Their photos are worth reposting for the few who won't read past the first paragraph. This is cool.

https://mijnrealiteit.nl/

Like the models before it it struggles with my use case of tracing circuit board features. It's great with a pony on the beach but really isn't made for more rote industrial type applications. With proper fine-tuning it would probably work much better but I haven't tried that yet. There are good examples on line though.

I used to be really (really really) into photography. I respect anyone working hard on a physical product, but this misses the mark on every front I can think of.

The real issue that photographers grapple with, emotionally and financially, is that pictures have become so thoroughly commodified that nobody assigns them cultural value anymore. They are the thumbnail you see before the short video clip starts playing.

Nobody has ever walked past a photograph because they can't inspect its digital authenticity hash. This is especially funny to me because I used to struggle with the fact that people looking at your work don't know or care what kind of camera or process was involved. They don't know if I spent two hours zoomed in removing microscopic dust particles from the scanning process after a long hike to get a single shot at 5:30am, or if it was just the 32nd of 122 shots taken in a burst by someone holding up an iPad Pro Max at a U2 concert.

This all made me sad for a long time, but I ultimately came to terms with the fact that my own incentives were perverse; I was seeking the external gratification of getting likes just like everyone else. If you can get back to a place where you're taking photographs or making music or doing 5 minute daily synth drills for your own happiness with no expectation of external validity, you will be far happier taking that $399 and buying a Mamiya C330.

This video is about music, but it's also about everything worth doing for the right reasons. https://www.youtube.com/watch?v=NvQF4YIvxwE

> Given consistent trends of exponential performance improvements over many years and across many industries, it would be extremely surprising if these improvements suddenly stopped.

I'm sure people were saying that about commercial airline speeds in the 1970's too.

But a lot of technologies turn out to be S-shaped, not purely exponential, because there are limiting factors.

With LLM's at the moment, the limiting factors might turn out to be training data, cost, or inherent limits of the transformer approach and the fact that LLM's fundamentally cannot learn outside of their context window. Or a combination of all of these.

The tricky thing about S curves is, you never know where you are on them until the slowdown actually happens. Are we still only in the beginning of the growth part? Or the middle where improvement is linear rather than exponential? And then the growth starts slowing...

Off topic, but where should one start learning writing physical simulation?

Several years ago I ran into this project [0] and got overwhelmed even the algorithm can be written in 88 lines of C++. I realized that out of all CS topics, physical simulation is probably the one I knew the less (not saying I'm a compiler/database expert or something, but at least I've implemented a toy compiler and some basic data structures used in database. When it comes to physical simulation my bran just draws a blank.)

[0]: https://github.com/yuanming-hu/taichi_mpm