It’s Liberation Day, folks, and I want to tell you a story that happened to me years and years ago. I was working on a linked-data, semantic-web project: structured data exposed on the web. This data was supposed to describe news, so there was this whole ontology, a shared schema, a structure for data shared across different organizations, meant to represent news items. Each news item had the where, the when, the how, the why, who the protagonists were, a set of things useful for representing the news.
The most fascist thing I ever heard
And out came this argument from the entrepreneur I was working for. He said: “But this way we get fake news, we get an overabundance of news from people who can’t afford to publish it.” Everyone was talking about fake news back then, and he said: “In my opinion there needs to be a central European body that establishes what is news and what is not.”
I told him this was the most fascist thing I’d heard in my entire life. And in tech you hear plenty of fascist reasoning. Letting a central body decide what is news and what is not is the ABC of propaganda. The authority decides what happens and what doesn’t, regardless of whether it actually happens, purely because it’s represented at the digital level that something did or did not occur.
In the name of fake news, in the name of spam, in the name of the slop we’re seeing now with artificial intelligence, it might seem easy to think the solution is to establish authorities. Authorities that seal things off and decide. And I’ll tell you straight: I prefer the slop. I prefer total chaos, because I don’t want the authority, the central authority. I don’t trust the American models, just as I don’t trust the Chinese ones. I don’t trust absolutely anyone. I want everyone to have the right to speak, to create, to publish, to say their piece, even at the price of this flood, this overabundance.
Alignment is also political
Specifically on artificial intelligence, we’re witnessing a centralization, and a consequent principle of authority, over what intelligence can and cannot do, and over what opinions it holds. And those are opinions instilled into the AI. On its own, that thing just produces language. If you have a hard time getting ChatGPT to say something against Israel, that’s no accident. Alignment is rightly done to make these machines as close as possible to our needs, but it’s also done in political and propagandistic terms, to concentrate authority and power. And that, in my view, is one of the dangers of the times ahead.
So, on Liberation Day, I want to encourage everyone getting close to artificial intelligence, the veterans, and everyone tinkering away: a lot of critical thinking. Handle these objects with tongs. Know that, little by little, it would be nice to free ourselves of these external providers that have influenced us for decades now, not only militarily and politically, but also economically and, above all, culturally. So much so that our language has become a string of English terms, as if we all worked in a corporation. Even when you go take a dump you have to be productive and measure how much you produced and in how much time. They won’t let you lose a single moment.
A lot of critical thinking, and above all a healthy resistance to excess control, excess centrality, and abuse of authority. That’s what I feel like saying. There’s much more, but I won’t push it too far into politics, otherwise the comments will tear me apart.
AI is becoming a political matter
Artificial intelligence, in the times ahead, will become a largely political subject anyway, in my opinion. In just a few years it went from scientific to technical to commercial to infrastructural, and the moment something becomes infrastructural it also becomes deeply political. On our own turf we’re moving too slowly.
There are some happy zones. I feel like mentioning one, and nobody asked me to: I survived the second year of keeping the Cheshire Cat alive thanks to a company called Seeweb, which granted me a sort of research year asking basically nothing in return. They have language models, GPUs, and they do everything the Americans do, but in Italy. Clearly on a much smaller scale, but they do it. And in the same way, I’d like to remember all the small businesses I’m collaborating with and bringing training to. Small, yes, but fierce. They study hard, they try to keep up.
In the times ahead, I think we have an opportunity. We’ve thoroughly lost the race on infrastructure and on autonomy over compute and the energy consumption tied to AI. That battle is gone, it went the way it went. But we can win a minimum of independence on the applications side, and push for there to be standards, so that we’re independent from the gates these big tech players will build. Gates that at first welcome everyone. Great news, subscriptions at 20 euros that actually cost them 2000, so that you end up locked inside. Look at everything that’s happened with Anthropic, with the whole business, folks. And all the people who pass through will tell you what a lovely flower it is.
