Interestingly, it makes significant mistakes, almost like it’s too lazy to read all the relevant materials it has access to; it makes stuff up. Still, it’s better than Siri… and it’s still an infant. It’s kinda fun too.
Me: You said that “If photons had rest mass, they would be affected by gravity and would not travel in straight lines at the speed of light, as observed.” But experiments have shown that a gravitational lens can bend the path of a photon. Are you incorrect in your assertion?
AI: You are correct that my previous statement was not completely accurate. I apologize for any confusion.
In general relativity, all forms of mass and energy, including photons, contribute to the curvature of spacetime and are affected by gravity. This means that photons can be deflected by the gravitational pull of massive objects, such as stars or galaxies, and this deflection can be observed as a gravitational lensing effect.
It’s not lazy; it just doesn’t actually know anything about physics. It has been trained to construct sentences and paragraphs about physics based upon applying a Bayesian model other sentences and paragraphs in its training set, and so it produced a response that is cromulent and not too misleading, but someone with virtually no understanding of physics could literally generate the same thing just by cribbing from Wikipedia or Scientific American articles or any other source of basic information. In the case of ChatGPT in particular, it isn’t even referencing information outside of its language training set; it is just structuring sentences that its language model determine are appropriate. It is really just a really powerful autocomplete function.
Since this is IMHO, I’m going to state my opinion that this is probably a waste of time for most people.
As a disclaimer, bear in mind that I am studying tensor math right now, at the age of 43, even though it will not benefit me in my career in any way. I’m someone with a passion for mathematics and lifelong learning; I am not going to try to dissuade someone from learning for its own sake.
However, what happened to me years ago, and I’ve seen it happen with other people, is that I got into pop sci books but got frustrated by the arbitrariness of it all. While the mathematics might form into consistent models (for the most part), the various natural language analogies and intuitive descriptions don’t. For the most part, you can’t draw inferences from these kinds of descriptions.
So I decided to try to study some of the mathematics of the actual models to try to get a more solid understanding.
Well, I wasted a lot of time, and didn’t get very far.
IME, it isn’t the kind of thing you can study on the side, casually. The “confidence gained per time spent” is very low.
I think you can study it on your own and gain confidence, but you need feedback, especially exercises with written-out solutions so you can check your work or get a hint.