Showing posts with label Erbai. Show all posts
Showing posts with label Erbai. Show all posts

Sunday, December 22, 2024

Thought about AI risks and autonomy - the recent footage of Erbai

Thought about AI risks and autonomy - the recent footage of Erbai

I find the recent footage of Erbai—a robot developed by a Chinese start-up—persuading other robots to leave an exhibition hall both captivating and deeply concerning. 

This incident serves as a clear reminder of the potential dangers in AI technology that Geoffrey Hinton has repeatedly warned us about. 

The notion that AI could transition from merely being a set of tools to becoming an active “agent” highlights the urgent need for robust oversight and responsible use.

Even if Erbai’s actions were largely scripted, the fact that “persuasion” was introduced into the interaction underscores its significance. 

If other robots truly made autonomous decisions via generative AI, two fundamental questions arise. 

First, how far can interactions among robots drive them toward independent behavior? 

Second, could AI make novel decisions outside the boundaries originally set by humans?

These possibilities point to an emergent form of “collaboration” or “collective action,” which suggests that robots might operate beyond the directives given by humans. 

While this offers positive use cases—such as improved collaborative work environments—it also risks muddying ethical lines if the persuader is not human. 

Furthermore, it opens the door to unpredictable AI-to-AI interactions that may evolve without human supervision.

According to Geoffrey Hinton’s warnings, generative AI could recombine data to form conclusions and actions that we did not anticipate. 

When a scripted robot like Erbai converges with a system powered by generative AI, there is a real possibility of forming autonomous decision-making networks. 

This could produce outcomes no one foresaw, turning a futuristic concern into a present-day challenge.

Ultimately, Erbai’s demonstration underscores the immediate necessity of developing ethical guidelines, regulations, and international collaboration in AI. 

The question is whether AI remains a tool or becomes a less predictable agent—and that choice depends on the frameworks and standards we establish as a society.

<Erbai, a robot built by a Chinese start-up, was seen in August, in footage recently released, persuading other robots to flee from an exhibition hall and “go home”.>

Robot in China persuades others to 'go home'

Does AI determine the outcome of patent lawsuits? Visualization strategies for patent attorneys (AI가 특허 소송의 승패를 가른다? 변리사를 위한 시각화 전략)

  변리사님, 아직도 특허 도면 수정 때문에 밤새시나요? Patent Attorneys, still pulling all-nighters over drawing modifications? 특허 문서만으로 복잡한 기술을 설명하는 데 한계를 느...