The important thing to understand about (rapid) Soviet industrialization and economic growth 1928-40 was that it was driven by sectoral reallocation from agriculture to industry, not rising productivity within these groups (agriculture in particular badly suffered).
Also the Tsarist industrial economy was bigger than that of France and close in productivity to the UK, it was just stapled to a continental empire of peasants.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
This paper's analysis ofsocial science abstracts over time. Economics is slightly left-of-center but has been roughly consistent since 1960. The rest were solidly left of center in 1960, grew dramatically moreso 1960-70, and have continued trending left since then.
Between 1960 and 1970 you had physical violent takeovers of many colleges by leftist radicals, who succeeded in creating fake leftist academic fields and thereby institutionally capturing academia over the course of generations.
Because social sciences academia relies on consensus for promotion, without much feedback from reality, once an intolerant clique gains sufficient cohesion and numerical dominance, which happened 1960-70, they can kickstart a positive feedback loop with no self-correction.
An admin for one of the biggest right-wing Facebook groups DM'd me with his impressions/experience with Facebook moderation and censorship (and gave me permission to post this thread). RW Facebook was big in 2016/17.
The big crackdown began in summer 2017; it did not take the form of bans for hate speech but rather all publicly-known admin accounts getting suspended for no reason, leading to the pages disappearing.
This included device bans which permanently destroyed most of the pages.
Thread on Apple's role in the closure of the Internet. From 2016 to 2023, Apple's App Store, half the mobile duopoly, went from a curated software marketplace to one of the most important content control systems on Earth.
In June 2016, Apple completely reorganized their App Store Review Guidelines into five pillars: Safety, Performance, Business, Design, and Legal.
Most of Apple's big decisions were not policy ones but specific removals that had a chilling effect on future discourse. In Aug 2018, Apple removed 5/6 Alex Jones podcasts for hate speech. This was done jointly with similar actions from Facebook, YouTube, and Spotify.
Thread on Anglosphere intelligence's role in the 2015-2023 closure of the Internet. Not a ton of evidence on the topic (obviously), so this thread isn't super dense. There was a huge surge in tech hiring of ex-FBI employees in 2018.
It is not inherently suspicious that ex-spooks go to Silicon Valley companies; many have expertise in cybersecurity and related fields. What IS suspicious is that so many flock to the content control/moderation roles (Trust and Safety etc).
For example, you have Meta product policy managers for disinformation (ex-CIA) and senior managers of Trust & Safety at Google (also ex-CIA).
Deplatforming of websites thread. This is when private web infrastructure actors (cloud providers, payment processors, DNS providers, DDoS protectors) coordinate to purge websites. Several layers of the web stack are oligopolies, so this can happen sans explicit coordination.
The first major case of infrastructure-level deplatforming was WikiLeaks in 2010, for releasing classified information obtained illegally (US diplomatic cables), referred to as "Cablegate," by AWS, EveryDNS, PayPal, Visa, Mastercard, and Bank of America.
The attacks on WikiLeaks were not ideological ones (they were committing actual crimes as a website), but the tactics (both deplatforming and extremely bogus sex "scandals" like sending love letters to a 19 year old), presaged later efforts during the Great Awokening.
In realistic resume-study settings, previous-generation (Sonnet 4) LLMs consistently prefer black to white and female to male applicants. These preferences are invisible in model chains-of-thought, which are thus unfaithful (don't correspond to actual decision process).
The same outcome but for sex rather than race.
Why does this matter? Well, LLMs enable resume spamming and then practically require their own use to filter them. With some effort, it is possible to remove these biases, as the authors of the paper show - but this effort needs to actually be made.