WASHINGTON, March 30 (Reuters) – The Center for Artificial Intelligence and Digital Policy, a technology ethics group, has asked the U.S. Federal Trade Commission to stop publishing new commercial releases of GPT-4. Its quick and human-like responses to queries.
In Thursday’s complaint to the agency on the group’s website, the Center for Artificial Intelligence and Digital Policy called GPT-4 “biased, deceptive and a risk to privacy and public safety.”
OpenAI, based in California and backed by Microsoft Corp ( MSFT.O ), released the fourth iteration of its GPT (generative pre-trained transformer) AI project in early March. Conversation, composition of songs and compilation of long documents.
The formal complaint to the FTC follows an open letter signed by Elon Musk, artificial intelligence experts and industry executives calling for a six-month moratorium on the development of systems more powerful than OpenAI’s newly introduced GPT-4, citing potential risks to society.
The group says in its complaint that OpenAI’s ChatGPT-4 meets the FTC’s standard of being “transparent, explainable, fair, and empirically fosters accountability.”
For example, OpenAI exposed private chat histories to other users, and one AI researcher found it possible to “take over someone’s account, see their chat history, and access their billing information without their knowledge.” complaint
Mark Rotenberg, president of CAIDP and a senior privacy attorney, said there were business pressures pushing the company to release a product that wasn’t ready.
“Open AI simply does not comply with FTC guidelines, and there are concerns that the product is unfair and deceptive,” said Rotenberg, one of more than 1,000 people who signed a letter urging a freeze on AI tests.
The committee urged the FTC to “initiate an investigation into OpenAI, allow further commercial releases of GPT-4, and ensure that the necessary safeguards are in place to protect consumers, businesses and the commercial marketplace.”
Report by Diane Bartz; Editing by Mark Porter
Our Standards: Thomson Reuters Trust Principles.