S2 E2: Oversaturations

 I tattooed fake skin for the first time since I started tattooing. I'm only 15-ish tattoos in, but  it's crazy how much you learn in a richer, more complete environment. The skin gives feedback to the machine, your clients muscle twitches, and your own hand shaking under the bright ring light. I can't imagine executing art in a better way. There's a robotics Sim2Real gap, and it's in tattooing as well. 

A brief delve into sim2real RL

In a situation where real, actual, live data is unavilable, we resort to the next best thing: simulation. Physics simulations for roboticists, and simulation skin for tattooers. Fake skin is an uphill battle in tattooing honestly. It's the worst of the both worlds, but a necessary evil. Honestly I think looking back, much of my fake skin work was not that useful, but I also don't think I could have done without it. I wonder what the current SoTA is for bridging the sim2real gap using few-shot reinforcement learning is. Unsure. Results inconclusive. But I definitely know there is absolutely nothing which I could have done in simulation could have prepared me for tattooing real skin, a safety-critical task might I add.

Unfortunately, it seems that the machine learning way of thinking has dominated my world-view :v. I feel like there's so much intuition in machine learning because... I've done a lot of human learning. You'd think it would be easy to tell someone how to learn. But the field is oh so saturated :(. I do not feel like parsing through 300 papers that propose iterations on the same thing and slap it into a new modeling environment. We gotta start being pickier with what papers actually get published lmao. It's time to gatekeep AI!! Also there's some insufferable people out there binded by their insecurities. 

Currents

I was also brutally recently reminded that not everyone has come to terms with the fact that they have insecurities (men). They are so insecure that they don't even want to admit that they are insecure. What a phenomenon. Yet another task for AI to tackle!! 

Alas, the quarter has ended. Kinda? I have nothing to do this last week, and yet I lowkey still gotta work doe. I'm getting into the work mindset though, brain go brrrrrrrrr. Sometimes it's good to just spend some time in my head and run around in circles. Every lap I learn something new, even if infinitesimal. Which im sure someone has written a beautiful rigorous proof about infinite sums. I love an integral. 

Last night, my friends and I studied the nonlinear behavior of this fascinating fidget spinner, made of spokes, WITH MORE THINGS SPINNING ON THE SPOKES. A simple and elegant idea, and yet it would invite so much chaos. We observed quasi-periodicity, we plotted phase plots, we screamed, we cried, we rejoiced. It is times like these that make me happy I live on a college campus, am pursuing a PhD, and get to stay up until 4:00 am thinking about bifurcations in chaotic systems, after drinking a Coors Light. Also two of my friends are dating!!! How wonderful :)

Recently have also been reading the Three Body Problem series. Hedonistically indulging in my no-touching-grass nerd tendencies. I've also been running though!!! So I feel less guilty :) playing minecraft for 5 hrs at night :). The series is so good though. I haven't been this entrenched in a sci-fi book series in a whileeeee. Honestly probably since the hunger games lmao. Okay, that's a lie. The last one was Hitchhiker's Guide to the Galaxy. Both bangers, and read on public transit :). 

Alright I'm going to massage my back. I wonder what the human learning analog to dropout regularization is. My learning rate lowkey kinda low tho. 

Saturation in ink, in fields, in markets, and in morale. Stay saturated out there.