Hacker News Reader

Comment by bjornsing | Hacker News Reader

I could have been a theoretical physicist, but got out after my master’s. Subconsciously I think I had already come to this (the OP’s) conclusion then.

Since then I’ve worked a bit in statistics and machine learning. There’s a saying in this field that captures my subconscious conclusion well: “all models are wrong, but some models are useful”.

I’ve often said that I would have loved to do a PhD in physics in 1910, but less so in 2010. The models physicists found in the 20th century were extraordinarily useful. I have little hope that I will see anything comparable in my lifetime.

Don’t get me wrong: More models will undoubtedly be found, and there’s beauty and honor in that work. But they will most likely all be “wrong”, and far less useful than e.g. the models that let us harness nuclear reactions.

View on HN

This is bit of a commonly misheld perception in my opinion.

It took 30-50 years for anything in the 1910s to become particularly useful. Even the photoelectric effect itself took over 30 years before the first devices (photomultipliers) were even developed.

The much celebrated quantum mechanics likely wasn't so useful either until 30-40 years later with electronics such as transistors.

General relativity didn't really find a use until GPS and that was over 70 years later.

The only field that contributed 'relatively' quickly during that time period was nuclear physics, but that field saw an extreme concentration of effort due to the war.

Meanwhile theoretical condensed matter physics today is rife with things that are on the cusp of being useful with 2D materials, superconductors, quantum simulators, etc

That depends a bit on the field. If you were doing particle physics (in particular using earth-borne accelerators), the second half of the 20th century would probably be best. But if you're into astroparticle or cosmology, the best time is literally right now. The 21st century has been one giant thing after another thanks to Sudbury, WMAP and Planck, the EHT, LIGO, the Pulsar Timing Array and tons of other experiments. When it comes to understanding the universe as a whole, there is no better time to work in physics than today. Similar goes for quantum and condensed matter, albeit with a focus on real world applications rather than understanding the universe.

I’m into usefulness. And I’m not convinced by your argument. I still doubt very much that I will see any new models as useful as those that transformed the world in the 20th century. I’d love to be wrong though.

Usefulness usually means real world use. It took the theory of general relativity 70 years before we saw a real world application make use of it. The theory that led to modern hard drives (GMR) went from new fundamental physics in the late 80s to full commercial use within a decade. Things like the blue LED (1990s) or frequency combs (early 2000s) were even faster and can be found truly everywhere now (unlike GR). There are far more useful models out there than the high level overview stuff you might get from an intro to fundamental physics class. And even if you're into explaining the universe rather than improving day to day life, as I said, there's tons of stuff going on right now if you're in the right field.

Reply on news.ycombinator.com