Month 3 - Exploring new frontiers

Aside from unplugging and decompressing, I spent this month focusing on what might end up being my next career move.

I decided I would start exploring three different but related paths and see what felt like a good fit for a potential next job.

AI Safety: Alignment & Governance

The first and most interesting path I followed was to enroll in BlueDot Impact’s AI Safety Fundamentals courses. They have 2 tracks, one about technical alignment and the other about policy and governance. Each is a ~3 month course involving weekly readings and discussion groups and then a month long project at the end.

I decided to jump in the deep end and enroll in both, as they both appeal to me, and so far I’m extremely engaged with the material and my fellow participants. It’s a relatively nascent field, and most people in the course are exploring a similar kind of career transition as I am, often coming from software development, data science, or STEM.

It looks like a common path to enter the field is start with some small independent research and/or find a fellowship for a few months at an existing research lab, and then depending on your path, join a major AI lab, think tank, or government organization.

That’s going to be my goal too.

To that end, I revamped my personal website: simonwisdom.com. I bought a $6/month VPS from DigitalOcean then installed Ghost on it, and copied this minimal theme. I followed a few guides to optimize the space and harden security on my VPS, and set up Cloudflare. Finally, I relocated my old Wordpress website (a blast from the past, I made it in 2014) to archive.simonwisdom.com.

I later updated the layout with my own style:

I plan on posting my thoughts and hopefully some related research there soon.

Cybersecurity

I have been reading that cybersecurity, particularly as it relates to protecting AI model weights or similarly powerful new technology, is extremely in demand. It’s another potential fit for my skills and interests, so I thought I’d dip my toe in.

I started by enrolling in Google’s Cybersecurity Certificate on Coursera. It’s 8 modules over 6 months of part time study, designed to get you in the door to a cybersecurity role. I made it through modules 1 and 2 in one afternoon, and decided that the teaching style was not for me (read: it was dead boring).

Instead I found tryhackme.com which looks to be similar content, but much more hands on. I learn best by doing, so this is a much better fit. I haven’t put much time into it yet, but it’s interesting enough so far that I’m going to put in some more hours and see where it goes. Another similar site recommended to me is hackthebox.com - this one is designed to test your skills once you’ve got some more knowledge.

Deep learning

The third track I plan on exploring is learning how to train neural networks; this will complement the other two tracks. I didn’t make any progress here yet other than to find a course that looks like a good starting point: Andrej Karpathy’s Neural Networks: Zero to Hero.

Starting next month I will be back on a more predicable schedule and I plan to resume my creative projects - ones that have colors and images and videos more suited to this blog. Stay tuned!