That artificial intelligence is advancing by leaps and bounds is not news. But since the advent of ChatGPT, and with the recent release of GPT-4, the evolution of language models seems to have jumped decades in just a couple of months. However, its development is not going as fast as the vast majority imagine, or at least that’s what Sam Altman says. The CEO of OpenAI assured in the last few hours that, contrary to what many might think, your company is not working on GPT-5nor does he intend to. At least immediately.
This was stated during his participation in an event organized by MIT. In a panel moderated by Lex Friedman, the businessman referred not only to the progress currently being made in his San Francisco, California lab, but also criticized the recent open letter promoted by Elon Musk to halt the development of more powerful models than GPT-4.
Note that OpenAI is not yet working on GPT-5, does not mean that you are not interested in doing it. Nor that his training has been slowed down by the aforementioned letter signed by more than 1,000 experts in artificial intelligence. What happens is that, always according to Sam Altman, the company is focused on multiple projects related to its language models. And that doesn’t necessarily mean releasing one new version of GPT after the other.
So if training a successor to GPT-4 isn’t in the startup’s immediate plans, why bring it up? Because Altman said that an early version of the letter promoted by Elon Musk claimed that OpenAI was already training GPT-5which was a lie.
Sam Altman denies that OpenAI already works in GPT-5
Asked about the letter in question, the CEO of OpenAI was blunt. “I agree that as capabilities become more and more serious, the safety bar needs to be raised. But unfortunately I think the letter lacks technical nuances about where we need to pause. An early version of the letter claimed that OpenAI is training GPT-5 right now, which is not true, at least not for some time. So, in that sense, it was a bit silly.“, he asserted.
In this way, Sam Altman tried to downplay the criticisms of those who claim that, in their current state, language models such as those of OpenAI can be dangerous. In fact, the businessman raised the ante and assured that today they are working on developments riddled with security alertsbut which were not even mentioned by the experts.
“We’re doing things on top of GPT-4 that I think have all kinds of security issues that are important to address that are totally off the charts. I think moving with caution and increasing rigor around security issues It is very important. But I think the letter is not the optimal way to address it,” challenged the OpenAI leader.
That being the case, then forget about seeing GPT-5 anytime soon. Of course, Sam Altman’s sayings raise many questions that still have no answers. The most important, without a doubt, is what is so “dangerous” that today they are developing from GPT-4 and what effects its implementation could have.
But the businessman remarked that OpenAI’s commitment to the security of its language models is total. In fact, he claimed that it took around 6 months from when they finished training GPT-4 to when they released it to the public.