Why I'm going back to school

"Why is there a linear algebra book on the table?" It was March 23rd and Mao-Lin had a question. I explained to him how I had applied and was accepted into the CUNY School of Professional Studies Master of Data Science program for the fall.

"When you graduate, is that when you'll be rich enough so I can stop working and become a lady of leisure?" Ever since the, like, second week of our almost four year relationship, he had been talking about how he always wanted to marry someone rich so he could stop working. Since he was dating a not-rich boy from Maine, the goal posts changed to me becoming rich enough for him to no longer have to work. Good day at work? Question about what that means for his lady of leisure dreams. Bad day at work? Question about if this will delay him becoming a lady of leisure. You name it, it always returned to the same question.

"We'll see. I have to learn linear algebra first though."

It had been two and a half months since Mao-Lin's mother passed away at 2:45am on January 1st after a seven day battle with RSU and pneumonia. Those seven days in December were amongst the most traumatic of our lives, both as individuals and as a family unit. She had Parkinson's and our entire relationship and worldview was oriented around caretaking for her. This is the unspoken nature of caretaking: your entire life, and I mean your entire life rightfully becomes oriented around providing care.

In an instant, everything changed and I found myself wondering what comes next. The nature of working in tech means you always need to be vigilant about trends and parse out what's a temporary fad versus what's here to stay. Crypto and blockchain? Fads, scams, money laundering rackets. The only people I know who've used crypto have used it to buy drugs online so what are we doing here?

You know what's not a fad and is very much going to change the nature of work over time? The latest developments in artificial intelligence. It's a bit overhyped at the moment and there are still a lot things to sort out, especially in the cost category. If I had to speculate wildly, the real tidal wave will happen when someone comes along with a model that's, idk, 70% as good and costs 10-20% of the amount you'll pay for the leading generative AI models. I bet a lot of rank and file companies would make this trade. This speculation is closer to the vibe check end of the spectrum than based on any real data.

Until cost barriers are over come and the models mature a bit more, the shift will happen over time and bit by bit. To keep up, you have roughly three options:

1. Completely BS it and hope no one notices: slap "AI Strategy" on your LinkedIn. No one knows what this means, which is important since you also don't know what it means. Sign up to speak on a panel with others involved in AI Strategy. Throw around some buzzwords, pay for a ChatGPT subscription. Completely yolo it. Hucksters gonna huck, I suppose.
2. Get involved in projects at your workplace: depending on where you work, there may be various flavors of machine learning or deep learning going on. They probably won't let you do prod code but they may give you access to a sandbox where you can play around. There are probably non-technical ways to get involved. If you let a model rip on company data without permission, though, you will (probably) be out of a job soon.
3. Actually learn the different components of the tech: this is the option I'm pursuing. It's much harder and takes much longer but you'll have confidence you know what you're talking about and, even if you hate the idea of coding and math, you'll still be able to have intelligent conversations with the folks building and training the models. As Andrew Ng, the famous deep learning Stanford researcher and founder of DeepLearning.AI, likes to say, "if you don't understand the math, don't worry about it."

I'm pursuing option number 3 because I do want to let a model rip on some data and see what happens, which you're very much free to do on your personal computer with public data Company data on company computer without authorization? Very bad. Public data on a personal computer? Have at it. You can download 47 viruses to your personal computer at home and corporate IT won't care so long as company data, hardware, or infrastructure isn't involved.

Back to the actual point: To prepare for all of the above and a stretch-goal desire to at least try and fine tune train one of the open source models, I've been taking courses on the following topics:

• Algebra 1 and 2
• Linear algebra
• Single and multi variable calculus
• Statistics and probability
• Python
• Deeplearning.ai specializations in:
  • machine learning (in progress)
  • Deep learning specialization (up next)
  • Natural language processing (after the above)

Along the way, I'll release projects that touch on skills from the above. My first will be a machine learning (single and multi variable linear and logistic regression with gradient descent) project focused on some electronic music data I have. Project details can be found in the Projects section, which will have details as I'm ready to post them.

As general rule, I won't post anything I can't explain with a certain degree of confidence. Muscle memory to remember which sequence of code to use when will take me longer but I'm ready to ramble about anything posted. To the extent there's a skill gap involved, I'll highlight them along the way.