Schneier on Security

Clive Robinson • October 4, 2025 11:55 AM

@ din’t doo nuffin!, ALL,

With regards,

“Covert Eavesdropping through Computer Mice”

This is not unexpected.

Because the sensors in the mouse work on rotary movement in the mouse, of ever increasing fine granularity.

And at some degree they will respond to any form of vibration.

In effect they provide a single bit response to the stimulus signal at or near the “zero crossing point”(ZCP). Each sensor provides a time marker DeltaT at the ZCP either for the DeltaX or DeltaY movement.

If the DeltaTs are sufficiently coincident then you can get both Phase and Amplitude information for each sample at DeltaT.

If you bung this info into a DFT, DWT, FFT or FWT algorithm or similar then you can get some of the acoustic information from the samples.

The thing about human speach is nearly all the information is in the “signal envelope” not the “signal frequency”. The signal frequency and phase however alows a narowband “tracking filter” to pull the desired narrowband information signal from the wide band audio spectrum.

Getting 20-50 Db of gain due to the filter bandwidth multiple has been used in the past to pull faint speach out of loud background noise sources.

Importantly if the noise sources are predictable like fan noise or even music from a radio etc. Then an “inverse signal” can be created to null it out.

At some point you have got about as much out as you can from basic spectral processing.

In the past I’ve explained that the difference between an LLM Digital Neural Network and a “Digital Signal Processing”(DSP) network is very minimal in function.

What corresponds as the difference is what is effectively “the spectrum” that is being looked for.

It’s multidimensional and “tokens” represent a much more select form of frequency filter in those dimensions.

As I’ve indicated before the spectrum response curve is beyond most peoples understanding but can be indicated as multiple phonetic and syntactic spectrums.

The result is that the AI can be seen as “speach predictive” in much the same way as “predictive texting” algorithms select their next word selection down to very few words.

These selections can set a “matched filter” to have a high degree of word selectivity.

If the LLM ML trains on a specific user it can fairly quickly tell a lot more about the user than just what they may be saying. That is “voice stress” can pick out all sorts of information about an individuals physical state.

Also it takes very little information to come up with settings to create a fake voice that most humans would consider to be a given person.

It’s why I advise nothing Personal, Financial, Emotional, healthcare, or social should be sent by any form of “electronic communications”.

Further that the fact you have this policy in place should be sent to any other parties and that they should acknowledge this.

Since a court case a number of years ago where a Government Agency falsely claimed I had spoken to them on the phone and they had “transcribed and deleted the voice recording” I was lucky enough to be able to show that via written contemporaneous correspondence I did not have a phone at that time and that the number they claimed had belonged to someone else not me (and I was able to find out who). And more importantly the written record they had taken part in actually contradicted their supposed transcription…

Now unfortunately if they have your voice recorded AI can reverse the process and turn a phony/fake supposed transcription into a “voice recording”. Worse the AI will be able to add “background noise” such as a radio station and mains hum as fake “time verification”…

This is a problem that people are going to have to face increasingly in the near future. Worse AI Voice Faking does not require significant resources, so can increasingly be done by people with “SoHo AI setups”…

Clive Robinson • October 5, 2025 3:31 AM

@ Winter,

Whilst,

“… a falling tide lowers all boats.”

It was pointed out to me when I was quite young and had got into the repair, building, and design of boats, that the tide was gentle and regular and the least of your worries…

It was the wind[1] that you should worry about.

Because it made the waves that rocked a boat, and broke across it’s decks and pressed it down some times to break deep down. Likewise the gales and tempests that broke a boat asunder and draged anchors and drove a ship onto lee shores. Then there were the cyclones that threw entire ships as easily as a child does a toy, and a sensible man like Admiral Beaufort decided were beyond the whit of man thus left off his wind scale[2].

I mentioned a couple of weeks back the old traditional warning of “Being beware of a bitter wind from the east”, and a version found in one of the later Sherlock Holmes books,

Sherlock Holmes:
Good old Watson. The one fixed point in the changing age. There’s an east wind coming all the same, such a wind as never blew on England yet. It will be cold and bitter, Watson. And a good many of us may wither before its blast.

(The short story it was from was based on Sir Arthur Conan Doyle’s experience of “The Great War” that later became known as World War I when the madness behind it rose again a quarter century later).

Thus what is coming is I think an “East Wind” in more ways than one.

I guess I’m going to have to fondle the seaweed[3] more frequently 😉

[1] Sailing has so many words and phrases that are all to do with the way a boat is effected by the wind it can take quite a while just to learn them, but understanding them can take a life time. And creaky as I am and my racing days long behind me, when I get the chance, like Einstein, I like to learn some more.

[2] Hydrographer Francis Beaufort devised his scale in 1805 based in part on a system developed by Daniel Defoe a century earlier. The wind scale is in effect a power scale of wind force, based on “observable effects” (there was no real way to measure air speed reliably back 200 years ago). He took it his scale from 0-12 as being the maximum that was realistically survivable. Since then ships of steel can survive Wind Force 12 so it got informally extended to 18 levels (0-17).

https://skybrary.aero/articles/beaufort-wind-force-scale

[3] There is an old belief that seaweed can predict if it is going to rain… Surprisingly to many these days it can work
https://www.weatherandradar.co.uk/weather-news/can-seaweed-be-used-to-predict-rain–2827cdb0-2b34-43a2-b94f-0c51cce05a6b

Because it sort of works as cross between a dew plate and sorber used for atmospheric water harvesting,

https://engineering.purdue.edu/ME/News/2022/atmospheric-water-harvesting-can-we-get-water-out-of-thin-air

Clive Robinson • October 5, 2025 12:33 PM

@ , ALL,

With regards,

“[Flock is] rolling out a new and troubling product that may create headaches for the cities that adopt it: detection of “human distress” via audio. As part of their suite of technologies, “

Yes it probably falls foul of “wiretap legislation” and similar. Such legislation is often the reason “CCTV Systems” although easily fitted with microphones directional or not, very rarely have them fitted in quite a few jurisdictions.

However as some know “High definition” and/or “High Frame rate” cameras can use any reflective object as a “Sampled Microphone”.

The only thing missing from making them a close analogue to “Laser Mics” is the coherent light source. The thing is that “Digital Signal Processing”(DSP) that implements “matched filters” and is what,

“Current AI LLM and ML Systems” are in reality”.

Has reduced the need for coherent light sources these days as the usage scope has gone up from just the “frequency domain” into various “Syntactic Domains”. So Just any ambient light source that has low “in band noise” is sufficient.

So daylight works, Even florescent and LED striplights that basically flicker at twice “Mains Hum” frequencies work with slightly more complex DSP to recover the “speech envelope”.

But even Current AI LLM and ML systems, can do this sufficiently to not just recover much of the speech envelop, but also go on to “synthesise clean voice signals” from maybe 30 seconds or so of an individual speaking at an entirely different time.

Great for “fitting up” people but not as “forensic” or other evidence… for which LLMs and ML are actively starting to ring the “death knell”.

Going on to getting the faux voice to sound like it was “original recording” with background noise is only marginally harder.

Whilst such technology was just theoretical a couple or three years ago and very expensive at the begining of this year… Even today it’s now available “privately” (think of such systems as like high end computer / smart phone vulnerabilities that the likes of some Israeli, Italian surveillence for rent companies)

Unfortunately such “deep fake systems” as they are very lucrative are happening rather rapidly. In the very near future some student or academic will “Open Source” the guts of such “Deep Fake Systems”. With one result being the rapidly diminishing development cost that wiil knock on into the “price for hire” fairly rapidly diminishing from there on in[1].

In a decade or less evidence that was once considered “gold standard” like CCTV and Wiretap will get significantly challenged in courts and at some point “gold standard” drops to at best “circumstantial” then,

“Not meeting the burden of proof for use in criminal courts.”

But a more mundane aspect nobody is talking about,

“Technology makes monkeys of future practitioners in a domain”.

I’ve been warning about this for some years here with the likes of Palantir. Their business plan is akin to that of a “Drug dealer outside a school” and perhaps a great deal worse.

It’s basically,

“Provide for free, or very cheap, get them hooked, then ramp up the price more and more. To the point only a criminal can feed their habits”.

We are seeing a new variant of this business plan in the drug trade with “Vaps and Spice” or similar illegal highs as adulterants that are more addictive than “Skunk” and similar,

https://www.independent.co.uk/life-style/health-and-families/vapes-spice-drug-vaping-effect-b2598985.html

What Palantir is doing is dropping their systems into cash starved police departments. To in effect replace investigators and other detectives that tend to have high pay due to having built up long experience. As detectives “age out” or “move out” to other jobs the police departments are not “moving up” or “replacement hiring” new investigators and detectives. The result is that the police department “dumbs down” and thus more and more reliant on Palantir to “make the numbers” the politicians demand.

But Palantir have a “side gig” they get the police departments to get the beat cops to type their reports into Palantir’s systems. This gives effectively for free Palantir a product they can “rewrap and sell on” to other Agencies at quite high prices.

But it has a side effect for the police department. It has the effect of not just taking cops off the street, it also dumbs them down.

The overall result is a rapid “deskilling of the police”. A known side effect of this is “Canteen Culture” and worse where at first racism and then brutality become more and more common in the frontline forces. Worse it does not matter the intentions of “new hires” they will fall in line or get alienated and have a way way higher risk of “Death on the job” as they won’t get support from other frontline officers, and their seniors will “send them in first” in dangerous situations effectively as cannon fodder…

We already see these things starting to “play out” and it will only get worse as more organised criminals realise that “Renting Cops” on the side is in effect a “cost of business” not unlike major shopping chains regard “shop lifting”.

A look into the history of Chicago PD gives an indicator of just how institutionalised such things can become and get worse as certain less than desirable officers get “promoted up”.

Which brings us onto the technology it’s self.

As you note,

“Flock has been pushing Raven, their version of acoustic gunshot detection.”

It’s actually not very good at what it’s supposed to do. The False positive rates cause issues of “slow or no response”. Which is what we’ve come to expect with CCTV over the years.

Worse for political reasons lots of money get pushed into these systems not just themselves. But Site and backhaul Rental, Maintenance, and basic running costs.

Put simply most “wide area security” just does not pay, and thus “lost opportunity costs” arise fairly quickly along with the “dumbing down” costs.

I’ve talked on this blog in the past about the “evolution issue”. Where “static technology systems” like CCTV work initially but end up failing hard.

Because the targets of the systems that are slightly brighter than average don’t stop their activities they just move them out of the eye of the static surveillance systems. Whilst those criminals that are basically as dumb as a stump get more easily removed to prison etc. In both cases they thus cease to be an active issue in the area so street crime numbers go down (though burglary and similar “off street” crime will probably go up).

But this has an effect on the technology system, as operators of the systems in turn “dumb down” remarkably as quickly for various reasons especially “cost cutting”.

Then the smarter crooks work out how to “out evolve” the “static system” and street crime goes up again. Also the criminal type changes from individuals to what are called “feral teen gangs”.

So often the street crime is not just worse in numbers but violence and similar.

As I and others keep noting,

“Technology is not a solution to a social problem.”

But politicians are “short-termists” they don’t want lasting solutions they want people to have reasons to “vote them in again”.

Thus technologists that are probably more dishonest than the targets of wide area surveillance, know who to sell their over priced and basically useless in the long term ideas too. Thus they can soak up rapidly ramping up tax dollars in extraordinarily large amounts…

So the waste of resources is inordinate before we even consider “legal costs” and “lost opportunity costs” of such technologists committing what is “theft of common property” that ordinary citizens hard earned tax dollars represent.

[1] Consider a couple of reasons such “private AI faux evidence” generators are dropping in price.

1.1 They are actually cheaper and more secure than “other techniques”.
1.2 The use cases for Current AI LLM and ML Systems are actually really quite low due to the very limited capabilities they have outside if “Parlor Tricks” (and “dog and cart shows are only novel once or twice).
1.3 The use price per hour of these Current AI LLM and ML systems makes them way way more expensive.
1.4 The work output of current AI LLM and ML Systems is actually so riddled with problems “vibing” is a nightmare waiting to be used against those using it.

All this in most cases makes the use of Current AI,

“Not worth the waste it creates”.

As humans basic abilities in most work cases still makes humans way “more cost productive” and may well remain so for for a very long time to come (due to problems like run resources that scalling and synthetic data can not solve). So Current AI LLM and ML Systems are not going to put bread on the table, let alone a decent meal. As for the AGI dream it is still very very very unlikely to happen.

However the use of “Specialised AI” is rather different…

The parlor trick Current AI LLM and ML Systems” that people don’t realise –or want to realise– falls under at best the old warning of,

“Jack (Nave) of all trades, is master of none.”

Applies so in general they are,

“Of less use than a jobbing labourer on a finished construction site.”

The same is not true for more specialised forms of AI some of which go back more than four decades. I’ve been implementing the likes of Expert Systems with additional Fuzzy Logic off and on since the 1980’s. They are a useful tool in the box when you are developing not just complex systems but control systems and even embedded systems (the trick is learning two things, how to constrain the input scope usefully, and what order to use them in or multiple times and why).

Clive Robinson • October 5, 2025 5:37 PM

@ ALL,

What do you do when your cloud turns into a puff of smoke?

And perhaps more importantly,

1, There are no backups
2, It contains the work of 3/4 million civil servants.
3, From the more important Government departments.

As the last paragraph in the article says,

<

blockquote>“Criticism continues to build regarding the government’s data management protocols.
“

<

blockquote>

https://koreajoongangdaily.joins.com/news/2025-10-01/national/socialAffairs/NIRS-fire-destroys-governments-cloud-storage-system-no-backups-available/2412936

The country concerned is South Korea, back last century I worked for a major Korean manufacturing company “Maxon Systems International”(MSI). The thing about South Korea is they have an interesting business and ethics culture, so I can see this getting interesting.

Clive Robinson • October 6, 2025 5:51 AM

@ Another reason to keep using IRC!, ALL,

Hmm,

“Discord has revealed that one of its customer service providers has suffered a data breach.”

This sort of thing happens so often these days… That, it rarely gets reported these days as it’s nolonger considered news[1].

However using the old press maxim of,

“If it bleeds it leeds!”

Means that under certain circumstances even the mundane gets elevated to the extraordinary by it’s “newness”, “novelty”, or “shock” factor.

Or even that it is so embarrassing for,

“Persons In Social Significant And Politically Embarrassing Situations”(PISSAPES).

Which appears to be in part the case due to,

“The attackers gained access to Government-ID images, and user details.”

So of real use for “ID Theft” and worse.

And all caused by a brain dead UK “think of the children” dog whistled bit of Surveillance Legislation of the “Online Security Act”(OSA). Note how it shares OSA with the “Official Secrets Act”…

I warned that OSA would harm more people than any potential children saved[2]… For this and several other reasons. I just did not expect at just 2months it to become public quite so fast.

So I wonder how many people have been harmed by having their details stolen from this “third party service provider” negligence, and thus how many of those were from other organisations using the “third party service provider”…

What’s the betting it will be via a “lost password” or equivalent issue that allowed the attackers full access to the third party service provider systems?

It is after all one of the oldest security get arounds… How many years is it now since unauthorised systems access started plaguing the “Certificate Authority”(CA) industry? Or other OnLine industries..

Time to sing the most well said part of the chorus of Pete Seeger’s,

“Where have all the flowers gone”

[1] Which is actually a significant problem, because of the,

“Out of sight out of mind issue”

That social engineering and several other attacks kind of rely on.

[2] At this point of time I think it’s safe to say that the Online Safety Act has saved all of “NO Children” for obvious reasons.

Clive Robinson • October 7, 2025 8:16 AM

@ ResearcherZero,

With regards,

“Over there water just comes out of a tap and no-one is exactly sure where it comes from.”

In theory Londoners should know where the water in London comes from… But look up “Thames Water” in the press and you might come to the conclusion that they can not tell the truth about anything…

But few as you say think more than the turn of a tap if they even think that far. It’s like electricity and gas “you just turn the tap”.

And Electricity and the Gas to generate it are the next to infrastructure concerns, and the heat and CO/CO2 are the next environmental concerns. All so people can make their own dancing hamster / cat videos or worse for no economic benefit what so ever. After all, who in their right mind, would trust AI that hallucinates / “soft bullshits” 1/3rd of the time or worse, and thus often takes double the effort to “check” to do their homework?

But ask a question that I know very few can actually answer

“If you had to walk 5miles each way with all the fuel and water you need to live how long before you died of the effort?

Once not long ago the US CDC and EPA used to ask these questions for FEMA, the DoD and other Federal Agencies for “Disaster Preperation and Planning” (you might have noticed Disasters are getting so common in and around the Americas these days, due to Mankind, or more strictly certain kinds of Capitalist activities).

As a rough guide NASA budgets for 5kg of water and 2kg of food per person per day (but not really any less if food is dehydrated because you have to rehydrate to eat it, so what you loose on one you add to the other).

In space energy is what burns you up with solar weather so it in effect gets delivered for free. But down here 10kg of fossil fuel per person per day is considered what is needed as a minimum for military camps.

Speaking of military camps they work on 25kg of water per person…

So each person needs 17kg or 37.5lb as a minimum per day but more like 37-50kg or 82-110lb for minimal replacement of infrastructure deliverables (then of course there is the waste to get rid of…)

I know of few people that can carry 35kg on their back for a brisk five mile walk. And whilst I once was fit enough to carry 120lb for ten miles, every day, Realistically I was not fit to do anything else. Worse the stress it put on my body hit me not long after and for nearly 40years so far (there’s a reason they retire army grunts by 45years of age).

The point is few think about what the price of “convenience” actually is in real terms.

If they did, they might have different less complacent views, and give voice to them.

So to be blunt, we need the hardware and infrastructure systems that are needed for Current AI LLM&ML systems to run, in the same way we need a hole in the head. Or more accurately a volcano under our feet, forest/bush fire, hurricane/tornado, earthquake, mud slide, or other supposedly “major natural disaster” every week or so…

Clive Robinson • October 7, 2025 9:59 AM

@ ResearcherZero, ALL,

AI becomes a major Cost to major corporate name.

The cost of cheating by using AI for making an Official Report can be high when you get caught…

As the catching entity is part of the Australian Government it will be a surprise to many that firstly it was uncovered and secondly they asked for some money back.

Deloitte to pay money back to Albanese government after using AI in $440,000 report

https://www.theguardian.com/australia-news/2025/oct/06/deloitte-to-pay-money-back-to-albanese-government-after-using-ai-in-440000-report

The article goes on to say,

“Deloitte will provide a partial refund to the federal government over a $440,000 report that contained several errors, after admitting it used generative artificial intelligence to help produce it.”

Oh and that the report with only very minor non substantive changes is being published.

Which makes me wonder if the report was actually “independent” or just a way to make political mantra look like it has substance (which I suspect based on RoboDebt court cases and enquires it does not).

Clive Robinson • October 8, 2025 6:52 AM

@ ResearcherZero, ALL,

Spying ain’t what it used to be!

In the UK there was a group of a half dozen Bulgarians doing some quite nasty things and other activities “in support of a hostile foreign power”. So they were effectively “Agents of a hostile power”. They eventually got convicted.

However in the court case arguments arose as to what the requirement was to be such an agent.

It appears that the result of this has been adverse to other cases,

https://www.bbc.co.uk/news/articles/cy8rl7e7xp3o

Specifically the requirement for the UK Government to have specifically named in the right way, the country concerned to be hostile to the UK.

The problem with China and UK Politics goes back more than a decade and can be found routed in the opinions and vehaviours of the then Chancellor of the Exchequer. Best known to UK Journalists and other Politicians as,

1, Gidiot
2, White Lines

The first due to the fact his given name was Gideon but he called himself George instead.

The second due to the fact that quite a few UK MP’s had good reason he was “sniffing through a £20 bill”.

In fact one MP stood in the house of commons during Questions Time and specifically asked if Gidiot’s proposed tax on the fizz in life would include coke.

Anyway Gidiot was apparently so enamored of China there were jokes about “Postal Brides” and the Chinese leader with Gidiot featuring as the package to be delivered as it were.

Since those times we’ve had the “5G Crisis” in various forms and you can see how the then UK “political follow on’s” resisted the promptings of the Trumpetter across the puddle but when it came to the Boris and the unimaginable horror of being under him during C19 lockdown, policy finally started to go the Trumpetter’s way.

Clive Robinson • October 8, 2025 9:35 PM

Memories are made of silicon, but limited by light speed…

To see why have a read of,

https://vitalik.eth.limo/general/2025/10/05/memory13.html

“Memory access is O(N^[1/3])”

Every so often in the past I’ve explained bits of the “memory problem” issues, and I’ve put off writing it all in one go (as it gets quite long).

So somebody else has done it instead (and their goes fame, fortune, etc, etc not 😉

However one thing not covered and is quite important is “Heat Death” out side of certain types of “Static RAM” most other types of space efficient RAM are not put on CPU chips. Which if you think about it could get the C-Distance right down. And as RAM tends to be static rather than dynamic per unit of surface area the amount of heat is way way smaller.

So intermixing smaller area per bit RAM onto a CPU chip inside the likes of the ALU can with care pay some interesting dividends.

Oh the other thing not mentioned is that where speed and thus C counts is at the “edges” of pulses. Due to the fact charge does not move in the wire but around it you are in effect pumping current into a capacitor and that takes a lot of energy. But only for the time it takes for the line to change state. This has many implications including “heat”. However reducing the amount a line changes state by is reflected in both current thus heat, and speed of the rise/fall. However reducing it makes things more susceptible to powersuply and signal like noise. Keeping the area small can help reduce the noise issues.

So it all gets a little complicated 😉

Clive Robinson • October 9, 2025 5:23 AM

@ ALL,

A day or so back I mentioned that the likes of Amazon brick Internet connected devices when the MBA Mantra espoused “surveillance profit” actually shows as a deficit.

That is they realise that the design actually leads to, financial loss and path to bankruptcy…

So they “Brick the devices” or worse just “walk away” from now highly dependent customers.

The worst offending items in this regard so far are “big name, home automation devices”.

Surprisingly to many “home automation” is a very old idea[1] and predates 8bit CPUs let alone home computers and most computer networking[2].

So you would think it would be a “mature market”?

In some respects it is if it’s done properly after suitable consideration and care. It can easily rival what is seen in Industrial facilities and major office blocks.

However, it’s mostly “done on the cheap” and there are lots of “latest thing” systems that are just “bones in a desert”. Just the most recent before “WiFi/Bluetooth with everything” took over it was “Internet of Things” using a poor-mans version of “Power over Ethernet”. But before that various “mains signalling” systems that sent data in the zero-crossing time of the mains waveform.

The cost of 2.4GHz ISM band WiFi/Bluetooth “System on a Chip”(SoC) devices is now less than the cost of putting a network or other comms socket in a device and doing the required “safety testing”… So is a natural for “start up founders” and “do it on the cheap” major corporations alike. All now chasing those MBA Wonk(ies) dreams of surveillance profits that made Google and Co.

And not unexpectedly for things done on the cheap loosing money and filling landfill with the E-Waste of broken dreams of house holders with “too-much but not-enough money” that want the good life of “electronic servants”.

The latest to dash such dreams on the scrap heap of E-Waste is Logitech who last week sent out an email to say its 100 Dollar “Pop smart home buttons” are nolonger supported which it will brick in less than a week (October 15 is “kiss you’re A55 good-bye day”).

https://www.reddit.com/r/logitech/comments/1nu2u7g/logitech_pop_buttons_ejunk_in_2_weeks/

But have you considered what happens when “Mick with a back hoe” takes out your Internet for a week or more?

But this consider a little further this rising trend of “no more support or functionality” has a more sinister side…

Any one else remember this from over a couple of years back,

https://spectrum.ieee.org/bionic-eye-obsolete

I repeatedly warn about “medical implants” as the FDA used to just “look the other way” about these issues.

But also, I’ve long warned about the horror that Smart Grids with their near useless for consumers “Smart Meters”. And so it is coming to pass bit by bit little by little people through no fault of their own are getting hurt.

And guess what… It’s going to be forced on a home near you in the very near future…

Ahh the joys of,

“Power at the push of a button.”

You think it will be convenient for you, but it won’t. But others know it will be convenient for controling you and it will come at a price from your wallet…

It’s why my home does not have “HiTec wizardry” of any kind connected to it. Whilst there are some fancy lighting for the art on the walls etc, it’s all under “old fashioned” control[3].

[1] There is a crazy example of this in a “stately home” the electric lights in the rooms were turned on and off from the “generator hall” a room that looked like a “more power Egor lab” from a black and white horror movie. It was the first domestic hydro electric facility. There was a “servant” in the hall that watched “pull bells” and threw the appropriate switch.

[2] So old in fact it was a mature idea back in the 1970’s… Whilst I was still technically at school but not having done my finals… I helped a friend who had an electrical contractor business “re-wire” his house (think internship/apprenticeship for “having a trade”). It was a major job, because it had all the lights, light switches, and power sockets came back to the now vastly expanded “consumer unit” board. So rather than traditional Power “rings” that sockets with built in switches were hung off of, his house was “Star Wired” and each individual circuit switched via relays. As an internship to a lifetime trade it taught me more in a couple of months than a couple of years of schooling had done. But as later became a trend, my life moved from the “trade path” to the “professional path” and I went to college as a step toward getting a degree and have spent most of my work life “hopping professions” and helping create some new ones along the way.

[3] As I’ve mentioned before the technology stuff I do have is for “research” and for good reasons it is “isolated” not just from the house power but all outside communications in ways few have heard of.

Clive Robinson • October 9, 2025 12:39 PM

@ Bruce, ALL,

Turing test fails for Audio.

Do you think you can tell the difference between a human voice and AI generated audio?

It appears increasingly people can not,

“Humans flunk the Turing test for voices as bots get chattier

Coin toss odds for spotting a deepfake, study finds. And that’s before the machines learn to sing “

https://www.theregister.com/2025/10/09/voice_clone_detection_study/

As with all these potential “Fit for social engineering” new technologies, the real question is,

“When will criminals use it to defraud or get around security?”

We already know “Politicians are using voice based AI to avoid having to speak to voters or employ staff”.

So the up and coming US 2026 mid terms may prove interesting…

Oh does anyone know if it’s illegal for AI to impersonate / replace a politician or just stand as a candidate?

Some years ago Futurama did an episode where “Nixon’s head” got re-elected for a third term because the legislation said,

“No body could stand for a third term.”

And Nixon pointed out he was a “Head without a body” therefore the rule did not apply.

Back then it was funny, now however technology has moved on sufficiently that people are asking,

“Are there loopholes through which AI could sneak?”

With a worried edge in their voices…

After all you could as an over arching controlling mind, come up with a hundred AI ChatBots enough to get them all voted and give you a majority elected. But in reality it’s just one human having a hundred fakes under their control…

The rules were never written for this potential outcome.

Clive Robinson • October 9, 2025 5:54 PM

@ lurker,

“It’s a problem with AI: too many people are taken in by the sizzle, and fail to notice there’s no steak, only cheap sausage meat.”

Yup and a lot of people happily chow down on “textured vegetable protein” provided you burn it right… So high in fibre that you or I might have chucked on the composting heap. Or in older times dried out to make corn cob pipes with.

Such is the mantra of “leave nothing on the floor” type neo-con free market economics that get fed to MBA students to “chow down on”.

But some years ago I was a consultant to a UK University that still had the “Local Council Education Dept” mentality.

Most importantly was,

“You can not spend any money unless you can show in a report that it saves MORE money than currently being spent…”

Pointing out the obvious that the policy was going to have three basic detrimental actions,

1, Lead to endemic fraud.
2, Create a catastrophic downward spiral.
3, Create bureaucratic waste as the significant employment activity.

Was either ignored without comment, by those at mid level management or treated as,

“I’ve heard it all before, now tell me how we get income growth from within the rules.”

By senior levels of management.

So you quickly realise you only have three choices,

A, Option 1 to get work done.
B, Option 1 to get personal upgrade
C, Both A&B to fake a project to pad your CV with and get out of this hell to the next bureaucratic hell.

But hey,

“Don’t tell anyone the ‘secret to my success’ is ‘take the money and exercise option’ and ‘burn those who are left behind’…

This was the “standard business plan” under “The Prime Minister, the right honourable Anthony Blair MP”.

And will be “built into the “Political AI’s” coming to a constituency near you real soon.

But if you look carefully you will see the real “success mitigator” is being your ability to surround yourself with with “willing fall guys” to protect “the dear leader” (or “girls” in Tony’s case, who are still there following the same but more ineptly carried out policy).