Unlock Editor's Digest totally free
FT editor Roula Khalaf selects her favourite stories on this weekly newsletter.
OpenAI said it had begun training its next-generation artificial intelligence software, however the startup walked back earlier claims it wanted to construct “superintelligent” systems that were smarter than humans.
The San Francisco-based company said Tuesday that it had begun production of a brand new AI system “designed to take our capabilities to the following level” and that its development can be overseen by a brand new safety committee.
But as OpenAI makes rapid progress with AI development, a senior OpenAI executive appears to be walking back earlier statements by its CEO, Sam Altman, that the corporate ultimately goals to create a “superintelligence” far superior to humans.
Anna Makanju, vp of worldwide affairs at OpenAI, told the Financial Times in an interview that the corporate's “mission” is to develop a synthetic general intelligence that “can perform cognitive tasks within the sense of what a human could do today.”
“Our mission is to develop AGI; I wouldn't say our mission is to develop superintelligence,” Makanju said. “Superintelligence is a technology that shall be orders of magnitude more intelligent than humans on Earth.”
Altman told the FT in November that he spends half his time researching “the best way to construct superintelligence.”
As OpenAI fends off competition from Google's Gemini and Elon Musk's startup xAI, the corporate can be attempting to persuade policymakers that it’s making responsible AI development a top priority after several senior security researchers resigned this month.
The recent committee shall be chaired by Altman and board members Bret Taylor, Adam D'Angelo and Nicole Seligman and can report back to the remaining three board members.
The company didn’t provide any information on what the successor to GPT-4, which powers the corporate's ChatGPT app and received a significant upgrade two weeks ago, might do or when it will launch.
Earlier this month, OpenAI disbanded its so-called Superalignment team, which was presupposed to take care of the safety of doubtless superintelligent systems, after Ilya Sutskever, the team's leader and co-founder of the corporate, resigned.
Sutskever's departure got here months after he led a surprise coup against Altman in November that ultimately failed.
The closure of the Superalignment team resulted in several employees leaving the corporate, including Jan Leike, one other senior AI security researcher.
Makanju stressed that work continues on the “long-term possibilities” of AI, “even in the event that they are theoretical in nature.”
“AGI doesn’t exist yet,” Makanju added, saying such technology wouldn’t be released until it was secure.
Training is step one in learning a synthetic intelligence model, which involves using an enormous amount of information and knowledge provided to it. After the model processes the information and improves its performance, it’s validated and tested before getting used in products or applications.
This lengthy and highly technical process signifies that OpenAI's recent model may not develop into a tangible product for a lot of months.
