Peter Klein argues that artificial intelligence does not require a new economics: AI-related hardware, software, subscriptions, firms, labor effects, and investment can all be analyzed using standard Austrian concepts of subjective value, marginal utility, capital, entrepreneurship, and economic calculation. He views current AI as a powerful general-purpose prediction technology that will displace some tasks while augmenting others, potentially increasing worker productivity and making complementary human judgment more valuable, while warning that incumbent AI firms such as OpenAI and Anthropic are already lobbying for regulations that could raise rivals’ costs. Klein ultimately rejects the idea that current or foreseeable AI can genuinely replace human action or entrepreneurship, arguing that LLMs exercise only “derived judgment” on behalf of human decision-makers and remain tools rather than autonomous economic actors.
Top 5 Key Topics
AI needs no “new economics”: Klein rejects claims that AI changes the fundamental rules of economics, comparing the hype to the “new economy” rhetoric surrounding the internet in the late 1990s and early 2000s. Nvidia, OpenAI, Anthropic, Tesla, and AI subscriptions can still be understood through ordinary theories of firms, prices, capital, marginal utility, entrepreneurship, and profit.
LLMs are prediction machines, not minds: Klein describes ChatGPT and similar systems as sophisticated pattern-recognition systems that predict words or other outputs from enormous training datasets rather than genuinely creating novel thoughts. He is skeptical that scaling narrow AI into artificial general intelligence will produce something equivalent to the human mind and suggests AGI may be logically impossible.
Competition, pricing, and regulatory capture: Klein contrasts closed-weight frontier models such as GPT, Claude, Gemini, Grok, and Kimi with open-weight alternatives such as DeepSeek, noting how quickly competitive positions can change. He argues AI subscriptions may currently be subsidized below their true computing costs and accuses OpenAI and Anthropic of supporting strict regulation partly because it would “raise rivals’ costs,” including restricting Chinese competitors.
AI will change jobs, not eliminate human labor: Like elevators, telephones, electricity, computers, and earlier automation, AI will eliminate certain tasks while creating and enhancing others, and Klein sees no economic reason to expect permanent mass unemployment. He argues workers equipped with AI can become more productive and earn more, although eliminating entry-level “grunt work” could create a genuine problem if people no longer develop the experience needed for higher-level professional judgment.
AI cannot be an entrepreneur: Klein distinguishes human “original judgment” from AI’s “derived judgment”: an AI might write or evaluate an excellent business plan, but a human must decide whether to become an entrepreneur, deploy the AI, accept its recommendation, and commit resources. He therefore concludes that AI cannot perform entrepreneurship or economic calculation in Mises’s sense, regardless of whether its behavior convincingly passes the Turing test.
