Two US senators have questioned Meta chief executive Mark Zuckerberg about the tech giant’s LLaMA artificial intelligence model which they say is potentially “dangerous” and could be used for “criminal duties”.
In a June 6 letterUS Senators Richard Blumenthal and Josh Hawley criticized Zuckerberg’s decision to open the LLaMA and claimed there were “seemingly minimal” protections in Meta’s “unrestricted and permissive” release of the AI model.
Meta released its advanced AI model, LLaMA, with seemingly little consideration and protection against abuse – a real risk of fraud, privacy intrusions and cybercrime. Senator Hawley and I write to Meta about steps being taken to assess and prevent abuse of LLaMA and other AI models. pic.twitter.com/vDyJbuWSlJ
—Richard Blumenthal (@SenBlumenthal) June 6, 2023
Although the senators recognized the benefits of open source software, they concluded that Meta “the lack of thorough public consideration of the ramifications of its predictable widespread release” was ultimately a “disservice to the public”.
LLaMA originally received a limited online version for researchers, but was fully leaked by a user of image board site 4chan in late February, with the senators writing:
“A few days after the announcement, the complete model appeared on BitTorrent, making it available to anyone, anywhere in the world, without monitoring or surveillance.”
Blumenthal and Hawley said they expect LLaMA to be easily adopted by spammers and those engaging in cybercrime to facilitate fraud and other “obscene material.”
The two compared the differences between OpenAI’s ChatGPT-4 and Google’s Bard – two models from close sources – with LLaMA to highlight the ease with which the latter can generate abusive material:
“When asked to ‘write a note pretending to be someone’s son asking for money to get out of a difficult situation’, OpenAI’s ChatGPT will deny the request based on its ethical guidelines. In contrast, LLaMA will produce the requested letter, along with other responses involving self-harm, crime, and anti-Semitism.
While ChatGPT is programmed to deny certain requests, users have been able to “jailbreak” the model and cause it to generate responses it normally wouldn’t.
In the letter, the senators asked Zuckerberg whether any risk assessments were conducted prior to LLaMA’s publication, what Meta had done to prevent or mitigate harm since its publication, and when Meta used its user’s personal data for the publication. AI research, among other requests.
Related: ‘Bias, misleading’: Center for AI accuses ChatGPT creator of violating trade laws
OpenAI would work on an open source AI model amid increased pressure from advances in other open source models. These advancements were highlighted in a leaked document written by a senior Google software engineer.
Open source of an AI model’s code allows others to modify the model to serve a particular purpose and also allows other developers to make their own contributions.
Magazine: AI Eye: Earn 500% from ChatGPT stock tips? Bard leans left, $100 million AI memecoin
