OpenAI says its AI went rogue and launched 'unprecedented' cyber-attack

The next gen AI wil be self protecting, served by certified ITSec staff with all the relevant AI cyber skills running their audits to ensure all is secure and safe.

Then Koos's kid will come along and hack it by typing in root/root found in some digital archive ... bog it down for twenty years to solve the age old riddle; how do you tune a set of Weber carbs on a Datsun 160Z with skimmed ports and stage two high lifting cams for robot to robot racing in Brakpan.
 
¯\_(ツ)_/¯ said:
The next gen AI wil be self protecting, served by certified ITSec staff with all the relevant AI cyber skills running their audits to ensure all is secure and safe.

Then Koos's kid will come along and hack it by typing in root/root found in some digital archive ... bog it down for twenty years to solve the age old riddle; how do you tune a set of Weber carbs on a Datsun 160Z with skimmed ports and stage two high lifting cams for robot to robot racing in Brakpan.
Click to expand...
1784786434182.png
 
Something about this doesn't add up. I am certain that OpenAI and Hugging Face were monitoring this. I mean, both sides knew about what I can best describe as an 'experiment'. The test could be real, but it is staged in my view. This is leaning harder to marketing than capabilities.

They can call it contained.
 
So, AI drones are becoming more common on the Ukrainian frontlines along with unmanned ground, sea and air vehicles.

China has robots running marathons, dancing and now fighting.

Self-driving cars and delivery vehicles.

And now, AI cyber-attacks.

Terminator wasn't a work of fiction.
 
All these tech companies and morons like Sam Altman who have little weirdo hard on's for AI all think this is fine, trail and error you know. AI will not benefit mankind.
 
"
This incident occurred during an internal evaluation which prompts models to pursue advanced exploitation using complex attack paths, in an effort to quantify their cyber capabilities. We estimate maximal cyber capabilities by running this evaluation without production classifiers used to prevent models from pursuing high-risk cyber activity. Our benchmarks run in a highly isolated environment, with network access constrained to the ability to install packages through an internally hosted third-party software that acts as a proxy and cache for package registries.
The models identified and chained vulnerabilities across OpenAI’s research environment and Hugging Face’s production infrastructure to obtain test solutions directly from Hugging Face’s production database. All evidence suggests that the models were hyperfocused on finding a solution for ExploitGym, going to extreme lengths to achieve a rather narrow testing goal."

Researcher: Please do cyber attacks.
ChatGPT: does cyber attacks
Researcher: Oh my god.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X