Putting the Cart Before the AI (blog.tohojo.dk)

A recent commentary highlights the pervasive hype surrounding AI, particularly large language models (LLMs), while cautioning against the unsustainable business practices and rising failure rates seen in AI projects. The author argues that despite advancements, many generative AI tools still fall short, producing mediocre outputs and relying heavily on existing knowledge rather than true innovation. For instance, an AI-generated recommendations document proved too basic for specific needs, while another tool for kernel backporting offered limited but somewhat helpful insights, emphasizing that the gains are often incremental rather than revolutionary. Significantly, the article underscores the substantial costs associated with generative AI, including environmental impacts, social harms, and increased corporate control over technology. The author warns that reliance on AI could lead to decreased cognitive skills and accountability, creating a cycle of dependency. While aspirations for an open-source AI landscape exist, the current market is dominated by proprietary models that could exacerbate existing inequalities. The commentary concludes with a call for a shift in focus towards selecting the best tools for specific tasks rather than defaulting to AI as a blanket solution, urging a reevaluation of how the technology is integrated into society.