AI is not “just a tool”

People are tired of being told that "AI is just a tool" because it's clearly incorrect. Labeling large language models (LLMs) as "just a tool" doesn't do anything to help or justify them. It's a simple phrase used to either downplay the complexity of the technology or to justify the use of unsafe products made by irresponsible companies.

Software engineers tend to say "it's just a tool" as an easy way to avoid dealing with the complicated aspects of LLMs. This phrase appears in tech organizing circles, justifications for using LLM-generated code in the Linux kernel, journalism articles, and more. AI is not simply a tool. It's an industry driven by individuals with questionable ethics.

LLMs, agents, and the huge growth of data centers are all part of AI. The irresponsible ways these terms are used right now mean more than just a tool. AI today means a lot more than just a tool. For instance, using a coding harness connected to a chatbot API to write code isn't just a tool; it's a paid product. The model you use today might be retired, and the cost will increase as tokens are subsidized and companies aren't focused on making a profit yet.

Companies building these products are either irresponsible themselves or rely on others who are, pushing for high market valuations and personal ambitions. That's not "just a tool." Sure, you could run a local LLM model on your computer to generate SQL for a report or create awful art for a D&D campaign with friends. But who made that model?

Who trained it? Where did the data come from? What were the environmental costs and the company's motivations? Let me explain what a tool is. A tool is something crafted or adapted to do specific tasks. A power drill is a tool, almost like "just a tool." It involved engineering, manufacturing, and resource extraction. The power drill is more efficient than a hand drill.

The technology behind motors, batteries, and plastic molding all came together to make it easier to bore a hole. While the power drill is also a product, the companies that manufacture it aren't talking about the potential to wipe out humanity. Using a power drill doesn't make it more likely to stunt critical skills in carpenters.

My power drill doesn't "sometimes make mistakes" and go the wrong way. It would be considered defective if it did. Emacs and Vim are just tools too. They're pieces of software made by people to accomplish specific tasks—editing text. They're not industries or companies. They don't require massive resources to use or create. They are not inherently dangerous.

Of course, someone could use Vim to write a computer virus, but that's the application of the tool, not a problem with the tool itself. If I type something like "try to hack into this server" into Vim, it won't commit a cybercrime. It's absurd to compare long-term coding harnesses hooked up to probabilistic LLM APIs to "just a tool."

When people say AI is "just a tool," they're avoiding responsibility for the dangerous and irresponsible use of a technology that's being handled in a dangerous way. By using AI products, we're demanding something we haven't even begun to understand the potential harms and benefits. The leaders of AI companies continue to say outrageous things, causing mass anxiety.

We need to stop perpetuating the "just a tool" rhetoric that enables the continued behavior of these companies and their leaders. So, please stop saying "AI is just a tool" because it couldn't be further from the truth and causes more harm than good.