{"id":36103,"date":"2026-11-06T12:08:19","date_gmt":"2026-11-06T10:08:19","guid":{"rendered":"https:\/\/ukrmedia.news\/?p=36103"},"modified":"2026-11-06T12:08:19","modified_gmt":"2026-11-06T10:08:19","slug":"shtuchnyi-intelekt-pryskoriuie-vlasnyi-rozvytok","status":"publish","type":"post","link":"https:\/\/ukrmedia.news\/en\/science-tech\/shtuchnyi-intelekt-pryskoriuie-vlasnyi-rozvytok\/","title":{"rendered":"Artificial intelligence has begun to accelerate its own development \u2014 study"},"content":{"rendered":"<p>Artificial intelligence is increasingly being used to create new, more powerful models, thereby accelerating its own technological development. An independent study by METR has shown that the American company Anthropic, the developer of Claude, is improving its systems at almost twice the rate it did previously. A significant part of this acceleration is driven by the AI itself, which helps engineers develop and test the next generations of models.<\/p>\n<div class=\"news-summary-box\"><strong>Briefly about the main points<\/strong><\/p>\n<ul>\n<li>Anthropic\u2019s models are improving at almost twice the rate seen in previous years.<\/li>\n<li>Researchers attribute this acceleration to the use of AI in the development of new models.<\/li>\n<li>Artificial intelligence is already helping with programming, conducting experiments and analysing results.<\/li>\n<li>METR has not identified any obvious obstacles to further accelerating development.<\/li>\n<\/ul>\n<\/div>\n<h2>AI helps create more powerful versions of itself<\/h2>\n<p>The findings were published on 5 November by the non-profit research organisation Model Evaluation and Threat Research (METR), which spent 11 weeks examining Anthropic\u2019s internal processes. The researchers gained access to model testing results, data on their training and the work of the engineering teams, and also conducted interviews with company staff.<\/p>\n<p>The analysis showed that, since October 2025, the rate at which Anthropic\u2019s models have improved their performance in standardised tests has increased by approximately 1.9 times. In fact, over the past year, the company has achieved progress that previously took around two years to accomplish. At the same time, the research focused primarily on tasks with clearly defined conditions and success criteria.<\/p>\n<p>According to METR\u2019s estimates, approximately 80 per cent of the current acceleration can be attributed to the use of artificial intelligence in research. This figure remains a preliminary estimate; however, the researchers have established that the acceleration cannot be explained by a simple increase in computing power for training models.<\/p>\n<h2>How artificial intelligence is speeding up the development of new models<\/h2>\n<p>Engineers <a href=\"https:\/\/ukrmedia.news\/en\/science-tech\/anthropic-ipo-dokhody-zbytky-2026\/\">Anthropic<\/a> AI is used to write programme code, conduct experiments, test hypotheses and refine algorithms. Tasks that previously required extensive work by specialists can now be carried out much more quickly thanks to the automation of individual stages.<\/p>\n<p>That said, this does not mean that artificial intelligence is creating its own successors entirely independently. Humans continue to set the direction of research, make key decisions and oversee the training of models. However, an increasing proportion of the technical work is being taken over by AI systems, and their new capabilities are being used to further improve the technology.<\/p>\n<p>It is precisely this mechanism that researchers regard as a prerequisite for recursive self-improvement \u2014 a process in which more powerful models help to create even more sophisticated systems. To date, no fully autonomous cycle of such development has been observed.<\/p>\n<h2>Researchers warn of the risks of uncontrolled acceleration<\/h2>\n<p>METR\u2019s experts have found no convincing evidence that a lack of data, computing resources or algorithmic limitations will significantly slow down the development of Anthropic\u2019s models over the next 12 months. At the same time, they acknowledge that it is difficult to assess all possible obstacles at this stage.<\/p>\n<p>The acceleration of development raises questions about the ability of companies and independent experts to verify the safety of new systems in a timely manner. If the pace of model improvement continues to increase, technological progress may outstrip the development of mechanisms to monitor them.<\/p>\n<p>The study also has limitations in terms of transparency. Anthropic removed a significant portion of internal data from the published report, which it considered commercially sensitive. As a result, independent experts are unable to fully verify all of METR\u2019s findings, although the organisation was granted access to the relevant information during the study.<\/p>","protected":false},"excerpt":{"rendered":"<p>\u0428\u0442\u0443\u0447\u043d\u0438\u0439 \u0456\u043d\u0442\u0435\u043b\u0435\u043a\u0442 \u0434\u0435\u0434\u0430\u043b\u0456 \u0430\u043a\u0442\u0438\u0432\u043d\u0456\u0448\u0435 \u0432\u0438\u043a\u043e\u0440\u0438\u0441\u0442\u043e\u0432\u0443\u0454\u0442\u044c\u0441\u044f \u0434\u043b\u044f \u0441\u0442\u0432\u043e\u0440\u0435\u043d\u043d\u044f \u043d\u043e\u0432\u0438\u0445, \u043f\u043e\u0442\u0443\u0436\u043d\u0456\u0448\u0438\u0445 \u043c\u043e\u0434\u0435\u043b\u0435\u0439, \u043f\u0440\u0438\u0441\u043a\u043e\u0440\u044e\u044e\u0447\u0438 \u0432\u043b\u0430\u0441\u043d\u0438\u0439 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0456\u0447\u043d\u0438\u0439 \u0440\u043e\u0437\u0432\u0438\u0442\u043e\u043a. \u041d\u0435\u0437\u0430\u043b\u0435\u0436\u043d\u0435 \u0434\u043e\u0441\u043b\u0456\u0434\u0436\u0435\u043d\u043d\u044f \u043e\u0440\u0433\u0430\u043d\u0456\u0437\u0430\u0446\u0456\u0457 METR \u043f\u043e\u043a\u0430\u0437\u0430\u043b\u043e, \u0449\u043e \u0430\u043c\u0435\u0440\u0438\u043a\u0430\u043d\u0441\u044c\u043a\u0430 \u043a\u043e\u043c\u043f\u0430\u043d\u0456\u044f Anthropic, \u0440\u043e\u0437\u0440\u043e\u0431\u043d\u0438\u043a Claude, \u0443\u0434\u043e\u0441\u043a\u043e\u043d\u0430\u043b\u044e\u0454 \u0441\u0432\u043e\u0457 \u0441\u0438\u0441\u0442\u0435\u043c\u0438 \u043c\u0430\u0439\u0436\u0435 \u0432\u0434\u0432\u0456\u0447\u0456 \u0448\u0432\u0438\u0434\u0448\u0435, \u043d\u0456\u0436 \u0440\u0430\u043d\u0456\u0448\u0435. \u0417\u043d\u0430\u0447\u043d\u0443 \u0447\u0430\u0441\u0442\u0438\u043d\u0443 \u0446\u044c\u043e\u0433\u043e \u043f\u0440\u0438\u0441\u043a\u043e\u0440\u0435\u043d\u043d\u044f \u0437\u0430\u0431\u0435\u0437\u043f\u0435\u0447\u0443\u0454 \u0441\u0430\u043c \u0428\u0406, \u044f\u043a\u0438\u0439 \u0434\u043e\u043f\u043e\u043c\u0430\u0433\u0430\u0454 \u0456\u043d\u0436\u0435\u043d\u0435\u0440\u0430\u043c \u0440\u043e\u0437\u0440\u043e\u0431\u043b\u044f\u0442\u0438 \u0442\u0430 \u0442\u0435\u0441\u0442\u0443\u0432\u0430\u0442\u0438 \u043d\u0430\u0441\u0442\u0443\u043f\u043d\u0456 \u043f\u043e\u043a\u043e\u043b\u0456\u043d\u043d\u044f \u043c\u043e\u0434\u0435\u043b\u0435\u0439. \u041a\u043e\u0440\u043e\u0442\u043a\u043e \u043f\u0440\u043e \u0433\u043e\u043b\u043e\u0432\u043d\u0435 \u041c\u043e\u0434\u0435\u043b\u0456 Anthropic \u0443\u0434\u043e\u0441\u043a\u043e\u043d\u0430\u043b\u044e\u044e\u0442\u044c\u0441\u044f [\u2026]<\/p>\n","protected":false},"author":207433346,"featured_media":36104,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"rank_math_description":"\u0428\u0442\u0443\u0447\u043d\u0438\u0439 \u0456\u043d\u0442\u0435\u043b\u0435\u043a\u0442 \u0434\u043e\u043f\u043e\u043c\u0430\u0433\u0430\u0454 \u0441\u0442\u0432\u043e\u0440\u044e\u0432\u0430\u0442\u0438 \u043f\u043e\u0442\u0443\u0436\u043d\u0456\u0448\u0456 \u043c\u043e\u0434\u0435\u043b\u0456. \u0414\u043e\u0441\u043b\u0456\u0434\u0436\u0435\u043d\u043d\u044f METR \u043f\u043e\u043a\u0430\u0437\u0430\u043b\u043e, \u0449\u043e Anthropic \u043f\u0440\u0438\u0441\u043a\u043e\u0440\u0438\u043b\u0430 \u0440\u043e\u0437\u0440\u043e\u0431\u043a\u0438 \u043c\u0430\u0439\u0436\u0435 \u0432\u0434\u0432\u0456\u0447\u0456.","rank_math_title":"\u0428\u0442\u0443\u0447\u043d\u0438\u0439 \u0456\u043d\u0442\u0435\u043b\u0435\u043a\u0442 \u043f\u0440\u0438\u0441\u043a\u043e\u0440\u044e\u0454 \u0432\u043b\u0430\u0441\u043d\u0438\u0439 \u0440\u043e\u0437\u0432\u0438\u0442\u043e\u043a \u2014 METR","rank_math_focus_keyword":"\u0428\u0442\u0443\u0447\u043d\u0438\u0439 \u0456\u043d\u0442\u0435\u043b\u0435\u043a\u0442","td_subtitle":"","subtitle":"","footnotes":""},"categories":[805620],"tags":[806764],"class_list":["post-36103","post","type-post","status-publish","format-standard","has-post-thumbnail","category-science-tech","tag-806764"],"_links":{"self":[{"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/posts\/36103","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/users\/207433346"}],"replies":[{"embeddable":true,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/comments?post=36103"}],"version-history":[{"count":1,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/posts\/36103\/revisions"}],"predecessor-version":[{"id":36105,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/posts\/36103\/revisions\/36105"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/media\/36104"}],"wp:attachment":[{"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/media?parent=36103"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/categories?post=36103"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ukrmedia.news\/en\/wp-json\/wp\/v2\/tags?post=36103"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}