Random Things I Liked, Dec 2022

I'd like to try a new challenge for myself to start building an archive of things I find interesting. Not sure if it's going to be a monthly, quarterly, or even annual thing, but the idea is to sit back and reflect on a short list of top articles, books, videos, talks, ideas, etc. once in a while.

Unfortunately, I seem to consume a lot of content, and naturally most of it is actually useless for me in the long run. However, when skimming over my notes, posts, or tweets from a year or two ago, I often find something interesting I completely forgot about. Getting back to that material with your refreshed experience and a new viewpoint sometimes gives you additional insight. That's precisely what it's going to be: a place to rehearse and review my past.

So why do it here? If I do it in personal notes or on social media, these links will get buried in the pile of other content, memes, shitposts, and deep personal introspections. Having a dedicated category on the website seems to be a perfect match for this goal. And if it happens to be useful to anyone else on the web, then it's good karma for me.

The Nikopol Trilogy by Enki Bilal

A random visit to the Science Museum in London. From all the things there my attention is caught by a small section of some retrofuturistic comic books, The Nikopol Trilogy. I am not the biggest fan of such fiction, but there is something fascinating about it. They say the original Blade Runner was somewhat inspired by this novel. Oh, and it also has a flying pyramide and the ancient Egyptian gods!

Quantum World playlist

I've been stumbling upon ScienceClic channel occasionally for a long time now, but recently I watched and re-watched some of their playlists, in particular this one about quantum physics. You know, we've all heard many times about space-time, waves, fields, etc., but explanations in these videos really clicked for me. Our constant speed in space-time and how it relates to relativity, spins, etc. Great content.

The Second Renaissance

Oh, this was particularly good. There seems to be so many parallels between the last few years and the Renaissance. Unprecedented changes in the world, the plague killing so many and making people think about living now. Rapid advances in technology bringing forward a plethora of new possibilities. The rise of multi-disciplinary creative persons and content makers.

The author compares the original literacy, i.e. invention of press and spread of reading and writing, with the modern tech literacy, i.e. being able to use computers to quickly learn new things and solve tasks.

It is indeed the best time to be a creator right now, the tools are a commodity, and you don't need crazy budgets anymore. Advances in generative AI, like Stable Diffusion, Midjourney, ChatGPT allow anyone to turn ideas into reality, almost for free. Cloud platforms give you so much for a few bucks. Delivery platforms like Gumroad let you reach and sell to so many customers.

It reminds me of how creating music turned from the super-expensive process, where you need label money to work in the studio, into a no-barrier routine available for anyone with a laptop.

UK Digital Government

Coming from Belarus, I'm impressed how UK is many steps ahead in the digital transformation. Starting from a special infosec division scanning all UK websites for vulnerabilities to The GDS Way describing how to choose programming languages or build Docker images for government websites. They even have their own design system, which they maintain and evolve via GitHub.

Oh, and official government and NHS websites are also super informative and helpful for anything you may need as a resident, from applying for visas to treating a particular health issue.

Deep learning courses

As I don't work with ML/AI on a daily basis, I only touched this subject occasionally in the past. For me, Huggingface course and fastai book+course were great to refresh my knowledge and learn something new.

I took Deep Learning with PyTorch a year ago and enjoyed it a lot, and it's definitely better in the "go into details" aspect, as it takes you through the depth of PyTorch, neural network building blocks, and the math behind it. However, Huggingface course, being more modern, teaches you a lot about Transformers architecture and NLP in general, which was a new thing for me, much appreciated.

The main insight: don't build your models from scratch, fine-tune instead. For me, as an engineer with some NIH syndrome and the urge to "build your own thing" (spending too much time and then not solving the original problem) this was definitely a helpful reminder.

John Carmack interview

This man is a legend, period. It's hard to pick one specific thing from this 5h interview, it's just so inspiring to listen to these two intelligent gentlemen discussing AI. Fuel your curiosity, man!

nothing,nowhere. show

After several years of sitting at home, going to live shows was a needed refresher.