The author argues they are not “anti-AI,” but specifically opposed to large language models (LLMs) and the hype surrounding them. They reject the core assumption that statistical word-prediction systems are a good basis for general-purpose tools that answer questions, generate text, or write code. Even if trained ethically and powered by renewable energy, they still believe LLMs are fundamentally flawed; commercial versions are worse because they also raise consent, copyright, and environmental concerns.
A major point of the essay is that the term “AI” is too vague to be useful. Historically, it has referred to many different technologies, including rule-based systems, machine learning, deep learning, and now LLMs. Because “AI” now often means only chatbots or next-token predictors, critics of LLMs are easily misrepresented as opposing unrelated technologies like image recognition, medical diagnostics, or translation systems that are not the real target of their criticism.
The author also rejects the label “anti-AI” because it is politically and conceptually muddled. They argue that some “AI safety” or “pause AI” campaigns are backed by industry interests and focus on hypothetical superintelligent systems rather than the real-world harms of current LLMs, such as labor replacement, poor performance, and environmental costs. The essay concludes that people should use precise terms—like LLMs, expert systems, or machine learning—rather than the overloaded word “AI.”
Article timeline
Build this article’s timeline
Extract dated events, key actors, and open questions using the same timeline analysis available in Signal.
Comments
Login to view and post comments.