Karthik Tadepalli disputes 2027 AI superintelligence timeline
The author argues that predictions of artificial superintelligence arriving by 2027 are unlikely, contending that recent evidence regarding AI research automation is frequently misinterpreted. The analysis suggests that experimental compute scaling and inflated internal estimates of labor productivity uplift better explain recent progress than simple recursive self-improvement models. Furthermore, the author points to DeepMind's competitive performance and a lack of critical reassessment among short-timeline forecasters as reasons to favor a more moderate 2029 timeline. Open article →