I've always thought doing large-scale biological or economic simulations would be an incredibly powerful tool if it can be done well enough. Imagine being able to run clinical trials for new drugs in a simulation in 1 day instead of 1 year. Of course this is currently science fiction, but I'm wondering if there are any companies or startups working on this kind of problem?
My current company has toyed with running cybersecurity breach propagation simulations, but I'm not sure if we've productized it.<p>I personally think there should be a "human genome" sized project to create an accurate physics engine and a chemistry engine (almost an operating system) on top of that to unlock material science and the pharma subset of that.<p>I hope at some point there are more companies following the path of DeepMind (Go, Starcraft) and OpenAI (Dota 2), but also developing the platform for others to iterate upon (like OpenAI's Gym). Perhaps even converting it into a revenue-sharing model where game enthusiasts can donate some of their idle GPU time to helping run the simulation would defer some of the setup costs.
I have a former coworker that went to System1 Bio[1] and he described the approach there as something like "using machine learning to reduce the search space of likely pharma treatments". It sounds tangentially related to what you are asking about.<p>[1] <a href="https://system1.bio/approach/" rel="nofollow">https://system1.bio/approach/</a>