5 ms·
I suppose we both underestimate how far we've progressed in simulating physics. Either we're utterly, totally wrong in anything related to astrophysics or we ha
by firebaze 5y ago
I suppose we both underestimate how far we've progressed in simulating physics. Either we're utterly, totally wrong in anything related to astrophysics or we have a basic grasp on fundamental stuff like aerodynamics.
If the latter, we don't need to spend billions on R&D for more efficient designs. We just devote the same comparatively few bucks spent on GPT3 and other ML models to inventing a new, more efficient airplane design.
If we did, maybe an improvement like the one in this linked article is bonkers. If so, we could invest in other assets.
- caconym_ 5y agoI think the issue is less about coming up with shapes that fly well in the sim (or even in the wind tunnel), and more about turning those shapes into practical aircraft. Computers can’t really help us with that (yet) except at a really granular level, with human supervision. As an example, take the longitudinal fuel bladder in this design. I don’t know how you design a version of that that isn’t subject to very scary longitudinal sloshing and doesn’t have unprecedentedly complex plumbing that will be very, very, very expensive and risky to validate for normal flight operations, let alone emergencies. And that’s, like, the most obvious problem with it, and I’m not even counting all the normal, mundane airplane problems the incumbents have already spent millions of man-hours solving.