Halfway through

The Army has been conducting firing exercises on Salisbury Plain, about 20 miles from where I live, so I’ve had a distant booming as background noise for the last week. It kind of sounds like a truck hitting a dip in the road, but every 30 seconds. Focusing on getting things done around that has been fun.

A few good developments this week, I got my final grades back from last semester so I’m now conclusively sitting on a decent first for the moment. In addition I now have the supervisor and the topic for my dissertation, I’ll be working on learning to play Tetris using AI over the summer. This is a good intersection of interesting vs. tractable in a problem. Tetris is NP-complete with perfect information, so it’s a perfect candidate for AI solutions. I need to think about my approach for the next few weeks and catch up on the relevant literature, but that should definitely keep me occupied over the summer, and hopefully lead to a decent submission for my dissertation.

This weekend I’ll be taking part in Hack the Burgh. It’s my first hackathon, and I’m tackling it with a team of other Bath MSc students, so hopefully we’ll have a good time and create something worth sharing. Aside from coursework deadlines, I haven’t had much practice working on delivering software, so this should be a really good time constrained exercise for me.

I’m just about reaching the midway of my master’s now. It’s been a brilliant course, very grateful to Bath for the level of teaching at an incredibly strenuous time. Hopefully the rest of it unfolds as smoothly as the first half.

Reading list

• On the Dangers of Stochastic Parrots 🦜 (pdf)
  • And this related twitter thread
• What is an “algorithm”? It depends whom you ask

  “The term “algorithm,” however defined, shouldn’t be a shield to absolve the humans who designed and deployed any system of responsibility for the consequences of its use. This is why the public is increasingly demanding algorithmic accountability—and the concept of impact offers a useful common ground for different groups working to meet that demand.”

• How our AWS Rust team will contribute to Rust’s future successes
• HERE IS THE ARTICLE YOU CAN SEND TO PEOPLE WHEN THEY SAY “BUT THE ENVIRONMENTAL ISSUES WITH CRYPTOART WILL BE SOLVED SOON, RIGHT?”
• Keynes was wrong. Gen Z will have it worse.
• Speed is the killer feature
• When marginal returns on data collapse, and when they do not
• Multimodal Neurons in Artificial Neural Networks
  • This has to be seen purely for the typographic attacks (writing ‘iPod’ on a piece of paper convincing the neural net that it’s an iPod), but very interesting that we’re seeing concepts responded to when presented in different modal ways! Absent concept research is another really interesting future research direction highlighted.
• It Can Happen to You
  • Interesting point about documentation for functions potentially listing time complexity to prevent pitfalls like this happening in future.