The notion that Artificial General Intelligence (AGI) will create new jobs to replace those it automates is a comforting one. However, this assumption relies on the idea that AGI will not be able to perform new tasks that humans can. But what if AGI can generalize to new, unmeasured dimensions of intelligence?

Research suggests that human intelligence can be broken down into a finite number of dimensions, with some estimates as low as 7-10 broad cognitive components. This is evident in IQ tests, which have shown that a handful of factors can explain 50% of the variance in human cognitive performance.

If we apply this same principle to AGI, it’s possible that current AI systems already demonstrate a similar ‘g-factor’, where performance on one benchmark translates to performance on other benchmarks. This would mean that AGI could potentially generalize to new tasks and dimensions, making it unlikely that humans will be able to find new niches that AGI cannot fill.

The implications of this are significant, as it suggests that AGI could replace not just current jobs, but also future ones. While it’s impossible to test every possible dimension of human intelligence, the fact that current AI systems are already showing signs of a ‘g-factor’ suggests that AGI may be more capable than we think.

Photo by Özlem on Pexels
Photos provided by Pexels