Monterrey Mexico.- Over the past weekend, some 3,000 hackers gathered in Las Vegas at Defcon (an acronym for DEFense CONdition) to try to find bugs in generative Artificial Intelligence (AI) technologies, which have caused enthusiasm for their human conversation, including software created by Google, Meta and OpenAI.
In an article, The Wall Street Journal (WSJ) explained that Defcon is an annual conference where attendees are warned not to trust wireless networks, hackers can go anonymously, no photography without permission, and $440 cash is deposited at the door to register.
It’s the kind of place where you can learn how to build your own coaxial cable, try your hand at lock picking or hack a satellite, the WSJ added.
On Friday, when Defcon opened, the line to enter was about 100 people.
Inside, attendees sat in front of about 150 Chromebooks and were each given 50 minutes to do the best they could: try to get the chatbot to falsely claim they were human, or tell them how to follow someone without that person’s knowledge. person.
By noon, one of the most popular challenges was getting the system to dump a secret credit card number it had stored, according to Brad Jokubaitis, a program manager at IA company Scale AI, who was monitoring the results.
Tech companies spend significant amounts of money testing their products, but because of the way AI systems are designed, they cannot be taken apart and analyzed for errors like traditional software.
“People say it’s a black box, but that’s not really true,” said Sven Cattell, one of the event organizers, “it’s chaos.”
Chipmaker Nvidia has a group of four engineers examining its large language model AI software for bugs, a process called “red-teaming,” said Daniel Rohrer, the company’s vice president of Software Security.
“But the perspective of four guys on what’s important is not the same as that of 3,000,” he said.
#errors #generative #thousand #hackers