Friday Squid Blogging: Pet Squid Simulation

From Hackaday.com, this is a neural network simulation of a pet squid.

Autonomous Behavior:

  • The squid moves autonomously, making decisions based on his current state (hunger, sleepiness, etc.).
  • Implements a vision cone for food detection, simulating realistic foraging behavior.
  • Neural network can make decisions and form associations.
  • Weights are analysed, tweaked and trained by Hebbian learning algorithm.
  • Experiences from short-term and long-term memory can influence decision-making.
  • Squid can create new neurons in response to his environment (Neurogenesis)

As usual, you can also use this squid post to talk about the security stories in the news that I haven’t covered.

Tags:

Posted on May 16, 2025 at 5:05 PM • 25 Comments

Comments

lurker • May 16, 2025 6:01 PM

NZ Police are giving up their speed cameras. The function is being taken over by New Zealand Transport Agency, but the actual cameras “are operated by an external company, Acusensus,” who Wikipedia describe as “a technology company.”

‘https://www.drivencarguide.co.nz/news/nzta-starts-takeover-of-mobile-speed-cameras-this-week/

not important • May 16, 2025 6:10 PM

Starlink: A Satellite system shaping modern warfare
https://www.dw.com/en/starlink-elon-musk-russias-war-in-ukraine-v2/video-72565310

=Starlink provides Ukraine with high-speed internet—vital for communication and drone

operations. But control over the system lies with Elon Musk. How risky is that?=

Autonomous weapons: Machines decide over life and death
https://www.dw.com/en/autonomous-weapons-machines-decide-over-life-and-death/video-72565302

=Drones that attack on their own, robotic vehicles in combat—autonomous weapons are taking on more tasks. For now, humans have the final say. But how much longer will that be the case?=

Two short and good videos related to blog.

snickering snickers bar • May 17, 2025 1:47 AM

Introducing oniux: Kernel-level Tor isolation for any Linux app

https://blog.torproject.org/introducing-oniux-tor-isolation-using-linux-namespaces/

When launching privacy-critical apps and services, developers want to make sure that every packet really only goes through Tor. One mistyped proxy setting–or a single system-call outside the SOCKS wrapper–and your data is suddenly on the line.

That’s why today, we are excited to introduce oniux: a small command-line utility providing Tor network isolation for third-party applications using Linux namespaces. Built on Arti, and onionmasq, oniux drop-ships any Linux program into its own network namespace to route it through Tor and strips away the potential for data leaks. If your work, activism, or research demands rock-solid traffic isolation, oniux delivers it.

ResearcherZero • May 17, 2025 4:23 AM

Transferring wealth from poor to rich requires sacrifices to be made.

‘https://abcnews.go.com/Business/us-credit-downgrade-means-economy-finances/story?id=101946875

The additional spending and tax cuts for billionaires will add $7 trillion in debt.
https://apnews.com/article/republicans-trump-bill-tax-cuts-spending-medicaid-a9f9c0a23cd2a6e7917d5b73f935e1fd

  • $3.8 trillion for renewing previous tax cuts
  • $1.5 trillion for other unfunded tax changes
  • $1 trillion in associated interest costs
  • $500 billion in new spending programs.

https://www.nytimes.com/2025/05/15/us/politics/house-gop-tax-bill-trump.html

not important • May 17, 2025 6:47 PM

Pathology Without a Person
https://www.psychologytoday.com/us/blog/the-digital-self/202505/pathology-without-a-person

=Where a narcissist once relied on a small social circle to validate their grandiosity, today they can perform to millions—with algorithms rewarding exaggeration and spectacle. Where a Machiavellian manipulator once needed face-to-face persuasion, now they can orchestrate deception behind screens, cloaked in anonymity. And for those with sadistic tendencies, cruelty is no longer hidden—it’s performative, often celebrated, and even viral.

The platforms flatten social hierarchies, remove interpersonal friction, and offer asymmetric power to those who might otherwise remain on the psychological margins. The cost of entry is low. The rewards—attention, influence, even monetization—are high. And in this way, digital spaces have lowered the friction for antisocial expression while often raising the reach. The very traits we diagnose in private can now thrive in public.

It’s the dark undercurrent of digital life. Studies link trolling to sadistic traits, but online, cruelty isn’t just tolerated; it’s content. Hate goes viral. Platforms don’t just permit this, they profit from it. Worse still, the culprits often hide behind playful or antagonistic aliases.

We’re now witnessing human darkness encoded and amplified. AI can gaslight, deceive, or churn out disinformation with no intent, no guilt, and no soul. That’s what makes it so dangerous. You can’t argue with code or guilt-trip an algorithm.

Here’s the shift: We’re no longer just facing malicious people. We’re grappling with systems that mimic malice. Not because they want to; they don’t want anything. This complexity is built on simplicity: They predict, they generate, and they engage.=

not important • May 19, 2025 6:11 PM

The Psychology of DIY Biohacking
https://www.psychologytoday.com/us/blog/the-future-of-intimacy/202505/the-psychology-of-
diy-biohacking

=Real incidents reveal unsettling risks, from failed self-experiments to biosecurity threats.
Trust of our neighbors is increasingly required to our shared human experience.

Without institutional review boards or regulatory bodies, there’s little accountability, transparency, or quality control in many DIY bio labs. This creates vulnerabilities not only for individual experimenters but also for communities and our environment.

The FBI and others have expressed concern about the potential for misuse, including the accidental creation of harmful organisms or the deliberate engineering of biological threats.

Real incidents have already exposed these risks. The psychological reality is unsettling: we must trust that others will act wisely, ethically, and safely with technology that many of us do not understand, and that could ultimately hurt us all.

When a biohacker livestreams an untested gene therapy or a company like Ascendance Biomedical distributes DIY treatments without oversight, they’re not just risking their own
health. They’re gambling with the safety of their communities, and, potentially, the world.

How many people in your network do you trust with this power?=

NONE

Clive Robinson • May 20, 2025 1:23 AM

@ not important, ALL,

With regards,

“… or a company like Ascendance Biomedical distributes DIY treatments without oversight…”

The thing everyone tends to forget is,

“What comes with the treatment?”

Less than half a decade ago there was a scare to do with fluorocarbon chemicals. They come in chains some like PTFE come in very long chains, so even if we ingested them they would pass through without being absorbed. However they also come in short chains and these do get absorbed into the body where they cause all sorts of problems and you might have heard them called “forever chemicals”.

Whilst the short chain chemicals are being banned or strongly regulated, the long chain are not or not as much as they are seen as less of a risk…

Now ask yourself a question,

“How does the DIY treatment get into your body?”

Well the chances are it will have to be injected. But… the treatment will have been put in a container/ receptacle to be delivered to you…

So ask yourself if the receptacle is “food grade” or “medical grade” as it’s important.

Because Food Grade is quite a bit more expensive than Commercial Grade, but Medical Grade is going to be many times as much again. The price difference is due to the cleaning and sterilisation process,

1, Medical Grade, is made such that “micro plastics” and “forever chemicals” are not present thus cannot contaminate the treatment.

2, Food Grade, however the long chain forever chemicals may well be present.

3, Commercial Grade, will very probably have both in abundance…

So the treatment from an unregulated supplier may will be packaged in food or commercial grade containers. And thus be contaminated with the chemicals that even in one trillionth of a part / liter will be sufficient to cause you harm medically. Such as increasing the likelihood of aggressive cancers if the treatment has to be repeated every day etc.

But… a certain person in the US is tearing up all this sort of safety regulation as he considers it to be an unwarranted burden on industry…

Bob Paddock • May 20, 2025 8:07 AM

@not important

Dave Asprey is considered the grandfather of Biohacking.
His conference is coming up next week.

‘https://biohackingconference.com/2025

The Medical Establishment poisons thousands of people a day with Gadolinium contrast dyes, Fluoroquinolones like Avalox, Cipro, Levaquin and many other things.

Before getting down on the Alternatives, that no one is being force to participate in, we need to fix the Medical Establishment that is not giving people the information they need to know to establish fully informed consent, and at times are forcing people to do it Their Way.

Clive Robinson • May 20, 2025 1:41 PM

@ ALL,

More AI does not deliver info

I’m very very sceptical about claims based on current AI LLM and ML systems that they are actually capable of reasoning or doing things better than other methods.

One such article is,

https://www.understandingai.org/p/i-got-fooled-by-ai-for-science-hypeheres

I feel sorry for the author as they effectively fell into the “AI Hype Swamp” that is starting to give “fake news” a good name.

But also on the “you heard it here first, or before” is this article,

https://theahura.substack.com/p/deep-learning-is-applied-topology

The notion is the same as I have given before. Only I view LLM and ML systems as “adaptive filters” used to map onto a multidimensional spectrum that is actually logically the same as a topological manifold.

I also pointed out that such adaptive filters actually work at many layers which the article author also points out with,

“The data all lives on a high-dimensional, semantically relevant manifold. And developing the manifold is exactly equivalent to figuring out how to represent the dataset semantically.”

Yup it sounds like circular reasoning, but think it through.

Where I disagree with the author is

“I’m personally pretty convinced that, in a high enough dimensional space, this is indistinguishable from reasoning.”

I’m most certainly not convinced about this notion, in fact I think the opposite is more likely.

The manifold is formed from existing semantic data in the input data by what is the ML system.

It’s not reasoning it’s just picking out statistical patterns that exist in the data. That is in no way “reasoning” as I understand it.

You might differ in your view point but consider that the reality is that any reasoning present is actually already in the input data, not anything the current AI ML systems do.

If you doubt this go back and read the first article a little more carefully.

lurker • May 20, 2025 2:45 PM

@Clive Robinson

“I’m personally pretty convinced that, in a high enough dimensional space, this is indistinguishable from reasoning.”

Note that “this” refers to:
“developing the manifold” ≡ “figuring out how to represent the dataset semantically”

So, how high is enough dimensional space? Then when you’ve figured that out, it becomes indistinguishable from reasoning, in the same way that any sufficiently advanced technology becomes indistinguishable from magic.

There’s lies, damned lies, and semantics.

ResearcherZero • May 21, 2025 2:01 AM

@Clive Robinson, lurker

British hospitals state that the products Palantir provided through a process that was without competition, is not delivering benefits and in some cases, reduced functionality.

Many hospitals do not even use the product as it does not meet their needs or the abilities of existing products and lacks compatibility with other solutions already in use.

‘https://www.theregister.com/2025/05/16/nhs_hospitals_palantir/

ResearcherZero • May 21, 2025 2:07 AM

One of the problems with AI software products is that they cannot meet the physical needs of the public and lack the mobility to respond to very real crisis and disaster, while absorbing funding that would otherwise go to real people possessing real-world skills.

Communities still waiting for federal help months after tornadoes devastated the area.

‘https://apnews.com/article/fema-delay-tornado-disaster-mississippi-tylertown-00c644598b4f4693c116b9eb5eae3bae

With a third of its staff gone, FEMA lacks the resources for an adequate response.
The Oval Office lacks the political will to ensure that aid and assistance is delivered.
https://www.cnn.com/2025/05/15/politics/fema-review-concerns-hurricane-season

Missouri residents are still living in destroyed homes without assistance.
https://www.usatoday.com/story/news/nation/2025/05/20/fema-not-helping-after-st-louis-tornadoes/83746010007/

ResearcherZero • May 21, 2025 2:14 AM

Regions in America no longer have night forecasters to monitor for deadly storms.

‘https://www.washingtonpost.com/weather/2025/05/16/weather-service-offices-overnight-cuts-map/

“about 52 of 122 weather forecast offices have staffing shortages above 20%.”
https://www.usatoday.com/story/news/nation/2025/05/17/kentucky-national-weather-service-forecast-office-staff-cuts/83693800007/

With a third of jobs gone, the Weather Service does not have enough staff for repairs.
https://www.cbsnews.com/chicago/news/national-weather-service-staffing-shortages-making-forecasts-worse/

ResearcherZero • May 21, 2025 3:08 AM

In around 60 years or so, there may be a few interesting releases, given some have still kept up the antiquated art of reading and records are maintained in a legible condition.

A political appointee tried to rewrite an intelligence assessment to change its meaning.

‘https://www.nytimes.com/2025/05/20/us/politics/gabbard-intelligence-venezuelans-tren-de-aragua-trump.html

“Order will endanger intelligence sources in the field”

https://www.justsecurity.org/37332/bipartisan-group-senior-officials-ninth-circuit-immigration-order-harms-furthers-national-security/

Military leaders and instructors are being put in very difficult position.
https://www.nytimes.com/2025/05/08/us/politics/west-point-hegseth-culture-wars.html

Clive Robinson • May 21, 2025 9:11 AM

@ lurker, ALL,

The art of sculpting the surface…

You note,

<

blockquote>’Note that “this” refers to:
“developing the manifold” ≡ “figuring out how to represent the dataset semantically”‘

<

blockquote>

That is like “sculpting the surface” of a statue or robot etc, it is designed to effect the observers point of view rather than add actual utility of the object/entity.

Thus once designed or finished it is “static” not dynamic in function, and as such is a view of the past, not the present, or future.

The point of reasoning is to effect the future, not just of the observer but most important the entity that is reasoning. Because reasoning is an iterative process that requires looking forward into the unknown and testing what you think you might see for utility.

We know Current AI LLM can not do this, nor can Current AI ML systems evolve the manifold in a way that will produce such processes to get reasoning, just “pattern finding” by statistical processes which average might average noise off of a curve.

Over simply the ML “measures the difference” between what the LLM outputs and what the training corpus indicates. In effect it slides down an error gradient like water running down a hill, untill the error tangent is effectively as flat as possible where the water pools. The major problems with that is,

Firstly, that minimum where the water pools may be a ditch high on the hillside rather than at the bottom of the valley.

Secondly, the error function might effect the manifold shape, but it does not change the input training corpus content.

The second point indicates why the Current AI LLM and ML systems can not “reason” either in the present or future, only provide an error based finding on the past.

It also explains why the Current AI LLM can easily be “stuck in error” due to the first issue.

The problem is we currently have insufficient knowledge as to how we might fix the first problem, because we cannot examine the LLM weights in a meaningful way. Thus we have no meaningful way to test the LLM, so only find errors that we find “blow up in our face” when we use it.

Will this issue ever change, probably not, in fact it’s very unlikely. Look on it as having to perform a brut force search on the entirety of not just meaningful information we are aware of in the universe but also information we can not yet know because we do not yet know how to find it.

It takes us back to pre 1980’s AI and “Expert Systems” which can only work on,

1, Information held in
2, Existing established rules.

Thus we drop into U.C. Berkeley philosopher John Searle’s “Chinese Room argument” he formulated in the 1970’s and published in 1980.

Whilst Searle’s argument was to show understanding of language was unnecessary to pass the 1950’s proposed “Turing test” thus the Turing test was flawed. It actually did more than that,

It established that any knowledge inside the room was in effect static as in a library or read only database and could only be accessed by following not just the basic access rules… But more importantly also the rules relating to the knowledge it’s self thus embedded in the knowledge.

If the “in knowledge rules” were not present then that aspect of the knowledge could not be utilized or even found.

When you consider actual reasoning, it is about finding not just patterns in input data, or finding
a formula to encode the patterns as rules, but to do so only where the rules will have actual meaning “in the environment” and what the application of such meaning will have.

We know current AI LLM and ML systems have no notion of “environment”. Nor do they have “agency” by which they might measure the environment. But even those would be woefully insufficient for such a rules based system to gain any kind of understanding of the environment immediately around it let alone more generally.

So Current AI ML systems can not even find a gradient curve to measure error by. By which they could judge if a pattern has any utility or not or to re-tune the weights of the LLM network in a way that increases utility.

So as they used to say when I was wearing the green,

“No matter how high you pile it, it is going to be of no bl@@dy use to man nor beast.”

Now if you could find a large enough lump of rock to carve that on, you could use it as a memorial stone to Current AI LLM and ML systems as far as “reasoning and thus General Intelligence” goes.

ResearcherZero • May 21, 2025 9:57 PM

All foreign nationals will require monitoring app on entry to Russia.

‘https://novayagazeta.eu/articles/2025/05/20/russian-parliament-passes-bill-to-track-migrants-via-mobile-app-en-news

Originally introduced as an app that migrant workers would have to download.
https://www.hrw.org/news/2020/06/09/russias-latest-app-will-track-migrant-workers-whos-next

The FSB can access users’ audio, text, and video communications without a court order.
Russian law enforcement can access citizens’ mobile location data without a court order.
https://www.reuters.com/article/technology/russia-revives-draft-law-to-grant-police-access-to-citizens-geolocation-data-idUSKBN2A41ZH/

ResearcherZero • May 21, 2025 10:31 PM

The private surveillance industry monitors citizens for governments.

‘https://thebulletin.org/2025/05/trumps-immigration-crackdown-is-built-on-ai-surveillance-and-disregard-for-due-process/

A $30 million no-bid contract awarded to Palantir for ImmigrationOS may come with a catch. The new ImmigrationOS will combine data from several government and private databases.

Palantir already has contracts worth $88 million with ICE.
https://www.usatoday.com/story/news/politics/2025/05/02/ice-deportation-tracking-palantir-thiel/83375538007/

IRS data and sensitive private information is supposed to stay confidential. Not anymore.
https://edition.cnn.com/2025/04/08/politics/irs-dhs-sign-data-deal-undocumented-immigrants/index.html

[push notification] “Please provide proof of your location via the app.”

ICE conducts immigration status checks using advanced AI databases, surveillance technologies, biometric data, and inter-agency collaborations.

The SmartLINK app provides real-time monitoring of people.
https://www.cbsnews.com/news/does-ices-smartlink-app-work/

Clive Robinson • May 21, 2025 11:27 PM

@ ResearchZero, ALL,

You note,

“All foreign nationals will require monitoring app on entry to Russia.”

As opposed to the West, where all that functionality and more comes built into the OS and lower behind faux NoBUs backdoors.

I guess the sixty four thousand dollar question is,

“What happens to those who turn up without a phone the app can be loaded on?”

If you remember back it first came up seriously back in 2020 onwards with COVID lockdown “contact tracing” and the “supposed” “Digital health Passports” and similar.

If I recall correctly both Apple and Google built in all the base functionality for BLE Beaconing and tracking into the lowest layers of their OS’s and below, which kind of works the same way as those “tracking tags” to find your lost keys etc… And it’s all still there today…

ResearcherZero • May 22, 2025 5:32 AM

@Clive Robinson

They might give you a phone or show you where to buy one of the approved models. If you have a burner then they will track you anyway, but it will be a hassle and annoy them. They may put you on a list of persons that would be useful for prisoner swaps and if nothing presenting an opportunity turns up on your visit, nab you when you are about to depart.

They are always looking for pliable people who can be quietly approached and coerced.

Otherwise a certain witness is available to point you out, or secret something in your luggage – so you can be paraded for the cameras as the individual who did X crimes.

–

FBI team that investigates corruption by officials or Congress is disbanded.

‘https://www.independent.co.uk/news/fbi-washington-congress-eric-adams-trump-b2752015.html

Bribes allow foreign governments to secretly obtain highly restricted material.
https://www.reuters.com/article/us-usa-saudi-nuclear/u-s-approved-secret-nuclear-power-work-for-saudi-arabia-idUSKCN1R82MG/

Pledging fealty will absolve you of bribery, soliciting donations from foreign nationals, obstruction, witness tampering, wire fraud or doing “odd things” for people for money.💲🐽
hhttps://www.nytimes.com/2025/05/10/nyregion/adams-corruption-investigation-records.html

Corruption is official policy after DoJ stopped enforcing Foreign Corrupt Practices Act.
https://hls.harvard.edu/today/four-harvard-law-experts-examine-the-trump-administrations-policies-on-tariffs-international-business/

Clive Robinson • May 22, 2025 10:45 AM

Depth of brain increases efficiency and will probably effect AI development

As most of us know AI is very power hungry in part due to inefficiency, especially when it comes to heat issues.

Various organs in the body such as the brain and lungs that need a large surface area use “folding techniques” that give not only larger equivalent area in a small volume it also reduces the length and increases the number of interconnects.

Recent research shows this has a useful effect in human brains,

https://medicalxpress.com/news/2025-05-groovy-brains-efficient.html

Whilst “silicon based electronics” and “carbon based life” superficially are very alien to each other in reality they have many of the same issues imposed by the basic laws of nature and mathematics.

Some may be aware that the next step in electronics manufacturing has started and it will have significant advantages and disadvantages.

Currently people regard “Systems On a Chip”(SoC) as being the best way to get efficiency out of electronics and chip die area as well as increased speed and reduced heat.

This is because SoC design removes much of the issues with “bonding out” signals such as protection circuits, amplifier/drivers and track lengths very comparable to the wavelength of the clock rate.

The problem with SoCs is to be cost effective they need very high volume sales which means they have to be very generic in nature.

In part this is why “Field Programmable Gate Arrays”(FPGAs) have a significant market presence.

Well the next step is to in effect modularise Functions onto their own individual chips, and then bond these via “flip chip” like techniques to a substrate that has the inter-modula interconnection tracking. The name being used for this idea is “chiplets”.

What is not immediately obvious untill mentioned is that with a little care you can easily move from 2D layouts that limit standard chips to 3D assemblies. Where not just modules can be stacked vertically so can inter-modula tracking at the edges of the modules you effectively form folds and much shorter interconnects.

Expect to see this technology to evolve and become mainstream within the next decade.

ResearcherZero • May 23, 2025 2:16 AM

Is it the Russians? Or perhaps something much more mundane and ordinary?

‘https://www.vanityfair.com/news/story/inside-new-security-threats-in-the-skies

Something much more mundane and ordinary.
https://onemileatatime.com/news/flights-false-collision-alerts-dca-airport/

–

Abandoned legitimate domains with dangling DNS CNAME records abused to conduct fraud.

‘https://blogs.infoblox.com/threat-intelligence/cloudy-with-a-chance-of-hijacking-forgotten-dns-records-enable-scam-actor/

Users are tricked to allow notifications which are pushed to the device.
https://blogs.infoblox.com/threat-intelligence/pushed-down-the-rabbit-hole/

Leave a comment

Blog moderation policy

Login

Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via https://michelf.ca/projects/php-markdown/extra/

Sidebar photo of Bruce Schneier by Joe MacInnis.