Trendaavat aiheet
#
Bonk Eco continues to show strength amid $USELESS rally
#
Pump.fun to raise $1B token sale, traders speculating on airdrop
#
Boop.Fun leading the way with a new launchpad on Solana.
What this makes me think of is human vision.
It's easy to forget, and most of us never even learned, that human vision doesn't really have great coverage either.
We have a very narrow, high-fidelity spotlight, surrounded by a wide, low-fidelity periphery.
The reason our vision feels a lot more "complete" than that to us is that the data our forebrain gets is actually the result of a lot of post-processing in other areas of the brain.
Once, while helping a doctoral student in perceptual psychology with some C++ code for her thesis research, I accidentally discovered how to erase parts of this post-processed model.
This create an effect where you could show something, a stimulus, to the eye, but make it invisible to the brain.
It was one of the strangest sensations I have ever felt. At first, I thought that I had introduced a bug, and the code that presented the visual test stimulus was no longer running, because I couldn't see it.
But when I walked through with a debugger, the stimulus was there. It was simply being erased from my visual cortex before my conscious mind ever "saw" it.
The PhD student looked at this, shrugged, said "weird" in an incurious tone of voice, and went back to the topic of her study. And I got the ick and scrapped my plans to ask her out on a date.
Anyway, I bring this up because it makes me suspect that the "panopticon" hardware approach was always doomed in practice, if even nature doesn't do things that way, and instead substitutes heavy post-processing for detailed raw input.
I suspect that, just like in our biological evolution, the evolution of our computer systems will hit a point where additional compute is a lot cheaper than better peripherals.
Although perhaps John is about to tell me this happened years ago and I just haven't been paying attention.
13.8. klo 00.15
There have been a lot of crazy many-camera rigs created for the purpose of capturing full spatial video.
I recall a conversation at Meta that was basically “we are going to lean in as hard as possible on classic geometric computer vision before looking at machine learning algorithms”, and I was supportive of that direction. That was many years ago, when ML still felt like unpredictable alchemy, and of course you want to maximize your use of the ground truth!
Hardcore engineering effort went into camera calibration, synchronization, and data processing, but it never really delivered on the vision. No matter how many cameras you have, any complex moving object is going to have occluded areas, and “holes in reality” stand out starkly to a viewer not exactly at one of the camera points.
Even when you have good visibility, the ambiguities in multi camera photogrammetry make things less precise than you would like. There were also some experiments to see how good you could make the 3D scene reconstruction from the Quest cameras using offline compute, and the answer was still “not very good”, with quite lumpy surfaces. Lots of 3D reconstructions look amazing scrolling by in the feed on your phone, but not so good blown up to a fully immersive VR rendering and put in contrast to a high quality traditional photo.
You really need strong priors to drive the fitting problem and fill in coverage gaps. For architectural scenes, you can get some mileage out of simple planar priors, but modern generative AI is the ultimate prior.
Even if the crazy camera rigs fully delivered on the promise, they still wouldn’t have enabled a good content ecosystem. YouTube wouldn’t have succeeded if every creator needed a RED Digital Cinema camera.
The (quite good!) stereoscopic 3D photo generation in Quest Instagram is a baby step towards the future. There are paths to stereo video and 6DOF static, then eventually to 6DOF video.
Make everything immersive, then allow bespoke tuning of immersive-aware media.
12,41K
Johtavat
Rankkaus
Suosikit