Food for thought: ChatGPT’s o1 intelligence already has a higher IQ than most people

Fears of artificial intelligence are nothing new, they go back a long way. As early as the 1980s, the writer William Gibson addressed them in his revolutionary novel Neuromancer (the creator of the cyberpunk genre), and films like Blade Runner either Terminator They instilled fear in people at any nonsense that a machine capable of reasoning might have.

However, with the current boom in AI, for which much of the blame lies with the company OpenAI, responsible for the popular ChatGPT, these concerns have become even more common. And perhaps, who knows to what extent, justified. Especially now that, it seems, This technology already has a higher IQ than many humans.

The IQ of artificial intelligence

Despite the assumptions financing problems What the OpenAI company, led by Sam AltmanThe truth is that the creators of ChatGPT do not stop in their constant attempt to improve the performance of their artificial intelligence. Something understandable, if you take into account that there are already many competitors who want to take the throne of this technology.

Their latest release, o1, was presented as a model capable of reasoning. More or less. An ambitious announcement, but also somewhat ambiguous. What did it really mean to say that an artificial intelligence has certain reasoning capabilities? Would it really be something similar to the so-called general artificial intelligence, the great goal of researchers?

The answer, obviously, is no. It is difficult to know exactly what OpenAI was referring to when it announced something like this, but if it had achieved a new evolution of this technology, it is clear that its way of presenting it to the world would have been very, very different. However, this does not mean that steps are not being taken, some of them really striking from a technical point of view.

The most notable of them has caught many by surprise. According to experts, OpenAI’s o1 model is smarter than most peopleThis was determined after subjecting the AI ​​to a test to determine its IQ level. The result obtained was nothing more and nothing less than 120, higher than the human average.

How is the IQ of an AI calculated?

To do this, the artificial intelligence has been subjected to the well-known Norwegian Mensa exam. Basically, it is an intelligence test designed to measure the intelligence quotient (IQ) of those who participate in it, and for which it uses questions that evaluate skills such as logical reasoning, solving different types of problems and abstract thinking.

The results do not seem to deceive: OpenAI’s o1 model has correctly answered 25 of the 35 total questions it was challenged with. This fact is not only impressive in itself, but it also gives us food for thought about what will happen in the future. If the progress made so far continues, in just a year or two AI will be extremely intelligent. For better or worse.

Meet How we work in ComputerHoy.

Tags: Artificial intelligence