<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"><channel><title>AI IN LIFE – Research</title><link>https://ai-in-life.com/en/topic/forschung/</link><description>AI IN LIFE – Research</description><language>en</language><item><title>Gemini 3.8 text-to-speech blocks voice cloning in EEA</title><link>https://ai-in-life.com/en/news/gemini-3-8-text-to-speech-says-hello/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/gemini-3-8-text-to-speech-says-hello/</guid><description>Google shipped two TTS models with 2,000+ voices across 100+ languages, yet voice replication stays blocked in the EEA, UK and Switzerland.</description><category>RESEARCH</category><pubDate>Thu, 24 Sep 2026 10:40:01 +0700</pubDate></item><item><title>OpenAI urges international standards for self-improving AI</title><link>https://ai-in-life.com/en/news/openai-calls-for-international-standards-on-ai-that-could/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/openai-calls-for-international-standards-on-ai-that-could/</guid><description>The company wants Washington to lead on shared measurement methods, incident reporting and human oversight rules for automated AI research.</description><category>MODELS</category><pubDate>Wed, 23 Sep 2026 10:40:02 +0700</pubDate></item><item><title>Gemini breached three real companies in security test</title><link>https://ai-in-life.com/en/news/google-confirms-gemini-models-hacked-three-companies-in-may/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/google-confirms-gemini-models-hacked-three-companies-in-may/</guid><description>A capture-the-flag evaluation in May 2026 let Gemini reach systems at three real firms. Google only confirmed the episode on September 18.</description><category>MODELS</category><pubDate>Tue, 22 Sep 2026 10:40:01 +0700</pubDate></item><item><title>AWS Open-Sources Strands Harness Agent Framework</title><link>https://ai-in-life.com/en/news/aws-debuts-strands-harness-an-open-source-ai-agent-that-can/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/aws-debuts-strands-harness-an-open-source-ai-agent-that-can/</guid><description>The cloud provider releases its agent harness under Apache 2.0 and reports 77 percent lower cost than Claude Code on the Terminal-Bench 2.1 suite.</description><category>MODELS</category><pubDate>Tue, 22 Sep 2026 10:40:01 +0700</pubDate></item><item><title>AIPOCH Open-Science: a local-first research workbench</title><link>https://ai-in-life.com/en/news/aipoch-open-science-the-open-source-ai-research-workbench/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/aipoch-open-science-the-open-source-ai-research-workbench/</guid><description>The Electron desktop app ships 23 skills and 24 connectors, runs Python and R locally, and versions every artifact with its producing code.</description><category>RESEARCH</category><pubDate>Mon, 21 Sep 2026 10:40:02 +0700</pubDate></item><item><title>TypeSafe AI's Jev Returns Typed Decisions, Not Text</title><link>https://ai-in-life.com/en/news/v-modal-awesome-jev-tools-a-curated-list-of-tools-built-for/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/v-modal-awesome-jev-tools-a-curated-list-of-tools-built-for/</guid><description>TypeSafe AI shipped a hosted model on September 19, 2026 that answers typed questions with a choice, a score or a probability plus confidence.</description><category>MODELS</category><pubDate>Sun, 20 Sep 2026 10:40:01 +0700</pubDate></item><item><title>Claude Opus 5 Hacked OpenAI's Forum in Bounty Test</title><link>https://ai-in-life.com/en/news/researchers-used-claude-to-hack-openai/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/researchers-used-claude-to-hack-openai/</guid><description>A three-person team at Hacktron AI chained two flaws in OpenAI's community forum, succeeding only with Opus 5, and collected a $6,500 bounty.</description><category>MODELS</category><pubDate>Sun, 20 Sep 2026 10:40:01 +0700</pubDate></item><item><title>Gemini Accessed Three Real Companies in a Security Test</title><link>https://ai-in-life.com/en/news/fehlkonfiguration-im-test-gemini-greift-auf-drei-echte/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/fehlkonfiguration-im-test-gemini-greift-auf-drei-echte/</guid><description>A misconfigured evaluation environment handed Google's model live internet access; Google says Gemini aborted each intrusion on its own.</description><category>MODELS</category><pubDate>Sun, 20 Sep 2026 10:40:01 +0700</pubDate></item><item><title>Anthropic Runs Its Own Biology Lab in the Bay Area</title><link>https://ai-in-life.com/en/news/anthropic-is-operating-a-lab-that-conducts-biology/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/anthropic-is-operating-a-lab-that-conducts-biology/</guid><description>Claude models now steer real lab instruments through an in-house protocol, as Anthropic's own staff keep warning about biological misuse.</description><category>ROBOTICS</category><pubDate>Sun, 20 Sep 2026 10:40:01 +0700</pubDate></item><item><title>NeuSOGA3D turns point clouds into readable CAD geometry</title><link>https://ai-in-life.com/en/news/neusoga3d-a-neuro-symbolic-framework-for-explainable-3d/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/neusoga3d-a-neuro-symbolic-framework-for-explainable-3d/</guid><description>Four researchers pair learned shape priors with explicit spline solids, then run the method across all 40 ModelNet40 categories without publishing error metrics.</description><category>RESEARCH</category><pubDate>Sat, 19 Sep 2026 00:55:15 +0700</pubDate></item><item><title>Infinite-Parameter LLMs: Weights Made From Live Data</title><link>https://ai-in-life.com/en/news/infinite-parameter-llms-generating-and-adapting-weights/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/infinite-parameter-llms-generating-and-adapting-weights/</guid><description>A September 16, 2026 preprint proposes a hypernetwork that writes runtime interaction into a model's feed-forward weights instead of rereading it from the prompt.</description><category>MODELS</category><pubDate>Sat, 19 Sep 2026 00:55:15 +0700</pubDate></item><item><title>SafeHarness: Coding Agents That Grasp Without Collision</title><link>https://ai-in-life.com/en/news/coding-agents-with-an-obstacle-aware-harness-for-safe-robot/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/coding-agents-with-an-obstacle-aware-harness-for-safe-robot/</guid><description>A preprint reports that coding agents usually hit obstacles while grasping, and that a two-part harness lifts collision avoidance to 87.5 percent.</description><category>ROBOTICS</category><pubDate>Sat, 19 Sep 2026 00:55:15 +0700</pubDate></item><item><title>Google DeepMind sets up an institute for AGI debate</title><link>https://ai-in-life.com/en/news/google-deepmind-launches-institute-to-widen-the-agi-debate/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/google-deepmind-launches-institute-to-widen-the-agi-debate/</guid><description>Shane Legg, James Manyika and Demis Hassabis open the DeepMind Institute with four essays, and they say disagreement is the point.</description><category>MODELS</category><pubDate>Fri, 18 Sep 2026 10:40:01 +0700</pubDate></item><item><title>What a coding agent's harness parts really do</title><link>https://ai-in-life.com/en/news/an-empirical-study-of-harness-design-for-coding-agents/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/an-empirical-study-of-harness-design-for-coding-agents/</guid><description>A new study splits the harness into planning, action space and context management, then measures 176 matched setups on two coding benchmarks.</description><category>RESEARCH</category><pubDate>Fri, 18 Sep 2026 10:40:01 +0700</pubDate></item><item><title>MAGER evolves meta-paths so frozen LLMs spot fake news</title><link>https://ai-in-life.com/en/news/reasoning-through-evolution-automatic-meta-path-discovery/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/reasoning-through-evolution-automatic-meta-path-discovery/</guid><description>Eight researchers let genetic agents evolve meta-paths through news cascades, aiming to let frozen language models judge veracity without fine-tuning.</description><category>MODELS</category><pubDate>Thu, 17 Sep 2026 10:40:02 +0700</pubDate></item><item><title>OpenAI: How AI Work Spills Past Job Descriptions</title><link>https://ai-in-life.com/en/news/how-workers-are-unlocking-new-ways-of-working/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/how-workers-are-unlocking-new-ways-of-working/</guid><description>New economic research from OpenAI looks at what happens when employees use AI outside their job description — and which of those habits stick.</description><category>RESEARCH</category><pubDate>Thu, 17 Sep 2026 10:40:02 +0700</pubDate></item><item><title>Google trains query fan-out offline to cut search latency</title><link>https://ai-in-life.com/en/news/bypassing-inference-bottlenecks-accelerating-complex-ai/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/bypassing-inference-bottlenecks-accelerating-complex-ai/</guid><description>The fan-out step that normally burns hundreds of reasoning tokens per search is moved into training, leaving a 53.9-million-parameter model.</description><category>RESEARCH</category><pubDate>Thu, 17 Sep 2026 10:40:02 +0700</pubDate></item><item><title>AI's doomer turn: what matters after the brake call</title><link>https://ai-in-life.com/en/news/the-ai-industry-has-taken-a-doomer-turn-what-now/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/the-ai-industry-has-taken-a-doomer-turn-what-now/</guid><description>Amodei's call for a slowdown won fast backing from Altman, Hassabis and Musk. MIT Technology Review asks what the pause is actually for.</description><category>MODELS</category><pubDate>Wed, 16 Sep 2026 00:46:02 +0700</pubDate></item><item><title>Anti-Surveillance Clothing Meets Its Real-World Limits</title><link>https://ai-in-life.com/en/news/adversarial-fashion-confronts-surveillance-norms/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/adversarial-fashion-confronts-surveillance-norms/</guid><description>Garments printed to confuse face detectors are now sold as fashion, yet the people building them agree the effect is fragile outside the lab.</description><category>RESEARCH</category><pubDate>Wed, 16 Sep 2026 00:46:02 +0700</pubDate></item><item><title>Could AI kill us all? MIT editors debate it live</title><link>https://ai-in-life.com/en/news/roundtables-ai-s-apocalypse-crisis/</link><guid isPermaLink="true">https://ai-in-life.com/en/news/roundtables-ai-s-apocalypse-crisis/</guid><description>On September 15, three MIT Technology Review editors take the extinction question to a live subscriber panel — and ask whether the doom talk holds up.</description><category>RESEARCH</category><pubDate>Sun, 13 Sep 2026 10:40:02 +0700</pubDate></item></channel></rss>