Schneier on Security

Clive Robinson • April 25, 2026 12:24 PM

@ Untitled, Jan van Prooijen,

With regards getting at “temperature sensors”.

It’s all to easy to do these days… Just remember “heat” is the ultimate form of pollution. All “work” is by definition “inefficient” and the “inefficiency becomes heat” by “radiation transport” etc.

The hard part used to be distance / accuracy untill fairly recently.

Now high power lasers used for welding and 3D cutting have solved that problem…

You just shine them on the temperature sensor casing and it’s internal temperature follows fairly quickly…

With regards,

“… the difficult part would be limiting the temperature jump to something realistic!”

Actually not as hard as you think have a look at Laser/IR Thermometers and FLIR cameras you can now get as add ons for mobile phones and Single Board Computers like Raspberry Pi’s.

All you do is “pulse modulate” the drive to the high power laser and use the IR / FLIR temperature sensor in a feed-back loop that controls the On to Off period of the high power laser.

If you don’t want to risk high power lasers, you can make a HERF (High Energy Radio Frequency) Gun using the magnetron out of a microwave oven to likewise generate heat “as a waste product” at the temperature sensor.

Though personally I’d give any High Energy RF system a very wide birth as they can be quite dangerous for a whole heap of reasons.

Clive Robinson • April 25, 2026 10:58 PM

@ ALL,

A few days ago I mentioned that there had been civil disturbance in a Market town in Surrey UK, a little ways south of South West London.

It started because of an “alleged” attack that Surrey Police had “asked for witnesses” got pushed by social media algorithms, and the unfortunate problem that UK seniors in politics and policing have pushed certain agendas onto the Police that were at best inadvisable and had in effect been created by demand from earlier media reporting that had given rise to other significant social unrest.

The UK BBC has released an update which indicates the alleged attack on a young woman that was “under police investigation”, that the police had “asked the public for witnesses” for as a “standard procedure”, the Police have not so far found any evidence for the attack,

https://www.bbc.co.uk/news/articles/c9370jqxy18o

What the incident does clearly show however is that there is an issue with Social Media Algorithms that “go viral world wide” and a growing degree of civil unrest that can turn violent, being stirred up by the algorithms. With a “Damned if you do Damned if you don’t” dilemma arising for civil authorities, looking for witnesses etc that in the long term is most likely to result in the withholding of any disclosure that in turn will hamper the open / public justice process.

Clive Robinson • April 26, 2026 6:21 AM

@ ALL,

A short while back I said that the US trying to keep Chinese AI development constrained by locking down on Nvidia hardware was going to fail in the near future due to the collaborative way the Chinese are developing their AI, unlike US AI Orgs.

Well when I said “near future” I did not think it would be such a short time to happen and get reported…

DeepSeek’s new models are so efficient they’ll run on a toaster … by which we mean Huawei’s NPUs

Now available in preview, DeepSeek V4 cuts inference costs to a fraction of R1

“Chinese AI darling DeepSeek is back with a new open weights large language model that promises performance to rival the best proprietary American LLMs. Perhaps more importantly, it claims to dramatically reduce inference costs and it extends support for Huawei’s Ascend family of AI accelerators.

Unveiled on Friday, DeepSeek V4 is available for download on popular model repos like Hugging Face, the company’s API, and web service in two new flavors. The first is a smaller 284 billion parameter Flash mixture-of-experts (MoE) model with 13 billion active parameters, while the larger of the two is a 1.6 trillion parameter model, 49 billion of which are in use at any given moment.

V4-Pro was trained on 33 trillion tokens and, if DeepSeek is to be believed, beats out every open weight LLM while rivaling the West’s best proprietary models across its benchmark suite.“

https://www.theregister.com/2026/04/24/deepseek_v4/

Obviously we will have to wait a while for confirming testing to be done.

But if true the center of US AI real money Nvidia will take a stock market hit which will ripple out…

Not just through the over hyped and incestuously round and round self investing hang offs like Oracle, but down through those supplying third line and beyond goods and services like building companies that would build Data Centers etc if “real money” rather than “investment speculation” for loans was there.

It will be also interesting to see how US Executive Politicians spin this as “a war won” when in fact it is another “war lost by capitalism” that has caused major damage to the US GDP and economy in general.