AI allegedly differs from prior technologies by automating any human task including new jobs it creates

The post reframes AI as fundamentally different from prior technologies by positing it can automate any human task, including those that arise as replacements for displaced work. This contrasts with historical automation, where new roles emerged in sectors like robotics or software. The argument rests on a recursive premise: if AI outperforms humans in existing jobs, it could also outperform them in any new jobs created, leading to a closed loop of displacement.

The author dismisses objections about AI’s limitations, such as creativity or emotional labor, as either temporary or economically irrelevant. For engineers, this implies that no task is inherently resistant to automation, even those requiring adaptability or judgment. The post suggests the burden of proof has shifted: rather than assuming AI *cannot* do something, the default should be that it *can*, given sufficient data and investment.

The economic implications are stark. If AI can recursively automate all work, traditional models of job creation through technological progress no longer apply. The post acknowledges a niche market for ‘human-made’ goods, but frames it as a rounding error in the global economy. For engineers, this challenges assumptions about long-term demand for human labor, even in roles that seem insulated from automation today.

The argument’s strength lies in its logical structure, but its weakness is the lack of empirical evidence for AI’s ability to fully replicate human adaptability. The post assumes that data collection and scaling will overcome any current limitations, but this remains unproven. Engineers should note that while the theoretical framework is compelling, real-world constraints, such as energy costs, regulatory barriers, or unintended consequences, could disrupt the scenario.

The post’s framing has practical consequences for tooling and system design. If AI can automate any task, engineers may need to prioritize systems that augment AI’s capabilities rather than those that rely on human oversight. This could accelerate shifts in software architecture, favoring modular, data-hungry models over bespoke solutions. However, the post does not address whether such systems are sustainable or scalable in practice.