AI is not “just a tool”

Summary

The speaker argues that describing AI as “just a tool” oversimplifies the technology and can excuse irresponsible use. In their view, “AI” includes more than an individual model: it encompasses LLM-based products and agents, the companies behind them, large data-center infrastructure, and AI features added to software.

They point to coding harnesses connected to chatbot APIs as an example of a product—not merely a tool. Such services depend on providers, may change or retire models, and could become more expensive if current token prices are subsidized. The speaker also raises questions about model training: who funds it, where the training data comes from, and what computational and environmental resources are involved. Running a model locally, they argue, does not remove those questions.

To clarify the distinction, the speaker compares AI with a power drill and text editors such as Emacs and Vim. These are purpose-built tools; if a drill malfunctioned, it would be considered defective. By contrast, LLMs can produce errors as part of their operation, and their use can involve broader societal and industry-level concerns. The speaker specifically criticizes long-running coding agents without adequate safeguards and says that calling them “just tools” shifts responsibility away from developers and companies.

The speaker further argues that the benefits currently promoted—such as faster software development or automatically generated presentations—do not yet outweigh the harms, in their assessment. They conclude that people should stop using the “just a tool” framing because it normalizes deployment they consider premature and irresponsible.

Additional Context

• Reviews, guides, or tutorials: None. The video is an essay reading and commentary.
• Main speaker/source: Brett Codes, reading his essay “AI Is Not Just a Tool.”