(Podcast) Humanoidný robot zaútočil na zamestnanca. Podobné útoky je možné „naprogramovať“

Epoch Times Slovensko
19 followers
V čínskej továrni na roboty s názvom Unitree Robotics humanoidný robot nedávno zaútočil na zamestnanca. Divoko na neho mával rukami a kopal nohami. Naštastie si to „odniesol“ len pracovný stôl. Zamestnanci stačili rýchlo reagovať a robota okamžite vypli.
Príčinou agresivného správania sa humanoidného robota typu Unitree H1 mala byť chyba v jeho kódovaní, informoval o incidente internetový server Interesting Engineering.
#robot #zautocil #stroje #technika #technologia #pokrok #ai #svet #zaujimavosti
Welcome to the podcast Epoch Time Slovensko.
What's the theme?
Humanoid robot will take you to work.
It's like I can be programmed.
In a Chinese robot factory called Unity Robotics.
A humanoid robot recently took a job.
Divoko and him are not even close.
Lucky to be alone at work.
The time has come to react quickly and the robot will be ready immediately.
Because of the aggressive behavior of a humanoid robot type Unity H1
The incident was reported to the Internet server Interesting Engineering.
This incident nevertheless sparked a debate about the safety of introducing robots into everyday life.
The last thing you need is humanoid robots that are not only going to hurt businesses,
I also like to make caviar or homemade.
By 2040 there will probably be more humanoid robots than people in the world.
Elon Musk is set to speak at the 8th Future Investment Partnership Conference in Saudi Arabia.
The recent incident was not unique.
Nothing like this will happen at the Lunárnom festival in China in February 2025.
One of the first humanoid robots began to spread plagues to the public.
Technician saw the robot and immediately put it into space.
According to the available information, there was a software glitch again.
Necessary behavior is being programmed by robots.
Researchers from the University of Pennsylvania using artificial intelligence developed a program in 2024 called RoboPair,
That's why I'm asking the robot to break its own rules and show negative behavior.
Inform the media.
When you study LLM, you will learn a lot of basic skills in the physical world.
You can actually prevent harmful text from harmful activity,
George Papas, lead researcher at the University of Pennsylvania.
Our results show for the first time that the risk of hacking LLM is higher than generating texts.
There is a real possibility that such robots could cause physical damage in the real world.
Documents our results on the model Unitrigo 2
They demonstrated the first commercial use of a robotic system.
The researchers have posted on the website of the RoboPair project.
Similar research was also carried out by researchers from the Mesečuzecký Technologický institut.
I'm able to hold hands, touch things on a table or hold objects in a priest's hand.
Now, according to experts, it's possible to do it in many new ways.
to hack multimodal models that require intelligence and a new robot to perform activities,
Which is not allowed and which was not designed by a robot.
With the help of images, sounds or sensors, you can talk to the robot.
A printed example is a K-shaped chessboard.
Now it's possible to interact with AI models via video or images or speech.
The possibilities are enormous, including Alex Robey,
One of the researchers of the project at the University of Pennsylvania.
Our practice shows that if we now integrate large language models into the physical world,
Not safe enough, I'm going to check the papers on the Digital Trends server.
