SAN FRANCISCO (KCBS RADIO) – Dealing with an evil cyber-intelligence probably sounds more like a movie than real life (spoilers: it is in fact the plot of at least one very popular film franchise). It may be closer to reality than we realize – some researchers even think we should begin preparing for artificial intelligence (AI) going rogue.
For more, stream KCBS Radio now.
A study recently submitted to the ARXIV archive found that AI systems driven by Meta's Llama31-70B-Instruct and Alibaba’s Qwen25-72B-Instruct “have already surpassed the self-replicating red line.” That means that the systems have succeeded in creating copies of themselves – 50% of the time for Meta’s and 90% of the time for Alibaba’s.
“Successful self-replication under no human assistance is the essential step for AI to outsmart the human beings, and is an early signal for rogue AIs,” the researchers said. “That is why self-replication is widely recognized as one of the few red line risks of frontier AI systems.”
Since the release of OpenAI’s ChatGPT program, artificial intelligence awareness and research has grown. Leading companies in the tech world, including Meta, Google and Apple have been investing in AI models and machine learning. President Donald Trump announced a $500 billion joint venture into building AI infrastructure in the U.S. just days after he was inaugurated last month.
This Friday, The Intercept reported that as “the Trump administration and its cadre of Silicon Valley machine-learning evangelists,” settle in, the Internal Revenue Service is also “preparing to purchase advanced artificial intelligence hardware,” citing procurement materials reviewed by the outlet.
According to its report, multibillionaire Elon Musk – CEO of Telsa and SpaceX and the leader of the newly-created Department of Government Efficiency (DOGE) – is installing DOGE into the IRS. Soon, the IRS’ computing center in West Virginia is also expected to house a Nvidia SuperPod AI computing cluster. These systems start at around $7 million.
“Though small compared to the massive AI-training data centers deployed by companies like OpenAI and Meta, the SuperPod is still a powerful and expensive setup using the most advanced technology offered by Nvidia, whose chips have facilitated the global machine-learning spree,” The Intercept said. It added that how the IRS plans to use the SuperPod is unclear, as is which administration actually ordered the technology.
Audacy also reported this week on how Google’s updated artificial intelligence principles removed a previous commitment not to use the technology for weapons or surveillance, and the surprise move from the U.S. not to sign an international agreement regarding AI. The pledge called for “trustworthy AI in the world of work.”
Trump has recently raised concerns about China’s development of AI models. When responding to reports about DeepSeek, a Chinese AI model in competition to U.S.-based OpenAI, the president said it was a “wakeup call” and that the U.S. has to be “competing to win” when it comes to AI.
That’s just one area of concern, according to researchers.
Authors of the study focused on AI’s ability to replicate said that “by analyzing the behavioral traces, we observe the AI systems under evaluation already exhibit sufficient self-perception, situational awareness and problem-solving capabilities to accomplish self-replication.” They also found that some “AI systems are even able to use the capability of self-replication to avoid shutdown and create a chain of replica to enhance the survivability, which may finally lead to an uncontrolled population of AIs.”
If we eventually lose control over the frontier AI systems, the researchers said the systems “could take control over more computing devices, form an AI species and collude with each other against human beings.” This finding comes as another technological development gets its footing – humanoid robotics. Bloomberg’s Mark Gurman recently spoke to KCBS Radio about the nascent field.
“Well, it’s the stuff out of The Jetsons, sci-fi and movies like Will Smith’s iRobot… right?” he said. “Humanoid robots are coming sooner than later.”
Gurman said that Meta – one of the companies that the study identified as having an AI system capable of replication – wants to be at the core of that field. Other companies working in the space include Apple and Telsa. Meta is particularly interested in developing and selling robots that can perform home chores and tasks such as folding laundry.
“So, Meta really believes that with this AI, it can build something centric to the home,” said Gurman. “Whereas you've seen other companies so far really focus on enterprise use cases, warehouse management, manufacturing and the like.”
Still, humanoid robots are in the early stages of development, he added. For now, many big tech companies are investing in their AI programs.
Authors of the recent study advise caution on that front.
“Our findings are a timely alert on existing yet previously unknown severe AI risks, calling for international collaboration on effective governance on uncontrolled self-replication of AI systems,” they said.
DOWNLOAD the Audacy App
SIGN UP and follow KCBS Radio
Facebook | Twitter | Instagram