Sci-fi author Neal Stephenson wants AIs fighting AIs so those most fit to live with us survive
- Reference: 1747373594
- News link: https://www.theregister.co.uk/2025/05/16/neal_stephenson_ai_evolution/
- Source link:
Stephenson coined the term “metaverse” in his 1992 classic “Snow Crash”, and his 1999 epic “Cryptonomicon” digital currencies and the encryption needed to make them possible. The prescience of those works, and the [1]cracking yarns he spins, mean he is in demand as both a novelist and thinker.
T. Rex had to worry about getting gored in the belly by Triceratops. By training AIs to fight we can perhaps preserve a healthy balance in the new ecosystem
In the latter capacity he recently participated in what he described a “a panel discussion on AI as part of a private event in New Zealand.”
He’s since [2]posted his opening remarks from the event, which open by noting understandable anxiety about the sudden arrival of generative AI.
Stephenson suggests remembering that we already share Earth with many non-human intelligences from the animal kingdom and have learned to get along with them.
[3]
“We’re used to thinking of them as being less intelligent than we are, and that’s usually not wrong,” he wrote, “but it might be better to think of them as having different sorts of intelligence, because they’ve evolved to do different things.”
[4]
[5]
His remarks offer ways to categorize non-human intelligences, and he ends up suggesting that the most useful AIs will be like sheepdogs – creatures that can do specific tasks better than a human. He thinks other AIs will be like dragonflies – largely oblivious to humans but doing certain things brilliantly well – while others will be like ravens or crows in that they are aware of humans but don’t care about us.
Stephenson thinks ChatGPT is like a lapdog, a class of intelligence that can’t survive without people and is “acutely tuned in to humans and basically exist to make life easier for us.”
[6]
The author thinks AIs which pose a threat to humans will also emerge.
“I am hoping that even in the case of such dangerous AIs we can still derive some hope from the natural world, where competition prevents any one species from establishing complete dominance,” he wrote. “Even T. Rex had to worry about getting gored in the belly by Triceratops, and probably had to contend with all kinds of parasites, infections, and resource shortages.
“By training AIs to fight and defeat other AIs we can perhaps preserve a healthy balance in the new ecosystem. If I had time to do it and if I knew more about how AIs work, I’d be putting my energies into building AIs whose sole purpose was to predate upon existing AI models by using every conceivable strategy to feed bogus data into them, interrupt their power supplies, discourage investors, and otherwise interfere with their operations.”
[7]OpenAI caves to pressure, keeps nonprofit in charge
[8]American cops are using AI to draft police reports, and the ACLU isn't happy
[9]Vernor Vinge, first author to describe cyberspace and 'The Singularity,' dies at 79
[10]India ditches its 'Google Tax', perhaps to tickle Trump and dodge tariffs
Stephenson doesn’t suggest that out of malice, instead articulating “a general belief that everything should have to compete, and that competition within a diverse ecosystem produces a healthier result in the long run than raising a potential superpredator in a hermetically sealed petri dish where its every need is catered to.”
Augmentation and amputation
The novelist admitted that AI-on-AI combat is unlikely in the short term, so turned his attention to other ways of curbing harms it might create.
To do so he invoked Marshall McLuhan’s observation that “every augmentation is also an amputation”.
[11]
He illustrated that concept by evoking conversations he has had with educators “who all report … their students use ChatGPT for everything, and in consequence learn nothing.
“We may end up with at least one generation of people who are like the Eloi in H.G. Wells’s The Time Machine , in that they are mental weaklings utterly dependent on technologies that they don’t understand and that they could never rebuild from scratch were they to break down,” Stephenson wrote, before asking “who is really the lapdog in a world full of powerful AIs?”
He thinks it’s easy to avoid becoming Eloi: All we need is “simple interventions such as requiring students to take examinations in supervised classrooms, writing answers out by hand on blank paper.
“We know this is possible because it’s how all examinations used to be taken. No new technology is required, nothing stands in the way of implementation other than institutional inertia, and, I’m afraid, the unwillingness of parents to see their children seriously challenged,” he wrote.
“In the scenario I mentioned before, where humans become part of a stable but competitive ecosystem populated by intelligences of various kinds, one thing we humans must do is become fit competitors ourselves. And when the competition is in the realm of intelligence, that means preserving and advancing our own intelligence by holding at arms length seductive augmentations in order to avoid suffering the amputations that are their price.” ®
Get our [12]Tech Resources
[1] https://www.theregister.com/2015/05/23/book_review_the_happiness_industry_seveneves_and_confessions_of_a_tinderella/?page=2
[2] https://nealstephenson.substack.com/p/remarks-on-ai-from-nz
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aCcMwFIqqNHmy7W4cilpPwAAAU0&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aCcMwFIqqNHmy7W4cilpPwAAAU0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aCcMwFIqqNHmy7W4cilpPwAAAU0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aCcMwFIqqNHmy7W4cilpPwAAAU0&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://www.theregister.com/2025/05/05/openai_keep_nonprofit_in_charge/
[8] https://www.theregister.com/2024/12/12/aclu_ai_police_report/
[9] https://www.theregister.com/2024/03/22/vernor_vinge_obituary/
[10] https://www.theregister.com/2025/03/26/india_drops_digital_ads_tax/
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aCcMwFIqqNHmy7W4cilpPwAAAU0&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[12] https://whitepapers.theregister.com/
Re: Handwriting
You've not seen mine.
Re: Handwriting
I had to relearn how to write legibly for various examinations in my forties... too much typing and not enough careful writing had rendered anything other than block capitals illegible.
(My junior school laid great emphasis on a good italic hand).
Robot Wars
AI Robot Wars competitions may revive an old pastime.
Re: Robot Wars
I have a vague recollection of a game/challenge involving two (or more) chunks of code executing concurrently within the shared memory space of a simple virtual machine with the objective being for (human) participants to produce a piece of code which could locate and then disrupt, corrupt, or otherwise disable its “competitors”.
Anyone else remember that, or did I just imagine it or come across it in a work of fiction….?
Re: Robot Wars
You didn't imagine it, I can't remember it's name though. There's also Robocode, where you program virtual tanks to fight each other.
Re: Robot Wars
[1]Core Wars .
[1] https://corewars.org/
SMARTR MMORPG v4.2 versus Old Traditional Style New World Order Great Gamers ‽
If I had time to do it and if I knew more about how AIs work, I’d be putting my energies into building AIs whose sole purpose was to predate upon existing AI models by using every conceivable strategy to feed bogus data into them, interrupt their power supplies, discourage investors, and otherwise interfere with their operations. ..... Science fiction author Neal Stephenson
Those would be stupid AIs offering nothing novel and worthwhile with such a mimicking of failed similar human actions with their perverse and subversive corrupt command and inept geo-political control of nationalised mainstream media streams, Neal.
That is definitely not going to be attractively addictive to SMARTR AIs which know how humans think how things and even how everything works and how a chosen few imagine how easily humanity is led to believe what Mass MultiMedia BroadBandCasts present them with as a current running virtual reality to follow with contextual scripting aiding and abetting future remote programming.
Things certainly aint like they used to be and there's no way anything is going back to any previous good time rather than quantum leaping forward into what can also admittedly be a dark and dangerous brave new world with SMARTR AI Disciplines should any proposed future human actions treat the future badly as an existential threat.
Don't poke the NEUKlearer HyperRadioProACTive IT Bear protecting urFuture Derivative Bull Markets. AI and IT do not take prisoners.
Variety is the spice of life
So goes the old saying.
I too have begun advocating an ecology of AI approaches, in order to achieve a Darwinian balance and prevent the risks from any one type dominating. As ever with my bright ideas, someone else got there before me.
But I wouldn't spend too much effort on malignant AI-on-AI. Especially considering the concurrent headline here that everybody is deploying AI but nobody is securing it, the superpower spooks are surely way ahead of us, with organised crime hot on their heels.
imagination - what some people are better at than others...
some people have more than others
Variety can help but not always...
frankly when someone who wrote cryptonomicon
you listen
you learn
or your dog food
JJ
Re: imagination - what some people are better at than others...
'or your dog food'
Semantically and grammatically this phrase doesn't parse with the rest of the statement. Presumably you meant to say: 'or you are dog food', which could be informally expressed as: 'or you're dog food'.
Whether our pearls of wisdom are handwritten or typed it does pay to at least think about the rules to ensure clear transmission of meaning.
I mean, think of the LLMs, poor things.
Re: imagination - what some people are better at than others...
"I mean, think of the LLMs, poor things."
damn them
Human on Human battles don't generally go well for non-combatants caught in the middle.
I'm not sure I see where AI Vs AI wouldn't be the same outside of the virtual field.
As for students taking writen exams, we've allowed education theory to go too far now with kids not requiring to write and now could rightly claim that hand written work puts them at a disadvantage. That and universities seeing virtual exam rooms as a way to cut costs and charge the same fees.
Rolling back enshittifications seems to be a very steep uphill battle.
"...and have learned to get along with them."
Some of them are even tasty.
Re: "...and have learned to get along with them."
Said the Great White to his mate as he munched on the tourist who'd waded out too far.
"Not too keen on that rubbery skin (wetsuit) some of them have, it gives me the shits".
Landmines are a very simple but quite effective form of artificial intelligence.
Handwriting
... is a beautiful thing.