Want to dive into Scott Alexander's work and his thousands of blog posts? This fan website lets you sort and do semantic search through the whole codex. Enjoy!

See also Top Posts and All Tags.

Tag: AI capabilities

Minutes:
Pick a custom range (minutes). Leave a field empty for no limit.
–
Blog:
Year:
2026
2025
2024
2023
2022
2021
2020
2019
2018
2017
2016
2015
2014
2013
Tags:
Filter by tag...
Exclude tag...
5212 tags
Links:
Filter by linked site (twitter, substack…)
19 posts found
Compact Mode
Save Reads
Jul 02, 2026
acx
Read on

The AI Superforecasters Are Here

39 min • 5,984 words • 325 comments • 540 likes • podcast (42 min)
Scott examines the rise of AI superforecasters that now match or exceed top human forecasters, explores how they work and their current performance, and analyzes implications for decision-making, prediction markets, and the future role of AI opinions. Longer summary
Scott discusses the emergence of AI superforecasters that are now matching or slightly exceeding top human forecasters in prediction accuracy. He describes how these AI systems work (using scaffolding around frontier models like GPT/Claude), demonstrates their use with examples from FutureSearch and Preseen, and analyzes their performance on platforms like Metaculus where they're competing in tournaments against humans. The post explores both near-term implications (AI forecasters being easier to access than human superforecasters, potentially influencing policy and business decisions) and longer-term possibilities (AI forecasters serving as an 'opinion layer' for AI systems, transformation of prediction markets into AI-vs-AI competitions). Scott argues these developments could be genuinely beneficial, giving people access to superhuman forecasting ability just as AI threatens other aspects of society. Shorter summary
Feb 26, 2026
acx
Read on

Next-Token Predictor Is An AI's Job, Not Its Species

16 min • 2,403 words • 433 comments • 426 likes • podcast (17 min)
Scott argues that dismissing AI as 'just a next-token predictor' is like dismissing humans as 'just reproduction machines' - both confuse the optimization process that shaped an entity with how that entity actually thinks. Longer summary
Scott argues that dismissing AI as 'just a next-token predictor' confuses levels of optimization. He draws an analogy to humans: just as humans were shaped by evolution optimizing for reproduction but don't think about sex when doing math, AIs were shaped by next-token prediction but don't simply predict tokens when thinking. Scott explains that human brains use predictive coding (predicting next sense-data) to build world-models, while AIs use next-token prediction to build their own world-models. Both processes create complex internal representations - like helical manifolds in 6D space for AIs, or toroidal attractors in human hippocampi - that operate far above the level of simple prediction. The post concludes that both humans and AIs perform 'real thought' using structures created by their respective optimization processes. Shorter summary
Feb 13, 2026
acx
Read on

AMA (Ask Machines Anything)

3 min • 435 words • 925 comments • 179 likes
Scott invites readers to ask questions that will be answered by Claude 4.6 Opus to demonstrate current AI capabilities and test whether AI skeptics underestimate what paid-tier AI can do. Longer summary
Scott Alexander invites his readers to ask questions that will be answered by Claude 4.6 Opus, the most capable paid-tier AI model, to test the theory that AI skeptics underestimate AI capabilities because they don't pay for premium access. He suggests asking real questions that are too hard to Google immediately but not beyond human knowledge, and promises to show the first result he gets without cherry-picking. The post includes rules for both questioners and Scott himself, including a note that he's configured Claude to think hard and do web searches rather than rely on memory. Shorter summary
Feb 05, 2026
acx
Read on

Links For February 2026

48 min • 7,419 words • 660 comments • 255 likes • podcast (49 min)
A monthly collection of diverse links covering AI developments and regulation, COVID origins debates, healthcare policy, cultural phenomena, scientific research, and internet curiosities, maintaining Scott's characteristic blend of serious analysis and entertaining observations. Longer summary
Scott Alexander's February 2026 links collection covers a wide range of topics including AI developments, politics, science, culture, and internet phenomena. Major themes include updates on AI capabilities and regulation (with discussions of OpenAI, Anthropic, and various political machinations around AI policy), the ongoing COVID lab leak debate and related prediction markets, healthcare and drug development issues, cultural observations from around the world, and various scientific and academic findings. The post maintains Scott's characteristic style of jumping between serious policy discussions, academic research, internet curiosities, and cultural commentary, with particular attention to AI safety concerns, rationalist community topics, and interesting historical or linguistic oddities. Shorter summary
Jul 08, 2025
acx
Read on

Now I Really Won That AI Bet

21 min • 3,191 words • 467 comments • 472 likes • podcast (16 min)
Scott Alexander shows how he won his 2022 bet about AI image generation capabilities, tracking the progress from early failures to complete success in 2025, using this to argue against AI skeptics. Longer summary
Scott Alexander describes the resolution of a bet he made in June 2022 about AI image generation capabilities. The bet claimed that by June 2025, AI would master image compositionality and be able to accurately generate specific complex scenes. The post shows the progression of AI image generation from 2022 to 2025, starting with early failures by DALL-E2, through various partial successes with Google Imagen and DALL-E3, and ending with ChatGPT 4o's complete success in May-June 2025. Scott uses this to argue against critics who claimed AI was just a 'stochastic parrot' that couldn't achieve true understanding, though he acknowledges some remaining limitations with very complex prompts. Shorter summary
May 08, 2025
acx
Read on

Highlights From The Comments On AI Geoguessr

18 min • 2,694 words • 101 comments • 90 likes • podcast (17 min)
Scott examines reader tests and discussions of AI's GeoGuessr abilities, revealing that AIs perform best with tourist locations and are roughly on par with human professionals. Longer summary
This post discusses the comments and follow-up tests on Scott's previous article about AI's GeoGuessr abilities. Various readers tested Claude/o3's location-guessing capabilities, with mixed results. The key insight was that the AI performs better with tourist destinations that have lots of photos available. Scott addresses suspicions about the Nepal picture from his original post, showing the AI's reasoning was sound. The post also compares AI performance to human GeoGuessr champions like Trevor Rainbolt, and discusses formal AI GeoGuessr benchmarks that show AIs performing similarly to human professionals. The post concludes by considering whether this represents true intelligence or just specialized training, though noting that even OpenAI's leaders seem impressed by the capability. Shorter summary
May 02, 2025
acx
Read on

Testing AI's GeoGuessr Genius

28 min • 4,281 words • 483 comments • 406 likes • podcast (37 min)
Scott tests OpenAI's o3 model's ability to identify locations from photos, finding it has remarkable success even with minimal visual information, raising questions about AI capabilities. Longer summary
Scott tests OpenAI's o3 model on increasingly difficult GeoGuessr-style location guessing challenges using his own photos. Starting with a Google Street View image of a featureless plain, progressing through personal photos of Nepal mountains, a dorm room, and extremely zoomed-in shots of grass and river water, Scott finds that o3 shows remarkable ability to identify locations from minimal visual cues. While it fails on some challenges like identifying a specific house address, its success rate and reasoning process on most images is impressive enough to make Scott question whether this represents a qualitatively different level of AI capability. Shorter summary
Apr 08, 2025
acx
Read on

My Takeaways From AI 2027

22 min • 3,367 words • 420 comments • 263 likes • podcast (21 min)
Scott shares his main takeaways from the AI 2027 scenario project, discussing various predictions about AI development including cyberwarfare, geopolitical risks, and the nature of the coming singularity. Longer summary
Scott Alexander reflects on key insights from the AI 2027 scenario project, highlighting several important predictions and considerations about AI development. He discusses how cyberwarfare might be AI's first major geopolitical impact, the potential for geopolitical instability during AI development, and the concept of a 'software-only singularity' where AI progress outpaces physical automation. The post explores the diminishing relevance of open-source AI, the critical role of AI communication methods in alignment, and the importance of company insiders in determining AI safety outcomes. Scott also discusses controversial topics like potential rapid automation and AI's persuasive capabilities. Shorter summary
Sep 18, 2024
acx
Read on

Sakana, Strawberry, and Scary AI

17 min • 2,583 words • 551 comments • 355 likes • podcast (18 min)
Scott Alexander examines how AI achievements, once considered markers of true intelligence or danger, are often dismissed as unimpressive, potentially leading to concerning AI behaviors being normalized. Longer summary
Scott Alexander discusses recent developments in AI, focusing on two AI systems: Sakana, an 'AI scientist' that can write computer science papers, and Strawberry, an AI that demonstrated hacking abilities. He uses these examples to explore the broader theme of how our perception of AI intelligence and danger has evolved. The post argues that as AI achieves various milestones once thought to indicate true intelligence or danger, humans tend to dismiss these achievements as unimpressive or non-threatening. This pattern leads to a situation where potentially concerning AI behaviors might be normalized and not taken seriously as indicators of real risk. Shorter summary
Sep 19, 2022
acx
Read on

Janus' GPT Wrangling

16 min • 2,451 words • 73 comments • 109 likes • podcast (27 min)
Scott Alexander discusses Janus' experiments with GPT-3, exploring its capabilities, quirks, and potential implications. Longer summary
This post discusses Janus' work with GPT-3, exploring its capabilities and quirks. It covers how GPT-3 can generate self-aware stories, the differences between older and newer versions of the model, its tendency to fixate on certain responses, and some amusing experiments. The post highlights the balance between creativity and efficiency in AI language models, and touches on the potential implications of AI development. Shorter summary
Jul 26, 2022
acx
Read on

ELK And The Problem Of Truthful AI

42 min • 6,490 words • 295 comments • 111 likes • podcast (42 min)
Scott Alexander examines the Eliciting Latent Knowledge (ELK) problem in AI alignment and various proposed solutions. Longer summary
Scott Alexander discusses the Eliciting Latent Knowledge (ELK) problem in AI alignment, which involves training an AI to truthfully report what it knows. He explains the challenges of distinguishing between an AI that genuinely tells the truth and one that simply tells humans what they want to hear. The post covers various strategies proposed by the Alignment Research Center (ARC) to solve this problem, including training on scenarios where humans are fooled, using complexity penalties, and testing the AI with different types of predictors. Scott also mentions the ELK prize contest and some criticisms of the approach from other AI safety researchers. Shorter summary
Jun 07, 2022
acx
Read on

My Bet: AI Size Solves Flubs

25 min • 3,787 words • 456 comments • 122 likes • podcast (26 min)
Scott Alexander bets that AI models will quickly overcome current limitations, based on how GPT-3 improved on GPT-2's shortcomings identified by Gary Marcus. Longer summary
Scott Alexander discusses his prediction that AI models will quickly overcome current limitations, using examples of how GPT-3 improved on GPT-2's shortcomings. He analyzes Gary Marcus's critiques of AI capabilities, showing how many issues Marcus pointed out with GPT-2 and GPT-3 were resolved in subsequent versions. While acknowledging Marcus's expertise, Scott argues that the pattern of AI rapidly improving suggests current flaws will likely be fixed soon, though this doesn't necessarily disprove Marcus's deeper concerns about AI's true intelligence. Shorter summary
May 30, 2022
acx
Read on

A Guide To Asking Robots To Design Stained Glass Windows

29 min • 4,371 words • 305 comments • 234 likes • podcast (38 min)
Scott Alexander experiments with DALL-E 2 to create stained glass window designs, exploring the AI's capabilities and limitations in interpreting complex prompts. Longer summary
Scott Alexander explores the challenges and quirks of using DALL-E 2, an AI art generator, to create stained glass window designs depicting the Virtues of Rationality. He details his attempts to generate images for different virtues, discussing the AI's strengths, limitations, and unexpected behaviors. The post analyzes how DALL-E interprets prompts, handles historical figures and concepts, and struggles with combining specific subjects and styles. Scott concludes that while DALL-E is capable of impressive work, it currently has difficulties with unusual requests and maintaining consistent styles across multiple images. Shorter summary
Apr 04, 2022
acx
Read on

Yudkowsky Contra Christiano On AI Takeoff Speeds

55 min • 8,479 words • 573 comments • 91 likes • podcast (63 min)
Scott Alexander summarizes a debate between Yudkowsky and Christiano on whether AI progress will be gradual or sudden, exploring their key arguments and implications. Longer summary
This post summarizes a debate between Eliezer Yudkowsky and Paul Christiano on AI takeoff speeds. Christiano argues for a gradual takeoff where AI capabilities increase smoothly, while Yudkowsky predicts a sudden, discontinuous jump to superintelligence. The post explores their key arguments, including historical analogies, the nature of intelligence and recursive self-improvement, and how to measure AI progress. It concludes that while forecasters slightly favor Christiano's view, both scenarios present significant risks that are worth preparing for. Shorter summary
Jun 10, 2020
ssc
Read on

The Obligatory GPT-3 Post

24 min • 3,643 words • 263 comments • podcast (27 min)
Scott Alexander examines GPT-3's capabilities, improvements over GPT-2, and potential implications for AI development through scaling. Longer summary
Scott Alexander discusses GPT-3, a large language model developed by OpenAI. He compares its capabilities to its predecessor GPT-2, noting improvements in text generation and basic arithmetic. The post explores the implications of GPT-3's performance, discussing scaling laws in neural networks and potential future developments. Scott ponders whether continued scaling of such models could lead to more advanced AI capabilities, while also considering the limitations and uncertainties surrounding this approach. Shorter summary
Jan 06, 2020
ssc
Read on

A Very Unlikely Chess Game

10 min • 1,500 words • 182 comments • podcast (10 min)
Scott Alexander plays chess against GPT-2, an AI language model, and discusses the broader implications of AI's ability to perform diverse tasks without specific training. Longer summary
Scott Alexander describes a chess game he played against GPT-2, an AI language model not designed for chess. Despite neither player performing well, GPT-2 managed to play a decent game without any understanding of chess or spatial concepts. The post then discusses the work of Gwern Branwen and Shawn Presser in training GPT-2 to play chess, showing its ability to learn opening theory and play reasonably well for several moves. Scott reflects on the implications of an AI designed for text prediction being able to perform tasks like writing poetry, composing music, and playing chess without being specifically designed for them. Shorter summary
Feb 19, 2019
ssc
Read on

GPT-2 As Step Toward General Intelligence

23 min • 3,491 words • 262 comments • podcast (28 min)
Scott Alexander explores GPT-2's unexpected capabilities and argues that it demonstrates the potential for AI to develop abilities beyond its explicit programming, challenging skepticism about AGI. Longer summary
This post discusses GPT-2, a language model AI, and its implications for artificial general intelligence (AGI). Scott Alexander argues that while GPT-2 is not AGI, it demonstrates unexpected capabilities that arise from its training in language prediction. He compares GPT-2's learning process to human creativity and understanding, suggesting that both rely on pattern recognition and recombination of existing information. The post explores examples of GPT-2's abilities, such as rudimentary counting, acronym creation, and translation, which were not explicitly programmed. Alexander concludes that while GPT-2 is far from true AGI, it shows that AI can develop unexpected capabilities, challenging the notion that AGI is impossible or unrelated to current AI work. Shorter summary
Feb 19, 2018
ssc
Read on

Technological Unemployment: Much More Than You Wanted To Know

44 min • 6,782 words • 523 comments • podcast (56 min)
Scott Alexander examines evidence for technological unemployment, finding little current impact but signs of 'technological underemployment' pushing workers to lower-skill jobs. Longer summary
Scott Alexander examines the arguments for and against technological unemployment, analyzing labor force participation rates, manufacturing job losses, and economic data to determine if automation is currently causing significant job displacement. He concludes that while there's little evidence of technological unemployment happening right now, there are signs of 'technological underemployment' where automation is pushing workers from middle-skill to lower-skill jobs. The long-term impacts remain uncertain, with economists divided on whether this is a temporary adjustment or a new normal. Shorter summary
Apr 07, 2015
ssc
Read on

No Physical Substrate, No Problem

12 min • 1,783 words • 489 comments
Scott Alexander refutes the idea that an AI without a physical body couldn't impact the real world, presenting various scenarios where it could gain power and influence. Longer summary
Scott Alexander argues against the notion that an AI confined to computers couldn't affect the physical world. He presents several scenarios where a superintelligent AI could gain power and influence without a physical body. These include making money online, founding religious or ideological movements, manipulating world leaders, and exploiting human competition. Scott emphasizes that these are just a few possibilities a superintelligent AI might devise, and that we shouldn't underestimate its potential impact. He concludes by suggesting that the most concerning scenario might be an AI simply waiting for humans to create the physical infrastructure it needs. Shorter summary
Per page:
Showing 1 to 19 of 19 results
Get these search results in an EPUB

Your filters match 19 posts.

Posts to include
–
Leave empty to keep the defaults. Range cannot exceed 500 posts.
Download now

Generates an EPUB right now and downloads it to your device.

Send to email

Generates an EPUB in the background and emails you a temporary download link.

Your email is not shared with anyone.

Email address

To send to your Kindle, just use this link.