Join the community of Machine Learners and AI enthusiasts.
Sign Up
And here's a thought: What if the rich, instead of hoarding these tools, actually had something more to offer? Could it be that they've honed their skills in the crucible of privilege? Because after all, you can't take intelligence from a Harvard-educated billionaire and simply give it to the next-door neighbor.
The Aurora in Argonne will process anyone's science for free after a costly upgrade. There are people on your side, and even those willing to monopolize realize that they never know when some advance helpful to them will be found without them paying a thing accidentally as a result of any scientific research.
That's the point of Project DIGITS, and it's hard to convince big players they aren't shooting themselves in their foot with moves like that, but they aren't, mostly. We will receive amazing open-source models that can be used freely by everyone, and they'll still be able to develop them more to make money off of a truly better product. I hope for big changes, but we didn't see a ton more inventions due to 3D printing, in perspective, since we've had SLA printing for some time, although it has been updated (I was involved with CLA) and improved for making rocket parts that are more durable and better able to be prototyped actually more quickly.
It's amazing the GPUs people have been given toward this already. 3090s are excellent. Somehow, people are getting H100s. I can't afford such things, although I could, if I asked for money for my ideas, like many corporations and even people do. They need to learn about compound interest, and how cheap computers actually are to make, with some good connections.
Well, because I've been around for a while, helping exactly what you're talking about, while helping the corporations, I'm asking for an HPC with a custom PCIe 6 5090 with extra RAM (probably some stacked SRAM), and a faster clockspeed, an H200 with considerations for the fact that I will use only one, maybe an L4 for low-power inference, two 7995WXs, a custom NPU similar to the WSE-3 in design, a ton of probably HBM3 and ZBT SRAM (depending on speeds necessary) right on the custom (black and gold ASUS with better pins all over) motherboard with adequate bus sizes all over, persistent memory to have things loaded all of the time that are used often even after shutdown and for scratch space, a phase-change liquid cooler for the CPUs (direct-die), GPUs (custom copper shrouds), Peltiers, industrial fans on the hot parts of the Peltiers and to bring through the custom air channels to the components running on another phase-change cooler for air rather than evaporating liquid, and two copper radiators to put any actually water-cooled parts (possibly the OPU, if they'll give me one, and a little silver button on the all black case with RBG around it, because it's not for gaming. If I get This HPC due to my efforts already in improving computing, which are too big to talk about, so it's likely, all the processing goes to science, with me willing to process anyone's AI for free. So, there's a reason why this relates. Some of us don't want to be paid in money, but in helping the world, and that's what I want, so giving such people the technology helps everyone.
What am I going to do on it? Well, it can game, but it can also run AI that people need. Primarily, it will process for BOINC, Folding@home, and a few AI projects that use distributed processing. I want people to send me ideas on how to use it if I get it, and I'm thinking the corporations will have a hard time turning me down, because they made HPCs and GPUs based on them that are quite profitable and I only wanted more science and AI processed, as well as suitable gaming for others (it's still not there yet, and will include more AI if they listen for a number of important reasons), which is improving, also using photonic computing now, which wouldn't be functional without my efforts.
That was a long read. You may not believe anything about me having influenced things, even though I scarcely mentioned what. My next goal is to have wireless GPUs via docking stations with visible light communication and IR (possibly powered that way too), along with increasing smart home wireless capabilities using light fixtures and the like, not unlike Li-Fi and Koruza, but with ordinary Wi-Fi and 5G redundancy beyond line-of-sight. So, you'll know if you see those crazy advances or the HPC. The idea is that people will be more prone to buy expensive GPUs, which can be shared by the whole house. Just pop in the new card somewhere in the center of the house, and it's installed. It should allow hot-swapping, and the idea is to either put it somewhere cool already (next to a vent, in one, or by a mini-split) or to put it where it will get hottest to make the TPV cells that use the light to make energy work even better, since we hope to save on energy by using the same light in lieu of needing to use the full amount from a socket.
Photonic-integrated circuits are the future. You can see photonics are functional for actual computing (not just NVIDIA's current photonic switches) in the LightOn The Appliance OPU. I've asked for a small version of this for particular tasks to pass off. Nobody will know what it is. That's because the computer will be good for an HPC of that size for probably ten years.
it's pretty exciting to see the newer, more powerful hardware coming down the pipeline. with better hardware becoming more prolific I hope we can see expansions of concepts like BOINC and Folding@Home and newer ideas as well. It's impossible for you or I to compete with BigAI, but with federated learning techniques and enough people I think it can be done, at least "good enough". It sounds like the kind of thing you might be interested in, yeah?
