Recently, there have been more voices bearish on the growth of AI, I think Julian Schrittwieser's recent blog post is the best response The public once again fails to understand the exponential growth of AI Julian is a researcher at Anthropic, a former senior software engineer at Google DeepMind, and the second author of the AlphaGo Zero paper. The excerpt from the article is as follows: The current discussion on AI progress and the so-called "foam" reminds me of the early weeks of the COVID-19 epidemic. After it has been evident for a long time that the timing and scale of the global pandemic were estimated through exponential trends, politicians, journalists, and most public commentators still view it as a distant possibility or localized phenomenon. Similar strange phenomena have emerged in the current state of artificial intelligence capabilities and further development... They come to the conclusion that artificial intelligence will never reach human level or will only have a minor impact for some unknown reason. And just a few years ago, it was completely science fiction to let artificial intelligence accomplish many of the tasks it does today! ... We can see a clear exponential trend, with Sonnet 3.7 performing the best, able to complete autonomous tasks for up to an hour with a 50% success rate. However, Sonnet 3.7 has been released for 7 months now, and mainstream models such as Grok 4, Opus 4.1, and GPT-5... Capable of completing tasks lasting over 2 hours! The latest internal testing reports from OpenAI and Anthropic claim that the model is capable of completing autonomous tasks for up to 7 hours Outlook: Even relatively conservative trend projections indicate that 2026 will be a crucial year for the widespread integration of artificial intelligence into the economy By mid-2026, mainstream models will be able to work autonomously for a full day (8 working hours). By the end of 2026, at least one model will match the performance of human experts across multiple industries. By the end of 2027, the model will frequently surpass experts in many tasks. This may sound too simple, but predicting the future by extrapolating straight lines on a chart is likely to provide better future models than most "experts" - or even most experts in practical fields! (Image: Comparison of the abilities of various AI models and industry experts)