Scott debunks the "all exponentials become sigmoids" argument against AI risk by showing how forecasters consistently predict premature flattening of exponential trends, and argues that without deep understanding of AI dynamics, we should expect current AI progress to continue for roughly as long as it's already been going. Longer summary

Scott argues against the "all exponentials eventually become sigmoids" talking point often used to dismiss AI capability concerns. While technically true that exponential growth must eventually level off, he demonstrates through examples (UN birthrate predictions, solar power deployment forecasts, and AI capability projections) that people consistently misidentify when this flattening will occur, often predicting it prematurely. He explains that while some technological progress does follow sigmoid curves (like airspeed records), predicting when a trend will flatten requires either deep understanding of the underlying process or, in the absence of such understanding, applying Lindy's Law - which suggests a trend will continue approximately as long as it has already lasted. Scott concludes by challenging AI skeptics to either provide detailed models explaining why AI progress will slow down, or explain why they're not using Lindy's Law as their default assumption. Shorter summary

Scott Alexander's April 2026 links roundup covers diverse topics including Venn diagram complexity, flag desecration laws, AI developments, political analysis, scientific studies, and various cultural curiosities. Longer summary

This monthly links post compiles interesting articles, studies, and observations from across the internet in April 2026. Major themes include AI progress and policy (including discussions of AI alignment, capabilities, and regulation), political developments (Trump administration actions, election analysis), scientific findings (from evolutionary psychology to medical treatments), and various cultural oddities. Scott provides brief commentary on each link while noting that he hasn't independently verified all claims and that commenters typically find errors in a few links per post. Shorter summary

Scott Alexander presents 51 links covering AI progress and safety, political developments, scientific research, cultural oddities, and ongoing philosophical debates about miracles and education reform. Longer summary

Scott Alexander shares 51 links covering diverse topics including AI developments (agents, safety, consciousness research), political news (Ukraine policy, UK politics, Trump administration), science updates (climate predictions, genetics, bacteriophages), cultural curiosities (Shakespeare superfan plastic surgery, Soviet naming conventions, flag cones), health research (Alzheimer's prevention, shingles vaccine reducing dementia, kidney donation), and philosophical debates (Hume's argument against miracles, the Fatima miracle discussion). The post maintains Scott's characteristic blend of serious analysis and quirky observations, touching on everything from Bach's descendants in Oklahoma to the mystery of why AI still struggles with laundry folding despite mastering protein folding. Shorter summary

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

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

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

Scott announces his collaboration with AI Futures Project's blog and their upcoming AMA, highlighting recent posts including one about AI time horizons that was validated by new OpenAI data. Longer summary

Scott Alexander announces he will be shifting most of his AI blogging to the AI Futures Project blog, where he has already co-written several posts. He highlights three recent posts, particularly one about AI time horizons that was validated by new OpenAI data showing faster horizon growth than previously estimated. He also announces an upcoming AMA with the AI Futures Project team on ACX. Shorter summary

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

Scott Alexander won his three-year bet on AI progress in image generation compositionality within three months, demonstrating unexpectedly rapid AI advancement. Longer summary

Scott Alexander discusses his recent bet about AI image models' ability to handle compositionality, which he won much sooner than expected. He explains the concept of compositionality in AI image generation, showing examples of DALL-E2's limitations. Scott then describes the bet he made, predicting improvement by 2025. However, newer AI models like Google's Imagen achieved the required level of compositionality within just three months. This rapid progress supports the idea that AI development is advancing faster than many predict, and that scaling and normal progress can solve even complex problems in AI. Shorter summary

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

Scott reviews recent changes in prediction markets covering the Ukraine war, nuclear risk, AI development, and other current events. Longer summary

This post covers several prediction markets and forecasts related to current events. It discusses changes in Ukraine war predictions, nuclear risk estimates, AI development timelines, and other topics like Elon Musk's Twitter acquisition and the French presidential election. Scott analyzes discrepancies between different forecasts and markets, and explores potential reasons for changes in predictions. Shorter summary

Scott Alexander updates readers on Ukraine war predictions, Insight Prediction's challenges, ACX 2022 Prediction Contest results, and various prediction market developments. Longer summary

Scott Alexander provides an update on prediction markets related to the Ukraine war, discusses the situation with Insight Prediction (a prediction market platform), shares data from the ACX 2022 Prediction Contest, and gives brief updates on various prediction-related topics. The post covers changes in probabilities for key Ukraine war outcomes, the challenges faced by Insight Prediction due to the war, analysis of the ACX prediction contest data, and mentions of new prediction market platforms and AI-related predictions. Shorter summary

Scott Alexander explores theories to reconcile contradictory views on AI progress rates, considering the implications for AI development timelines and intelligence scaling. Longer summary

Scott Alexander discusses the apparent contradiction between Eliezer Yudkowsky's argument that AI progress will be rapid once it reaches human level, and Katja Grace's data showing gradual AI improvement across human-level tasks. He explores several theories to reconcile this, including mutational load, purpose-built hardware, varying sub-abilities, and the possibility that human intelligence variation is actually vast compared to other animals. The post ends by considering implications for AI development timelines and potential rapid scaling of intelligence. Shorter summary

Scott Alexander warns against forming strong heuristics based on limited data, using examples from AI research, elections, and campaign strategies to illustrate the pitfalls of this approach. Longer summary

Scott Alexander discusses the dangers of forming strong heuristics based on limited data points. He presents three examples: AI research progress, election predictions, and campaign strategies. In each case, he shows how people formed confident heuristics after observing patterns in just one or two instances, only to be surprised when these heuristics failed. The post argues against treating life events as moral parables and instead advocates for viewing them as individual data points that may not necessarily generalize. Scott uses a mix of statistical reasoning, historical examples, and cultural references to illustrate his points. Shorter summary