# The Human Cost of Exponential Curves: A Reflection on ASI Risk > Coverage of lessw-blog **Published:** March 01, 2026 **Author:** PSEEDR Editorial **Category:** risk **Content tier:** free **Accessible for free:** true **Word count:** 435 **Tags:** AI Safety, ASI Risk, LessWrong, Eliezer Yudkowsky, Exponential Growth, Global Coordination **Canonical URL:** https://pseedr.com/risk/the-human-cost-of-exponential-curves-a-reflection-on-asi-risk --- In a candid retrospective published on LessWrong, a researcher issues a public apology to Eliezer Yudkowsky, highlighting the critical gap between the pace of AI development and the speed of human response. In a recent post on **LessWrong**, a contributor explores the personal and professional consequences of underestimating Artificial Superintelligence (ASI) timelines. The piece, titled _"My personal apology to Eliezer Yudkowsky for not working on ASI risk in 2022 itself,"_ serves as both a confession and a warning about the deceptive nature of exponential technological growth. **The Context: The Difficulty of Predicting Velocity** For years, Eliezer Yudkowsky and the broader AI safety community have warned that AI capabilities would improve faster than society's ability to regulate or align them. This creates a high-stakes environment where timing is everything. The author of this post validates that concern through the lens of individual regret. Having decided against focusing on ASI risk in 2022, the author assumed there was ample time before critical thresholds were met. This assumption relied on a linear or moderately polynomial view of progress, rather than the steep exponential curve that actually materialized. **The Gist: A Retrospective on Lost Time** Writing from the vantage point of 2025/2026, the author expresses deep regret for the three-year hiatus. The core realization is that the speed of AI advancement-specifically the jump from systems like DALL-E and PaLM to current iterations-was faster than anticipated. The author admits to underestimating two distinct factors: * **The rate of technical progress:** The capabilities of AI systems scaled far more aggressively than their internal models predicted. * **The stagnation of coordination:** While technology accelerated, the human capacity for global coordination regarding safety protocols remained difficult and slow. The post highlights a painful asymmetry: technical problems yield to compute and data, but coordination problems (getting the world to agree on safety standards) remain stubbornly human. The author notes that while awareness has spiked in specific tech enclaves like San Francisco and Hacker News, the broad mainstream mobilization they predicted in 2022 has not materialized to the necessary degree. **Why This Matters** This reflection is significant because it encapsulates the "fog of war" in AI development. Even informed observers struggle to calibrate their actions against non-linear progress. It serves as a signal that the window for meaningful intervention may be narrower than many experts previously calculated. The post underscores that waiting for clear signs of danger before acting may result in irreversible lost time. [Read the full post on LessWrong](https://www.lesswrong.com/posts/CpiF7QBvcs2fiysfD/my-personal-apology-to-eliezer-yudkowsky-for-not-working-on) ### Key Takeaways * The author expresses regret for a three-year delay (2022-2025) in working on ASI risk, citing it as a potentially life-long error. * Exponential AI progress is intuitively difficult to predict; the author underestimated how quickly systems would evolve from 2022 baselines. * Global coordination on AI safety is significantly harder and slower than technical advancement. * Predictions made in 2022 regarding mainstream AI risk awareness have not fully materialized by 2026, despite high engagement in niche communities. * The post reinforces the urgency of AI alignment work, validating warnings from figures like Eliezer Yudkowsky. [Read the original post at lessw-blog](https://www.lesswrong.com/posts/CpiF7QBvcs2fiysfD/my-personal-apology-to-eliezer-yudkowsky-for-not-working-on) --- ## Sources - https://www.lesswrong.com/posts/CpiF7QBvcs2fiysfD/my-personal-apology-to-eliezer-yudkowsky-for-not-working-on