A New Way AI Learns to Solve Tricky Problems

You've probably heard a lot about AI programs (called Large Language Models or LLMs) that can write text or create images. But a new company called Ineffable Intelligence is taking a different approach. They've teamed up with Google Cloud to build something they call 'superlearners'.

These superlearners aren't focused on mimicking human conversation or art. Instead, they're designed to solve complex challenges through a method called 'reinforcement learning'. Think of it like teaching a child — they learn by trying things, getting feedback, and adjusting their approach based on what works and what doesn't. This allows the AI to develop surprising and clever ways to tackle problems it hasn't seen before, potentially finding much more efficient solutions.

The company's founder, David Silver, was a key person behind AlphaGo, an AI that famously beat the world champion at the game of Go. That AI used a similar learning technique to master a game many thought was too complex for computers. Now, they're applying that idea to build AIs that can learn and adapt across many different scenarios without needing a massive database of human conversations to start with.

While this might sound very technical, the goal is practical. If these superlearners prove effective, they could be used to optimise all sorts of systems – from managing complex logistics for a trucking company, to finding new ways to develop medicines, or even improving energy grids. For Australian small businesses, such advancements could eventually lead to smarter tools that help automate planning, predict problems, and boost efficiency in ways we're just starting to imagine.