I need to check my notes because I'm not sure if I've mentioned before that I've been working at an AI startup for the past couple of years. That's part of the reason I stopped posting as much on social media. Working at a small startup had a bit of a chilling effect I guess, I definitely didn't want to be fired for being critical of anything, and after working at an org like NCMEC for a decade there were naturally a lot of, "well at my old job we used to do it this way" style thoughts that it's probably best that I kept to myself anyway.
I have to admit that when I joined my new org I wasn't what you would call a "true believer" in AI, I wasn't even mildly interested to be perfectly honest. I somehow managed to luck into a job in the hottest field in tech at the moment. This is a completely new experience for me, my LinkedIn profile gets a lot of traffic these days and I'm positive it's because of where I work because I haven't updated it in years. So I wanted to kind of take a moment here now that I'm feeling that I've rounded a corner, or leveled up my understanding of AI in general and LLMs in particular to document the process and note down how my feelings have changed.
So, how did I end up here? A few years ago I realized that I'd outgrown my previous job (big fish in a small pond situation) and it was time to move on to new things, or stagnate and die. So I took the plunge and put my feelers out. The market was extremely hot at the time so I only ended up interviewing at ~3 places before someone extended me an offer. The pay increase was significant, and they were asking me to step down from manager/team lead to individual contributor, so way less responsibility, I took it obviously.
I knew that going from large non-profit to a for-profit startup was going to require a big context switch on my part but I don't think I could have prepared myself for the speed that things move. When I joined, I thought it was kind of weird that the company hired me under one name, but all of the internal resources were still sporting an older branding from the previous iteration of the company. I tried to get my bearings by digging through source code (as one does) and looking up the developers. The first thing that jumped out at me was that the vast majority of the old developers were gone, some very recently, giving me the impression that I was a backfill. The other odd thing was that most of these projects were not very old. Shortly after that I learned what "Pivoting" was, something startups do while they are trying to find their way, making stuff and seeing what sticks. Apparently, I'd joined right in the middle of pivoting away from one thing and into the news search engine business, so that was the first pivot.
I was originally hired as a C# developer to work on .net core projects that supported an AI powered search engine that was using LLMs to detect political lean (among other things) in news articles. Everything I worked on was "high priority", and then suddenly I was asked to work on a Java Spring Boot API to support a new service that would be rating podcasts for their content, very similar to what we were doing with news, and that was the second pivot.
After that I was asked to learn python and work on anther, totally different API to support training large language models and serving them, so that was the third pivot. This one seems to be sticking, I guess we'll see.
It took about three years to go through the three pivots, the first one I was just getting my bearings, figuring out the company. By the second pivot I'd gained confidence in what we were doing, enough that I didn't run screaming for the hills when they asked me to learn Python at the beginning of the third pivot. After that I started to really dig into LLMs and get my hands dirty, and somewhere along the way I realized that LLMs really are the next big thing, and even though it already feels like they are everywhere, it's only a matter of time before they are truly everywhere. I imagine in the future most children's toys will run some kind of LLM, and we won't even bat an eye at it.
It took a while for me to settle on a predictive position about where I think things are going, are we going to crack gen ai and super intelligence? When I joined the company, I was firmly in the "nah" camp, but now I'm definitely a "strong maybe" but even if we don't get there, what we already have today even if it stopped improving at the breakneck pace that it's making now, is already extremely revolutionary, and we're only scratching the surface of what's really possible here.