Scientists worldwide are getting powerful new AI helpers this week. The ICLR 2025 conference in Singapore revealed plans for AI systems that can dream up science ideas and check if they work. These agentic AI tools act like robot scientists, using special math and logic to solve hard problems.

In London, Clarivate launched AI Agents that help researchers with boring tasks. One agent acts like a super-fast librarian, finding all the best papers about any topic. Another helps school leaders decide which science projects to fund. Teachers can even build their own AI helpers using simple picture blocks instead of computer code.

Big tech companies are making AI play nice together. Google’s Agent2Agent system lets different AI programs share information, like a science club where robots team up. This helped create CodeScientist, an AI that found 19 possible discoveries - with six being real breakthroughs.

Schools are joining the fun too. At the University of Maryland, students made AI that protects rainforests by predicting where trees might get cut down. Another project helps doctors by reading X-ray pictures faster than humans can.

Google’s AI co-scientist is like a smart lab partner that never sleeps. It helps human researchers think up experiments and even plans how to do them step-by-step. Over 100 schools are testing this tool to study diseases and climate change.

People are making sure these AI helpers stay safe and honest. New tests like PaperBench check if AI can properly copy important science work. Companies promise humans will always make final decisions, with AI just giving suggestions.

As AI does more science, experts are creating rules to stop mistakes and cheating. Special computer programs now watch for AI errors in medical studies and climate reports. The goal is to have AI and humans work together to solve big mysteries faster than ever before.