Artificial intelligence (AI) has been something that people have discussed almost since the beginning of computers. In the last couple years, AI systems like ChatGPT have really made a splash and have been advancing at some pretty insane rates. That being said, how do they compete against the intelligence of actual humans?

This is not something that is very easy to answer. After all, while there are plenty of tests out there that claim to measure intelligence, none of them can measure all aspects of what the human mind can do.

For example, some people are exceptionally good at math, but really struggle with social situations or ’emotional intelligence.’ Others are great at special reasoning, but can’t understand even modest scientific concepts. Most people, of course, are right around average at most things.

The point here is that we can’t even come up with one definitive way to measure the intelligence of humans, so it is impossible to actually measure the true intelligence of AI. Of course, that doesn’t mean we should try, or that intelligence tests of various types have no value.

In the past, when it came to testing the intelligence of a computer system, most people used the Turing Test. This was a fairly simple concept where in order to pass the test, the machine needed to be able to convince a human that the machine was actually a human.

Depending on exactly how this is scored, systems like ChatGPT have already passed this test.

Another way to measure AI intelligence was developed by AI researcher Francois Chollet, who created the ARC-AGI (Artificial General Intelligence) test. This test is intended to test how well a system is able to develop new skills that are outside of its training data.

When this test was given to the ChatGPT-3 system (several years ago), it scored a 0%. Not very impressive. When ChatGPT upgraded to GPT-4o, it still did poorly at just 5%.

More recently, however, the new o3 system from OpenAI (the organization behind ChatGPT) smashed all previous records on this test, attaining a score of 82.8%. This puts it in line with many humans. Even more important than the fact that it scored so well is just how quickly its scores shot up. It went from a 0% just a few years ago up to over 80% today.

It is very likely that these scores will continue to increase in the coming years, and before long, AI will be able to beat all humans.

The rate of development by AI is a little scary.

If you enjoyed that story, check out what happened when Email This Subscribe For Emails