i bet every1 reading this is a boomer pft can all my zoomers in the chat press 1
|Farza||Nov 3, 2019|
TWAS A GOOD WEEK.
Went to a metal concert on Tuesday. My 2nd fav band of all time. Here’s their best song.
I went hard in the mosh pit. I got punched right in the face. I currently have a black eye. I lost my glasses in the process and they got completely crushed. I then found them on the ground between one of the songs. Fixed them. I went back in the mosh pit. Then I lost the left lens.
10/10 concert would do again.
On Monday I randomly started watching the first fast.ai lecture. For those who don’t know, FastAI is the thing everyone mentions when it comes to learning machine learning. It’s 100% free and open-source. The course is literally built by experts + students over time through suggestions and even direct pull requests to the course repository. I looooooove this model for learning. It forces the student to create instead of just consume. The key to learning for me creating. Lectures, bland HW assignments, and copying what the professor does is consuming. Actually using the knowledge to make something happen in the real world is creating.
Basically, on paper, FastAI has all the elements of what I envision to be a “perfect” course.
I had skipped through it last year.
But, this time I wanted to do it for real because I’ve had some new ideas related to coding + education similar to the FastAI model and wanted to try out FastAI first to sorta understand the current state-of-the-art of this style course.
Honestly. I’ve been liking the course way more than I expected.
I’m going super slow. A lot of the course is like a review for me, but there is also a sufficient amount of new stuff that keeps me interested. So much new stuff. UNets, one cycle learning, how to properly use Jupyter Lab, PyTorch, new datasets, NLP (something I’ve never touched), and more. Plus, the homework assignments are dope and teach you a lot even if you’re already familiar w/ a lot of the concepts.
I’ve also started watching Andrew Ng’s course on machine learning recently, one of the most popular online courses of all time w/ over 2.6 million people enrolled. Honestly, most of the stuff in this course isn’t that important. It’s all math that no one really needs to know anymore. But, idk, I’ve been really enjoying it - walking through the theory step by step and working it out on paper. It’s been fun.
Readings of the Week
I read “Talking to Strangers” this week by Malcolm Gladwell. 7/10 book. The book pretty much asks the question: When we meet a stranger, why is it that we almost never understand them properly? It uses historical examples where Hitler was able to deceive Chamberlain, recent examples of cops where they shot people for doing pretty much nothing, and examples from the CIA where Cuban double agents were discovered in the CIA. It breaks down all these scenarios and digs deep into how it all went down. At the core, the book is trying to drive home this point that strangers are EXTREMELY confusing, yet we as humans always assume we understand them.
I read “A Gentleman in Moscow” by Amor Towles last week. 9.5/10 book. I’m not even gonna lie this is probably one of my favorite novels of all time now. Just go pick it up from your local library or something :). I promise you won’t waste your time.
I also read a cool new computer vision paper from ICCV 2019 about DeceptionNets. Been trying to read more papers lately!
Libraries are an Unseen Opportunity
Been doing a lot of reading into the local budget of my city and the budget of my county.
I wrote some tweets talking about it here. Basically, there is a fuck ton of money flowing into the city from governments and you can very easily look at these budgets and understand where exactly the money is going and how it’s allocated. It literally says RIGHT THERE what each department spent on software in 2019 and what they're budgeted to spend on software in 2020.
Here’s an example w/ the city’s police department where they're cutting back the budget on software.
This is probably a good signal for you to not expend too much effort on selling software to the Pembroke Pines police department and instead find another department who is actually increasing their budget for software.
For example, sell to the Pembroke Pines Technology Services department instead where they are literally doubling their budget for software:
Anyways, I also spoke with my past teacher from high school (shoutout to Ms.Hughes) who stressed to me that libraries are an unseen opportunity. Instead of selling software to schools, sell it to libraries! They have a shorter sales cycle and also have more incentive than schools to try new things. All libraries want is to have more people engaging with them, either physically or online. More engagement = higher budgets since libraries compete for w/ other departments (ex. parks) for capital.
It makes a lot of sense….
I spent a lot of time this week going to lots of different libraries in my area and trying out all the different software that’s offered. It’s all… pretty meh. Most of it is garbage educational software for kids. And I’m sure those companies are making a fuck ton of money.
Def an unseen opportunity. They also have MASSSSSSIVE BUDGETS. Just Broward Country has a budget of $52M for its 38 libraries. The USA has around 117,000 libraries in total.
So much could be done to supercharge libraries.
I wonder if we can make libraries cool again. Not sure. Just been brainstorming right now. Nothing concrete!
90% chance I start making some machine learning related YouTube videos. For fun ofc. I’m pretty bored.
Note: Not doing this to get YT famous LOL. If I wanted to get YT famous I’d make dumb ass videos like this one titled “Going Through The Same Drive Thru 1,000 Times”.
I think my first video is either going to be a simple ML tutorial or a simple explanation on gradient descent. Unsure! Stay tuned.
I put out a tweet saying I’ll help anyone learn ML for free here. (Note: if you’re interested, just reply to this email w/ the word “potato”).
I got a lot of DMs, LOL.
I’m actually really excited to teach people some new stuff.
Why am I doing this?
1) I like helping people.
2) I like making friends.
3) I like seeing other people win.
4) I want to figure out what’s “missing” in the world of ML education and fill in those gaps myself.
I dropped my brother off to college at FIU the other day and decided to stick around. I walked through campus and chilled in the library. It reminded me:
1) That I do indeed miss college. Learning full time about random shit while collaborating w/ other people doesn’t happen many other places. In fact, it’s a huge thing I miss right now.
2) That college is a waste of time for most of the people there. I randomly sat with a bunch of students studying for their Anatomy class and dude like 2 out of the 12 students there were hyped to learn and succeed. Everyone else was fucking around on Instagram for the entire 6 hours I was there. These people shouldn’t be going to college. College sucks if you don’t know what you’re passionate about. Stop wasting your damn money, you idiots. Figure out what you love, then pay an institution like a college to help you find a career in it.
3) College coffee shops are the best.
Just taking it week by week.
I’ve been feeling pretty useless by my own measurements lately.
Sometimes you feel useless.
So it goes…
Though. I am really enjoying learning all the new stuff in ML, reading papers, reading random books. It’s fun. Probably going to keep doing that for the entire month of October. Hyped to teach people some ML to.
I also started doing coding challenge questions. I’m not super set on working for a bigger company like Apple or Google yet, but I think it’d be nice to have the ability to get to those kinda places easily. I’ve bee doing a lot of research into Apple’s and MSFT’s ML departments and their soooooo cool. Coding challenges are a huge part of getting to work in big tech. If you can’t sort a binary tree you get rejected right there, LOL. Plus, tbh, their hella fun. I usually wake up and just do a challenge as like a brain warm-up.