The human brain is not a machine

The temptation to see the human brain as a kind of machine has been around for a long time. Marvin Minsky, a pioneer in artificial intelligence, used to provocatively call humans “meat machines”. Going further back, one analogy from the pre-computer era described the brain as an “enchanted loom where millions of flashing shuttles weave a dissolving pattern”, as the science writer Michael Pollan describes in his new book A World Appears.

So it is no surprise that some people at the forefront of AI now believe their models could soon become conscious. If the brain is akin to a computer, then why wouldn’t a super-powerful computer develop consciousness too?

This is fuzzy territory for technologists to wade into, not least because nobody can agree on what consciousness is, let alone how and why it arises. Indeed, Pollan concluded his 280-page attempt to unravel the “hard problem of consciousness” by admitting that he knew less at the end than he did at the start.

But rather than our new machines ascending to reach truly humanlike qualities, my fear is that we will steadily lower that bar by losing faith in who we are and becoming more like machines ourselves. In the spirit of Minsky, a recent article by the economist Tyler Cowen argued that there was “no ghost in the machine” but that “perhaps more importantly, there is barely a ‘ghost’ in your own human machine” because our brains make a lot of decisions without our being conscious of them.

I think Cowen’s tongue was firmly in his cheek when he said we should “downgrade our own sense of self”, but the same cannot be said for Sam Altman of OpenAI, who has defended the energy consumption of AI models by pointing out that it “takes a lot of energy to train a human” too. “It takes about 20 years of life – and all the food you consume during that time – before you become smart.”

Elon Musk, meanwhile, sees some inherent human characteristics as weaknesses to be warded against. “Beware the empathy exploit,” he has written on his social media platform X – an “exploit” being a term from cybersecurity to describe a piece of code that takes advantage of a vulnerability in a system. “Empathy is good and right when thought through (deep), but can be deadly to civilisation when simply stimulus-response (shallow).”

This habit of comparing humans to machines invites us to see ourselves as sub-optimal alternatives to intelligent robots or AI agents: too weak, too emotional, not indefatigable enough. Although many of those claims are not yet actually true (many humanoid robots need to recharge after a few hours, for example), it is already possible to see this mindset seeping in for some employers.

Take Standard Chartered’s chief executive, who recently spoke about replacing “lower-value human capital with the financial capital and investment capital we’re putting in”. His remarks caused a widespread backlash. But I think many of us have accidentally become complicit in this framing by adopting certain machine-like expectations for ourselves.

We buy endless books about how to maximise our productivity. We strap gadgets to our wrists in order to measure and “optimise” ourselves. Even when we buy books about how to stop, they seemingly have to be couched in the promise of greater productivity. Alex Soojung-Kim Pang’s book Rest, for example, has the subtitle: “Why you get more done when you work less.” Indeed, even our language of resistance often depicts us as machines. We say we need to “switch off” or to “disconnect” in order to “recharge our batteries” before we “burn out”.

But humans are not very much like machines – and that includes our brains. A number of scientists now argue that the “computer” metaphor is not a useful way to understand its immensely complex, interrelated biological workings, which are more akin to “the murmurations of starlings”, as the neuroscientist Luiz Pessoa has put it.

As for consciousness, some believe it arises from the dialogue between the body and brain. In this framing, our mortality and ability to feel are not flaws or “exploits” to be warded off, but the secret to who we are.

It doesn’t have to be a competition, of course. Humans and AI systems are both powerful, but in fundamentally different ways. In an ideal world, we would use these tools to extend our reach to achieve new things.

But if we allow ourselves to be compared to machines, I fear we will come to expect too much of ourselves in some ways, and too little in others. FINANCIAL TIMES