NuseReal

Science News: September 2026

Global warming will exceed limit, UNEP says in a report that maps path to get back below danger zone "If we act now we can limit overshoot," the report said. "Not a safe world, but safer than remaining above" 1.5 degrees for the long term (LINK) — 2 September 2026

— Earth is soon going to exceed the safe temperature threshold agreed upon by the world in the 2015 Paris climate accord because of warming from the burning of coal, oil and gas, according to a bleak first-of-its-kind report issued by the United Nations Environment Programme Wednesday. The world is now at 1.4 C (2.5 F) above pre-industrial levels, some scientists have calculated. The measure is based on a 20-year average, rather than a single year's temperatures. Wednesday's report finally admitted that the world will cross the 1.5 threshold "in the next few years." The report said the most optimistic scenario has warming peaking at 1.8 C (3.2 F) around mid-century if countries pursue ambitious climate goals, but that governments' current policies should push temperatures to 2.6 C (4.7 F) above pre-industrial levels by 2100.

An AI capable of doubt can optimize scientific discovery: A team at EPFL have developed a framework that trains Large Language Models to create AI systems that can find the best possible setup or "recipe" for a scientific experiment (LINK) — 1 September 2026

— Bojana Rankovic and Philippe Schwaller at EPFL's Laboratory of Artificial Chemical Intelligence have developed a method that combines an LLM with a a Gaussian process, a kind of 'doubt detector", the probabilistic model commonly used in Bayesian optimization." Rather than asking the LLM to choose experiments directly ("direct prompting"), GOLLuM trains it using the Gaussian process's way of scoring uncertainty and performance of each option. Because of this, as the LLM learns from past experiments, it adjusts how it "organizes" the search space, the "master list" of every possible choice for an experiment.

— The researchers evaluated GOLLuM across 23 benchmark tasks covering organic synthesis, analytical and process chemistry, materials and catalysis, and molecular property optimization. Each optimization run began with ten low-performing observations, and the team used the same GOLLuM configuration across all benchmark tasks rather than tuning it separately for each problem, showing its versatility. Across the 23 benchmarks, GOLLuM ranked first on average, finding high-performing experimental conditions more consistently than Bayesian optimization with expert-designed descriptors.

New autism study deals blow to RFK Jr. theory: no evidence that exposure during pregnancy to traces of pesticides commonly found on fruits and vegetables is linked to autism-related traits in children (LINK) — 1 September 2026