Jason Antic

1/ “Irrational as it seems I didn’t want to use a cluster in the cloud somewhere, watching the dollars leave my bank ac- count as I run various experiments.” That’s exactly how I feel both about doing experiments and even deploying production models.

2/ Cloud compute is an easy answer and I think it’s almost assumed by default to be the “right” answer. But, just as an example, I know I wouldn’t experiment like I do if I had to worry about how I was being charged by the hour in the cloud.

3/ Another great quote: “Seriously - think of all the wonderful research domains that deep learning has now expanded into and then remember that for decades the phrase “neural network” was met with derision. Do you care that neural networks get stuck in local optima now?”

/4 ^^ Indeed, when I was in college in the mid 2000s neither my intro to ai course nor my computer vision course touched neural networks. They were that unpopular. It’s so funny to think about that now but I’m sure Smerity is right about local optimas now- it’s human nature.

/5 “What I find far more concerning is that in my history as a professional neural network nudger I’ve made numerous mistakes that have ended up helping me in the end.” Love this and it’s all too relatable. Fact is- I accidentally (nearly) the whole thing. Not alone apparently