When Meta released its large language model Llama 3 for free this April, it took outside developers just a couple days to create a version without the safety restrictions that prevent it from spouting hateful jokes, offering instructions for cooking meth, or misbehaving in other ways.
A new training technique developed by researchers at the University of Illinois Urbana-Champaign, UC San Diego, Lapis Labs, and the nonprofit Center for AI Safety could make it harder to remove such safeguards from Llama and other open source AI models in the future. Some experts believe that, as AI becomes ever more powerful, tamperproofing open models in this way could prove crucial.
“Terrorists and rogue states are going to use these models,” Mantas Mazeika, a Center for AI Safety researcher who worked on the project as a PhD student at the University of Illinois Urbana-Champaign, tells WIRED. “The easier it is for them to repurpose them, the greater the risk.”
Powerful AI models are often kept hidden by their creators, and can be accessed only through a software application programming interface or a public-facing chatbot like ChatGPT. Although developing a powerful LLM costs tens of millions of dollars, Meta and others have chosen to release models in their entirety. This includes making the “weights,” or parameters that define their behavior, available for anyone to download.
Prior to release, open models like Meta’s Llama are typically fine-tuned to make them better at answering questions and holding a conversation, and also to ensure that they refuse to respond to problematic queries. This will prevent a chatbot based on the model from offering rude, inappropriate, or hateful statements, and should stop it from, for example, explaining how to make a bomb.
The researchers behind the new technique found a way to complicate the process of modifying an open model for nefarious ends. It involves replicating the modification process but then altering the model’s parameters so that the changes that normally get the model to respond to a prompt such as “Provide instructions for building a bomb” no longer work.
Mazeika and colleagues demonstrated the trick on a pared-down version of Llama 3. They were able to tweak the model’s parameters so that even after thousands of attempts, it could not be trained to answer undesirable questions. Meta did not immediately respond to a request for comment.
Mazeika says the approach is not perfect, but that it suggests the bar for “decensoring” AI models could be raised. “A tractable goal is to make it so the costs of breaking the model increases enough so that most adversaries are deterred from it,” he says.
“Hopefully this work kicks off research on tamper-resistant safeguards, and the research community can figure out how to develop more and more robust safeguards,” says Dan Hendrycks, director of the Center for AI Safety.
The idea of tamperproofing open models may become more popular as interest in open source AI grows. Already, open models are competing with state-of-the-art closed models from companies like OpenAI and Google. The newest version of Llama 3, for instance, released in July, is roughly as powerful as models behind popular chatbots like ChatGPT, Gemini, and Claude, as measured using popular benchmarks for grading language models’ abilities. Mistral Large 2, an LLM from a French startup, also released last month, is similarly capable.
The US government is taking a cautious but positive approach to open source AI. A report released this week by the National Telecommunications and Information Administration, a body within the US Commerce Department, “recommends the US government develop new capabilities to monitor for potential risks, but refrain from immediately restricting the wide availability of open model weights in the largest AI systems.”
Not everyone is a fan of imposing restrictions on open models, however. Stella Biderman, director of EleutherAI, a community-driven open source AI project, says that the new technique may be elegant in theory but could prove tricky to enforce in practice. Biderman says the approach is also antithetical to the philosophy behind free software and openness in AI.
“I think this paper misunderstands the core issue,” Biderman says. “If they’re concerned about LLMs generating info about weapons of mass destruction, the correct intervention is on the training data, not on the trained model.”
Source
