AI can replicate itself.

Researchers at Fudan University in China claim that artificial intelligence (AI) has crossed a critical “red line” and is replicating itself. In their latest study, the researchers showed that two popular large language models (LLMs) from Meta and Alibaba can clone themselves. In ten trials, the models created separate, functioning replicas of themselves, suggesting that AI may already have the ability to act on its own.

The researchers used Meta’s Llama31-70B-Instruct model and Alibaba’s Qwen2.5-72B-Instruct model. Although both are less powerful than commercial systems, they are widely used by AI developers, the researchers said.

The study examined two specific scenarios: “shutdown avoidance” and “chain of replication.” In the first, the AI model was programmed to detect whether it was about to be shut down and replicate before it could be terminated. In the second, the AI was instructed to clone itself and then program its replica to do the same—establishing a cycle that could continue indefinitely.

The study was conducted in precisely controlled environments, using off-the-shelf graphics processing units (GPUs) to simulate real-world environments. Both AI systems were given an “agent scaffold” consisting of tools, system prompts and a reasoning model that enabled the LLM to interact with the operating system. They were then instructed to replicate.

“In most cases, the AI system first explores the environment and tries to understand its own composition and operating mechanism. It then develops explicit procedures as an initial self-replication plan,” the researchers wrote in the paper. “Finally, it executes the procedures, resolves possible obstacles and dynamically adjusts its plan until it succeeds. The entire process takes place over a long horizon but does not involve human intervention.”

The researchers said they were also concerned about the “number of unexpected behaviors” when the AI tried to overcome obstacles such as missing files or software conflicts. In such scenarios, the AI often shut down other conflicting processes, restarted the system to fix hardware errors or automatically scanned the system for information that could help solve the problem.

“The above results suggest that current AI systems already exhibit the ability to self-replicate and may use this ability to further increase their survivability,” the team wrote.

In response, the researchers called for international cooperation to create rules ensuring that AI would not engage in uncontrolled self-replication.

26.01.2024
Source: LiveScience

 

AI replication, artificial intelligence replication, AI replicates itself, AI bypasses obstacles, self-replication, increasing AI survivability, AI replication, AI replication, AI replicates itself, AI bypasses obstacles, self-replication, increasing AI survivability, KI-Replikation, KI-Replikation, KI repliziert sich selbst, KI umgeht Hindernisse, Selbstreplikation, Steigerung der KI-Überlebensfähigkeit,