Posts

I was listening to a recent episode of the Machine Learning Street Talk podcast (a very fine podcast!). Specifically: How Do AI Models Actually Think?, Laura Ruis. Fantastic episode! Just great. I need a re-listen. Early in the conversation they touch on (I’m paraphrasing, probably wrongly) whether LLMs think/reason as Douglas Hofstadter suggests, abstracting via a collection of analogies. It’s agreed they do. The host touches on Hofstadter’s often repeated quote: ...

Use it or loose it. We used to do manual labor which had the dual benefit of getting the things we needed (hunt->food, work->money, etc.) and keeping us in reasonable physical (and mental) condition. No longer for many of us, so our bodies atrophy. To fight the entropy, many of us go to the gym. We simulate the labor we used to do in order to keep our bodies in good condition and reap the rewards (energy, look/feel better, longer life, etc.). ...

The hard part of stochastic optimization is the evaluation function. You get whatever you’re optimizing-for or toward and it’s always a trade-off, even if you can’t see it at first. This got me thinking, there must be tons of problems that we cannot optimize easily because we don’t have good (cheap) fitness functions where we could use an LLM to step in. I know I’ve read papers on something like this in the openendedness literature. ...

I was reading Perdurabo about Aleister Crowley. In his youth he was accomplished mountain climber and the book touches on a suite of his climbing achievements in the early chapters. This got me thinking. Mountain climbing is a challenging hobby where the climbers “collect” peaks. That is, they pick hard maintains to climb, that may or may not have been climbed by others, and climb them for fun. It is about personal accomplishment. ...

I’ve probably read Ayn Rand’s The Fountainhead about a dozen times over the years. I’m not really into the philosophy of Objectivism, but I’m a fan of the individualism in the story. Really, I’m a sucker for simple hero stories (think Ender’s Game, Dune, etc.) Not sure I’ll read the book again for a while. It’s comforting to re-read, but I’ve had enough for now. Anyway, there’s a scene that I think back on often. ...

I have Misophonia. At least, I strongly suspect I do, self-diagnosed (!). Misophonia is a neurological condition where specific sounds trigger strong negative emotional and physiological reactions. Common trigger sounds include chewing, breathing, or repetitive noises. People with misophonia may experience intense anger, anxiety, or panic when hearing these triggers, often leading them to avoid situations where they might encounter these sounds. The main daily triggers for me are: Chewing. Breathing. Tapping. Loud walking (stomping/scraping/etc). Lisps. Chewing/slurping/eating noises though. Every meal is hard. ...

It occurs to me that we can use LLMs to give better recommendations. I want this for books. The goodreads recommendations are crap. The ton of “new hot books” newsletters I subscribe to are crap. I want a daily email that makes highly specific recommendations made by an LLM based on: The books I am reading and have just read. The authors of those books. The genres of those books. The general areas of interest/study of those books. The specific details in my background. I typically read on topics that interest me right now and I typically read a cluster of books on a topic. ...

I’m reading “Mind-Reach” by Russell Targ and Harold Puthoff on kindle. It’s ostensibly on the topic of remote viewing as investigated by two physicists at SRI International in the 70s. My default, like most, is that it’s total bunk. The investigations are presented somewhat rigorously. They are trying. Each time they drop stats, it’s an odds estimate (e.g. this result is 1:1,000,000), my spidey sense tells me “p-hacking” and why aren’t you reporting the negative results as well? ...

I had an idea: It seems that with the rise of interest/hype in “tech” we see a similar rise in “spiritualism” (for lack of a better word). I recall during the late 1990s with the rise of the internet/.com boom, that Aliens/UFOs was all the rage and the X-files was one of the biggest shows on TV. Hmmm. I see now with the AI-boom hype/LLMs/ChatGPT and the rise in conspiracy theories, UAPs, telepathy tapes, graham hancock, etc. ...

Tyler Cowen was interviewed again by Dwarkesh Patel which was recently released (Jan 10): Tyler Cowen - The #1 Bottleneck to AI progress Is Humans I listened to it on release and a number of points got me. One was Cowen’s comment on diminishing returns when it comes to AI. Here’s a cartoon of diminishing returns from Wikipedia, full credit: He may have made this point before/elsewhere, but it’s the first time I’ve tripped over it. ...