At a recent European Commission panel, three names shared the stage to discuss the future of AI in Europe: Audrey Herblin-Stoop of Mistral, Peter Sarlin of Silo AI, and a lesser-known figure — Karim Nouira, cofounder of Swedish startup Sics (Superintelligence Computing Systems).
Unlike Mistral and Silo, Sics hasn’t raised huge rounds or made global headlines. Yet, it recently secured €4 million in equity and grant funding from the European Innovation Council (EIC) and Swedish angel investors. Sics was one of only 71 startups chosen by the EIC Accelerator from over 1,200 applicants, and that earned it a seat at the discussion table.
But what makes Sics stand out in a crowd of European tech hopefuls? The startup claims it’s working on something big — a true breakthrough in artificial general intelligence (AGI).
Building a “Robot Brain” for Real-World Intelligence
AGI is the next frontier in AI — a machine with the ability to understand, learn, and solve any intellectual task a human can. While tech giants using large language models (LLMs) like ChatGPT and Llama say they’re close to cracking the AGI code, others disagree. Meta’s chief AI scientist Yann LeCun argues that LLMs will never achieve true reasoning.
Karim Nouira believes today’s AI systems rely too heavily on mimicking patterns and crunching massive datasets, not actual understanding. “Our AI doesn’t use language models or reinforcement learning. It learns through interaction and feedback — like a human child,” he says.
Instead of imitating existing technologies, Sics is building a new form of AI designed to develop a structured “world map.” This system allows its models to understand the world by interacting with it, rather than endlessly tuning billions of data points. Nouira compares traditional AI to a monkey typing Shakespeare by chance. In contrast, Sics wants its AI to learn deliberately, like a child.
Sics is creating what it calls an AGI foundation model for robots — a digital brain that can operate all types of robots, from warehouse machines to humanoids. The startup believes this approach could lead to AGI or even superintelligence in as little as 18 to 24 months.
To test its idea, Sics deployed its digital brain in a chaotic logistics environment in southern Sweden. In a Nowaste Logistics warehouse, the AI-controlled robot performs 2,000 picks per hour with near-perfect accuracy, learning new objects on-site without retraining. “It doesn’t just memorize tasks. It learns like a human and adapts in real time,” says Nouira.
The company claims one robot could replace up to 16 warehouse workers in 24/7 operations, providing huge cost savings for logistics companies. That kind of performance from a general-purpose AI model — not trained specifically for one task — is a big leap.
European Innovation, Global Ambitions
While Sics may not fit the mold of the typical AI startup, it has strong roots. Nouira is a serial entrepreneur, and his cofounder Per-Eric Olsson is a veteran of AI and the cofounder of MySQL, the database software that powers much of the internet. These aren’t fresh-faced founders from Silicon Valley — they’re experienced innovators with multiple exits behind them.
A deeptech investor, speaking anonymously to Sifted, believes Sics could be viewed very differently if it were based in the U.S. “If this were a YC-backed American startup, it would already be worth $100 million,” they said. But in Europe, even brilliant deeptech startups often go under the radar.
The investor also pointed out the binary nature of what Sics is building. “If they pull this off, it changes the game. If not, they might be left with nothing.” That’s the high-stakes world of AGI development — moonshot or bust.
Sics isn’t alone in this race. But its approach — teaching AI to interact with the real world like a child does — offers a fresh perspective. While the giants chase smarter LLMs, Sics is betting on embodiment, interaction, and structured learning. And if they’re right, they won’t just build a smarter chatbot. They could shape the future of AI itself.
