Comment by 20k
15 hours ago
>of course this is a simplified visualization
Its titled a "physically accurate black hole"! Please don't do this, if its intentionally not physically accurate and heavily simplified! The general public on here don't know better, and it takes someone else with years of experience simulating black holes to dig through your code to show what you've simplified. There's like a few hundred people who have the right experience worldwide to know that you've got some of your equations incorrect
The issue is when you're communicating doppler, if you want to correctly show the change in doppler, you can't make the intensity or colour mapping arbitrary
There's two ways to communicate accurately an example linear change in emissive power:
1. Use a perceptual brightness system, so that a linear change in power represents a linear perceptual change in perceived brightness. This accurately communicates to a viewer the underlying shift in power emission, if you want to show off how a radio telescope perceives a black hole for example
2. Use an emitted power <-> sRGB mapping (with tonemapping), so that a linear change in power is perceived as if it were a physical light getting brighter. This is more "what would this look like to a person?"
At the moment the change in your image brightness doesn't correspond to anything, which means that its not really communicating any science, and it doesn't correspond to doppler. Which is fine if its just illustrative and that's clear, but again vanishingly few people will know this is a visualisation not a physically accurate render
It's AI garbage - OP asked the LLM for "physically accurate", Claude gave them this, and they didn't know any better.
>and they didn't know any better.
i think an "astrophysicist at Harvard's Black Hole Initiative studying quasars and black hole environments" who "work[s] with raytracing/radiative transfer simulations professionally"probably has a decent understanding of what's going on...
You'd think, and yet, they happily slapped the Harvard BHI name on something that doesn't do what it says on the tin. I think ignorance is preferable to deliberate deception, so I assumed ignorance.
3 replies →
> Its titled a "physically accurate black hole"! Please don't do this, if its intentionally not physically accurate and heavily simplified!
It also doesn't exert the gravity an actually accurate black hole should :) And also, even on a phone with two cameras, they only cover a part of sphere – while a real black hole would bend light coming from all 4pi directions. Some simplification is generally required ("the map is not the territory") and I don't see "simplified" as being opposite to "accurate".
Colormap choice: it can be a useful mode to show proper linear power -> linear perceived brightness mapping – true! But that's only one of many possible choices, and scales like logarithmic are very common in scientific images.
>I don't see "simplified" as being opposite to "accurate".
Are black hole accretion disks generally red in real astrophysical cases?
They're most accurately represented as being blue. Red is fine for visualisations because it looks cool (which is fine!), but blue is the physically accurate colour. Visualisations claiming to be accurate like this presenting red black holes is becoming an increasingly large science communication problem
The way that the colour and brightness are mapped together is also not correlated correctly, and some of the equations just aren't right. The doppler isn't correct at all here, partly because its simulated incorrectly (with the camera problems), and partly because the mapping done at the end doesn't correlate to any reasonable brightness
Stuff like this: Its not accurate at all. It isn't simplified, its just wrong - which is totally fine for a visualisation, if that's the intent. But:
>scales like logarithmic are very common in scientific images.
Sure. What scale are you using for the doppler brightness here? Because as far as I can tell, its a piecewise mapping of the form ~pow(x, 2.2), after being run through a non smooth function `afmhot` - a function which doesn't have a linear brightness response. So the brightness in this image doesn't really correspond to anything related to doppler or brightness, and the colour also doesn't correspond to anything involving colour (or the doppler). The entire image's look is basically accidental
Here's a specific comment from the event horizon telescope talking about exactly this problem with specifically the afmhot transform being used, because its important for visualisation:
https://github.com/achael/eht-imaging/issues/85
Which shows some of the problems here
“Physically accurate” indeed raises the bar significantly.
You’ve already done great work here. That said, the feedback seems to come from someone who spent considerable time analyzing your work. Even if only a few of the suggestions are ultimately valuable, that’s still a meaningful contribution and worth considering.
Sure, most of the raised questions are reasonable potential feature additions to be made – things like rotating black holes, or linear brightness-to-screen-pixel mapping. They just aren't "inaccuracies", in my opinion.
Always funny when people claim that scientific term apply to their software and then disregard any founded criticism in a non scientific way.
Its frustrating because these issues are all fixable, and some of them are just bugs, but there's a mentality in this kind of science communication that ehh it doesn't matter because its just for the public. This does a real disservice to what people are trying to actually achieve in educating the public. Its also, unsurprisingly, a lot of hard work and validation to get from visualisation-grade, to physically accurate, and its a lot easier to not do that
4 replies →
Now after reading the OPs responses and how defensive they are of some high quality constructive criticism, I don’t even want to go back and play more with the visualization.
>I don't see "simplified" as being opposite to "accurate".
Is it accurate to model a cow as a sphere?