Stealing an idea I've seen elsewhere, I'm sharing a list of things I really enjoyed writing in 2024.
(The list of what I enjoyed reading would run for ages, so, no need to go there …)
To set the stage: in January I launched Complex Machinery, my newsletter on AI, risk, and related topics. This dominated my writing for this year but I still managed to squeeze in some blog posts.
So, in no particular order:
CM#002: "It's still a wild animal" -- This issue dove into a common problem with LLMs: a little something I call The Random™.
"Twenty percent" -- I'm … mildly obsessed with N-sided marketplaces. This was a quick blurb about OnlyFans showing the world how to run a marketplace where everybody wins.
CM#018: "Robot jobs" -- Sussing out when AI is a good fit, using the humble self-service kiosk as a guide.
CM#022: "Does this count?" -- Can we trust when a computer says that something has happened? Maybe … but we should have a think. And leave room for doubt.
"Yes, if: risk management as your creative guide" -- Addressing common misconceptions about risk management. In short: it's not all about saying "no" to every opportunity. It's about charting a path to "yes."
CM#010: "The looming AI debt wall" -- Speaking of risk, there's the question of whether AI can ever live up to its promises. Will it make good before that debt comes due?
"Knowing what to measure" -- A rare(-for-me, semi-)technical piece on choosing metrics for your ML/AI models. If you measure the wrong thing during R&D, expect your model to do the wrong thing in production.
CM#013: "Blue screen for Armageddon" -- A look at the CrowdStrike incident, which was a harsh lesson in risk and complexity. When everything's connected, one small disturbance can have widespread impact.
CM#017: "Stacking the deck" -- My recurring theme of "the rise of robots on Wall Street tells us a ton about what'll happen when AI enters other fields."
"You're terrible at AI" -- A riff on Meb Faber's "You are not a good investor," applied to AI. (It's not just a rant. I offer up some ideas to make it better…)
What's next?
For starters, 2024's last Complex Machinery lands later this week. You can subscribe now to get it in your inbox .
And there's even more to come in 2025! Plenty of new ideas. New projects. Plus, I'm dusting off some old drafts from my blog. (Some drafts are 2-3 years old and yet still hold true today. Scary...)
Knowing what to measure
Choosing a metric for your ML/AI model is as important as that metric's score.
Auto manufacturer VW has experienced a data leak
Be careful where you store sensitive data