The CEO of OpenAI, Sam Altman, emphasized the need for regulation of powerful artificial intelligence models during a panel with US lawmakers. Altman expressed concerns about the risks associated with AI and its potential to interfere with election integrity. He addressed a Senate judiciary subcommittee hearing, stating that OpenAI was founded on the belief that AI can improve various aspects of life but also carries serious risks.
The proliferation of increasingly sophisticated AI models in the market has raised concerns among critics and industry experts, who warn about the potential societal harms, including misinformation and prejudice.
Altman acknowledged the benefits of AI for society and stated that OpenAI’s generative AI has the potential to tackle significant challenges like climate change and curing cancer. However, due to the risks involved, Altman emphasized the critical role of regulatory intervention by governments in mitigating those risks.
During the hearing, Senator Richard Blumenthal played a recording of an AI version of himself reading remarks generated by ChatGPT. Blumenthal stressed the importance of testing AI systems and disclosing known risks before their public release. He highlighted that AI technology is no longer just a research experiment but a real and present force.
Senator Cory Booker acknowledged the global expansion of AI and the challenge of regulating it effectively. Senator Josh Hawley emphasized the transformative nature of AI, its implications for elections, jobs, and security. He stated that the hearing was a crucial first step in understanding what Congress should do.
Altman proposed guidelines and disclosure requirements for companies providing AI models, particularly in the context of elections. He suggested the consideration of licensing and testing requirements for AI model development. Altman also proposed the establishment of a regulatory agency that could license the most powerful AI systems globally and enforce compliance with safety standards.
Altman emphasized the importance of companies having the right to choose whether their data can be used for AI training, while publicly available web data would be fair game. He mentioned that AI models capable of persuading or manipulating beliefs could be subject to licensing.
Christina Montgomery from IBM, who also participated in the hearing, advocated for a precision regulation approach rather than focusing on regulating AI itself. This approach would involve establishing rules governing the deployment of AI in specific use cases.
Post Your Comments