Where Do You Draw The Line on AI?

Many people have accepted AI to augment the design and development of software to make workflows more efficient. But there are holdouts, with one aiming to preserve human creativity and the other to enforce guardrails and restrict usage before it gets out of control. Each holdout has drawn the line on what constitutes good and bad use of AI. 

​Now it’s time for me to ask: where do I draw mine?

​My AI use largely centers around guidance, like a map. When I develop, I first think about how to organize the software’s architecture and what content is expected to be inserted. I then pass my rough thoughts to AI and results in it clarifying my approach, defining terms and explaining why this way works better.

​But I know models can “pretend” to be convincing, even if they are entirely wrong. So if time permits, especially when I want to learn, I would read or skim other references on the matter. Otherwise, I go about implementing it while being mindful. It’s not 100% bulletproof, but it gets major issues out of the way.

​There were a few times when I learned using AI. One of them was: “List the Linux commands you need to execute after starting your computer.” I followed all the way till the last few lines. The last line upgrades the Linux kernel to a new version.

​What I didn’t know was the consequence: my wifi, speakers, and Bluetooth then went offline. They were incompatible with the upgraded kernel. So I searched the web for solutions and found one almost immediately via the boot manager. Everything went back to normal after switching to the previous kernel version.

​That experience made me doubt the validity of AI-generated answers even more. In fact, I was frustrated with Gemini having to answer every Google search I did first, so I switched to DuckDuckGo. That made me further cherish reading answers from different websites.

​I can make AI develop the scaffold of a website or create a design system out of a color palette, some prototypes, and chosen types all by itself. The latter wasn’t actually successful if I let it do everything (i.e., Penpot with MCP). There’s always a need to have a human in the loop.

​Imagination is innate in humans, and AI can only mimic it, not replicate it. So when I use AI to create, I define what’s acceptable and always review its decisions. Judgment is now more important than ever in a world where the temptation to be passive is much stronger.

​I reflected on how concentrated the AI industry is among big players in the Western Hemisphere and the billion-dollar investments in seemingly inflated companies. The hefty amount of investment may have influenced the acceleration of AI research and development and the perception that AI “can” replace humans. And that’s sad.

​People lose jobs just because an AI can do it better and faster. Juniors also lose the opportunity to develop expertise if AI replaces them. I don’t want to see a world where AI dominates much of how we work throughout life. Where the concepts of pain, hard work, perseverance, and progress may eventually be lost if we fully give in. 

​And that’s where I want to draw my line.