Countries - and state-scale corporates - should each be working on their own Strategic Fact Reserve. If they’re not already.
Let me trace the logic…
Assume for a moment that AI is upstream of productive work. It’ll start off that people just go faster with AI, and then it’ll become essential. I remember hearing that legal contracts got longer when word processors were introduced (I’d love to find the reference for this); you couldn’t handle a contract without a computer today. Same same.
So if I were to rank AI (not today’s AI but once it is fully developed and integrated) I’d say it’s probably not as critical as infrastructure or capacity as energy, food or an education system.
But probably it’s probably on par with GPS. Which underpins everything from logistics to automating train announcements to retail.
I just mentioned GPS. The EU has its own independent satellite positioning system called Galileo. Which makes sense. It would be an unfortunate choke point if GPS chips suddenly cost more for non-US companies, say. Or if European planes couldn’t safely land into perhaps Greenland, due to some plausibly deniable irregular system degradation.
Diplomacy, speak softly and carry a big stick, right?
With AI? It’s far-fetched but maybe degrading it would knock 20 points off the national IQ?
But from a soft power perspective…
We’ll be using AIs to automate business logic with value judgements (like: should this person get a mortgage? Or parole?) and also to write corporate strategy and government policy.
No, this isn’t desirable necessarily. We won’t build this into the software deliberately. But a generation of people are growing up with AI as a cognitive prosthesis and they’ll use it whether we like it or not.
However. Large Language Models Reflect the Ideology of their Creators (arXiv, 2024):
By identifying and analyzing moral assessments reflected in the generated descriptions, we find consistent normative differences between how the same LLM responds in Chinese compared to English. Similarly, we identify normative disagreements between Western and non-Western LLMs about prominent actors in geopolitical conflicts. Furthermore, popularly hypothesized disparities in political goals among Western models are reflected in significant normative differences related to inclusion, social inequality, and political scandals.
Our results show that the ideological stance of an LLM often reflects the worldview of its creators.
This is a predictable result? AIs are trained! Chinese large language models will give China-appropriate answers; American models American! Of course!
What if you’re a Northern European social democracy and your policy papers (written by graduates who are pasting their notes into ChatGPT to quickly write fancy prose) are, deep down, sceptical that, yes, citizens really will adhere to the social contract?
All of the above will not matter until suddenly it really matters.
Which means it’s important to retain independent capacity to stand up new AIs.
What you need to build a new AI: expertise; supercomputers; training data.
The capacity for the first two can be built or bought in.
But training data… I think we’re all assuming that the Internet Archive will remain available as raw feedstock, that Wikipedia will remain as a trusted source of facts to steer it; that there won’t be a shift in copyright law that makes it impossible to mulch books into matrices, and that governments will allow all of this data to cross borders once AI becomes part of national security.
There are so many failure modes.
And that’s not even getting into the AI-generated scientific research papers…
Or, technically, what if the get-out of using reasoning models to generate synthetic training data for future more advanced modes - which works really well now - is domain-specific, or stops working after a couple cycles: AI models collapse when trained on recursively generated data.
Or what if, in the future, clean training data does exist – but it’s hoarded and costs too much, or you can only afford a degraded fraction of what you need.
What I mean to say is: if in 2030 you need to train a new AI, there’s no guarantee that the data would be available.
Everything I’ve said is super low likelihood, but the difficulty with training data is that you can’t spend your way out of the problem in the future. The time to prepare is now.
I’ve been speaking from the perspective of national interests, but this is equally a lever for one trillion dollar market cap corporate against another.
Coming back to GPS, somebody who realised the importance of mapping data very, very early was Steve Coast and in 2004 he founded OpenStreetMap (Wikipedia). OSM is the free, contributor-based mapping layer that - I understand - kept both Microsoft and Apple in the mapping game, and prevented it mapping from becoming a Google monopoly.
ASIDE #1. Shout out to fellow participants of the locative media movement and anyone who remembers Ben Russell’s stunning headmap manifesto (PDF) from 1999. AI desperately needs this analysis of possibilities and power.
ASIDE #2. I often come back to mapping as an analogy for large language models. There are probably half a dozen global maps in existence. I don’t know how much they cost, but let’s guess a billion to create and a billion a year to maintain, order of magnitude. A top class AI model is probably the same, all in. So we can expect similar dynamics.
OpenStreetMap was the bulwark we needed then.
Today what we need is probably something different. Not something open but - perhaps - something closed.
The future needs trusted, uncontaminated, complete training data.
From the point of view of national interests, each country (or each trading bloc) will need its own training data, as a reserve, and a hedge against the interests of others.
Probably the best way to start is to take a snapshot of the internet and keep it somewhere really safe. We can sift through it later; the world’s data will never be more available or less contaminated than it is today. Like when GitHub stored all public code in an Arctic vault (02/02/2020): " a very-long-term archival facility 250 meters deep in the permafrost of an Arctic mountain." Or the Svalbard Global Seed Vault.
But actually I think this is a job for librarians and archivists.
What we need is a long-term national programme to slowly, carefully accept digital data into a read-only archive. We need the expertise of librarians, archivists and museums in the careful and deliberate process of acquisition and accessioning (PDF).
(Look and if this is an excuse for governments to funnel money to the cultural sector then so much the better.)
It should start today.
SEE ALSO: I hope libraries managed to get a snapshot of GPT-3 (2022).
Auto-detected kinda similar posts:
I often wonder what it felt like for the ancient Greeks, circa 800BC, to be wandering in the ruins of the previous Mycenaean civilisation. These cities they can no longer build; staring at writing they can’t read. The Greeks had to re-discover literacy.
I think perhaps they wouldn’t have known what they were looking at.
I know that dinosaurs aren’t our ancestors but… it’s adjacent? We live in a world of humbled kings. Birds were once dinosaurs. Don’t you suppose that, given the chance, they would again rouse themselves?
When they look at you with their beady eyes, what are they thinking?
The dinosaurs didn’t have a techno-industrial civilisation though. Not that we’d know. That’s the thought experiment of the Silurian hypothesis, as previously discussed: "would it be possible to detect an industrial civilization in the geological record?"
Squid. Octopus. Cuttlefish. There have been a bunch of books recently about cephalopods as terrestrial aliens… the quiet implication being that they could one day rise to civilisation.
For instance:
What if cephalopods already had their complex societies?
Read: The Silurian Hypothesis: It was the Cephalopods by Dr. Klaus M. Stiefel.
This is a magical article. (Hat tip to Clive Thompson’s Linkfest.)
The argument is that cephalopods are super sophisticated:
"Cuttlefish hunt by seemingly hypnotizing crabs by generating highly psychedelic moving stripe patterns on their bodies"
And tool users:
"There is even an octopus named after its tool use, the coconut octopus (Amphioctopus marginatus)."
And language users:
"there is a very sophisticated communication system between cephalopods, based on color and pattern changes in their skins"
But they get eaten too much by fish.
So what about BEFORE FISH?
Sometime in the late Cambrian, the first geological epoch with diverse animals with complex bodies, the Nautiluses evolved, primitive cephalopods with less well-developed nervous systems and larger numbers of simpler tentacles …
From these still somewhat primitive animals evolved the giant shelled Ammonites, and the modern cephalopods (the “Coleoids”), which include the cuttlefish, octopi, squid … the animals with fast propulsion via contraction of a water-filled body cavity, fine motor control of their arms and hunting tentacles, keen senses, good memory, skin-pattern-based communication, social lives and sophisticated mating rituals.
And so:
A crucial window where cephalopod civilization could have occurred is the time between when mentally high-performing cephalopods came to their own, and the time when aquatic vertebrates really took over.
That window is between the ammonites in the Triassic, and the emergence of modern fishes in the late Cretaceous… 55 million years.
It is a wild idea that we’re not saying: oh yeah these cephalopod things are in the early days of having a civilisation. But instead: yeah they had a civilisation and then it went and now these are the rump beasts.
Maybe the crab-hypnotising skin was biotech once upon a time.
Maybe that’s the Silurian techno-signature we’ve been looking for.
The ascent of planetary civilisation and then a quiescent period all the way back to sub-sentience.
It’s very Last and First Men (Wikipedia) – Olaf Stapledon’s 2 billion year future history from 1930 about rise and fall of eighteen human species (we’re the first).
If you recall, the Eighth Men escape Venus by engineering a new human species to inhabit Neptune, but it collapses, and the Ninth Men splinter into all kinds of beasts.
Certainly strange vestiges of human mentality did indeed persist here and there even as, in the fore-limbs of most species, there still remained buried the relics of man’s once cunning fingers. For instance, there were certain grazers which in times of hardship would meet together and give tongue in cacophonous ululation; or, sitting on their haunches with forelimbs pressed together, they would listen by the hour to the howls of some leader, responding intermittently with groans and whimpers, and working themselves at last into foaming madness.
(That’s the full text at Project Gutenberg.)
It takes 300 million years but they claw their way back via a rabbit-like species to become the Tenth Men. Then all die in a plague.
It works out. By the end of the book (SPOILERS) humans are vegetarian, there are 96 sexes, and they live to a quarter million years of age. They have six legs and an eye on the top of their heads, and the race is telepathic. They can join themselves together into a huge Neptune-wide telescope, just by looking up.
I think we often pattern-match to “progress” because that (a) matches what the exponential looks like from our perspective in the Anthropocene, and that means we’re inclined to look for the progress of octopuses; and (b) we centre ourselves, humans, in the historical story, because of course.
But maybe we’re not the main character here.
So it’s odd to think about humanity as a temporary flourishing between the peaks of somebody else’s civilisation, whether it’s cephalopods or dinosaurs.
Birds are just little Napoleons, exiled on their St. Helena of deep time, before they make their vengeful return. Octopus patiently biding their time until the fish clear off again. And here we are, just keeping the seat warm.
Meanwhile whoever - whatever mysterious force - first assembled the yeasts, those microscopic Drexler assemblers; it was a techno-industrial society with such abundance that its means of production still litter the Earth today; all of us await its eventual return.
Hello! This is a both my summary of 2024 and also the “Start here” post for new readers. Links and stats below!
According to Fathom, my 5 most popular posts in 2024 were (in descending order):
Here are some more that might be worth your time. 20 most popular in 2024.
Lots of AI.
It’s not all AI, I promise.
My personal faves often aren’t always the most popular… I’ve collected my favourite, more speculative posts on topics such as:
Find them here: 14 speculative posts in 2024.
Also check out the now decade-long Filtered for… series. Each is a handful of links and a bunch of rambling, uh, interconnectedness.
I like em.
In 2024 you’ll find posts such as:
Here’s the whole Filtered for… series (111 posts since 2014). 2024 posts at the top.
Here’s an essay:
From the other side of the bridge (3 May) is my talk from UNFOLD at Domus Academy in Milan. It’s a defence of AI hallucinations but, you know, designer-y.
I log my speaking appearances over on my studio site. Here’s the 2024 list. You’ll find a talk on YouTube and three podcasts to listen to.
Things I made outside work!
Between the clock and client work I didn’t have time to kick off anything new, but I continued digging on a few projects:
If you’re interested in the tech stack behind this blog, check out the new colophon.
PREVIOUSLY!
Other ways to read:
Or just visit the website: interconnected.org/home.
If you read a post you like, please do pass it along and share on whatever discords or socials you’re on!
I like email replies. I like it when people send me links they think I’ll enjoy (they’re almost always correct). I especially like hearing about how a post has had a flapping-of-the-butterfly’s-wings effect somehow, whether personally or at your place of work.
I like talking to people most of all. I started opening my calendar for Unoffice Hours about 4 years and 350+ calls ago and it’s still the highlight of my week. Learn more and book a time here.
Stats for the stats fans.
My current streak: I’ve been posting weekly or more for 249 weeks. Five years this coming March!
February 2025 also brings up this blog’s 25th birthday. No plans yet on how to celebrate.
I continue to be a terrible predictor re which of my posts will be popular, and that’s very freeing I think? Though my usual reflection: I have the most fun with the shortest, loosest posts, and I should let myself post like that more often.
Anyway.
I love having this public notebook. I get so much out of it – new thoughts from the writing, new friends from the reading. So thank YOU. I appreciate you being here, dear reader.
More posts tagged: meta (16).
Auto-detected kinda similar posts: