OpenAI CEO releases "reassurance pill": GPT5 will not be trained in the short term

**Source: **Financial Association

Edit Zhou Ziyi

Image source: Generated by Unbounded AI tool

According to the latest statement from the CEO of artificial intelligence startup OpenAI, OpenAI still has not trained GPT-5.

OpenAI CEO Sam Altman said at a conference hosted by India's "Economic Times" on Wednesday (June 7). "We have a lot of work to do before we can launch this model."

"We're working on new ideas that we think are needed, but we're certainly not close to starting (training GPT-5) yet."

Just a few months ago, after many industry executives and academics expressed concern about the rapid development of Sam Altman's large-scale language model, the AI startup promised at the time that it would not develop GPT-4's successor GPT for "some time". -5.

Won't do this anytime soon

At the end of March, more than 1,100 industry executives and experts, including Tesla CEO Elon Musk and Apple co-founder Steve Wozniak, signed an open letter calling for "all AI labs immediately suspend training of AI systems more powerful than GPT-4 for at least 6 months."

Responding to the letter a few weeks later, Altman said it "leaves out most of the technical details that we need a moratorium on." While Altman agrees that caution and an increasingly stringent approach to security is important, he doesn't concede that the letter is the best way to address the issue. He also insisted that OpenAI has not started training GPT-5 and does not plan to do so "for some time."

Altmam on Wednesday pushed back against some of the most vocal concerns about artificial intelligence, saying the startup was already assessing potential dangers through more meaningful measures such as external audits, red-teaming and security testing.

Red-teaming is a testing method that refers to the practice of rigorously challenging plans, policies, institutions, and assumptions by adopting an adversarial approach. OpenAI uses it to discover hazards in language models and mitigate them.

Altman also added, "When we finished GPT-4, it took us more than six months to be ready to release it." He meant to show that even if the company has completed the training of GPT-5, it will take some time. Ready to publish it.

In an earlier interview, Altman also said, "The only regulatory requirements we have are for ourselves and for larger companies."

multi-country tour

Altman's trip to India reflects a step in his active participation in meeting and building trust with lawmakers and industry figures around the world.

At the meeting, Altman also actively urged lawmakers to seriously consider the potential misuse and other negative effects of artificial intelligence in order to set up countermeasures to minimize accidents.

On Tuesday (June 6), Altman proposed during his visit to the UAE that the establishment of an organization similar to the IAEA to oversee AI. Founded in 1957, the IAEA is an independent intergovernmental international organization dedicated to the peaceful development of nuclear energy.

Altman pointed out that the IAEA is a good example. In the face of a very dangerous technology, after setting up protective measures, it can take both opportunities and safety into consideration.