{"id":25987,"date":"2025-08-11T15:16:00","date_gmt":"2025-08-11T12:16:00","guid":{"rendered":"https:\/\/forklog.com\/en\/vitalik-buterin-criticizes-excessive-autonomy-in-ai\/"},"modified":"2025-08-11T15:16:00","modified_gmt":"2025-08-11T12:16:00","slug":"vitalik-buterin-criticizes-excessive-autonomy-in-ai","status":"publish","type":"post","link":"https:\/\/u1f987.com\/en\/vitalik-buterin-criticizes-excessive-autonomy-in-ai\/","title":{"rendered":"Vitalik Buterin Criticizes Excessive Autonomy in AI"},"content":{"rendered":"<p>Ethereum co-founder Vitalik Buterin has criticized the trend towards creating &#8220;overly agentic&#8221; AI models. He called for increased human oversight to enhance quality and safety.<\/p>\n<blockquote class=\"twitter-tweet\">\n<p lang=\"en\" dir=\"ltr\">Echoing something <a href=\"https:\/\/twitter.com\/karpathy?ref_src=twsrc%5Etfw\">@karpathy<\/a> recently said, it does frustrate me how a lot of AI development is trying to be as &#8220;agentic&#8221; as possible, when actually creating *more* paths for human input both creates a better output (now for quite a while going forward) and is better for safety.<\/p>\n<p>\u2014 vitalik.eth (@VitalikButerin) <a href=\"https:\/\/twitter.com\/VitalikButerin\/status\/1954855962566639795?ref_src=twsrc%5Etfw\">August 11, 2025<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;It frustrates me that many AI developers strive for maximum system autonomy, although in reality, more opportunities for human control not only improve results (both now and in the future) but also enhance safety,&#8221; he wrote.<\/p>\n<\/blockquote>\n<p>Buterin responded to a post by former Tesla AI director Andrej Karpathy, who noted that large language models are becoming too independent due to optimization for long-term tasks. For instance, in programming, they spend a long time analyzing even the simplest requests.<\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;While this is justified for prolonged tasks, such an approach is less applicable to active iterative development or quick checks before script execution. I regularly interrupt AI with commands: &#8216;Stop analyzing. Limit yourself to this file. No additional tools. No redundant solutions,'&#8221; he emphasized.<\/p>\n<\/blockquote>\n<p>Buterin supported his position, stating that excessive AI autonomy reduces efficiency. In his view, open models with editing functions are much better than those creating content from scratch.<\/p>\n<p>The Ethereum co-founder sees potential in the <span data-descr=\"a system designed for exchanging information between the brain and an electronic device, such as a computer\" class=\"old_tooltip\">brain-computer interface<\/span>, which tracks user reactions to generated content in real-time and adapts accordingly. This would allow AI to consider intentions and expectations. Current models, he says, often ignore such nuances, reducing their usefulness.<\/p>\n<h2 class=\"wp-block-heading\">Community Reaction<\/h2>\n<p>Besides Buterin, many users supported Karpathy&#8217;s position.<\/p>\n<blockquote class=\"twitter-tweet\" data-conversation=\"none\">\n<p lang=\"en\" dir=\"ltr\">I cancelled cursor yesterday for this reason. Constantly thinking 5 steps ahead in the wrong direction. I\u2019m back to manually copy + pasting from chatgpt\/claude\/gemini for now.<\/p>\n<p>\u2014 barry farkus \u23e9\ufe0f (@paimon2cool) <a href=\"https:\/\/twitter.com\/paimon2cool\/status\/1954226543275696414?ref_src=twsrc%5Etfw\">August 9, 2025<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;I canceled <span data-descr=\"AI assistant for developers created by Anysphere Inc.\" class=\"old_tooltip\">Cursor<\/span> yesterday for this reason \u2014 it constantly thinks five steps ahead in the wrong direction. I&#8217;m back to manually copying responses from ChatGPT\/Claude\/Gemini,&#8221; noted AI enthusiast barry farkus.<\/p>\n<\/blockquote>\n<p>Software engineer Kaan Karakas <a href=\"https:\/\/x.com\/KocKaanKarakas\/status\/1954229297150935342\">added<\/a> that existing AI-based models tend to overcomplicate. According to him, &#8220;this is good for deep analysis, but it slows down simple checks.&#8221;<\/p>\n<p>However, not everyone agreed with the AI expert&#8217;s viewpoint. A user named Conor stated that Karpathy is discussing default behavior, which can be changed.<\/p>\n<blockquote class=\"twitter-tweet\" data-conversation=\"none\">\n<p lang=\"en\" dir=\"ltr\">You are commenting more on the default behavior. As always, you can prompt and be explicit with the models to do precisely what you want (eg literally paste this tweet in your system instructions and be like, watch out for this for this ask)<\/p>\n<p>\u2014 Conor (@jconorgrogan) <a href=\"https:\/\/twitter.com\/jconorgrogan\/status\/1954233091251200086?ref_src=twsrc%5Etfw\">August 9, 2025<\/a><\/p><\/blockquote>\n<p> <script async src=\"https:\/\/platform.twitter.com\/widgets.js\" charset=\"utf-8\"><\/script><\/p>\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p>&#8220;You can clearly instruct the model in the prompt about what you want (for example, literally paste this tweet into the system instructions and write something like: &#8216;pay attention to this when executing the request&#8217;). I&#8217;m not against default behavior being more autonomous, but yes \u2014 if your tasks go beyond these limits, you&#8217;ll need to give clear instructions,&#8221; he commented.<\/p>\n<\/blockquote>\n<p>On August 8, OpenAI <a href=\"https:\/\/u1f987.com\/en\/news\/openai-unveils-flagship-ai-model-gpt-5\">released<\/a> its flagship AI model GPT-5.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Ethereum co-founder Vitalik Buterin has criticized the trend towards creating &#8220;overly agentic&#8221; AI models. He called for increased human oversight to enhance quality and safety. Echoing something @karpathy recently said, it does frustrate me how a lot of AI development is trying to be as &#8220;agentic&#8221; as possible, when actually creating *more* paths for human [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":25986,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"select":"","news_style_id":"","cryptorium_level":"","_short_excerpt_text":"","creation_source":"","_metatest_mainpost_news_update":false,"footnotes":""},"categories":[3],"tags":[1751,200],"class_list":["post-25987","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news-and-analysis","tag-ai-agents","tag-vitalik-buterin"],"aioseo_notices":[],"amp_enabled":true,"views":"95","promo_type":"","layout_type":"","short_excerpt":"","is_update":"","_links":{"self":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/25987","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/comments?post=25987"}],"version-history":[{"count":0,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/posts\/25987\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media\/25986"}],"wp:attachment":[{"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/media?parent=25987"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/categories?post=25987"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/u1f987.com\/en\/wp-json\/wp\/v2\/tags?post=25987"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}