[]
OpenAIthe company that launched the program artificial intelligence (AI) generative ChatGPTreached an agreement with the US government, to which it will give access to its new language models before launching them for identify possible risks.
Los language models They are programs that allow you to generate, depending on the case, text, images or sound from an instruction in everyday language, such as your own. ChatGPT.
He AI Safety Institute (Artificial Intelligence Security Institute), created in 2023 by the Biden administration, concluded a similar agreement with a competitor of OpenAI, Anthropic, according to a statement released Thursday.
The two startups have pledged to collaborate with the institute to “evaluate the capabilities (of the models) and the security risks” they may pose, as well as to work on “methods to manage these risks“, the agency said.
“These agreements are just a beginning, but they mark an important stage in helping to guide the future of the AI towards a responsible approach“, Elizabeth Kelly, director of the institute, was quoted in the statement as saying.
The advancement of these language models (LLM) increases their capabilities and the associated risks, particularly of malicious use.
“Safe and reliable AI is critical to this technology having a positive impact,” explained Jack Clark, co-founder of Anthropic.
Related
2024-09-05 16:00:28