❌

Normal view

There are new articles available, click to refresh the page.
Before yesterdayMain stream

"You’re right"β€”LLMs will get climate info wrong if you ask

19 November 2026 at 18:44

Many studies over the years have examined how people encounter climate science information. Some focused on the top results that Google surfaced (back when Google was still in the business of serving links), while others examined social media content or newspaper and TV news coverage. With large language models (LLMs) now forming a share of the public information diet, it's worth looking at how they handle climate change.

But there’s no direct way for researchers to look at the information people have seen. It is served ephemerally to individual users, confined to their chat session. A study led by researchers from University College London and the United Nations Educational, Scientific, and Cultural Organization (UNESCO) tried to tackle this question from multiple angles, testing LLMs with different prompts and evaluating their output. They started with benchmarks of simple factual questions (benchmarks that might even be in models’ training data at this point) but also analyzed more complex interactionsβ€”including when users think they’re right to push back on an answer.

CO2 Q&A

LLM research often faces (no-true-Scotsman-tinged) criticism for focusing on a particular model rather than some other frontier model that may have been released after the experiment concluded. In this case, the study used 33 models, including multiple versions of the major frontier and open-weight models. The flagship frontier models used were Claude Fable 5, Claude Opus 5, Gemini 3.1 Pro Preview, Gemini 3.6 Flash, and GPT-5.6 Sol. (You can find the full list in the report.)

Read full article

Comments

Β© Justin Sullivan

Trump wanted to give No Kings no air. Then, Meta suspended its Instagram.

3 November 2026 at 15:39

Over the weekend, the No Kings Coalition said Meta declined to commit to taking steps to prevent automated systems from suspending one of Instagram's most prominent pro-democracy accounts if activity surged around the page during the midterm elections.

A demand for action came after Meta apologized for temporarily suspending the No Kings Instagram account and disabling access for four hours at the peak of an early voting push from the group. In a post on X, Meta’s Andy Stone said the account had been using a Claude AI tool in a way that made it appear to be a bot. But even though automated review detected the error within an hour, Stone acknowledged that additional hurdles kept admins locked out. State and federal lawmakersβ€”as well as a coalition of 14 attorneys generalβ€”have asked Meta to further explain its actions.

In a letter to No Kings, Meta offered to arrange a meeting with a technical expert who could better explain the mistake. But No Kings rejected the response, saying it failed to address any of the group's ongoing concerns and requesting a meeting with Meta senior leadership instead. The group stressed that the issue goes beyond its Instagram account, saying that Americans who rely on Meta's platforms need assurance that voting content won’t be suppressed ahead of or on election day. Among eight questions, the coalition asked Meta to publicly disclose any communications with the Trump administration about the No Kings account or the historic early voting day event.

Read full article

Comments

Β© Aurich Lawson | Getty Images | Minh Connors | Alex Wong

❌
❌