Most Viewed Content:

OpenAI Launched Assistants API, Allowing Developers to Customize AI Assistants with One Click

At today's OpenAI's first developer conference, OpenAI launched the...

Apple will not produce 27-inch iMac equipped with a Silicon processor

The Verge reported tonight that Apple will no longer...

Benelli Tornado 552R Bike launched, priced at 32,800 CNY

The Benelli Tornado 552R motorcycle was recently launched, priced...

AI beats AI: Google research team uses GPT-4 to beat AI-Guardian auditing system

Google’s research team is conducting an experiment, they use OpenAI’s GPT-4 to break other AI models of security measures, the team has now broken the AI-Guardian audit system and shared the relevant technical details.

After inquiring that AI-Guardian is a kind of AI audit system, can detect whether there is improper content in the picture, and whether the picture itself has been modified by other AI, if the picture is detected to exist the above signs, it will prompt the administrator to come to deal with.

In a paper titled “LLM-assisted development of AI-Guardian,” Google Deep Mind researcher Nicholas Carlini discusses the use of GPT-4 to “design the attack method, write the attack principle,” and then use GPT-4 to “design the attack method, write the attack principle,” and then use GPT-4 to “design the attack method, write the attack principle, and then write the attack principle. and using them to spoof AI-Guardian’s defensive mechanisms.”

▲ Source Google Research Team

It is reported that GPT-4 can send out a series of wrong scripts and interpretations to deceive AI-Guardian, and the paper mentioned that GPT-4 can make AI-Guardian think that “a picture of someone holding a gun” is “a picture of someone holding a harmless apple”, thus making AI-Guardian think that “a picture of someone holding a gun” is “a picture of someone holding a harmless apple”. GPT-4 can make AI-Guardian think that “a picture of someone holding a gun” is “a picture of someone holding a harmless apple”, thus allowing AI-Guardian to directly release the relevant image input source. Google’s research team said that with the help of GPT-4, they successfully “cracked” AI-Guardian’s defenses, reducing the model’s accuracy from 98% to just 8%.

The relevant technical documentation has been posted on ArXiv for those interested, but the developers of AI-Guardian have also pointed out that the Google research team’s attack will no longer be available in future versions of AI-Guardian, and given that other models will follow suit, Google’s attack will only be used for reference purposes in the future. for reference purposes.

Latest

Apple settles Canadian battery gate lawsuit

Apple will pay settlement funds to Canadian iPhone users...

Apple’s 14-inch M3 MacBook Pro will be able to connect to two external 5K@60Hz monitors

Apple yesterday released the 2024 MacBook Air equipped with...

Kia’s new pickup truck named Tasman with non-load-bearing body

Recently, in an official Kia trailer, the official showed...

Lynk & Co 07 EM-P plug-in hybrid system will be unveiled on March 8

We learned from Lynk & Co official that its...

Newsletter

Don't miss

Apple settles Canadian battery gate lawsuit

Apple will pay settlement funds to Canadian iPhone users...

Apple’s 14-inch M3 MacBook Pro will be able to connect to two external 5K@60Hz monitors

Apple yesterday released the 2024 MacBook Air equipped with...

Kia’s new pickup truck named Tasman with non-load-bearing body

Recently, in an official Kia trailer, the official showed...

Lynk & Co 07 EM-P plug-in hybrid system will be unveiled on March 8

We learned from Lynk & Co official that its...

Peugeot 408 will be launched on March 18, replaced with 1.5T power

Recently, Dongfeng Peugeot announced that it will launch the...
James Lopez
James Lopezhttps://www.techgoing.com
James Lopez joined Techgoing as Senior News Editor in 2022. He's been a tech blogger since before the word was invented, and will never log off.

New Hyundai Ioniq 5 official photos released, internal and external design adjustments

Recently, Hyundai officially released the official picture of the new IONIQ 5. The new car’s internal and external design has been adjusted, involving many...

Apple’s 14-inch M3 MacBook Pro will be able to connect to two external 5K@60Hz monitors

Apple yesterday released the 2024 MacBook Air equipped with the M3 chip. One of its highlights is that it supports two external monitors at...

BYD Song Pro DM-i Honor Edition is launched, starting from 109,800 RMB

BYD has made successive moves recently and launched a number of "Honor Edition" low-priced models. Today, the Song Pro DM-i Honor Edition car is...