely don’t know if I believe that society as a whole is better for it. I think fundamentally, all of these issues come down to the fact that we are developing systems that we didn’t evolve to live with. Our brains are the same brains that were optimized for a hunter-gatherer lifestyle. While superintelligence has theoretical benefits, I completely agree with you that we are not thinking enough about the issue of cognitive and social compatibility with our little ape brains.
Frankly, you can find a lot of claims in the literature (and I believe some of them). But how many of these multimodal systems are currently used in the clinic? That’s the only metric that matters. I’m not even disagreeing with the premise that multimidal systems should be able to improve prognostic power in theory. But I am curious how well these systems work in practice.