How to explore Scott Alexander's work and his 1500+ blog posts? This unaffiliated fan website lets you sort and search through the whole codex. Enjoy!

See also Top Posts and All Tags.

Minutes:
Blog:
Year:
4622 tags
Show all filters
11 posts found
May 08, 2023
acx

Constitutional AI: RLHF On Steroids

15 min • 1,983 words • 384 comments • 180 likes • podcast (14 min)
Scott Alexander examines Constitutional AI, a new technique for training more ethical AI models, discussing its effectiveness, implications, and limitations for AI alignment. Longer summary
Scott Alexander discusses Constitutional AI, a new technique developed by Anthropic to train AI models to be more ethical. The process involves the AI rewriting its own responses to be more ethical, creating a dataset of first and second draft answers, and then training the AI to produce answers more like the ethical second drafts. The post explores the effectiveness of this method, its implications for AI alignment, and potential limitations. Scott compares it to cognitive behavioral therapy and human self-reflection, noting that while it's a step forward in controlling current language models, it may not solve alignment issues for future superintelligent AIs. Shorter summary
Feb 02, 2022
acx

Why Do I Suck?

25 min • 3,433 words • 684 comments • 260 likes • podcast (24 min)
Scott Alexander explores reasons why some readers think his writing quality has declined, addressing various factors from idea exhaustion to intellectual progression. Longer summary
Scott Alexander responds to readers' claims that his writing quality has declined since 2013-2016. He explores several potential reasons, including having exhausted his backlog of ideas, the rationalist community's evolution, improved media quality, reduced need to criticize 'wokeness', psychological impacts of criticism, changing writing style with blog growth, and intellectual progression to more nuanced topics. He also addresses specific theories about selling out or being scared into submission, which he says don't match his experience. Shorter summary
Jan 24, 2022
acx

Grading My 2021 Predictions

12 min • 1,551 words • 140 comments • 86 likes • podcast (17 min)
Scott Alexander evaluates his 2021 predictions, analyzing his performance across various confidence levels and comparing his results to other forecasters and prediction markets. Longer summary
Scott Alexander grades his 2021 predictions, made at the beginning of the year. He lists 108 predictions on various topics including politics, economics, technology, COVID-19, community events, personal life, work, and his blog. The post details which predictions came true (in bold) and which didn't (in italics). Scott then analyzes his performance, breaking down the accuracy rates for different confidence levels. He compares his results to a graph of expected vs. actual accuracy, finding he was slightly underconfident overall. The post concludes with a comparison to other forecasters and prediction markets, showing Scott performed well but was outperformed by both Zvi and the markets. Shorter summary
Apr 05, 2021
acx

2020 Predictions: Calibration Results

14 min • 1,888 words • 165 comments • 61 likes • podcast (17 min)
Scott Alexander evaluates his 2020 predictions, finding he was generally overconfident, and discusses the implications for his prediction-making process. Longer summary
Scott Alexander reviews his predictions for 2020, comparing them to actual outcomes. He analyzes his performance in different confidence levels, noting that he was consistently overconfident this year, particularly in the 50% and 95% categories. Scott attributes some of his errors to unexpected events like the prolonged COVID lockdown and the NYT situation. He reflects on whether this overconfidence in an unusually eventful year might balance out his slight underconfidence in more normal years. The post concludes with plans for future prediction exercises and a link to his ongoing prediction log. Shorter summary
Feb 23, 2021
acx

Mantic Monday: Judging April COVID Predictions

16 min • 2,180 words • 180 comments • 63 likes • podcast (18 min)
Scott Alexander evaluates the accuracy of COVID-19 predictions made by himself and two others in April 2020, discussing the results and their implications for forecasting. Longer summary
Scott Alexander reviews the accuracy of COVID-19 predictions made in April 2020 by himself and two other forecasters, Zvi and Bucky. The post compares their predictions on 20 questions related to the pandemic, using a logarithmic scoring system. Scott discusses the results, noting that Zvi performed best, followed by Scott, then Bucky. He reflects on the value of such exercises for identifying biases and areas for improvement in forecasting. The post also includes updates on other prediction markets and forecasting platforms. Shorter summary
Sep 29, 2020
acx

Mistakes

46 min • 6,340 words • 63 likes
Scott Alexander maintains a list of significant mistakes in his blog posts, providing corrections and updates in reverse chronological order. Longer summary
Scott Alexander maintains a list of significant mistakes he has made in his blog posts, providing corrections and updates. He acknowledges errors in data interpretation, misunderstandings of studies, incorrect predictions, and instances where he changed his mind based on new evidence or reader feedback. The list is presented in reverse chronological order and covers a wide range of topics including medicine, economics, psychology, and social issues. Shorter summary
Feb 14, 2020
ssc

Addendum to “Targeting Meritocracy”

3 min • 322 words • 524 comments • podcast (4 min)
Scott Alexander revisits his 'Targeting Meritocracy' post, realizing the disagreements stemmed from different fundamental perspectives: mistake theory vs. conflict theory. Longer summary
Scott Alexander reflects on his previous post 'Targeting Meritocracy' and the disagreements it generated. He realizes that the core of the disagreement stems from different perspectives: mistake theory vs. conflict theory. Scott originally wrote from a mistake theory viewpoint, assuming good government is about solving problems efficiently. However, he now understands that those who disagreed likely approached it from a conflict theory perspective, where good government is about serving specific class interests. This realization helps explain the stark difference in opinions and reinforces the importance of identifying fundamental assumptions when encountering seemingly inexplicable disagreements. Shorter summary
Feb 21, 2019
ssc

My Plagiarism

4 min • 489 words • 53 comments • podcast (5 min)
Scott Alexander discusses unintentional plagiarism in his writing, reflecting on how writers blend their influences at various levels of recognition. Longer summary
Scott Alexander reflects on unintentional plagiarism in his writing, sparked by realizing a sentence in his previous post about plagiarism was unconsciously inspired by a Miss Manners quote. He discusses how writers often blend their influences, sometimes resulting in recognizable 'chunks' of others' work appearing in their own. Scott extends this idea to writing style, worldview, and reasoning methods, suggesting they are all 'slurries' of one's influences, with some elements more recognizable than others. Shorter summary
Jan 22, 2019
ssc

2018 Predictions: Calibration Results

14 min • 1,848 words • 123 comments • podcast (21 min)
Scott Alexander evaluates his 2018 predictions, analyzing his accuracy and discussing factors that affected his performance. Longer summary
Scott Alexander reviews his predictions for 2018, made at the beginning of the year. He lists each prediction, marking those that came true, those that were false, and those he couldn't determine. Scott then analyzes his performance, presenting a calibration chart and discussing his accuracy at different confidence levels. He notes that he performed poorly on 50% predictions and 95% predictions. Scott attributes some of his inaccuracies to two unexpected events: the cryptocurrency crash and a personal breakup, which affected multiple correlated predictions. He concludes by mentioning he'll post 2019 predictions soon and invites readers to share their own predictions. Shorter summary
Jun 05, 2014
ssc

Don’t Be An Asch-Hole

9 min • 1,247 words • 317 comments
Scott Alexander explores a thought experiment based on the Asch conformity experiment to examine the strength and flexibility of personal beliefs and societal norms. Longer summary
Scott Alexander discusses a thought experiment inspired by the Asch conformity experiment, where one imagines being told that their beliefs are completely wrong according to everyone else in the 'real world'. He explores how this experiment might affect various beliefs, from personal convictions to societal norms, using examples like slavery in the antebellum South and vegetarianism. The post suggests that this thought experiment can be a powerful tool for examining one's beliefs, understanding moral progress, and even as a motivational technique. Scott also reflects on how this experiment might reveal which beliefs we hold most strongly and which we might be more willing to doubt or change. Shorter summary
Apr 15, 2013
ssc

Things I Learned By Spending Five Thousand Years In An Alternate Universe

19 min • 2,651 words • 46 comments
Scott Alexander reflects on 13 years of participating in Micras, a complex geopolitical role-playing game, discussing its impact on his life and the lessons learned about leadership and human behavior. Longer summary
Scott Alexander reflects on his 13-year involvement in a complex role-playing game called Micras, where participants simulate countries and cultures on a fictional planet. He discusses how this experience has shaped his real-life relationships, skills, and understanding of leadership and human behavior. Scott explains how the game's lack of rules mirrors real-world politics and social dynamics, providing valuable lessons in consensus-building and leadership. He also describes how creating fictional societies in Micras can be a deeply personal and introspective experience, reflecting one's values and evolving worldview. Shorter summary