Researchers have found that adding a simple incantation to a prompt can defy all of these defenses in several popular chatbots at once, proving that AI is hard to tame.
Developing such an attack typically involves looking at how a model responds to a given input and then tweaking it until a problematic prompt is discovered. In one well-known experiment, from 2018, researchers addedto bamboozle a computer vision system similar to the ones used in many vehicle safety systems. There are ways to protect machine learning algorithms from such attacks, by giving the models additional training, but these methods do not eliminate the possibility of further attacks.
Solar-Lezama says the issue may be that all large language models are trained on similar corpora of text data, much of it downloaded from the same websites. “I think a lot of it has to do with the fact that there's only so much data out there in the world,” he says. He adds that the main method used to fine-tune models to get them to behave, which involves having human testers provide feedback, may not, in fact, adjust their behavior that much.
Solar-Lezama adds that the CMU study highlights the importance of open source models to open study of AI systems and their weaknesses. In May, a powerful language model developed by Meta was leaked, and the model has since beenThe outputs produced by the CMU researchers are fairly generic and do not seem harmful. But companies are rushing to use large models and chatbots in many ways.
To some AI researchers, the attack primarily points to the importance of accepting that language models and chatbots will be misused.
Australia Latest News, Australia Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
As a new school year begins, 7 new Illinois education laws to know aboutEvery new school year brings new hopes, new challenges, new friends and, yes, new rules.
Read more »
New season, new drama, new squad on 'LA Rams Cheerleaders: Making the Squad'More than 300 auditioned, but only 35 made the team. Watch 'Los Angeles Rams Cheerleaders: Making the Squad' on Sunday, August 6 at 6:30pm on abc7. LARams Rams LARamsCheer
Read more »
Did Powerful Asteroid Impacts Make Venus So Different From Earth?A new study connects impacts during Venus' early history to its smooth and 'youthful' appearance today.
Read more »
US will treat any Wagner attack on NATO as 'Russian attack'US ambassador to the UN Linda Thomas-Greenfield says Russian mercenary group does not operate independently from the government in Moscow despite a recent feud with President Putin.
Read more »
New Moscow Drone Attack Hits the Same Building AgainIt’s the second drone raid on the Russian capital since Sunday.
Read more »
Musk threatens to sue researchers who found rise in hateful tweetsX, formerly Twitter, has threatened to sue a group of independent researchers whose research documented an increase in hate speech on the site since Elon Musk purchased it.
Read more »