𝔊𝔴𝔢𝔯𝔫@gwernAug 4Always shocking to see how thin people were up until extremely recently. Even in 1993, you can barely see the obesity epidemic. That wasn't that long ago! Heck, I could've been in some of those shots for all I know.
9,790
428
4.4%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 21Bleg: does anyone know how to submit paper PDFs automatically to Sci-Hub better than GS?
Prefixing 'sci-hub.tld' to gwern.net doesn't work, librarian.libgen.gs/scimag/librari… is too hard to manually submit everything with & too complex/multi-step for me to script with curl.
9,102
168
1.8%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 8If we look at the fastest-growing sectors the past few decades, they are also notorious for tech CEO control... "one preferred-share class, one CEO—one vote"? 🤔
𝔊𝔴𝔢𝔯𝔫@gwernAug 22Reifying 'nodes' by 'steps in deep supply chain to a discrete product' (no 'ah but cars are even deeper because they use lots of chips!'), only thing that comes close to chips (which eg get flown through dozens of borders before finishing) I can think of is pharmaceuticals.
8,798
193
2.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 16Finally got working display of a new feature: backlinks or reverse citations—list pages that link 𝘵𝘰 the current page. Standard wiki feature, missing before.
(Not recursive but unfortunately Obormot has been too swamped/covid-y to update JS, so this will have to do for now.) pic.twitter.com/cdbm2swyxS
8,713
107
1.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 21(Quakers, Ashkenazi Jews, ———, and furries are the skeleton key to history.
Few understand this.)
8,401
219
2.6%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 4Roll to disbelieve. Their claims are mutually contradictory. If you truly had 30%+20%+possibly 11% penalties to ROI, you would not have the same representations long-term: compounding at >50% higher ROI (seriously?!) means you eat the industry.
7,821
251
3.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 11TIL that ~25% of Hanson & Schubert's poll-respondents don't realize that being on Twitter (20% US pop on Twitter), following Hanson/Schubert (0.015% of Twitter), or responding to a poll (<2.7% of followers, cf '1% rule') makes them (0.0004%) very weird.
7,811
266
3.4%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 10(Might I suggest purchasing some beef - something off the shoulder perhaps? Braised in a white wine sauce... Rumps are also good. All that exercise and grain.)
7,547
95
1.3%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 4(As someone who was there before 'rationalists' were a thing, I disagree with this thread and I think it says more about you and possibly just the Bay Aryans than about rationalists or AI safety research in general.)
𝔊𝔴𝔢𝔯𝔫@gwernAug 19More productive than the time I explained to the Amish dude repairing my bow that I needed it back within a month because I was going to the UK to talk about AI risk, and had to try to explain to him what AI was and why it might be a problem. (He didn't seem convinced.)
6,776
347
5.1%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 21(The problem with waiting for Google Scholar to index stuff is that AFAICT, sometimes it takes years or never happens. Less than ideal. Unsure why; I speculate it may be bad metadata in some cases but in others, they 𝘩𝘢𝘷𝘦 a DOI which should be all you need, so why? 😕.)
6,327
67
1.1%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 12Austin's comment there is painfully out of date. Of the cool applications in the past year (GPT-3, Codex, DALL-E, CLIP...), and successful companies (Covariant...), and algorithms (self-supervised zoo), which 'required' 'people perfectly labeling data by hand'?
6,193
155
2.5%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 27Extended to links/files in other directories...
What's the point of this? Well, consider a paper like Bol et al 2017 which drugged cats with 4 things, not just catnip. Now it can be properly tagged, and viewed in the tag index page for each of the 4 substances! 𝘕𝘪𝘤𝘦. pic.twitter.com/9cfRsMoMnU
5,785
127
2.2%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 24Does anyone still defend that? The idea of a point-mutation seems obviously false - quite aside from the huge evidence of polygenicity for anything interesting or animal capabilities, 100kya doesn't even take you back to, say, Khoisan divergence (and I'm pretty sure they talk)...
𝔊𝔴𝔢𝔯𝔫@gwernAug 23So, how much did returns go up? It must've been a lot, right?
5,285
145
2.7%
View Tweet activity
𝔊𝔴𝔢𝔯𝔫@gwernAug 4Until they have sibling comparisons, stuff like hairstyle *strongly* points to family-level confounds. (No, they did not perfectly control for SES, you never control for SES or family-level confounds until you use equal-by-construction pairs like siblings.)
5,210
35
0.7%
View Tweet activity
Engagements
Showing 31 days with daily frequency
Engagement rate
2.2%
Aug 31
4.1% engagement rate
Link clicks
1.6K
Aug 31
8 link clicks
On average, you earned 52 link clicks per day
Retweets without comments
0
Aug 31
0 Retweets without comments
On average, you earned 0 Retweets without comments per day
![](data:image/svg+xml,<svg xmlns="http://www.w3.org/2000/svg" width="128" height="128"><rect fill-opacity="0"/></svg>)
