On AI

I’ve had a bit of time to reflect on where my career has taken me over the years and one thing that comes up consistently is machine learning and expert systems. No one was calling it AI back then and there was still a sense of the past AI winter looming.

I was working on a government contract in the web3 days (the first web3, as in semantic Web technologies) using some inferencing software mostly in Python and Java. Tim BL was working on RDF and some other tagged info languages for metadata on Web pages. Turns out that this was all not really necessary in the end now that we have AI to summarize and do feature extraction from documents.

Another startup that I was at did a thing where we tried to predict which investors would invest in a dealroom scenario. This was done using a set of ML models that came out of the box in a Python library with some simple supervised learning techniques.

None of this was that groundbreaking but I suppose it was worth looking into along the way. No one suspected that the same basic techniques would change so dramatically with the introduction of modern transformers and large data sets.