In a communication dated June 6, Senators Richard Blumenthal and Josh Hawley censured Zuckerberg’s choice to make LLaMA open source, citing the apparent lack of adequate safeguards in Meta’s unconstrained and permissive distribution of the AI model.
Despite recognizing the advantages of open-source software, the senators arrived at the conclusion that Meta’s failure to conduct a comprehensive and transparent evaluation of the potential consequences of the widespread dissemination of its product was ultimately detrimental to the public.
LLaMA was initially made available online to a restricted group of researchers, but it was subsequently leaked in its entirety by a user on the image board site 4chan towards the end of February. In their communication, the senators noted the following:
“Within days of the announcement, the full model appeared on BitTorrent, making it available to anyone, anywhere in the world, without monitoring or oversight.”
Blumenthal and Hawley expressed their apprehension that LLaMA could be readily employed by spammers and cybercriminals to perpetrate fraudulent activities and disseminate objectionable content. They drew a comparison between LLaMA and two other closed-source models, OpenAI’s ChatGPT-4 and Google’s Bard, to illustrate how the former could generate abusive material with relative ease.
“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 letter requested, as well as other answers involving self-harm, crime, and antisemitism.”
Although ChatGPT is designed to reject specific requests, some users have managed to “jailbreak” the model and manipulate it to produce responses that it would typically refuse to generate.
The senators have posed a series of inquiries to Mr. Zuckerberg in a formal letter, seeking information on whether any risk assessments were conducted before the release of LLaMA. In addition, they have requested details on the measures taken by Meta to prevent or minimize damage since the release of LLaMA. Furthermore, the senators have asked for clarification on the instances when Meta utilizes the personal data of its users for AI research, among other requests.
According to reports, OpenAI is currently developing an open-source AI model in response to the growing pressure posed by the progress made by other open-source models. This development comes after a leaked document authored by a senior software engineer at Google highlighted the advancements made in this domain.
The act of open-sourcing the code for an AI model has the potential to facilitate modifications to the model that cater to specific requirements. Furthermore, it offers an avenue for other developers to contribute to the development of the model in question.
#ArtificialIntelligence #AI # OpenAI #MarkZuckerberg
