Posted on 05/12/2025 10:07:46 AM PDT by DFG
Last week, a U.S. congressman announced a plan to introduce a bill that would mandate producers of high-performance AI processors to track them geographically in a bid to limit their usage by unauthorized foreign actors, such as China. Senator Tom Cotton of Arkansas then introduced a legislative measure later in the week. The bill covers hardware that goes way beyond just AI processors, and would give the Commerce Secretary power to verify the location of hardware, and put mandatory location controls on commercial companies. To make matters even more complicated, geo-tracking features would be required for high-performance graphics cards as well.
The bill covers a wide range of products classified as 3A090, 4A090, 4A003.z, and 3A001.z export control classification numbers (ECCNs), so advanced processors for AI, AI servers (including rack-scale solutions), HPC servers, and general-purpose electronics of strategic concern due to potential military utility or dual-use risk. It should be noted that many high-end graphics cards (such as Nvidia's GeForce RTX 4090 and RTX 5090) are also classified as a 3A090 product, so it looks like such add-in-boards will also have to add geo-tracking capabilities.
The first and central provision of the bill is the requirement for tracking technology to be embedded in any high-end processor module or device that falls under the U.S. export restrictions. This condition would take effect six months after the legislation is enacted, which will make the lives of companies like AMD, Intel, and Nvidia harder, as adding a feature to already developed products is a tough task. The mechanism must allow verification of a chip's or device's physical location, enabling the U.S. government to confirm whether it remains at the approved endpoint. Yet, exporters would be obliged to keep track of their products.
(Excerpt) Read more at tomshardware.com ...
How much does a very short range GPS spoofer cost?
Any less heavy-handed way to do this?
The Left and Leftist gov’t is so oppressive.
Another unfunded mandate on tech. I often wonder what percentage of telecom bills are due to all the LE backdoor requirements.
They’re coming for your bitcoin.
Agree, not a fan of this. Article is a little fear mongering though. The bill doesn’t require GPS, and only applies to exports, not GPU/Video cards in the US. That said, if nvidia does end up having to put tricking in their chips, they probably won’t make a separate chip for the US domestic market and we all get stuck with govt tracking.
Trying to keep tech like this out of China is a hard problem, but this isn’t the right approach.
A Red Chinese agent can steal a chip, send it off to Red China where AI can reverse engineer it.
Where there’s a will, there’s a way.
The only thing Red Chinese can’t steal is honesty.
The Chinese with careful programing can get the performance they want with their own chips.
Factories and other large buildings have metal roofs.
Server racks have lots of GPS-blocking metal.
So your CPU is going to be tracked like a cell phone. Talk about an evasion of privacy. Just think were this can go in the end.Even Ray Bradbury could not have imagined it. Just wait till they demand it be installed in shoes.
So this idiot in Congress wants to place ankle bracelets on chips? What government agency is going to monitor where all of these chips are? What happens if one drops off the radar? What are they going to do? Just another dumb idea from Congress.
No they can’t.
Why don’t they require a Mission Impossible self-destruct circuit in addition to the tracker circuit? They could put a StarLink transceiver in, too, so we can tell what the Chicoms are doing.
I wish these guys would spend more time on fixing the budget and less on this kind of stuff.
This is a very important initiative. It is not directed against just the Chinese. These very powerful GPUs can be thought of similar to high speed centrifuge technology for uranium enrichment. It is in everyone’s interest they be controlled, but it does not go far enough by half. Most people have no idea at all yet of what is coming down the pike at us shortly as far as advanced AI is concerned. The problem, which surprised almost all experts in the field, was that it turned out that creating human and transhuman level intelligence was a lot easier than we thought, and it is happening a lot faster than expected.
That's very interesting. Please explain. Are there any references readable by the layman?
Everyone in the field (almost everyone) knew we could eventually create super intelligence, things just moved faster that most people thought possible. The English AI philosopher Nick Bostrom wrote about super intelligence in 2014. (He was far from the only one, I am just using him as my example.) He originally thought it was more likely to come after the first third of the century (2033) than before. Very recently he has said it might come as soon as a year or two, and I think he is right for many reasons. Here is a link for that claim: https://x.com/slow_developer/status/1923840949794341359
Bostrom’s book is a very good introduction to the subject. He is very, very good. Not infallible, but very good. Link for free version from Internet Archive: https://archive.org/details/superintelligence-paths-dangers-strategies-by-nick-bostrom The problem is this intelligence explosion happens at the speed of an extremely powerful super computer.
This is one of his more important insights: “At some point, the seed AI becomes better at AI design than the human programmers.
Now when the AI improves itself, it improves the thing that does the improving. An intelligence explosion results—a rapid cascade of recursive selfimprovement
cycles causing the AI’s capability to soar.” Superintelligence p. 95-96
He wrote that in 2014. Now here is the the really far out thing: OpenAI CPO Kevin Weil says we will essentially reach this point in a year or so. Link: https://x.com/slow_developer/status/1924403279007519115
Things are moving very fast in a very bad direction. I am really putting all my earthly hope on Trump suddenly announcing a deal with China to freeze AI development, because if we don’t, we are not very far away from developing something a lot more dangerous to humanity than the H Bomb.
Put one key sentence in the wrong place. Annoyingly. :) Here is how it is supposed to go:
“This is one of his more important insights: “At some point, the seed AI becomes better at AI design than the human programmers.
Now when the AI improves itself, it improves the thing that does the improving. An intelligence explosion results—a rapid cascade of recursive selfimprovement
cycles causing the AI’s capability to soar.” Superintelligence p. 95-96
The problem is this intelligence explosion happens at the speed of an extremely powerful super computer.
Thanks for the response. That is an awful lot to digest.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.