Excitement and Existential dread...that's what I felt until this week about AI Engineering. My dread is gone, replaced with a faith in our role as the architects of the future. Let me tell you why.
I'm a professional teacher, and I'm very honest with my students. I would tell them "I don't know what world I'm sending you into." Almost like a commander sending troops into battle against AI and an industry in flux. I wanted a better story, and I was open to either the death of engineering or a future for our trade. I landed on the latter, in no small part thanks to the wonderful speakers at AI Engineer Miami. I want to leave you with hope, and here is how I landed there.
There's a common thinking that AI can replace human work. This is not false. But AI can't replace human ingenuity and judgement, taste as it's sometimes called. What makes AI useful is its ability to take our taste as input, and combine it with our collective knowledge and synthesize a plausible product. We can outsource the production of code, of images and other artifacts. We can't outsource thinking. Even though the products become more plausible every day, there's clear proof that depending on AI taste leads to bad results.
AI has bad taste for an intrinsic reason, one embedded into its structure. Taste, and ultimately the ideas it spawns are distinct from knowledge. Our intuition comes not only from our knowledge, but from interconnection with ourselves, our world, and the experience of others. AI can parse many things, but human ideas can't be fully tokenized. They are a physical process that starts in our minds and unfolds outside of a computer through our lived experience. We have always been the conduit between internal knowledge and external reality.
This knowledge is in abundance, and AI is a channel to access the collective knowledge of humanity. AI is a book that writes itself, and in an abstract sense it looks inwards towards it's training, to its pages for generative features. This mirrors how we operate, but we have our own features that AI doesn't replicate. Humans swim through the non-discrete world of ideas, and we act as a layer or conduit for ideas to interact through other humans. This happens through languages that LLM's and multi-modal models can parse, but the true understanding and genesis of ideas, like the problems themselves live outside of a computer. Even giving AI a body to act in the world doesn't give it compassion and love to direct that action, and that's how we solve the human problems that software is ultimately about.
Our future then is not as developers, development as we know it is dead. I saved the click bait for the middle. Our future is as AI architects. The work then becomes turning our judgement, intuition, and ideas into processes that AI can follow. We work back and forth with AI through these defined paths to plan and focus on quality over the promise of quantity. It should constrain our output where we have less vision, and amplify our output where we have the most understanding.
I'll end by saying we are all prepared for this future, even if you've yet to wrap your head around AI. The skills of development are the same as the skills for being an architect. They're just weighted differently. AI is a coder and a knowledge base, and so coding as the primary skill that defines a developer is not the primary skillset that defines an architect. Rather its ideation, review, critical thinking, problem solving, communication, and ultimately asking questions and applying critique and skepticism to direct AI's iteration. Its what we've always done differently and what has always separated a developer from an architect. If you write a skill and tell AI to be a developer, it will follow those instructions to a point. But it takes more than 200 instructions and 100k tokens to define competency as a developer.
So don't despair, have hope in the future of our profession. And start now. Know what an agent is, know how it works, be able to communicate this process to others. Start working back and forth with AI, refining plans, and critically reviewing the results. Mentor each other in these tasks, there's room at the table. AI is the finest tool, a lever to lift our dreams into reality. The finest physical tools leave no imprint in the product, and in the same way the work of AI and humans should ultimately be a human product, and solve human problems.
Top comments (0)