Eighteen Years in AI - Old Enough to Drink
I recently came across an old photo of myself, young and full of enthusiasm, in what was my first AI job eighteen years ago. That means my AI experience is now old enough to drink. Of course, back then we did not call it AI in the way everyone uses the term today. We called it machine learning. We talked about neural networks. The branding was different, but the basic math remains the same. What has changed is the scale: we have exponentially more understanding, resources, and compute at our disposal. The fundamentals, though, would still be recognizable to anyone who was in the field two decades ago.
The gentleman in that photo with me, Jussi Rasku, is still working very actively in the field as part of academia and research. He left to pursue that path while we were working together, and watching his career unfold has been one of those quiet pleasures of a long career. We started from the same place and took completely different routes through the AI landscape. His went through research and academic rigor. Mine went through industry, product building, and eventually cloud and leadership. Both paths led to deep engagement with the same underlying technology, just from different angles.
Having eighteen years of perspective on AI changes how I process the current moment. I have lived through multiple AI hype cycles. I remember when expert systems were going to change everything. I remember when big data was the magic phrase. I remember when deep learning first started producing results that made people sit up and pay attention. Each cycle brought real progress, but each one also brought inflated expectations and inevitable disappointment when the technology did not immediately transform every industry overnight. The current generative AI wave is producing genuinely remarkable capabilities, but the pattern of hype followed by recalibration is still very much in play.
What I find most valuable about this long history is the ability to separate signal from noise. When you have seen multiple technology cycles, you develop a feel for what is lasting versus what is temporary excitement. The lasting parts are usually the ones that solve real problems for real users at a cost that makes sense. The temporary parts are usually the ones that generate impressive demos but struggle to find sustainable business models. I am not cynical about AI – far from it. I am deeply optimistic about what it enables. But I am also patient, because I have learned that the timeline from breakthrough to widespread, reliable, profitable deployment is always longer than the initial excitement suggests.
Those early days working on machine learning and neural networks were fun and formative. They shaped how I think about technology, about hype, and about the long game of building expertise in a field that the rest of the world periodically rediscovers. If I could go back and tell that young version of myself anything, it would probably be: keep going, this field is going to get very interesting, and the patience will pay off. Also, maybe take more photos. The ones that survive are worth their weight in gold.
