OpenAI, the company behind ChatGPT, announced Tuesday the creation of a new oversight committee, as the departure of several officials raised concerns about its desire to reduce the potential dangers of intelligence artificial (AI).
According to the company, the committee is expected to begin work as OpenAI begins training its new AI model, which is expected to exceed the capabilities of GPT-4, the current system at the heart of ChatGPT. Its boss, Sam Altman, will sit within the group.
“We are proud to create and deliver industry-leading AI models in both capabilities and security and, at this key moment, welcome the robust debate” it generates, said l company in a press release.
The committee, which will be composed of members of its board of directors and managers, must, over the next 90 days, work to evaluate and strengthen existing security and processes around the development of advanced AI.
OpenAI also assured that it will receive advice from external specialists during this period, including the former American head of cybersecurity, Rob Joyce, who was head of this division within the National Security Agency (NSA ), as well as John Carlin, former head of the Department of Justice.
The committee will take stock of the security protocols already in place within OpenAI and put forward recommendations for possible improvements or additions.
At the end of the three months, a report from the committee will be presented to the board of directors before being published.
The launch of this committee is intended to be a response to the departure of several senior officials from the group, concerned about the lack of priority given, according to them, to the security of artificial intelligence by OpenAI.
Earlier this month, the company dismantled its team responsible for reducing long-term risks induced by AI.
Announcing his departure, team leader Jan Leike criticized OpenAI, accusing it, in a series of tweets, of prioritizing “beautiful new products” to the detriment of the essential work of securing the AI.
“In recent months, my team has been going against the tide,” he declared.
The company also created controversy when its latest iteration included a voice very close to that of American actress Scarlett Johansson, even if OpenAI assured that it had not sought to copy the Hollywood star.