Artificial Intelligence – part 3: The Hype, the Dangers, and the Resistance - Socialist Project

TEXT START: This is the third post in a multipart series aimed at cutting through the fog of artificial intelligence (AI) hype in order to help us understand some of the real dangers we face from AI use and highlight hopeful avenues of effective resistance.

The Dissection

The article performs a useful demolition of AI marketing. It documents agent failure, hallucination, rogue behavior, security incidents, consumer frustration, weak productivity evidence, misleading layoff narratives, and unsustainable pricing. Its political purpose is explicit: prevent AI fear from disabling worker resistance.

But it attacks the loudest version of the jobs-apocalypse story, not the Discontinuity Thesis itself. It treats current deployment friction as evidence against the terminal trajectory.

The Core Fallacy

The article confuses present capability, present adoption, and present employment data with structural limits.

A 30% task-completion rate, coding agents making developers 19% slower, or firms blaming ordinary layoffs on AI may prove that current agents are unreliable and that corporate claims are inflated. They do not prove that AI cannot eventually achieve durable cost and performance superiority across cognitive work.

The article also assumes that human oversight preserves human employment. Under DT logic, oversight is itself a target for automation, consolidation, and reduction. Reliability problems create a lag; they do not constitute a permanent moat.

Hidden Assumptions

• If layoffs are not directly caused by deployed AI, AI has not yet altered labor demand. This ignores hiring freezes, entry-level pipeline destruction, wage suppression, and anticipatory substitution.
• Current agent failure rates will remain stable as models, tooling, verification, and workflow design improve.
• Human checking and correction are cheap enough to preserve mass employment.
• Failures in open-ended coding and computer-use tasks generalize to all future AI applications.
• Short-term labor-market stability after roughly three years disproves long-term disruption.
• Resistance can maintain stable human-only economic domains at scale.
• Unprofitable AI companies cannot remain strategically dangerous. Capital can subsidize adoption while firms pursue control, market share, and labor replacement.

Social Function

Classification: partial truth functioning as transition management and, when overextended, ideological anesthetic.

The article is valuable because it strips away investor propaganda and exposes real technical and financial weaknesses. Its anesthetic effect begins when those weaknesses are presented as evidence that the underlying displacement mechanism is imaginary. It risks making workers complacent precisely by proving that the first wave is clumsy.

The Verdict

The article is accurate about the present: AI agents are overhyped, costly, unreliable, and dangerous. It is inadequate about the future: it mistakes the broken machinery of early deployment for proof that the factory cannot eventually run.

Under DT, these are lag defenses—technical failure, corporate deception, institutional caution, and consumer resistance. They may delay P1 and soften or redistribute P3. They do not defeat P2. If AI reaches durable superiority and firms can coordinate around it, mass employment dies regardless of whether the first generation of agents deletes databases, crashes websites, or makes programmers slower. The article is therefore a credible autopsy of AI hype, but an incomplete diagnosis of capitalism’s terminal mechanism.