Mark Tyson Social Links Navigation News Editor Mark Tyson is a news editor at Tom's Hardware. He enjoys covering the full breadth of PC tech; from business and semiconductor design to products approaching the edge of reason.
pjmelect It annoys me when I see comments like this as if the USA is the whole world, while a ban maybe implemented in America the rest of the world will take no notice. Reply
USAFRet pjmelect said: It annoys me when I see comments like this as if the USA is the whole world, while a ban maybe implemented in America the rest of the world will take no notice. Not saying Sanders is 100% correct, but maybe the world should come to some sort of agreement, like what happened with nuclear weapons. Reply
timsSOFTWARE USAFRet said: Not saying Sanders is 100% correct, but maybe the world should come to some sort of agreement, like what happened with nuclear weapons. If they really wanted to stop AI, they'd need to stop developing more powerful computer hardware and building datacenters. Do that, and they'd stop the AI as well. Fundamentally, AI algorithms are just a different way to build a function. In math, or in traditional programming, you have a function a human created by hand, that takes some explicitly listed inputs, performs some operations on the inputs, and outputs a result. With neural networks/deep learning, you replace writing the algorithm by hand, with finding a close enough approximation through iterative adjustment. Instead of named function parameters chosen by a human, you have a vast quantity of unnamed ones, and instead of doing different math operations, you just do multiplies and approximate all other operations just by having a large enough quantity of weights. It enables solving otherwise intractable problems – like replicating human thought patterns as they apply to writing – at scale by having enough compute. Trying to stop AI development is just not realistic – it's a natural evolution of computer science, and it's not even a new idea/way of solving problems, as most of the ideas behind it are relatively old; it just wasn't feasible to deploy at scale until now. Reply
MrK26 USAFRet said: And these local models get their data from…where? The model IS the data. It's self-contained, that's what the training phase of a model is. You don't need to get data from anywhere; there are already open-weight (think open source) local models anyone can download. E.g., here's a qwen3.8 model (about as good as Anthropic's Opus 4.6) and you can run without an internet connection: https://huggingface.co/huihui-ai/Huihui-Qwen3.8-27B-abliterated GPUs to run this are up to $1k now, but there are millions and millions of gamers with the power to download, start it on a computer locally, and it'll tell you anything you want to know. We're already done here folks, lol. Way past, best get to figuring out how to use the tech to harden systems, like my company does. Reply
MrK26 USAFRet said: Cool. So we can now dispense with all these companies building these big DC's? Honestly, probably won't need them, yes. Reply
USAFRet MrK26 said: The model IS the data. It's self-contained, that's what the training phase of a model is. You don't need to get data from anywhere; there are already open-weight (think open source) local models anyone can download. E.g., here's a qwen3.8 model (about as good as Anthropic's Opus 4.6) and you can run without an internet connection: https://huggingface.co/huihui-ai/Huihui-Qwen3.8-27B-abliterated GPUs to run this are up to $1k now, but there are millions and millions of gamers with the power to download, start it on a computer locally, and it'll tell you anything you want to know. We're already done here folks, lol. Way past, best get to figuring out how to use the tech to harden systems, like my company does. I completely get what you're saying. But also, this is running on data current as of today . Next week, next year…. Reply
Kicapan07 Old senile Luddite at it again. Scares of shadows as per usual. Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/tech-industry/artificial-intelligence/SPONSORED_LINK_URL
- https://www.tomshardware.com/tech-industry/artificial-intelligence/sanders-proposes-20-year-prison-sentence-for-ai-devs-who-plow-ahead-with-artificial-superintelligence-plans-penalty-on-par-with-illegally-developing-rogue-nuclear-weapons#main
- https://www.tomshardware.com/membership
- Dumpster diver builds home lab Proxmox server from weekly landfill runs for SSDs, HDDs, GPUs, and even RAM — weekly e-waste raids net multi-drive Proxmox server
- Snag a huge $1,000 saving on this RTX 5090 OLED gaming laptop from HP, now $3,699 — 16-inch HyperX rig delivers 1600p gaming with ultra-fast 240Hz refresh rate,
- Windows 11 can now reinstall itself from the cloud — Cloud Rebuild revives dead systems without secondary boot media, wipes broken installation and downloads a
- How XPUs Meet a World-Class AI Factory
- NVIDIA to Acquire Hugging Face
Informational only. No financial advice. Do your own research.