Multimodal Models

I still experience a moment of excitement, whenever I start on a new ml project and look at my prediction results for the first time, not sure whether this will be a good model or just garbage. I am not talking about metrics, but just about the actual prediction outputs. There is this thing that so many people working with algorithms may now take for granted, that actually sometimes touches me in this particular moment, when the predictions do ‘make sense’: A machine actually learned to produce something meaningful....

January 15, 2021