OpenAI starts training a new AI model to power ChatGPT – San Diego Union-Tribune
OpenAI starts training a new AI model to power ChatGPT – San Diego Union-Tribune
Author: San Diego Union-Tribune
Published on: 2024-05-28 09:56:42
Source: Technology – San Diego Union-Tribune
Disclaimer:All rights are owned by the respective creators. No copyright infringement is intended.
Artificial intelligence company OpenAI said Tuesday that it has started training its newest AI model that will fuel the popular ChatGPT chatbot.
In a statement released on its website, OpenAI said this new model, which would replace GPT-4 technology, will bring the company closer to achieving “AGI,” or artificial general intelligence, a hotly contested idea that refers to computers matching the power of human brains.
OpenAI also announced that it is forming a new committee, including CEO Sam Altman, to evaluate the safety and security of its products, and pledged to release recommendations sometime after a 90-day review of its technology.
“We welcome a robust debate at this important moment,” the company said.
The San Francisco-based company is one of the leading players in artificial intelligence, with widely used tools such as ChatGPT and image creator Dall-E. The technology behind them has been adopted by large companies such as Microsoft, one of OpenAI’s biggest financial backers.
Tuesday’s announcements come amid turmoil for the San Francisco-based company as it faces mounting lawsuits, high-profile staff departures and concerns that AI technology will spread election disinformation, eliminate jobs and infringe on copyrights.
Earlier this month, the company released an update to its latest AI technology, called GPT-4o or “GPT-4 omni,” that enhances ChatGPT’s ability to listen and respond by voice. Actress Scarlett Johansson said one of the voices OpenAI released, called “Sky,” was “eerily similar” to hers, and that after declining overtures by Altman to voice the assistant, she was “shocked, angered and in disbelief” to find it sounding nearly identical to her speech.
OpenAI said the voice was not hers, and documents shared with The Washington Post show the company hired a voice actor. Johansson has threatened to file legal action against OpenAI.
The company already faces lawsuits from news organizations, including local papers and the New York Times, saying that the way OpenAI trains its chatbot technology – by ingesting vast amounts of online data such as news stories, Wikipedia articles and even Reddit forums – is infringing on the copyrights of news organizations.
OpenAI co-founder Ilya Sutskever, who co-led a team focused on the societal threats of AI, resigned from the company in May amid concern from critics that OpenAI was dismissing the dangers of its technology while chasing profit. His team co-lead, Jan Leike, also departed this month.
Sutskever was one of the board members who voted to oust co-founder Altman in November. Altman was reinstated a few days later, and Sutskever and the other directors who voted against him resigned from the board.
The new safety committee will be led by board Chairman Bret Taylor, Quora CEO Adam D’Angelo, former Sony Entertainment executive Nicole Seligman and Altman, OpenAI said.
The announcement of its formation comes just days after two former OpenAI board members said in an Economist article Sunday that more regulation of AI is needed because companies like OpenAI can’t be trusted to govern themselves.
Verma writes for The Washington Post.
Originally Published:
Disclaimer: All rights are owned by the respective creators. No copyright infringement is intended.
