“Imagination is perishable, act on it immediately”
I think about this quote by Chris Willamson a lot, especially on days when I’m unable to write, like during a long hike in Shenandoah today.
This quote got me to think about AI, agentic systems, and the role of humans in the modern world. Many of us, especially software engineers, are afraid that we will eventually lose our jobs to AI because it will be able to do it better. A recent study by Pew found that 71% of Americans believe AI will lead to fewer US jobs. This percentage was higher among people under 30. I personally don’t think this will happen in the slightest. A software engineer’s role is not writing code, Claude can do that much faster and perhaps better. It is creating products and imagining solutions. If viewed right, AI is a tremendous asset to your career, not a replacement.
Imagination is a funny word. In Dutch, my native language, “inbeelding” has at least two meanings. I’m sure the same is try for the English “imagination”. The first meaning is purely mechanical like imagining a 3D geometric construction. Trying to imagine how many 3D objects intertwine to make a larger whole is very hard, a task many humans (including this one) have some trouble with. The same is true for Large Language Models (LLMs). They have no map of the 3D world, they cannot mentally picture lines, planes and curved structures. Therefore, Ais cannot come up with truly novel geometrical constructions in mathematics. They are very good at going through the mathematical literature, reasoning about theorems and equations, and proving or disproving conjectures, but a human must have come up with that conjecture in the first place. Ais cannot create truly novel mathematics because they can’t imagine it — they don’t know what would be important to progress the field.
The second meaning is about envisioning the future. Imagination is about what CAN be, not what is. AI agents have no goal, we need to give them one. Agents don’t know what to build, a point that has been made over and over by others. However, once they acquire said goal, agents are much faster at executing it, especially at writing code. Imagination is a fundamental driver of innovation. We, as humans, can envision a way to make the future better, and then create products and solutions that make us get there. AI is great at the latter, but the former is deeply human experience. Always has been and always will be.
If imagination is a fundamental driver of creation, pain is another. Pain is also a deeply human experience. It is often said that “necessity is the mother of invention”, well, pain causes necessity. Extreme heat made our schools, factories, places of work, and homes unlivable? Someone invented air conditioning. Managing large code bases among multiple developers who are making changes to it simultaneously? Linus Torvalds invented Git. Anesthesia was literally invented to avoid pain during surgery.
Large language models cannot feel pain. Neither can AI agents. One reason is that they do not interact with the physical world. But even if we put AI in robotics and let robots experience the world, I doubt they will feel pain in the same way humans do. For one, robots don’t have a biological nervous system that transmits pain signals, but they are also not necessarily unhappy carrying out the same task over, and over, and over again. Humans hate repetition.
In short, AI has no will to make the future happen because it does not experience as humans do. Imagination and pain are deeply human activities. They drive innovation. So, if you define your job as performing something technical like writing code, AI will most likely do it better soon. But if your purpose is to make a better future for humans, AI and robots will never take it.