Researchers have developed a way to tamperproof open source large language models to prevent them from being coaxed into, say, explaining how to make a bomb.
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.
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.
Open Source Meta Security Research
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
Landlords like me can make this city better, block by blockAs the owner of several rental properties in Philadelphia, I’ve learned that being a landlord is a way to build community.
Read more »
The Reimagined Chateau Opens at NemacolinNew accommodations, a new nightspot, and new amenities enhance the luxe factor.
Read more »
Bill Belichick's slew of media jobs a 'PR makeover': NFL insiderNew girlfriend, new view of the press and a new image.
Read more »
Captain America: Brave New World Repeats MCU’s Oldest Recasting Trick With Harrison Ford’s RossRobert Downey Jr. as Tony Stark in Iron Man 2 and Harrison Ford as Thunderbolt Ross in Captain America: Brave New World
Read more »
The magnet trick: New invention makes vibrations disappearDamping vibrations is crucial for precision experiments, for example in astronomy. A new invention uses a special kind of magnets to achieve this -- electropermanent magnets. They consist of a permanent magnet and a coil. In contrast to electromagnets, they do not have to be permanently supplied with energy.
Read more »
See it: Simone Biles to perform new trick at 2024 Olympics in ParisSimone Biles is set to perform a new trick at the Paris Olympics that has never been done before. And, if she pulls it off, it will mark something even greater…
Read more »