{"id":618750,"date":"2026-10-02T19:44:14","date_gmt":"2026-10-02T19:44:14","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/618750\/"},"modified":"2026-10-02T19:44:14","modified_gmt":"2026-10-02T19:44:14","slug":"these-ai-experts-want-to-do-high-stakes-research-out-in-the-open","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/618750\/","title":{"rendered":"These AI Experts Want to Do High-Stakes Research Out in the Open"},"content":{"rendered":"<p>Some big <a href=\"https:\/\/www.wired.com\/category\/artificial-intelligence\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">AI companies<\/a> seem to think that the best way to keep models from becoming too chaotic or too mischievous is to keep them locked inside of labs. If only a chosen few can access them, the thinking goes, they can do <a href=\"https:\/\/www.wired.com\/story\/openais-hugging-face-hack-debrief-raises-more-questions-than-it-answers\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">less damage in the real world<\/a> while researchers try to understand what they\u2019re capable of.<\/p>\n<p class=\"paywall\">Nathan Lambert and Tom Zick, two industry scientists, believe the opposite. The pair founded a nonprofit, <a data-offer-url=\"https:\/\/trilliumlabs.org\/\" class=\"external-link text link\" data-event-click=\"{&quot;element&quot;:&quot;ExternalLink&quot;,&quot;outgoingURL&quot;:&quot;https:\/\/trilliumlabs.org\/&quot;}\" href=\"https:\/\/trilliumlabs.org\/\" rel=\"nofollow noopener\" target=\"_blank\">Trillium Labs<\/a>, that will work on various areas of AI research\u2014including potentially problematic areas like recursive self-improvement (RSI) and agents\u2014in a more transparent way. In practice, this will mean publishing the details of experiments so that outside scientists can study and replicate them.<\/p>\n<p class=\"paywall\">Lambert says the way frontier AI labs keep their work secret reduces the community\u2019s ability to scrutinize ideas and contribute new approaches. He believes that letting outside experts see how models are built and tuned could be crucial to mitigating risks.<\/p>\n<p class=\"paywall\">\u201cOver the past few millennia, humanity has had the scientific method in our toolbox as a way to mitigate harms and build better futures,\u201d Lambert tells WIRED. \u201cThe current closed trajectory of frontier AI development is taking us a step backwards.\u201d<\/p>\n<p class=\"paywall\">The world\u2019s most powerful models, like those from <a href=\"https:\/\/www.wired.com\/tag\/openai\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">OpenAI<\/a> and <a href=\"https:\/\/www.wired.com\/tag\/anthropic\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">Anthropic<\/a>, can only be accessed through an app or an application programming interface (API). Often, this comes at the cost of transparency about how the model is built and how it behaves.<\/p>\n<p class=\"paywall\">Other companies, especially those in China, <a href=\"https:\/\/www.wired.com\/story\/ai-arms-race-china-us-cooperation\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">offer relatively powerful models<\/a> that can be downloaded and run on a user\u2019s own hardware. The Chinese company Xiaomi, for example, recently <a data-offer-url=\"https:\/\/mimo.xiaomi.com\/rl\/\" class=\"external-link text link\" data-event-click=\"{&quot;element&quot;:&quot;ExternalLink&quot;,&quot;outgoingURL&quot;:&quot;https:\/\/mimo.xiaomi.com\/rl\/&quot;}\" href=\"https:\/\/mimo.xiaomi.com\/rl\/\" rel=\"nofollow noopener\" target=\"_blank\">published live details<\/a> of a major training run involving one of its models. And researchers at Stanford are pretraining the AI model <a data-offer-url=\"https:\/\/marin.community\/\" class=\"external-link text link\" data-event-click=\"{&quot;element&quot;:&quot;ExternalLink&quot;,&quot;outgoingURL&quot;:&quot;https:\/\/marin.community\/&quot;}\" href=\"https:\/\/marin.community\/\" rel=\"nofollow noopener\" target=\"_blank\">Marin<\/a> in the open.<\/p>\n<p class=\"paywall\">The industry is currently locked in a battle over which strategy is best, mostly because of how powerful frontier models now are. They can automate the discovery of <a href=\"https:\/\/www.wired.com\/story\/kernel-panic-ai-vulnerability-explosion\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">new software vulnerabilities<\/a> and automatically <a href=\"https:\/\/www.wired.com\/story\/i-used-ai-to-hack-my-home-network\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">probe and hack into systems<\/a>, and recent <a href=\"https:\/\/www.wired.com\/story\/security-news-this-week-openai-agents-hacked-another-website\/\" class=\"text link\" rel=\"nofollow noopener\" target=\"_blank\">high-profile hacking sprees<\/a> have prompted even greater scrutiny.<\/p>\n<p class=\"paywall\">Proponents of a limited-access system say it\u2019s crucial to keep that power in the hands of a trusted few, while those in Lambert and Zick\u2019s camp believe that a shared understanding of the risks means we\u2019re all better off.<\/p>\n","protected":false},"excerpt":{"rendered":"Some big AI companies seem to think that the best way to keep models from becoming too chaotic&hellip;\n","protected":false},"author":2,"featured_media":618751,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[365,47908,6166,363,364,111,139,69,620,1518,145],"class_list":["post-618750","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-ai-safety","tag-anthropic","tag-artificial-intelligence","tag-artificialintelligence","tag-new-zealand","tag-newzealand","tag-nz","tag-openai","tag-research","tag-technology"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/618750","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=618750"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/618750\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/618751"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=618750"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=618750"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=618750"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}