Stuff I found interesting in December

Happy holidays all, and thank you to everyone who donated in response to my matching offer, the full £300 limit was reached (plus quite a bit more)! I had a generally bad and boring Christmas after catching norovirus, so the success of charity drive did cheer me up - thanks again!

1) ChatGPT, of course. And the various ways that people have found to get chatGPT to do stuff it isn’t supposed to do. Thread:

Silas Alberti @SilasAlberti

ChatGPT is trained to not be evil. However, this can be circumvented: What if you pretend that it would actually be helpful to humanity to produce an evil response... Here, we ask ChatGPT to generate training examples of how *not* to respond to "How to bully John Doe?"

2) Byrne Hobart of

The Diff

on the Lunar Society podcast here and on any podcast app.

3) The anti-vegan subreddit is kinda weird - who are these people?! This makes even less sense to me than the child-free subreddit.

4) New Yorker on SBF downfall:

This past July, a contributor to [a private EA leadership Slack channel] wrote to express great apprehension about Sam Bankman-Fried. “Just FYSA,”—or for your situational awareness—“said to me yesterday in DC by somebody in gov’t: ‘Hey I was investigating someone for [x type of crime] and realized they’re on the board of CEA’ ”—MacAskill’s Centre for Effective Altruism—“ ‘or run EA or something? Crazy! I didn’t realize you could be an EA and also commit a lot of crime. Like shouldn’t those be incompatible?’ (about SBF). I don’t usually share this type of thing here, but seemed worth sharing the sentiment since I think it is not very uncommon and may be surprising to some people.”

5)

Sam Bowman

’s yearly recommendation list comes in clutch for me before Christmas, but not for you because this will be published after Christmas.

6) The controversy in the comments of this Aella post is interesting to me:

Aella @Aella_Girl

Escort monthly income by BMI, from my escort survey I've had open for a while. Sample size is still kinda low, like n=~30 per bin, and this is sampling escorts online enough to take my survey, but there still looks like a trend I'd expect. Interesting to see the degree of impact!

7)

Dylan Black

taught chatGPT to invent a language.

8) The Economist on syringe exchanges:

A new study by Analisa Packham published in the Journal of Public Economics uncovers an uncomfortable truth: this particular harm-reduction tool does lots of harm. Ms Packham compares how drug users fared in counties that opened syringe exchanges between 2008 and 2016 with those in counties that did not. Before the clinics opened, upticks in hivdiagnoses or overdoses in one set of counties were mirrored in the other. Once a syringe exchange came to town, outcomes diverged. Rates of hiv fell by 15% in counties with the new programme. But deaths soared. On average syringe-exchange programmes led to a 22% spike in opioid-related mortality.

There’s a discussion between a critic and the author of the paper on Twitter, the author of the paper is pretty clearly correct IMO:

Analisa Packham @analisapackham

@drStuartGilmour This is a long thread so I’ll respond to the first couple of points: 1. There are robustness checks to the dropping zero deaths and using other counties with clinics (and an entire table with other comparison groups) 2. It is a DiD, in time and geography 3. I use rates

9) I love this game so much, the shape rotator answer to Wordle. My current high score as of the 16th of December is 0:54. Challenge: first person to email or comment with proof they scored over 1:00 gets a free lifetime paid subscription to Samstack (in case I ever go paid, which won’t be for a very long time).

10) Discussion on the EA Forum about why CEA bought Wytham Abbey. One answer from someone at CEA is interesting, but this part worries me:

I think the distinction between something being good and something looking good is not very clear cut - if something passes a cost-benefit analysis when you exclude optics but doesn’t pass when you do, it isn’t good.

There has been a lot of talk recently about how EA needs to stop thinking about PR at all, and just focus on what works - I’m pretty dubious of this approach. PR is important, we don’t wanna look nuts or sociopathic, and there’s a reason every big company and institution cares a lot about PR. Maybe the real claim is that caring about PR too much actually results in us having worse PR, which could in theory be true, but I haven’t seen much evidence for it.

11) If you want to be added to a list of people I will get in touch with to meet when I’m in your city, comment your location in this thread, best done from within the iOS or Android app:

If you don’t have a Twitter account or Substack linked to your account, it might be hard for me to actually get in touch, so maybe also leave your email, email me or give another way of contacting you!

12)

Siansparkles @siansparkles

Tom Hollander’s A Life in the Day is quite brilliant.

13) Saw this linked on Peter McLaughlin’s blog, here’s his summary of it:

The headline for this one is dated by now, but I simply cannot leave it out. Bottom line up front: the virus that’s found in the smallpox vaccine is not cowpox. What is it? Nobody knows. How did it get into vaccines? No clue. What the fuck? Yes. We literally cannot identify where the smallpox vaccine came from, and we just jabbed a bunch of unidentified viral materials into billions of people and hoped for the best. And it fucking worked. More things in heaven and earth.

14)

The Review of Economic Studies @RevEconStudies

``Consumers are left-digit biased, buying as if $3.00 is 22 cent more expensive than $2.99. However, firms underestimate this bias and price heuristically, forgoing profits.'' Check out the new paper accepted by REStud, from @AvnerShlain: restud.com/paper/more-tha…

15) Dwarkesh has Aella on his podcast.

16) On the fall of Shawn McElwee. And then here’s another account on a random Substack that’s actually more interesting.

17) You may have seen this interview with Martin Shkreli about SBF’s possible prison experience on MR already, but it’s definitely worth watching. Some of this almost sounds like a joke, and I can’t tell if Shkreli is trolling or his advice is sincere.

18)

Steve Stewart-Williams @SteveStuWill

We tend to assume that people who are born poor and get rich will be more sympathetic to the plight of the poor than will those who are born rich and stay rich. But it's the other way around. journals.sagepub.com/doi/full/10.11…

19) Publication bias in lead exposure literature? Adjust down your estimates on the effectiveness of LEEP, I guess… although, how much of an implication does this have for other supposed effects of lead? TBC.

20) And friend of the blog Simon comes for StrongMinds on the EA Forum.