Sam Altman and his OpenAI outfit think AGI is only a few years away. Trouble is, most researchers who actually know what they’re doing aren’t buying it.

Former Google researcher and now Cohere founder Nick Frosst said: “The technology we're building today is not sufficient to get there.”

What we’ve got now are word predictors and pixel guessers, not sentient minds. They mimic, they don’t understand. And while they can win games with limited rules, like AlphaGo, they can’t handle the real world, which is full of chaos, nuance and things that don’t come in training data.

A survey of AI researchers from the Association for the Advancement of Artificial Intelligence showed more than three-quarters do not think today’s methods will lead to AGI. That means something big is missing. A fundamental new idea, not just another GPU and a pile of tokens.

The real problem is no one can agree on what intelligence actually is. Measuring human brains is tricky enough, but comparing them to machines is even worse. The best we’ve got are benchmarks that show models getting better at maths and code, which is a bit like saying a calculator is getting clever.

Harvard cognitive scientist Steven Pinker said: “There’s no such thing as an automatic, omniscient, omnipotent solver of every problem. These systems aren’t miracles, they’re “very impressive gadgets.”

Even philosophers are chiming in Ca’ Foscari University's Matteo Pasquinelli said that AI is still completely dependent on people. “AI needs us: living beings, producing constantly, feeding the machine. It needs the originality of our ideas and our lives.”

So while Silicon Valley is still lighting candles for AGI, what it’s actually building is a glorified autocomplete. Until someone invents something genuinely new, that’s all it will ever be.