More importantly,
A lengthy thread I wrote just to write. Feel free to ignore.
1/ The future of consumer GPUs is probably less ASICy than people think. I think we're nearing the end of growing non-general purpose compute/cache die utilization in consumer GPUs. Here's why:
More importantly,
Beyond that, adaptive sampling research usually focuses on general cases. But consumer GPUs use RT for e.g. games, Blender, etc. These aren't super general. The engines
Better exploitation of temporal data and ray tracing statistics can make a lot of the rays cast
Frankly, I love tensor cores on consumer cards. Cheap training hardware is great! But I
The reason
But you don't need THAT much compute for these.
The big concern is that for DLSS and AI denoising, you usually need Nvidia's help training a model for your work. This may make them inaccessible for small devs and
Sure, there are general NN denoisers & upscalers, but they're not demanding, and they suck in terms of predictability, flexibility, (and, hot take: promise,) etc., so it's hard to see demand for these ASICs growing long-term.
The answer is simple: the software is still being written. Nobody knows how quickly better RT algorithms
More from For later read
Nice to discover Judea Pearl ask a fundamental question. What's an 'inductive bias'?
I crucial step on the road towards AGI is a richer vocabulary for reasoning about inductive biases.
explores the apparent impedance mismatch between inductive biases and causal reasoning. But isn't the logical thinking required for good causal reasoning also not an inductive bias?
An inductive bias is what C.S. Peirce would call a habit. It is a habit of reasoning. Logical thinking is like a Platonic solid of the many kinds of heuristics that are discovered.
The kind of black and white logic that is found in digital computers is critical to the emergence of today's information economy. This of course is not the same logic that drives the general intelligence that lives in the same economy.
Help! What precisely is "inductive bias"? Some ML researchers are in the opinion that the machine learning category of \u2018inductive biases\u2019 can allow us to build a causal understanding of the world. My Ladder of Causation says: "This is mathematically impossible". Who is right? 1/
— Judea Pearl (@yudapearl) February 14, 2021
I crucial step on the road towards AGI is a richer vocabulary for reasoning about inductive biases.
explores the apparent impedance mismatch between inductive biases and causal reasoning. But isn't the logical thinking required for good causal reasoning also not an inductive bias?
An inductive bias is what C.S. Peirce would call a habit. It is a habit of reasoning. Logical thinking is like a Platonic solid of the many kinds of heuristics that are discovered.
The kind of black and white logic that is found in digital computers is critical to the emergence of today's information economy. This of course is not the same logic that drives the general intelligence that lives in the same economy.
