Looking back on 2023

The end of the year is good time to reflect on what went well and what didn't over the past twelve months, and what changes we hope to make in the year to come. Here's my list:

1. Presentations. I gave 11 talks this year to a variety of audiences (conferences, workshops, internal presentations, external seminars). Some went better than others. The main culprits for my bad talks are (still) trying to say too much in the time allotted, and failing to pitch well to the specific audience. It is particularly challenging to convey the broad strokes of the research to everyone present at a level that interests them while still going into enough depth to satisfy the few experts in the audience. The best presentations I gave involved audience participation using QR code polls - even including one or two over the course of an hour-long talk is a great way to get the audience to stop, think, and start paying attention again. The best talks I attended spent most of the time explaining the problem set up and broader context and very little time on the speaker's own contribution.

2. Publications. Midway through the year it seemed like I was going to put out fewer papers than usual. Then in November and December I ended up being swamped with finalising several manuscripts all at once (hence a reduced blogging frequency). The final tally is nine original manuscripts completed this year. Is this too many? Many decry the publish or perish culture, the endlessly increasing rate at which papers are being published, courtesy coauthorships, salami publishing, and whatnot. At least in my case, I think I have made a meaningful contribution to every paper I have coauthored this year, but I need to strike a better balance between deep work on new research directions and easier (but still time-consuming) work on existing areas of expertise.

3. Upskilling. I played around with using AI tools like StableDiffusion (text to image), LLama (text generation), whisper (speech to text), and a few different web-based academic paper summarisation / recommendation tools. Given the tendency of large language models to hallucinate and spit out falsehoods, it's hard to trust them when seeking new knowledge (e.g. summarising or suggesting new papers to read), but I've found them quite useful for rephrasing ideas in an amusing way or making cool images for talks (see below).

Happy 2024!