Rodrigo Barbado Esteban

A research preview of Operator, an agent that can use its own browser to perform tasks for you.

Spain is nowhere near being a leader in AI, yet we already have an AI regulatory agency with 80 employees paid with public funds. From the creators of "Cookies killed your parents, so we regulate them" comes the latest blockbuster in Spain: "We have no f*cking idea what to regulate about AI, but let's start by paying 80 hefty public salaries." Spain doesn’t have any frontier AI companies, but the country rushed to establish the first AI agency in Europe: the Spanish Agency for the Supervision of Artificial Intelligence (AESIA). The organization will have a president, a director, two subdirectors, a secretary general, and 10 departments. Imagine a startup launching with 80 employees but without any clear objectives or tasks. But since this is a public agency and the money is “everyone’s”, nothing happens! And soon enough, 80 won’t seem like much—they’ll grow to 200 employees. That’s how we roll here. It makes me sick 🤮 1. Bad States Are Like the Cookie Monster The worst kind of state isn’t one that doesn’t help its citizens. It’s the kind that’s so bloated it has to invent bureaucracy to keep all its bureaucrats busy. That’s a sign of decline. Imagine redirecting that money—call me crazy—towards lowering taxes, making it easier to start businesses, improving public education, or keeping people healthier with better public healthcare. But no, let's use it for AI regulations. Let’s throw a wrench into the wheels. 2. Europe’s Regulatory Obsession Will Be Its Downfall The orientation of this agency in Spain seems to align perfectly with the EU's regulatory obsession. This obsession is slowly creating two types of AI: a crippled one for Europeans and another, fully functional, for the rest of the world. For instance, OpenAI delayed its entry into the EU with Sora for that very reason, and the same is happening with many other AI platforms. Some Europeans might laugh about not having access to Sora or future ChatGPT versions, but when it starts piling up, it will be horrific, sad, and absolutely painful how far behind Europe will fall compared to the rest of the world. 3. The Agency’s HQ? A Castle My admiration for the palace where this regulatory agency will be headquartered (all paid with our taxes 💸). It’s truly beautiful. Imagine what a fantastic museum it could be. I picture myself walking through its halls, admiring fossils... but nope, maybe I will have to visit it after being reported for making an AI video of Trump kissing Pedro Sánchez in Christmas sweaters. Let me tweak an Indiana Jones quote slightly: That should be in a museum! 4. Bureaucracy Breeds Bureaucracy: Parkinson’s Law “Work expands to fill the time available for its completion” + “Bureaucrats create jobs for other bureaucrats”. That’s why I value small teams. If your team can’t share a family-size pizza and feel full, it’s too big. The future director of Spain’s new AI regulatory agency will earn €156,000 a year. Add to that the cost of the 80 employees, plus maintaining the palace HQ, and we’re looking at several million euros annually from taxpayers. Considering there aren’t many companies in Spain whose core business is AI, it’ll take years to generate even a fraction of that amount in value. In fact, if we’re not careful, more money will be spent on regulation than the industry will generate. 🤣 And, of course, the more it’s regulated, the less it’ll generate. Fun fact: the director will earn more than Spain’s Prime Minister, Pedro Sánchez. So if your kids tell you they want to be future presidents, tell them no... the future in Europe belongs to regulators and bureaucrats! (Until they bring Europe to ruin). By the way, wouldn’t it be enough to upload the AI EU Act PDF into Grok? 🤣 Make an AI bot regulate itself... "Grok Regulator 2.0". 5. Throwing Wrenches into the Wheels If 20-year-olds are already scared to start businesses because of GDPR and cookie policies, imagine when they get fined (hypothetically) for using ChatGPT with a “non-regulated” version. 🤣 Let’s regulate. Keep regulating. At the bottom of the ocean, we can regulate the rusty remains of the ship we’re on so that it doesn’t bother the fish. 6. Will We Be Forced to Emigrate? For me, it’s clear: - As an entrepreneur, having sold Magnific and happily working at Freepik, I don’t plan on starting anything new in the short/medium term. But if I did, and it was AI-related, the current situation in Europe/Spain and this trend of throwing wrenches into the wheels would definitely make me consider moving abroad to launch it. I’m older now, so I probably wouldn’t, but if I were 20-30 years old, I wouldn’t think twice. - As an investor, it’s also clear: why invest in an AI company in Spain or Europe, with the risk that the regulators will crush it? It’s far more attractive—and the tech is more advanced—to invest in the US or, call me crazy, even China. Here in Europe, as usual, we FIRST regulate and THEN wait to see what we’ll need to regulate. It’s all backward: no foundational models, no core AI companies, just a handful of players (like Freepik/Magnific). But we already have a regulatory agency... and not even clear laws or guidelines about what’s legal or not! 7. There’s No Other Option but to Bow Our Heads Anyway, I’ll stop ranting because in Spain, there are literally just a handful of us working on AI, and from those 80 regulators, maybe 20 will end up being assigned just to monitor Magnific/Freepik 🤣. We’d better bow our heads and accept the reality that China and the USA will leave us behind, kicking a can down the road. If this text goes viral, they’ll come after us. Maybe I should just keep my mouth shut, or these regulators will come after us with everything they’ve got… So, I take back everything I said. Regulation is amazing. Regulating is so much better than creating value, technology, projects, products, or services. Regulating a lot will make us strong. Do 20 burpees and 20 regulations every day, and you’ll be ripped. First, let’s establish a solid regulatory framework that encourages people to take the plunge because they’ll have a clear idea of all the fines they might face. This is FOOLPROOF and a HUGE HELP to entrepreneurs, researchers, and AI scientists. And thanks to regulation, it will be impossible for a Malicious AGI to emerge because if it’s very, very prohibited, it just won’t happen here. Maybe the Americans, who don’t regulate anything, will create a Malicious AI and deal with the consequences… but we won’t let it board a plane and come to Spain to cause trouble. We’ll have that STRICTLY prohibited. The future is colorful. It’s bright. Spain, leader in AI. Forever and ever. OpenAI is trembling. THE SPANISH AGI IS COMING. ... Imagine being China or the USA and laughing your ass off watching another power (Europe) throw wrenches into its own wheels, leaving the entire AI field wide open for you to do whatever you want 😂

New verified ARC-AGI-Pub SoTA! @OpenAI o3 has scored a breakthrough 75.7% on the ARC-AGI Semi-Private Evaluation. And a high-compute o3 configuration (not eligible for ARC-AGI-Pub) scored 87.5% on the Semi-Private Eval. 1/4

The original video from the talk by @ilyasut at the Neural Information Processing Systems (December 10 - 15, 2024, Vancouver)