As I noted in a previous post my brain cells and time on the weekends have been lacking. I just finished the 3rd week of my program at Insight AI, and it’s going great. I have been building out my version of a recent research paper generating realistic human (restaurant) reviews, and it works!
I have to admit in the middle of week 2 it hit me that there was a real likelihood that I might fail. I thought I was going to have to pivot my presentation and explain what I learned, or be satisfied with subpar results, but after many hours of attempting different configurations and letting them run for 10+ hours at a time things started to come together.
It’s such an empowering feeling being able to read a research paper published by academics, and being able to turn that into working code. I truly feel like I can build an initial version of any AI application now (of course within current technological possibilities) given a small amount of time to research the problem, gather/clean the data, and experiment with different machine learning algorithms.
My project works, but I’m going to spend this week cleaning up the code/fixing the small issues, and then put it out into the world so anyone can use it. I’m also training an even better version of my review generator. Even with an expensive computer/GPU that I’m renting, it takes almost 48 hours to get through the dataset one time, and I ideally would retrain it over the data 10-20 times. I’ll probably let it run on the computer for another 7 days straight, and just get the best results I can.
On the personal front I’ve been working on the whole work/life balance thing. The first week of Insight I started getting 5-6 hours of sleep a night (and with my extra caffeine intake not very good ones at that), and continued into week 2 until I realized that my project/fitness/happiness were all suffering. I forgot something from my years in the trading world: 1 focused hour where I’m alert and present crushes 2-5X that amount of hours where I’m just clocking it in.
I think increasing my sleep hours (and skipping my morning workouts on days when I had to work late the night before) directly correlated with my project starting to work. Instead of waking up in a hazy fog to an alarm, I would feel refreshed and suddenly ideas of how to fix the bugs in my project would come to me. It also helped me be more relaxed during the inevitable ups and downs of any creative work.
I went on my favorite short hike in the Bay Area (Rancho San Antonio) to chill with the farm animals, as well as a birthday party at a Go-Kart facility. Tonight I’m going to see the new version of “It”, which was one of the movies that truly scared me as a kid. Tomorrow morning back to educating my program in writing restaurant reviews!
Below is me post-race after jacking up my adrenaline with these Go-Karts that were at least twice as fast as the ones from my Michigan youth.
Andrew Manser 