AI Chip Architectures(jepeake.com) |
AI Chip Architectures(jepeake.com) |
Companies like EnCharge, Mythic, etc.
And much more efficient devices like RRAM, MRAM, and FETs. Like FE-FETs with AlScN.
The stuff just coming out of research or still in research is more exciting in terms of the potential for truly huge efficiency and performance boosts.
Taalas is interesting also because of it's efficiency and speed. Guess it was just purchased by AMD.
If they are in some analog half-open state they have both resistance and current, which means waste heat.
Also transistors in modern chips are already so small that you can literally count the number of electrons flowing through them. At the smallest scale, nature becomes discrete. So if there are something like 128 electrons, you only get 7 bits of 'analog' to play with anyway.
Recently on HN: https://news.ycombinator.com/item?id=49352470
Multiplication may make sense for the math, but it's just not practical in hardware. There has to be a better, much simpler operation, that can give you equivalent results. Yes, you may have to re-think and re-train everything, but eventually it should make current architectures obsolete.
The software architecture LLMs run on was chosen precisely because hardware can brute force it and brute force has become the method of choice ever since, because it is easier to scale.
And here you're saying "what if everyone was wrong and what we need is just more brute force?", without looking at the reason that makes brute force necessary to begin with.
Current architecture can't tell the difference between fact and fiction, so to speak. (Which I think is a better way to say it rather than truth and falsehood; facts have provenance and context, not just "true" or "false".)
In all cases it was meatbag. In the 1st case gave chatGPT 0.35. The remaining ones were much lower.
EDIT: I didn't ask just to generate some text. The original prompt was very specific, which is most of the time the way I use LLMs anyways. Probably it catches better the "hey just generate this email...".
When the density of italicized words is about 1:1, I'm out.
Might be that we come at this page with different experience and knowledge. I'm not seeing any specifically hard to read AI here.
As it turns out, you still need to be a good writer in order to prompt an AI to write well.
Ignorant people are always certain.
Not slope tho!
One of the best deep-dives on AI chip architectures that I've ever read.
Other posters are welcome to their opinions, but I (again, personally) thought it was great!
My brother bought one. I challenged him to find something that actually used the AI chip on the laptop. He couldn't - it always read at 0% in task manager.
As a consumer, I'd rather have that much more graphics card, hard drive space, or RAM for the price than a specialized chip thats almost never used and, when it is used, only saves you a little electricity.
I'm not saying AI has remotely solved this problem, not even close. But plenty of philosophy, especially in the 1950s+, from analytic philosophers that these kinds of statements aren't coherent.
Symbolic and formal AI has failed as a general path to intelligence, heuristic bash with deep learning won. There needs to be some kind of actual breakthrough on the actual structure of how we formalize things - we can either describe processes and just let them run (physics, CS), or verify well (math), but if you look at any messy, real-life domain, how "knowledge" is encoded is a bajillion heuristics that would not really be the "rules" of the domain. The heuristics being linear algebra rather than SAT equations doesn't change how bad the "knowledge" is compared to what humans idealize as "knowledge" or even empirically, the kinds of things humans cut at for "knowledge".
It is a very interesting concept and the company Q.Ant has delivered working products and software. Apparently it can greatly speed up certain training applications and future designs will solve some of the current problems.
The energy and performance advantages seem to be worth the R&D. If some form of optical memory to replace DRAM in the chain could be designed, that would solve several conversion costs between optical to digital.
Either way it is pretty cool to do math this way, basically just using light.
e.g for a Mach-Zehnder interferometer measuring the phase shifts requires many photons. A single photon arrives or doesn't arrive at the detector. Even if you could perfectly ensure it hits detector at the same time as the reference photon and subsequently detect with 100% quantum efficiency, that still only gives you 1 bit of phase information, not an analog value.
Digital can be done more efficiently with adiabatic or asynchronous circuits. The limiting factor then is speed and chip area rather than joules-per-and-gate.