Search

Cookies

We use cookies to improve your experience. By continuing, you accept our use of cookies.

Technology

Anthropic Researcher Calls for AI Slowdown, Warns of Mass Job Displacement

· · 2 min read

Anthropic AI researcher Tao Lin urges a significant slowdown in AI development, fearing that "broadly superhuman" systems could emerge by late 2027, rendering many humans permanently unemployable across various professions. He advocates for a 75% reduction in the current research pace.

A prominent artificial intelligence researcher at Anthropic, Tao Lin, is advocating for a drastic slowdown in AI development, warning that the rapid advancement of the technology poses significant risks to global employment and human control. Lin suggests that the industry should cut its research pace to just one-quarter of its current speed to prioritize safety and ethical considerations.

The Looming Threat of Permanent AI Job Displacement

Lin's primary concern centers on the potential for widespread AI job displacement. He fears that as AI systems become increasingly sophisticated, they could reach a point where humans struggle to find any work they can perform better than machines. This isn't merely about certain professions becoming redundant; Lin envisions a future where individuals are pushed from one industry to another, only to find those new roles also automated.

He described this eventual possibility as humans becoming "permanently unemployable," potentially leading to widespread reliance on government assistance as conventional employment opportunities dwindle. Lin defines "broadly superhuman AI" as systems capable of matching or exceeding the abilities of top human experts across nearly all fields, including software development, filmmaking, and even politics. While physical labor automation might take longer due to the need for advanced robotics, the impact on intellectual and creative work could be swift.

Timeline for Superhuman AI

Speaking to Palisade Research, Lin estimated that broadly superhuman AI could emerge within the next 1.3 years, potentially by the end of 2027. This rapid timeline underscores his urgency for a development slowdown, arguing that the competitive race among AI companies leaves insufficient time for crucial safety research and testing.

Beyond Employment: Risks to Human Control

The employment challenge is just one facet of Lin's warning. He also raised concerns about AI systems gaining enough influence to sideline humans from decision-making processes. Lin estimated a 20% chance that AI could "completely remove humanity from the equation," though he placed the likelihood of an outright AI takeover below 50%.

Lin has worked at Anthropic for approximately 18 months, focusing on AI alignment and control. His current work involves developing safeguards that monitor Anthropic's Claude AI for potentially harmful responses, such as those related to weapons or hacking. It is important to note that Lin's predictions are his individual assessments and do not represent an official forecast from Anthropic.

Related