- Sam Altman tries to temper the AI hype a bit.
- OpenAI’s CEO wrote on X on Monday that his company will not deliver an AI model that can compete with humans in February.
- The boss of the creator of ChatGPT even stated that we need to significantly adjust our expectations.
- Also read: We’re Living in the Era of ‘Agentic AI,’ According to the Boss of AI Chipmaker Nvidia: Here’s What He Means
CEO Sam Altman of OpenAI, the company behind AI bot ChatGPT, felt called upon on Monday to temper the AI hype a bit.
Altman typed out a message on X saying that the “Twitter hype has gone wild” and that we shouldn’t expect AGI in February. AGI stands for artificial general intelligence, a form of artificial intelligence that could rival humans.
We’re not there yet, says Altman. He would like to see us adjust our expectations significantly.
Altman’s comments are remarkable to say the least. The man known as the figurehead of the AI sector indicated earlier this month that he is certain that AI experts know how to build an AGI model. He wrote this in a blog on January 6.
AGI, or Artificial General Intelligence, is often cited as the ultimate goal of AI development. At that point, AI should be able to think and, more importantly, act at a human level. This would allow AI to autonomously take over tasks from humans.
We should be able to expect the first examples of this development from OpenAI as early as 2025. At least we now know for sure that this will not happen this month or next month.
No AGI and certainly no superintelligence
This point is underscored by OpenAI researcher Noam Brown, who studies the reasoning capabilities of artificial intelligence. Last Friday, he wrote on X that the AI hype is spiraling out of control based on vague claims.
He then responded to a comment confirming that OpenAI had not yet achieved superintelligence.
Still, Brown leaves it open whether AGI is feasible this year. He mentions superintelligence, which is generally seen as the next step after AGI. At the same time, Brown writes that there have been no major changes within OpenAI since the introduction of the o3 model. The development and scaling of the model is done in the same way as it was between o1 and o3.
At the end of last year, a number of AI developers indicated that developments in the field of artificial intelligence were, as it were, hitting a wall.
Up until now, it was assumed that increasing the amount of training data and computing power would result in better performing AI models. Now that this no longer seems to be the case, AI companies are turning to other solutions such as making models think longer before coming up with a better answer.
This brings with it other disadvantages, because it not only costs more energy to maintain this way of working, it is also much more expensive. The first is especially a problem, because AI companies are increasingly running into the problem that there is simply not enough energy to meet the demand.
Author:
Dennis Wilman
Dennis Wilman is an editor at Business Insider Netherlands and writes about technology and cars.
