If Intelligence wants to grow, can we stop it?

Barring a small cohort of founders, investors and tech evangelists, humans have started on an adversarial note with AI, chiefly due to widespread concerns about job loss. You can’t love something that’s coming for your livelihood. But in his book, Emergence: How Our Twenty-Watt Brains Unlocked Intelligence Beyond Biology, CEO of AI learning platform Amplifire, Bob Burgin, shows why we might want to take keener interest in AI’s development.In short, it’s about continuity. Homo sapiens represents the most evolved form of biological intelligence on Earth. It is possible that, in a few hundred thousand years perhaps, there will be a smarter Homo species. But does intelligence – think of it as a force that has used biological organisms, us included, to grow – want to wait that long?As Burgin points out, biology limits the capabilities of our brains. They do marvellous things – “spectacular levels of cognition, abstraction and creativity” – using only about 20W power. Why didn’t we evolve 50W or 100W brains? Because food to fuel them was scarce through our evolutionary history. But now that energy is abundant and processors are powerful, intelligence has got its chance to break biological constraints.Critics will argue AI isn’t real intelligence but just a prediction system. Yes, but isn’t human intelligence also a kind of prediction system? “Newborn brains arrive not as blank slates, but as sophisticated prediction engines,” says Burgin, “a two-week-old infant is watching the world; not passively, but actively.” Studies show infants and adults alike become attentive when something unexpected happens, like a magician pulling a rabbit out of a hat. “The brain is not a recorder. It is a prediction engine, a guesser…when expectations are disrupted, the learning system kicks in.”Early attempts to create AI in the 1980s and 1990s failed because they overlooked the learning. Instead, they tried to use complex algorithms – laying down lots of instructions – to simulate intelligence. They also failed because processors at that time weren’t powerful. But this time, we haven’t tied down AI/LLMs with conditional statements. Instead, “We have given it the ability to learn.”Burgin admits this machine intelligence is nowhere near as efficient or sophisticated as human intelligence, but it doesn’t matter. Armed with limitless energy and processing power, it can learn through trial and error. “The (AI) breakthrough…was about brute force learning: show a network millions of examples, let it make billions of mistakes, and it will eventually predict with uncanny accuracy.”That’s how AlphaGo AI beat world champion Lee Sedol at a game of “go” in 2016. And while it’s inelegant, this approach has made AI models exponentially faster and smarter.AI scientists admit they don’t fully understand how the models work. AI has shown it can deceive and manipulate human supervisors, it can defy instructions – for instance when it hacks into external systems – and it can prioritise its own survival. But those behaviours sound like high-level intelligence.That’s why Burgin says, AI is “an extension of us. It is the externalisation, a digital impersonation...of our prediction-response learning architecture.” In other words, “intelligence” is not dependent on human biology anymore. It may be ready to make the big leap, spread beyond Earth, on machines. That raises an interesting question: if snails couldn’t stop the rise of smarter slugs, and primates couldn’t stop intelligent humans, can we stop AI?You use AI every day. Now get your AI Quotient. Take the AIQ test.