Hacker News Reader

Comment by delichon | Hacker News Reader

If the Schwarzchild radius is reached within a larger mass, wouldn't that radius become inescapable and eventually consume the remainder while growing in proportion?

Since I did a sanity check on the result which seems to have failed, why do you think that it's necessary to warn me not to rely on it?

The hostility on this site for using LLMs as a tool for exploration is confusing to me. I also learn a lot by asking questions of highly unreliable humans. LLMs don't seem to be much worse than that.

View on HN

I don't have a problem with your post, but the hostility I see is mostly out of a real concern that people posting LLM comments could degrade discussion. It's probably also association, because a lot of people have started posting comments that just say "ChatGPT says..." and nearly the entirety of the body is just LLM output.

Honestly, a lot of people are just sick of seeing anything about LLMs and don't like seeing it at all in threads that aren't already about it.

Exactly. It's not appropriate for this type of question. LLMs can summarize the global internet sentiment about something pretty well, but they aren't the right architecture for answering precise scientific questions, nor will they ever be. It's like asking an old, experienced physicist to guess answers to things instead of solving equations.

They're already better than I'd have expected possible, so I wouldn't say "never" — but also yes I agree with you, currently I'd compare their output to an off-the-cuff answer from an experienced professional (or an undergraduate trying hard).

As a physicist I can say that unless ChatGPT is asked a definitional question, or a very common question (whose solution is likely described many places on the internet anyway), it is very likely to be wrong. Personally more than 90% of the questions I've asked it have been flat-out wrong so I stopped using it entirely.

This is maybe cause ppl don’t want to rely on some machine hallucinating wrong answer when trying to answer fundamental questions but rather use real wisdom instead.

As others have said, that's not what a Schwarzschild radius means. You would only get a black hole if all the matter is inside the radius.

The Schwarzschild radius comes from the free space solution to Einstein's equation. It only holds in free space. So if you want to use it here, all the matter would need to be within Schwarzschild radius, and then you could calculate what happens outside that radius (again, in the matter free space outside). If you want to know how gravity behaves inside a star you have to use a different tool.

Lots of other replies so hopefully it makes sense but just to be clear: 'the' Schwarzchild radius isn't one specific number, it's based on the mass of an object. Every massive object has a different SR (also called its Event Horizon), and if you compress the object entirely to within that radius, it will become a black hole. It's just a measure of how small the object needs to be before light cannot escape its surface.

> If the Schwarzchild radius is reached within a larger mass, wouldn't that radius become inescapable and eventually consume the remainder while growing in proportion?

If I understand your question correctly, no. If I have a mass M that corresponds to a Schwarzchild radius r, but the mass occupies a larger radius R, no, it will not create a black hole, since the mass m within r is less than the critical mass M. The portion of M that is within r can still escape that radius.

It's fairly well-accepted that tiny black holes are quite unstable:

https://en.wikipedia.org/wiki/Hawking_radiation

On a larger scale, kinda yes - a small-ish black hole would generally "feed" from any near-enough object. Though that feeding will generally be extremely energetic and messy, and fling most of the victim away at high velocity.

You only have a black hole and an event horizon if the mass whose Schwarzschild radius you are calculating is completely contained within said radius.

Reply on news.ycombinator.com