Monthly links – July 2022

Interesting links I’ve found during July 2022.
2022-07-01 – ⁠2023-10-01 finished

Low Impact Training: the Peloton of rowing machines. Clear conflict of interest: they sell you $2000+ rowing machines with resistance bands. Perhaps useful for some people, but we evolved for (some) impact, bones get weaker with too little impact / stressors. Found researching Jeff Bezos’s trainer, Wesley Okerson, whose Instagram doesn’t actually anything LIT… #antifragile

Midjourney: text-to-image AI. It’s currently inferior to GPT-3 and Google Imagen: with Midjourney I was unable to get output anywhere near what GPT-3 and Imagen output (surely cherry-picked, but I tried 12+ variations with each Midjourney attempt). I read that training a GPT3 or Imagen model costs ~$300k and MidJourney being " a new research lab focused on new mediums and tools for empowering people." doesn’t have the same resources. That said, I found it incredible that after some prompt-writing reading followed by trial and error I was able to create the images I got, and even more incredible is what I found others are already able to create today (see picture below). Immediate application ideas: create images for articles (I’ve already done it for some of these entries), book covers, game art (NPCs, locations, items) for D&D and other tabletop or card games, inspiration for writing classes. For GPT-3 and Imagen: ads. I wonder if Midjourney will end up catching up in terms of features and being acquired given that it’s already open to everybody whereas Google and OpenAI feel it’s too dangerous to make these tools open to everybody. This also feels like it’s really revolutionary and where second-order effects are hard to predict because humanity uses images to communicate in some places. Found through Mik. OpenAI clearly a misnomer. #ai

How a Secret Service agent or bodyguard sees the world: #perspective

This is what 5% feels like: make sure you’re calibrated and then build an intuitive sense of probabilities, like the poker player Phil Laak does in the article, to minimize consciously thinking about probabilities and falling into all the cognitive biases you already know you have but can’t really defend against.