AI vs. IA Timeline

People are currently terrified that an LLM might write a bad email for them (a visible error), yet they comfortably allow a social media algorithm to dictate their political reality or self-esteem (an invisible influence).

Let’s be brutally honest. If your email spam filter works, or if Netflix somehow knows you secretly love Sci-Fi space movies, you have been interacting with sophisticated Artificial Narrow Intelligence (ANI) daily for at least ten years.

Computer scientist Larry Tesler said, “AI is whatever hasn’t been done yet.”

Over the past 10 years, I’ve watched this fascinating psychological phenomenon play out: The "AI Effect." As soon as an AI problem is actually solved and becomes useful, like Google Maps routing your car, we immediately stop calling it "AI." We rebrand it as just "computing," a "feature," or an "algorithm".

We have historically reserved the scary label of "AI" only for things that remain mysterious. We got comfortable with invisible AI because it followed the philosophy of "Intelligence Amplification" (IA), it didn't try to be us; it tried to help us.

Then came Generative AI, and They Broke the Contract

The current friction we feel isn't because the AI is too smart; it's because the interface is too dumb.

As the timeline illustrates, we are witnessing a violent collision between two historical approaches. We have spent 80 years building an "Artificial Brain" but totally neglected the human needed to wield it effectively.

We currently have "God-like" models trapped in a 1960s-style chatbot text window (Same since 1966: ELIZA (Chatbot) Created by Joseph Weizenbaum).

Think about it: We are trying to drive a Ferrari engine (GPT-5.2) using the steering wheel of a Model T Ford (an empty text box).

"Prompt engineering" is not a profound new human skill. It is simply the manual crank we have to use right now because the starter motor hasn't been invented yet. It exists only because the AI is blind to your context and workflow.

The Practical Path Forward

We need to stop treating AI like a person we have to chat with and start treating it like a dynamic layer that lives inside our workflows. The future workforce doesn't need more "Oracles" that give advice; we need "Agents" that do the work; planning travel, drafting code, and awaiting our approval. This is why Microsoft Copilot is the most useful AI tool out there. It is embedded where I work.

The critical human skill of the next decade isn't "prompting." It's “Being Curious”: the ability to do your work in a workflow while at the same time wondering what your agent is doing or how it can help.

It’s time to close the Interface Gap. We don't need smarter models nearly as much as we need better cockpits. Let’s build something human.

What do you think?

Are you finding that "chatting" with AI in a separate window is becoming a bottleneck in your actual real-world workflows?

What's the most annoying thing you have to repeatedly explain to an AI today?

References

McCulloch, W. S., & Pitts, W. (1943). A logical calculus of the ideas immanent in nervous activity. Bulletin of Mathematical Biophysics.

Turing, A. M. (1950). Computing Machinery and Intelligence. Mind.

McCarthy, J., Minsky, M., Rochester, N., & Shannon, C. E. (1955). A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence.

Rosenblatt, F. (1958). The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain. Cornell Aeronautical Laboratory.

Weizenbaum, J. (1966). ELIZA—a computer program for the study of natural language communication between man and machine. CACM.

Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature.

Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet Classification with Deep Convolutional Neural Networks. NIPS.

Vaswani, A., et al. (2017). Attention Is All You Need. NIPS.