AI Is a Tool, Not a Measure of Character
Criticism of generative AI often becomes contempt for people who use it. The fair standard is human agency, not tool purity—what you contributed, corrected, and learned.
The debate around generative AI has developed a bug: criticism of the technology keeps spilling over into contempt for the people using it. There are legitimate concerns about low-quality content, job displacement, consent, attribution, and the copyrighted work used to train these systems. These are not imaginary problems invented by people who fear the future. The U.S. Copyright Office has published a multipart report covering AI-generated work, digital replicas, and the use of copyrighted material in model training.1 We should continue arguing about those issues until we find better answers. But none of them explains why an ordinary person deserves ridicule for using AI to write, draw, learn, or communicate.
When criticism becomes contempt
The hostility becomes harder to defend when AI helps someone cross a gap they have struggled with for years. A person who has trouble writing may still have something important to say, just as someone who cannot draw may still have a vivid visual idea. AI can also help people improve their grammar, explore unfamiliar subjects, organize scattered thoughts, or ask questions they might be embarrassed to ask another person. In one controlled experiment, professionals using ChatGPT completed writing tasks 40 percent faster while the measured quality of their work increased by 18 percent.2 Another study of 5,179 customer-support agents found the largest gains among novice and lower-skilled workers, suggesting that AI can help people learn from practices that were previously concentrated among more experienced peers.3 That makes this less a debate about creative purity and more a question of who is allowed to benefit from a useful abstraction.
Abstraction is how we build better tools
Software engineering has been building those abstractions since long before anyone tried to put a chatbot inside a refrigerator. A compiler lets us write software in languages designed for humans and then translates that work through intermediate representations into machine code.4 We use libraries, frameworks, debuggers, autocomplete, documentation, and search engines because writing every application in processor instructions would be a terrible use of a Tuesday. Nobody calls a programmer fraudulent because Clang handled the machine code. The abstraction does not remove the programmer; it changes the level at which the programmer applies knowledge and judgment. AI may be another layer of abstraction, which means the useful question is not whether it performed part of the work, but what the person still contributed.
Assistance still requires judgment
That distinction matters because not every use of AI deserves the same defense. Typing one sentence, accepting the first result, and declaring yourself a visionary is closer to operating a vending machine than completing a creative process. Meaningful work still requires intent, judgment, correction, rejection, and enough understanding to recognize when the machine has confidently wandered into traffic. Research into AI-assisted short-story writing found this same tension: access to AI ideas improved individual stories, especially for people with lower baseline creativity, but also made the collection of stories more similar overall.5 AI can help an individual produce stronger work while still flattening the larger creative landscape when everyone accepts the same defaults. The evidence is neither a victory parade nor a funeral; it is a reminder that the quality of the result depends heavily on how the tool is used.
Judge the work, not the tool
A better standard is human agency, not tool purity. UNESCO’s guidance for generative AI in education recommends a human-centered approach that protects agency while allowing the technology to support learning, creativity, and research.6 We can apply the same principle outside the classroom by asking practical questions: Did the person bring the idea, make meaningful decisions, correct weak output, and learn during the process? When the answer is yes, the tool helped extend that person’s abilities rather than erase them. When the answer is no, or when AI is used to deceive, plagiarize, impersonate, or remove people solely to reduce costs, criticism is justified. This standard lets us confront harmful uses without appointing ourselves to the Software Purity Tribunal every time someone asks a machine for help.
Linus Torvalds made the same point bluntly when pushback treated the kernel as an anti-AI project: AI is a tool, like the others engineers already use, and a useful one.7 The useful follow-up is not whether someone used it, but whether the result holds up under ordinary engineering standards.
We should stop confusing unnecessary difficulty with human virtue. Tools have always helped people compensate for weaknesses, move past limitations, and work at a level they could not reach alone. AI can become a crutch, but a crutch is not inherently shameful; sometimes it is the thing that helps someone begin moving again. We can criticize exploitation, protect creators, demand honesty, and teach people not to surrender their judgment without humiliating everyone who uses the technology. A world where only naturally gifted, traditionally trained, or already-confident people are allowed to create is not protecting creativity. The tool is not a measure of character; what we choose to do with it is.
Sources
[1] U.S. Copyright Office, Copyright and Artificial Intelligence. The Office’s multipart study addresses digital replicas, copyrightability, and generative AI training.
[2] Shakked Noy and Whitney Zhang, Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence, Science, 2023. The experiment reported a 40 percent reduction in completion time and an 18 percent increase in output quality.
[3] Erik Brynjolfsson, Danielle Li, and Lindsey R. Raymond, Generative AI at Work, NBER Working Paper 31161. The study examined 5,179 support agents and reported larger productivity improvements among novice and lower-skilled workers.
[4] Clang Documentation, Command Guide (compilation stages). Clang documents the path from source parsing through intermediate representation, code generation, assembly, and linking.
[5] Anil R. Doshi and Oliver P. Hauser, Generative AI Enhances Individual Creativity but Reduces the Collective Diversity of Novel Content, Science Advances, 2024. The study found improvements in individual creative writing alongside greater similarity among AI-assisted stories.
[6] UNESCO, Guidance for Generative AI in Education and Research. The guidance promotes ethical, meaningful uses that preserve human agency.
[7] Joshua Morris, The Problem Wasn't the AI, joshuamorris.info. Note on Torvalds’s LKML framing of AI as an ordinary engineering tool, and judging assisted work by correctness, usefulness, security, and maintainability.
About Joshua Morris
Joshua is a software engineer focused on building practical systems and explaining complex ideas clearly.