OpenAI CEO Fears the Threat of AI

Sam Altman at a WSJ event. Photo: Bloomberg

"We need to be cautious. People should be happy that we're expressing concerns about AI, especially when the models they develop can be abused for disinformation campaigns," Sam Altman, CEO of OpenAI, said on March 17. "AI is increasingly writing complex programming code that can be used for cyberattacks."

Altman stressed that regulators and society need to keep up with technology to deal with and prevent negative consequences for humans. However, he affirmed that AI is "the most wonderful technology that humanity has developed".

The warning comes shortly after OpenAI unveiled its GPT-4 model, less than four months after the release of ChatGPT's GPT-3.5 model. Alman called the new version "not perfect," but it scored near-perfect on a US high school math test and can code in most current programming languages.

There are many concerns about the risk of AI replacing humans, but Altman insists that AI can only work under the direction and supervision of a human controller. "This is a tool that is under control. What I worry about is that someone will not put safety restrictions on what goes into AI. Society doesn't have much time left to figure out how to manage this," he said.

Billionaire Elon Musk, co-founder of OpenAI, has repeatedly warned that AI and artificial general intelligence (AGI) are more dangerous than nuclear weapons. Altman admitted that the GPT-4 version uses "deductive reasoning" instead of memorization, leading to strange responses from users. "I always remind people about the illusion problem, where AI models present content that they consider to be factual information, but it is completely fabricated content," he said.

According to VnExpress