I was thinking a reinforcement learning algorithm could be a trained in a simulated environment to find a cure for cancer, testing how every which complex process of a new medicine or even nanotechnology might react to its environment. Am I way off?
In: Biology
No. The human body is orders of magnitude more complex to model mathematically than spaceflight dynamics.
Also mathematical models are useless unless you validate them with empirical tests, and unlike autonomous spacecraft, there are ethical issues involved in using real human beings to validate what would need to be a very invasive model.
Rocketry is taking inert objects and pushing them around, mathematics is great at accounting for that.
Simulating the body is much harder because it contains life, things that do what they want and go where they want in enough instances to not make it possibly to fully simulate a body and what reactions it would have any time in the near future.
Math is great at telling you how things happen *after* choices are made….not great at telling you what will happen *before* choices are made. The more choices you add, the worse and harder it gets.
> can biologists simulate the human body
Forget about simulating the human body, we can’t even simulate MOLECULES properly, not even cells, tissues, organs, or even human. By properly, I mean, in a cost effective manner.
That’s why the current research is less about simulation, but trying to figure shortcuts with AI, so we can figure things out without doing the complete simulation:
https://deepmind.com/blog/article/alphafold-a-solution-to-a-50-year-old-grand-challenge-in-biology
Physics is consistent and predictable. If you launch a rocket *that way* and cut the engine, then it just keeps going *that way*. That’s super easy to model.
Biology is a hot mess. Yes, we know “how things work”…but if a thousand people bump their legs on the corner of the exact same table, while moving at the exact same pace, they’ll still all develop slightly different bruises. That’s *really hard* to model.
That would not be practical, or even possible, the human body is probably several trillions of times more mathematically complex than space flight. And I’m probably still lowballing it by several orders of magnitude.
Additionally, the human body isn’t well understood. We could, by hand, calculate the position of Jupiter at any point in the future for millions of years because we understand orbital mechanics well enough to do that. But we don’t understand the human body well enough for that even if we did have the computing power to model it.
The simple answer is because it’s not that simple.
We could simulate nearly every organ’s function. What we couldn’t do is simulate the genetic differences between people and how the introduction of a new drug will react with each person’s physio-chemistry. Even if we made the perfect rendition of such a person, that simply doesn’t account for the variety amongst any two given people.
For example, if your cholesterol is high because you’re genetically predisposed to it, we’d have to create a “dummy” you to test out a drug on. What about medication you take to control your cholesterol? What about your diet? And of course, it’s not simply a matter of your genetics but how you’ve treated your body. What other illnesses have you had and how will *their* legacies react with a new drug for an entirely different problem.
Basically, we’d have to create a dummy version to cover, at the very least, the 10,000 most common physiological states for humans.
Yes you could. Almost certainly in a couple of hundred years almost all drug development will be done virtually in the way you describe.
The reason we can’t do it now is that we still don’t know a lot about how human biology works. Oh sure we know that the heart goes pump, pump, pump. But we have only just unlocked a computational method for figuring out how proteins will fold into 3 dimensions (and I’ll betcha that work is a little wrong some times and there will need to be fixes made to it over the coming decades).
There is a ton about how the brain works, our genes, and a ton of chemicals and proteins in our bodies that we don’t really know much about and it will be the work of decades or centuries to figure it all out.
Once that is done we will be left with a mammoth computational problem but one that computers of the distant future will probably be powerful enough to figure out.
But let me mess with your head here. Lets say you need to figure out if an anti-depressant is going to work. You start up your perfect computer simulation and you add the drug and observe how the perfectly simulated human brain in the computer works. If your simulation really is perfect, is the brain inside the computer alive and self-aware? If not then doesn’t it mean the simulation is missing something? If the simulation is missing something doesn’t that mean there is some potential for its results to be wrong?
If that brain is alive, is it ethical to create digital people to experiment upon? How is it any different than experimenting on living non-digital people?
As other posters say, biological creatures are extremely complicated (and vary widely between individuals), while a rocket in space is one of the easier to analyse physical models.
However, now computers are incredibly fast and sophisticated, ‘computational biology’ is becoming a thing. We are able to do useful work with simulations, but there are some limitations, so we still need to do lots of testing, but this is all changing very quickly.
Modeling and simulating spaceflight is relatively simple and straightforward because it is based on physical principles of movement that are well understood and have been well understood for some time. To the extent that generally the only thing that throws off our models is when a new, unaccounted-for variable (such as a large asteroid that hadn’t been previously detected crosses an orbital path and throws things off with its gravity) enters the fray. Then you can adjust for that new variable and you’re back in business.
With biology, an individual’s genetic makeup and individual body chemistry play a part in nearly everything, from drug and nutrient reactions, to organ function, everything. Infinitely more difficult to model.
Yes, but that level of technology/simulation wouldn’t be available until…. Maybe 300-1000 years in the future
Latest Answers