Sam Altman, president of Y Combinator
Patrick T Fallon | Bloomberg | Bloomberg | Getty Images
OpenAI CEO Sam Altman said he agreed with parts of an open letter from the Way forward for Life Institute signed by tech leaders similar to Tesla CEO Elon Musk i Apple co-founder Steve Wozniak, who called for a six-month halt to AI research, but added that the letter “misses many of the technical nuance of where we want a break.”
Altman made the remarks on Thursday video performance at MIT an event dedicated to business and artificial intelligence.
OpenAI creates ChatGPT, an AI bot that may create human answers to questions asked by the user. The bot has sparked the AI craze within the tech world. Microsoft uses OpenAI technology in its Bing chatbot, and Google recently launched its competitor Bard.
“I feel moving with caution and increasing rigor by way of safety is actually essential,” continued Altman. “A letter doesn’t seem to me the optimal way to address it.”
In March, Musk, Wozniak and dozens of other scientists called for a direct halt to training “experiments” involving large language models that were “more powerful than GPT-4,” OpenAI’s flagship large language model, or LLM. Greater than 25,000 people have since signed the letter.
OpenAI’s GPT technology gained international attention when ChatGPT was launched in 2022. GPT technology is the premise Microsoftchatbot Bing AI and triggered an avalanche of investment in artificial intelligence.
“AI labs and independent experts should use this break to jointly develop and implement a set of common security protocols for advanced AI design and development which are rigorously audited and overseen by independent third-party experts,” the letter reads.
“I also agree that the chances are getting increasingly more serious, that the safety bar needs to be raised,” Altman said on the MIT event.
Earlier this yr, Altman admitted that AI technology had made him “a little scared.” Questions on the secure and ethical use of artificial intelligence have been raised on the positioning White Houseon Capitol Hill and in conference rooms across America.
“We’re doing other things besides GPT-4, which I feel have various security issues which are essential to address and were completely omitted from the list,” the OpenAI director said.