Generative AI might end up being worthless – and that could be a good thing (theconversation.com)

Media and investors have rushed to assume generative AI will be wildly profitable, but this piece argues a different, plausible future: genAI could prove economically worthless. High run-time compute costs, questionable monetization (even paid ChatGPT tiers reportedly lose money), and an $800 billion estimated revenue shortfall for major firms mean the playbook of burning cash to gain users may fail. If ad-funded models become the default, “enshittification” could degrade quality without closing the gap between expenses and revenue, and investor enthusiasm could wane — drying up seed funding and hobbling startups. Technical and legal realities deepen the risk. Large models memorize and reproduce copyrighted content (one model reportedly recalled 42% of the first Harry Potter book), exposing firms to expensive lawsuits and settlement attempts (an attempted $1.5B-style payout to authors was challenged in court). Open-source releases like Meta’s Llama and unexpected competitors (e.g., a Chinese open model that briefly rocked markets) make free, locally runnable models commonplace, undercutting commercial pricing power. The net effect: genAI may become a “toxic asset” — useful but not monetizable — which could stall ambitious deployments, shift value back to accessible “good enough” tools, and limit big-tech dominance. For the AI/ML community this means revisiting business models, legal strategies, and the ethics of data sourcing, while possibly preserving wider access and slowing centralization.