Newsletter
Catastrophic leadership is actually really hard and other things
December 13, 2024
Hello,
Here’s everything since my last little missive to you:
A big series of articles largely around the psychology of groups and leadership will take precedence over the next few weeks as I (re)familiarise myself with the content we teach at my new job.
Updated On managing magic mushroom experiences to include a section specifically on risks of harm. Most of the literature, as you can expect, is on clinical use, and the literature that isn’t is pretty vague (again, as you’d expect from survey data recruiting from drug-use forums). But some clear points emerge—relative to other drugs, especially alcohol, psychedelics are astonishingly safe, and become even safer with careful, thoughtful, and better yet supervised use. Which is something you might have anticipated, given the rise of legal jurisdictions and use in clinical settings. But the ‘enduring changes’ that make them so appealing for clinical use are exactly the thing we should be taking care to think about, because there’s no guarantee these enduring changes need to be a good thing.
Catastrophic leadership is actually really hard
Excerpt: There’s this cluster of classic social psychology experiments from the 50’s through the 70’s that you’ll be presented with in documentaries and whatnot whenever groups of people are behaving crazily. You’ve probably heard of some of them. Milgram’s ‘shock’ experiments, or Zimbardo’s prison experiment, or Asch’s conformity tests, and so on. This is the first in a two part series. Here we’ll talk about the classic experiments, and show that the kinds of catastrophic group dynamics people trot out to illustrate them are actually really difficult to achieve.
Main idea: For group dynamics to produce really bad behaviour, you really need to work at it. You have to train your authority figures to be cruel, prevent dissent or disengagement, and intervene all the time to stop people fixing things. It’s hard.
The Vatican’s Secret Saint-Making Process, using the young Acutis as a case-study:
From the moment his Cause was officially opened, Acutis was referred to as potentially “the first millennial saint”. He has been nicknamed “God’s influencer” and “the patron saint of the internet”.
–
OpenAI is training embodied agents. Not precisely true, but certainly something close to this. Sora, the text-to-video AI, makes troublingly real videos, but OpenAI has bigger plans. Sora doesn’t just generate artificial worlds, it also can play characters within those worlds. This seems like a plausible way to simulate some of the embodied-ness most people think is required to create true conscious agents. See also my comments here,
–
AI model capabilities don’t just scale with size, but with efficiency. I don’t pretend to understand lots of this but I did see:
In other words, around three months, it is possible to achieve performance comparable to current state-of-the-art LLMs using a model with half the parameter size.
Assuming anything like that is true, we should see much better performance from smaller models—maybe even ones you can run on your own PC. Be nice to see the back of OpenAI et al. See also the paper.
–
The diminishing returns of research:
The downward historical trend of research productivity has been used to suggest that there are severe permanent diminishing returns of knowledge production. We argue that a substantial portion of the trend is a transitional composition effect resulting from self-selection in researchers’ ability and the expansion of the researcher sector … Our results suggest that the average ability of researchers has fallen substantially.
Interesting for any considering academia. Is being better a benefit? Or will you be swamped out?
–
52 of someone else’s fun links. Most I found interesting. Some highlights:
If you run one specific, but illegal, database query on a set of widely used health data, you can access Tony Blair’s entire personal medical history.
Film studios now add CGI effects to behind the scenes footage to hide how much CGI has been used to make the film.
In China, there are a registries of haunted apartments. If you’re willing to live somewhere with a sinister history, you can get a discount of 30%.
People whose surnames start with U, V, W, X, Y or Z tend to get grades 0.6% lower than people with A-to-E surnames. Modern learning management systems sort papers alphabetically before they’re marked, so those at the bottom are always seen last, by tired, grumpy markers. A few teachers flip the default setting and mark Z to A, and their results are reversed.
In 2024, around 10% of Anguilla’s GDP will come from fees for its .ai domain name.
Discerning readers will notice a few I posted myself as marginalia.
–
I hope you found something interesting.
You can find links to all my previous missives here.
Warm regards,