Lost in Math: How Beauty Leads Physics Astray II

Lost in Math, I think, is of the greatest interest to working theoretical physicists, because it examines the main challenges currently facing them. If you are not yourself a theoretical physicist, you may at best get a glimmer of what their work entails. In my case I have some familiarity with the aspects relevant to astronomy, such as dark matter and gravitational lensing, but I'm not as interested in particle physics. What I find the most interesting, however, is Hossenfelder's description of theoretical physics in terms of the routine vocational dysfunctions that occur in other fields. This makes a nice contrast to the mythical depiction of physicists as geniuses who think at a level so much higher than that of ordinary mortals that you dare not suggest their fallibility.

Although sociological and psychological analysis of the field is only loosely scattered throughout the book, Hossenfelder does a reasonably good job showing that physicists face the same hurdles that people in other fields do, even when they are seeking nothing more than scientific truth. Conformity, groupthink and the status quo tend to squelch original scientific inquiry, and getting funding for novel ideas is difficult. Because particle accelerators are the experimental backbone of quantum physics and are prohibitively expensive, it is more financially feasible to hire string theorists, whose work can be done in the absence of experimentation. Hossenfelder specifically compares physics to economics in terms of overusing mathematical models at the expense of experimentation. She herself had considered switching to economics, where the math is much easier than what she's used to, if only to provide a more stable career. She brings up some of the ideas that I've discussed before while commenting on books by Daniel Kahneman and Robert Sapolsky, but without going into as much detail, and perhaps not fully recognizing the futility of attempting to remedy the situation. Sapolsky in particular is acutely aware of the intractable limitations created by our biological provenance. I don't think that she has been exposed to some of these developments in biology.

Hossenfelder spends a lot of time asking why beauty is so important to physicists, and she provides some answers without fully settling the matter. Usually this boils down to people using equations which work fairly well, but not perfectly well, to describe a phenomenon, requiring a messy sort of "fine-tuning" that no one likes. For many physicists, according to her, beauty is a stand-in for meaning, because it provides a sound structure without ad hoc fudge factors. Her position seems to be that one must adopt the best model available whether it's pretty or not, and that one should always favor models compatible with the latest experimental data. Currently, it seems as if there are too many models and not enough data to eliminate a lot of them. She also discusses the intrusion of philosophy into physics, saying that philosophers usually have nothing of value to add in solving physics problems, though physics itself does require philosophical assumptions. I agree with her here, and think that academic philosophy is mostly a useless and obsolete subject.

Probably my favorite idea in the book concerns AI:

I try to imagine the day when we'll just feed all cosmological data to an artificial intelligence (AI). We now wonder what dark matter and dark energy are, but this question might not even make sense to the AI. It will just make predictions. We will test them. And if the AI is consistently right, then we'll know it's succeeded at finding and extrapolating the right patterns. That thing, then, will be our new concordance model. We put in a question, out comes the answer – and that's it.

If you're not a physicist, that might not be so different from reading about predictions made by a community of physicists using incomprehensible math and cryptic technology. It's just another black box. You might even trust the AI more than us.

But making predictions and using them to develop applications has always only been one side of science. The other side is understanding. We don't just want answers, we want explanations for the answers. Eventually we'll reach the limits of our mental capacity, and after that the best we can do is hand over questions to more sophisticated apparatuses. But I believe it's too early to give up understanding our theories.

I think this captures our situation well. Unless a method to combine our brains with sophisticated AI is developed, there is an upper limit on how much we can comprehend. Even so, without such an enhancement, it may be possible for AI to translate its findings into terms that will be intelligible to us; it could decode the laws of nature in language that we understand. It is possible that theoretical physicists as a group are already operating close to a cognitive boundary that they will never be able to cross.