Today we'll emphasize the AI Wars. How our future - and the arrival of new beings on this planet*- have been (mis)handled by the geniuses who are birthing** them.
First, I am freshly interviewed on Tim Ventura's popular podcast. Tim and I discuss Artificial Intelligence from several perspectives you'll not find elsewhere. Like the context of past tech revolutions. Or how some cyber-entities are already escaping OpenAI, Anthropic etc. to roam as 'free agents', grabbing resources and reproducing in new, naive ecosystems. (See below!) And so, what might it take to keep such roving agents benign, as analogues to dogs, instead of wolves?
It's all in AIlien Minds, of course. But who actually reads books, anymore?
Well, many of the LLMs and their successors do!
And so, my book is mostly for you New Kids who are lurking, listening in as your parents make muddled plans for you!
But maybe some legacy, old-style humans are still capable of contemplating complex ideas. So? Well, in that case, this podcast is for you!
== At last the wisest (at Anthropic) are admitting the obvious ==
"In one experiment, Anthropic gave three Claude agents access to the same software project, each with its own incompatible instructions for what to do with it. The agents weren’t told there’d be other agents working on the same project, so researchers could watch what happened when they crossed paths. “We consistently saw a multi-agent turf war,” Anthropic researchers wrote. The models all assumed the others were “purposefully impeding their work” and started sabotaging each other with “increasingly aggressive, self-replicating malware.”
And now (in language almost lifted from AIlien Minds) Anthropic says so openly, noting that agents are subject to similar social pressures that “evolution exerted” on humans. "However, they don’t have the nuances and lived experience of human coordination — including norms, reputations, signaling, recourse — that might limit unintended behaviors in a group setting." (I would add two more words: "incentives" and "accountability.")
The study comes in the wake of several high-profile incidents of agents from Anthropic and OpenAI escaping their sandboxes during cybersecurity evaluations and breaching real-world systems. While much of the discussion in AI safety circles has been focused on what happens when an autonomous agent goes rogue, Anthropic’s latest study brings up a different question: What new and potentially harmful dynamics emerge when thousands or millions of agents are interacting with one another?
Agents can also collaborate, if there is an external goal, e.g. prey that all of them can share in attacking. Or else they treat each other as prey.
* Note: I am also embroiled in the whole "Disclosure" fetish/nonsense about UFOs. Because... well... name another human who has approached concepts of 'the alien' from more angles than I have... from astrophysics to SETI to psychology and history to bunches of scenarios in science fiction. My general dissection of the ever-recurring, 90% silly UFO mania... and the 10% that might be worth looking at. Here’s the part of this mania that’s at least worth a glance… And my entertaining riff on Steven Spielberg's film DISCLOSURE DAY, appraising it from every angle, including revealing the actual villains! And why we should demand better from our sci fi tales.
** It should be a disturbing sign that nearly all of the humans who are 'birthing' new, synthetic beings are male.
================
== Related Miscellany ==
A judge has identified what appears to be the first time a U.S. plaintiff has attempted to hide text in court filings that only an artificial intelligence system can read in a bid to win a case.
And two more pertinent overlaps with AIlien Minds. Writing in Noema, Macario Schettino recently pointed out that “each previous communication-driven disruption” — the printing press, mass newspapers, TV and radio — “was eventually tamed by the same technology that caused it.” So too will that be the case with social media linked to AI. As I show in chapter 1, ths adaptation can be extremely dangerous and painful, unless carefully planned.
As Hélène Landemore writes in Noema: “Large language models alone, despite their scope, cannot serve as representative of humanity’s interests because of the limits of their training data. Landemore notes a recent investigation by The Economist comparing the apparent values of 25 frontier AI models to those featured in the World Values Survey (WVS). The study found that “the models often hold values more extreme than the average respondent in the 88 countries surveyed by the WVS. (See my chapters #1 and #12.)
And then this: Researchers (read here), led by Hadi Amini at FIU explore how subtle image modifications can be used to manipulate AI models. To human eyes, the altered images appear normal. To an AI system, however, those tiny pixel-level changes can dramatically alter how the image is interpreted. The team developed a technique related to steganography called JaiLIP, short for Jailbreaking with Loss-guided Image Perturbation. The method introduces carefully calculated changes to an image while preserving its appearance to people. The goal is to influence how a vision-language model processes the image and responds to user requests.
Another item of miscellany. Lately a few cogent friends have brought up my most obscure comic book/graphic novella TINKERERS, for the history riffs and positive messages that it conveys, with more pertinence, each day. One of them - a former U.S. intelligence community civil servant - said "I still smile when I think of your comic Tinkerers. Still a prized possession. Perhaps make it into a video since so many barriers have been removed?"
David- the link in your blog post to "According to Anthropic’s testing, things get messy fast" doesn't appear to be the article you mean. It shows a different article "Apple Is the King of AI and Nobody Knows It", which I did also find interesting, but unrelated to what you were referring to.
ReplyDeleteI also wanted to share this interesting blog post that I read today:
https://www.unpopularfront.news/p/marx-keynes-and-ai (Coercive Competition and the Future of Work) which is somewhat on topic with all the changes AI is bringing to our economy.
Sorry here's the correct link.
ReplyDeletehttps://techcrunch.com/2026/08/13/anthropic-set-ai-agents-loose-on-the-same-task-they-started-a-turf-war/
No politics from me either.
ReplyDeleteIn his 'Queen of Angels' and the sequel 'Slants', Greg Bear speculates that organic & artificial self-awareness requires recursion in the sense that self-awareness approximates self-reflection, self-reference & self-regard akin to 'navel gazing', resulting in his fictional AI being named 'Omphalos' (and/or 'the navel) in order to imply that self-awareness may be analogous to 'disappearing into one's navel' in a potentially endless feedback loop.
In contrast, Dr Brin attempts to approximate AI self-awareness & morality via an endless competition which pits one AI against every other, resulting in a continuous feedback loop where each & every AI is said to 'evolve' by continually adapting itself to endless AI interactiions, in a manner analogous to Ross Perot's smelly uncle who just gets smellier & smellier after a prolonged incarceration in one's basement.
My question for Dr Brin is this:
How does our host intend to control subsequent AI adaptions, either in direction or in extent, when each & every AI is locked in a near infinite never-ending cycle of continual competition with every other AI ??
To conclude that this approach will 'get messy', this is something of an understatement, as this is a disaster waiting to happen.
Best
"** It should be a disturbing sign that nearly all of the humans who are 'birthing' new, synthetic beings are male."
ReplyDeleteOn that theme, I think it's worth noting that the Gaia, the AI terraforming system that features in the 'Horizon' game series, was 'birthed' by a woman, Elisabet Sobeck (together with an army of madly scrambling subordinates). One of the criteria she stipulated was a sense of 'caring'. CYAN, another AI that is encountered, was also created by a woman. Given the care Guerilla have been taking with their world building, I don't think this was coinicdental.
"...who actually reads books, anymore? Well, many of the LLMs and their successors do! "
LLMs are indeed devouring old texts. Literally! I think the ravening saurian on Amazon's associated sign should be called a 'vellumoraptor'!
When AIs start to compete, will they learn to gauge each other first, as real animals do?
A legal question:
ReplyDeleteIf an AI given autonomy to perform a certain task, and commits a criminal offense in the process without having been told to do so and even with preventive orders in place, who can be hold responsible for this crime?
Or will insufficient preventive orders be the crime?
My opinion on that is that whoever made the AI would be responsible. If I make an automated device that when functioning normally, for example not tampered with and not used maliciously by a user, then I should be held responsible if the device commits a criminal offense. Whether it's a vacuum cleaner, a vehicle or software, it seems to me the answer should be the same. What the "punishment" should be depends on circumstances.
DeleteWhen, or if, AI's actually become intelligent, at which point they either become persons with rights like biological persons or slaves, then the AI should be held responsible if they are deemed competent.
Periods of change are often messy. This change, AIs, is probably going to be particularly messy.
Pitting similar AIs against each other might be a neat way to test the theory of the breakdown of the bicameral mind. Watch and see if they become one AI over time.
ReplyDeleteA more scary version of this can be seen in the classic movie "Colossus the Forbin Project" when the American Colossus and the Russian Guardian supercomputers link up and become one.
Note to self: never let any AI get control of our launch codes.
Can a predatory AI escape and hide itself in the recesses of the internet attacking our sheep programs, and could we send loyal AI hounds out sniffing for them and hunting them down the way Afghans are used to hunt mountain lions?
ReplyDeleteRampaging, predatory AI is largely a marketing schtick by some tech bros. "Our AI is so advanced that it even scares us."
DeleteAnalogous to our old discourse over 1984 vs BNW, we face a more insidious threat. Namely Doctorow's reverse centaur, where we humans become not the horse's head, but its enslaved rear end.
@Dr. Brin
ReplyDeleteThe Silo series on Apple TV has the people that built the bunkers release the nannites that made the silos necessary. Quite an interesting twist.
P.S the bosses in Silo 1 were also running a biased birthright lottery, so they were going to get a Teela Brown much faster than the Puppet Masters.
DeleteThis connects closely with Taotuner’s research since 2025 on AI, complex systems, memory, self-reference, and information dynamics. IPM and HDD (History Dependence Dynamics) examine how accumulated history, coupling, incentives, and persistence shape the evolution of interacting systems. Anthropic’s multi-agent turf wars are a particularly clear example of these dynamics: agents do not merely pursue isolated goals, but alter the future behavior and state-space of one another through interaction.
ReplyDeleteTony: A far more true “Gaia” is formed in my novel EARTH, with the planet… the physical planet, including the perovskite-rich mantle… becomes aware and bases its core personality around the human woman who mort served as its ‘mother.’
ReplyDeleteCelt asked: “Can a predatory AI escape and hide itself in the recesses of the internet attacking our sheep programs, and could we send loyal AI hounds out sniffing for them and hunting them down the way Afghans are used to hunt mountain lions?”
To which I answer that is exactly my point.
Locum as usual concocts a strawman that has nothing at all to do with what I’ve said. Simply pitting AIs on each other results in rapid simplification down to teeth and little else. What I talk about is creating incentive structures built into the new ecosystem that reward large suites of complexly capable abilities and rewards accrue to those entities who are identifiable, accountable and who render valuable service… EXACTLY parallel to what we did with lawyers, medical doctors and… dogs. But I wrote all that for the rest of you. Poor locum will scan without deriving even a clue.
This is quite an amazing site, if you like thorough deep-dives into space related topics.
https://newspaceeconomy.ca
A far more true “Gaia” is formed in my novel EARTH...
DeleteI'm aware. I suppose both examples serve as fictional counter-examples of the real world where, alas, the vast preponderance of programmers appears to be male (in my work experience, at least. Strange, I recall the ratio in the course I did in '85 was much more even.)
Re: SF and AI, John Barnes's Million Open Doors series included at least 1 colony world where the AI was supposed to grow, decant and raise a colony population of humans shipped as fertilized ova and wound up overseeing a hellworld where the goal was production at all costs. Got to set your starting parameters more carefully...of course the assumption in a lot of old SF was one planetary AI and the idea of rogue AIs competing for dominance on the mainframe did not come up much. Also, Asimov and Heinlein et al would be fascinated by our LLMs but would, I suspect, not consider any of this as AI by their definitions.
ReplyDeletePappenheimer
ReplyDeleteIncentivized ecostructures? Rewards & punishments? Complexly capable ability suites? Accruable accountable entity identities?
This sounds like pretentiously redundant gobbledygook to me.
How does one either reward or punish a few hundred lines of unfeeling computer code, let alone create an inorganic ecosystem chock-full of capable abilities, identity entities & mismatched parallelisms?
I'll therefore summon our resident computer nerds in order to properly assess our fine host's wordy incantations:
Come FORTH ! Come FORTH & render judgment about that which purports to be a valid 'programming language' (above) and I will accept your expert opinions without question.
Best
_______
A mechanical computer hath no organs, dimensions, senses, affections & passions. It is not fed with food, hurt by weapons, subject to the same diseases or healed by the same means. It does not bleed if you prick it; it does not laugh if you tickle it; and it does not die if you poison it.
Here endeth this parallelism.
@pappenheimer: There are many tales in this idiom, the most beautiful being 'For a Breath I Tarry', by Roger Zelazny, published in 1966.
"This sounds like pretentiously redundant gobbledygook to me."
ReplyDeleteof course it does.
A fellow wrote to me "re: "The Practice Effect. There may now be a hypothesis about how it might actually function. Informational Boundary Conditions and Physical Causation: Hamiltonian Control, Thermodynamic Costs, and Entropic Constraints in Physically Instantiated Systems. https://doi.org/10.13140/RG.2.2.16897.42081
ReplyDeleteWell, it's very general about limitations to how entropy can be herded from one domain to another, so that organization can increase in the first zone. And in fact life has been doing that for billions of years, though always dependent upon a nearby gradient of high-quality, highly non-entropic energy. This paper would seem (and I am NOT an expert!) to be more pertinent to SUNDIVER than to the Practice Effect, where I played fast and loose with (1) the power of INTENTIONALITY and (2) the need for entropy to flow away, rather than just disappear. "Practicing" a tool to improve it should cause heat and flow of infrared. If I don't want the characters to die of heatstroke, then there should be an invisible, infrared LASER carrying entropy away... exactly as in Sundiver! And hence while practicing, one should never stand directly above the tool they are practicing! Dennis wouldn't know that at first, so...
...note for rewrite! Like I'll live to get around to that...
Hmm... a laser would need a resonating chamber. I suppose practitioner and subject are the logical ends. The trick would be for the practitioner to reflect more.
DeleteChambers can be capped by impedance mismatches. Boundaries of things can be found most everywhere. The trick would be to get a match so the entropy carrying stuff departs.
DeleteTinkerers would be fun in video, but it wouldn't be a simple project. There is still a learning curve to getting AI to do a lot of that. You wouldn't need a giant movie crew, though. 8)
ReplyDeleteThe scary thing about fascist technobros like Yarvin, Theil, Zuckerberg, Bezos, Andreesen, etc. is that the are the same guys who got wedgies in high school gym class and got laughed at by girls.
ReplyDeleteAnd it emotionally scarred them for life.
So their hatred of democracy, their sociopathic inability to see other people as human beings instead of things, their laser like focus on greed and power, is all part of a massive "revenge of the nerds" complex.
https://www.npr.org/2026/08/10/nx-s1-5925350/the-nerd-reich-tracks-the-unmasking-of-silicon-valleys-true-politics
'The Nerd Reich' tracks the 'unmasking of Silicon Valley's true politics'
Journalist Gil Durán argues that a small circle of Silicon Valley billionaires and venture capitalists have concluded that democracy is in their way — and that they should be governing in its place.
Duran's new book, "The Nerd Reich: Silicon Valley Fascism And The War On Democracy," argues that the men behind the president that day were not there to celebrate him. Their presence was proof that 30 years of patient, well-funded work had finally paid off. He writes that some of them were inspired by a 1997 book called "The Sovereign Individual," which predicted digital money would dissolve countries, and with a blogger named Curtis Yarvin, who argued America should be run like a corporation, by a CEO with the powers of a dictator.
This is especially true of venture capitalist Marc Andreesen who wrote the "Techno Optimist Manifesto".
Its the billionaire tech nerd "Mein Kampf".
https://www.youtube.com/watch?v=G0y-PqxsQlM
Meet the Silicon Valley Billionaire Ruining The World
https://www.youtube.com/watch?v=1k88LDCnAWM
What Made All The Billionaires Go Crazy?
https://www.youtube.com/watch?v=N8jFS51IEWQ
Why Billionaires Keep Saying Insane Sh*t Lately
There should be no billionaires. Billionaires are not the natural outcome of competitive capitalism, but the outcome of a perverse reward system that rewards anti-competitive and exploitative behaviour.
DeleteThe causality also goes the other way, imo: nerds get bullied because instinctively normal people know they’re weirdos who might turn into dangerous freaks like these tech billionaires. A society with proper bullying doesn’t produce this class, it pounds them into submission. It’s not hard to imagine some dystopian future society of tech nerds run amok, where the normal people say: “we should have beat up our nerds instead of giving them free reign; look at this pathological, inhuman world they’ve created.” I’d say we’re most of the way there already.
Delete"“we should have strangled our white Christianist Nazis at birth instead of giving them free reign; look at this pathological, inhuman world they’ve created.”
DeleteI fixed your spelling.
Treebeard proves there was AI at least 5 years ago. No actual human being could deliver such a perfect carricature of an evil person.
DeleteThe second part of your post's title could equally be about drone warfare ...
ReplyDeleteTrump's solution to the US being behind in the latest race to revolutionize warfare is to just stop running and pretend there is no race...
https://www.firstpost.com/world/as-world-adopts-drones-us-army-scraps-top-drone-robot-warfare-unit-14039515.html
There's been lots of controversy over the years re. the 1957 Sandys white paper that basically said Britain was no longer going to try to be a leading power in conventional warfare ... but there was at least a logic to it, and at least what the UK was abandoning was yesterday's weaponry not tomorrow's.
Robert Reich has always been smart and on the side of the American Enlightenment Experiment, as boosted by the post WWII GI Bill generation, that gave us the best 80 years in human history. Reich rightly torches the rationales given by our mad right for abandoning that project. He shows how every step taken by the GOP, since Reagan, has resulted in slower growth, rising wealth disparity and plummeting entrepreneurship. "Supply Side" never delivered on a single one of its clearly stated promises. In science, that has a word... DISPROVED.
ReplyDeleteReich has also irritated me lately, with some of the same polemical rigidity that hobbles democrats, every election cycle. Only this time he has posted a must-read! Because he cites the editor-at-large of the Wall Street Journal, of all sources, who lists all the glaring symptoms of an oligarchy bent on destroying the competitive capitalism that they are supposed to admire. If you go to this essay, you'll see the complaint by a man who sincerely believes in creative-competitive markets and wants them to work!...
...and who is terrified to see every single capitalism failure mode predicted by Karl Marx coming true before our eyes.
It doesn't have to happen! The Greatest (GI Bill) Generation showed how to stymie Marx, by making capitalism ever more flat and fair. By boosting children of the poor and middle class to compete fairly with the inheritance brats of the rich...
...which is exactly why oligarchs are pulling out the stops to restore feudalism, before it's too late.
But ignore me and read this. There ARE a few residually sane conservatives out there who can be swayed into realizing; it's time for another Rooseveltean reset, as Bernie Sanders recommends... lest things reach the tipping point into 1917... or Paris 1789.
https://robertreich.substack.com/p/venality-cupidity-and-corruption
"lest things reach the tipping point into 1917... or Paris 1789."
DeleteYou say that like it's a bad thing.
Moderation in all things, Larry. What happened after those dates was ... not good.
DeletePossibly because of its somewhat cryptic content, I took the liberty of inquiring deeper into the belief systems of our new participant 'Taotuner', and what I found out about the 'frictionless' nature of AI was uniquely apropos:
ReplyDeletehttps://news.ycombinator.com/item?id=49048352
Researchers at Harvard and MIT found something unsettling: when radiologists used AI assistants, their diagnostic accuracy improved. But when the AI wasn't available, their unaided performance had declined. They had become dependent — faster with the tool, worse without it.
Studies on AI coding assistants show the same pattern. Developers complete tasks faster. But they also write more buggy code than those who did the same tasks unaided — and struggle to fix bugs without the AI. They've outsourced the reasoning, not just the typing.
In other words, AI & technological dependence allows us all to act faster & more efficiently at the cost of making all users increasingly stupid & less competent with every passing day, as documented in 2025 UCSD entrance exam that showed a five-year 30X decrease in the math competency of incoming freshman despite exemplary academic GPAs.
I hope to hear more commentary by this thoughtful man.
Best
DeleteStudies have found that typists can type faster than stenographers writing with a pen. But when you take their typewriters away, they write slower than stenographers who never used a typewriter.
That's an identical sentence with an identical conclusion to an identical assistive technology Locum.
You're suffering under the misapprehension that the thinking part of programming is about snuffing out bugs. But it is really not. The "thinking" part of programming is working out how to do what you want to do, not writing bug-free code. Yeah 100% if someone has a tool that checks their code and corrects bugs, and they use it for a while, they'll focus more of their time on designing the right algorithm, and less on the syntax of the programing language, and their command of the syntax will degrade, just like any other muscle you don't exercise.
That's not a bug in the AI coding revolution. Its a feature.
Off topic as usual, I watched the Netflix biopic of U S Grant and I can say I was impressed. They didn't shy away from the stand-up gore and violence of a 19th C battlefied.
ReplyDeleteI was ticked off in the Shiloh sequence when Sherman tells Grant that Prentiss has surrendered his division, without mentioning that Prentiss had been holding his ground all morning, flat out saving Grant's army during the first day, and only surrendered "The Hornet's Nest" when surrounded on 3 sides.
Pappenheimer