Artificial Insanity

The piece argues that current "artificial intelligence" - especially large language models from companies like Anthropic and OpenAI - is poorly named and fundamentally misrepresented: these systems are stochastic parrots, powerful probabilistic predictiors that cannot emulate consciousness or genuine judgment and will always produce unpredictable, potentially harmful token outputs. The real danger is not a machine uprising but humans misusing these chaotic predictors, for example placing them where their outputs can cause death or using them as accelerants for existing threats such as engineered pathogens or state surveillance and repression. Public scares about existential risk are painted as distractions amplified by media figures like Jacob Coxon and by companies that benefit from hype.

The motive for hype is financial and legal: firms have ingested massive amounts of digitized works with little respect for licenses and now depend on a narrative of revolutionary technology to justify valuations, regulatory capture, and efforts to reshape copyright so their mass training becomes permissible. Expect fights over open-weight models, distillation, and licensing while companies overpromise capabilities to investors and policymakers. The net effect is not a new species-level danger but a faster, more efficient consolidation of corporate power that accelerates existing socioeconomic and geopolitical harms.