Can a tiny startup fix what’s wrong with AI? Literal Labs thinks so

Literal Labs’ AI is so efficient it makes neural nets look like steam engines. 250x faster than XGBoost. 52x less energy than deep learning. If they’re right, we’ve been overengineering AI for years.
Image credits: Literal Labs

AI is everywhere now. It helps pick your next movie, catch fraud, and even spot diseases. But there’s a problem: most AI models are huge, use lots of power, and are hard to understand. Like, really hard. Even the people who build them can’t always explain how they work. That’s a bit scary, especially if AI is making decisions about your health, your money, or your job.

A small company from Newcastle, UK, says they have a better idea. Literal Labs wants to replace today’s AI with something simpler, faster, and easier to explain. Sounds nice, right?

They’ve just raised £4.6 million to do it. That money will help them grow their team, finish building their first product, and start working with real customers. Their focus? Devices that run AI without needing a supercomputer—or a massive electricity bill.

The secret sauce? Something called a Tsetlin Machine. It’s a type of AI based on logic, not layers of artificial “neurons.” That means it works more like clear step-by-step rules, not like a brain simulation. Literal Labs says this makes it easier to understand why the AI made a decision.

Think of it this way: if a neural network is a black box, the Tsetlin Machine is more like a checklist. You can see exactly how it got to an answer.

And it’s fast, really fast. Literal Labs says their model is 54 times faster and uses 52 times less energy than typical AI systems. Compared to another popular tool, XGBoost, it can be up to 250 times faster. Big numbers. Sounds impressive. But we’ll believe it when more outside experts confirm it.

The CEO of Literal Labs is Noel Hurley. He used to lead the CPU division at Arm, the chip company behind most phones and tablets. So, he knows a thing or two about efficient computing. He says neural networks are “too expensive and too power-hungry” for many real-world uses. Especially in places where you can’t count on fast internet or strong batteries.

He’s not wrong. Today’s AI models are great in the cloud, but they struggle in the real world, on small devices, in hospitals, or on farms. They also can’t explain their decisions clearly. That’s a problem if something goes wrong.

Literal Labs is trying to change that. It’s still early days, the team has only 12 people. But they’re growing. They’ve also hired a new CTO, Leon Fedden, who used to lead AI work at AstraZeneca, the big drug company.

Of course, this isn’t the first time someone’s tried to make logic-based AI work. It’s been around for decades. But maybe now the timing is right. People want AI that’s fast, clear, and doesn’t drain the battery.

Will Literal Labs succeed? Maybe. Maybe not. But at least they’re asking good questions, and offering something different in a world full of black-box machines.

And honestly, it’s refreshing to see an AI company that wants to simplify things instead of making them more confusing.

Fabrice Iranzi

Journalist and Project Leader at LionHerald, strong passion in tech and new ideas, serving Digital Company Builders in UK and beyond
E-mail: iranzi@lionherald.com

Leave a Reply

Your email address will not be published.