Scott Locklin on Quantum Computing: More Support for my Cocktail Party Theory of Science


Scott Locklin has a new post up on quantum computing, Quantum Computing as a Field is Obvious Bullshit. I have to agree, quantum computing [and AI] is largely bullshit, and we should probably nuke and pave the whole field. On a side note, I wonder whether this post functions as an extended reply to this comment thread on Greg Cochran’s West Hunter blog.

Which reminds me, I really should update my cocktail party theory of why science doesn’t work anymore. Locklin makes a much longer and more detailed argument than I did that a major problem for science in the twenty-first century is that too many scientists never actually build anything with their own hands to see whether their ideas work. In the case of quantum computing, the problem is as simple [and as hard!] as aligning all of the optical elements in the system well enough that you don’t introduce errors. A foolish idea has crept into science that such things are the tasks of mere technicians!

The great scientists of the past were often obsessed with problems that would now be derided as mere engineering, however the challenge of applying the powerful ideas of science to the real world often informed further theoretical advances. Since I’ve worked for years in manufacturing, I think you also learn a lot by trying to have someone else follow your instructions, and still make the thing work every time. Your idea of what a big problem is changes once you try to make it happen, either in the lab or the factory, and that is exactly what modern science doesn’t seem to want to do.

I love this image because of the silly conceit that AI research is the example of the hardest thing people do

I love this image because of the silly conceit that AI research is the example of the hardest thing people do

I am of course exaggerating for effect, but you really should try on Locklin’s argument for size. It is fundamentally similar to the reason I don’t worry much about automation taking jobs anytime soon, because I have tried to do it myself, and I know how hard it really is. McDonald’s has been applying automation for almost 80 years, which is why they can successfully put kiosks in their stores to replace cashiers. Lots of people see the kiosks, and wrongly conclude that all such jobs will disappear in a few short years. It just isn’t that easy, because there is a very deep foundation of streamlining, elimination of waste, and optimization of workflows in the background that such people do not see.

Others can imitate what McDonald’s has done in a shorter timeframe, but someone still has to understand what needs to be done and do the work to make it happen. The greatest of fools assume that AI will do this work too.

Locklin also makes an argument that is similar to my reaction to Paul Romer’s Nobel Memorial Prize in Economics:

…the total number of Einsteins in the world, or even merely serious thinkers about physics is probably something like a fixed number. It’s really easy, though, to create a bunch of crackpot narcissists who have the egos of Einstein without the exceptional work output.

It isn’t at all clear that we have really benefited by vastly increasing the number of people who work in research. We should redirect many of the bright and technically minded people who do dead end science like quantum computing into more applied fields. Hell, we would probably be better off if we just convinced them to actually apply what they are doing now. They might learn something useful.