The biggest mistake analysts make about Artificial Intelligence (AI) is thinking it’s one thing, moving in a single direction. (As in this illustration from Microsoft.)
This has caused them to focus exclusively on the battle among the top models – OpenAI, Grok, DeepSeek, Gemini, Co-Pilot, etc. – and miss what’s happening underneath.
I prefer to think of AI as a city. The big LLMs are the Interstates, but the real action is taking place on feeder roads, and a lot of life happens in what seem like quiet cul-de-sacs.
The human senses of seeing, of hearing, of touching, tasting, and smelling, these are all different AI disciplines right now, and efforts at integrating them can look like a mess. But breakthroughs are happening in all these areas, and more, which is why the AI city has so much life in it.
What’s clear to me is that there’s always a trade-off, among the size of a training database, the processing power available, and the output being delivered, that is incredibly useful. What we used to call a “good hack” – a program that achieved its goals with a minimum number of steps, and that was clearly documented – is coming back into vogue.
DeepSeek represents a triumph of this need for efficiency, but it’s not alone. There are opportunities for good hacks among all the AI disciplines, and success in one area often leads to breakthroughs in another. Or the creation of a new discipline.
AI Employment
AI was seen last year as a threat to employment. But it turns out to be just the opposite.
Expertise in every area of life, and of work, can be turned toward AI. Every specialty in the real world can assist in some form of AI development. The result isn’t some grand panjandrum replacing people at their highest level of creativity, but a growing set of tools that can help creative people achieve more and can make non-creatives a little more productive.
This is the way technology has always advanced. Jobs aren’t lost. They merely move into higher levels of abstraction. The process is easier to see as technologies themselves lose their usefulness. Tools like Google that were miracles 20 years ago are now nothing next to search engines like Perplexity. We also find that, instead of small tools evolving into massive tools, we must sometimes start over with a blank sheet of paper.
The earliest example of this in my career came in the rivalry between Microsoft and Apple. Microsoft insisted on backward compatibility. Apple would throw that aside every decade or two and start over, replacing the Apple II with the Mac, evolving toward iOS. Did it matter in the end whether your Windows NT machine could run your 1983 version of WordStar? No, in the end you just tossed that in an emulator and moved on, allowing the old stuff to be gathered up by the archivists and filed away.
Every tech revolution, in the end, turns out to be an evolution. The old ways may be buried like ancient civilizations, but they still inform us. This is how it will be with AI.
