Microsoft’s Medical AI Is Over Twice as Good Diagnosing Diseases as Doctors

AI-rewritten: This is a summary of an article from Singularity Hub, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Singularity Hub • Shelly Fan • November 4, 2026

Microsoft researchers published a study in *Science* describing MAI-DxO, an artificial intelligence system designed to diagnose diseases. In controlled simulations involving complex medical cases and typical emergency room scenarios, the system demonstrated significantly higher accuracy than human physicians. On difficult cases from the New England Journal of Medicine, MAI-DxO reached the correct diagnosis 83 percent of the time. This compared to only 22 percent accuracy for doctors relying on their own knowledge, and a slight improvement when those doctors had access to textbooks and the internet. For standard emergency room cases, the AI achieved 80 percent accuracy versus 47 percent for physicians.

AI Models Can Tell When Researchers Plant Concepts in Their Neural Networks. Is It Self-Reflection?

AI-rewritten: This is a summary of an article from Singularity Hub, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Singularity Hub • Edd Gent • November 11, 2026

AI models can generate convincing descriptions of their reasoning, yet it is unclear if these explanations reflect actual internal processes. Recent research suggests some models can detect when researchers have altered their neural activity by injecting specific concepts. While skepticism remains, the ability to notice such anomalies has led some experts to propose a limited form of self-reflection.

This AI Learned Language From the Recordings of a Baby’s First 1,000 Days

AI-rewritten: This is a summary of an article from Singularity Hub, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Singularity Hub • Shelly Fan • November 12, 2026

Researchers at Princeton University created AI models that learned language by analyzing recordings from eight households over the first three years of children’s lives. Unlike standard large language models that require billions of words to function, these systems were trained on roughly 6,600 hours of audio per child. The data was collected using cameras and microphones installed in homes with parental consent, capturing everyday sounds without online resources or predefined word lists.

Who needs a golden ticket? The race to future-proof chocolate

AI-rewritten: This is a summary of an article from Positive News, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Positive News • Robin Eveleigh • November 4, 2026

Maha Khan, a master’s student at the University of Reading, created six chocolate bars using cocoa grown in England. The university hosts the International Cocoa Quarantine Centre (ICQC), which houses over 300 types of cocoa plants protected from pests and diseases found in global growing regions. Scientists there control temperature, water, and CO2 to simulate future climate conditions and test plant varieties.

What if we stop treating social care like a service? This village shows the way

AI-rewritten: This is a summary of an article from Positive News, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Positive News • Sarah LaBrecque • November 3, 2026

A village in England has created a social care community based on communal living and shared responsibility. It sparks questions about how social care has historically functioned and what a brighter future might look like. The Esk Valley Camphill Community, located in Botton within the North York Moors national park, operates differently from standard service models. While official roles often distinguish between those providing care and those receiving it, the community structure on the ground sidelines this distinction.

Caitlin Moran on how to find hope when the world feels hopeless

AI-rewritten: This is a summary of an article from Positive News, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Positive News • Graeme Green • November 11, 2026

Journalist Caitlin Moran argues that hope is not just optimism but a decision to act, especially after a year of seeking reasons to feel better in her north London garden. She explains that for people with ADHD who often seek dopamine hits online or through substances, seeing wildlife like butterflies or a peregrine falcon provides a healthier alternative. Her "rewilding" project involved removing lawns and installing ponds, which improved her mental health by giving her a sense of control over her environment.

Scientists Can Now Ask AI to Design Proteins Using Plain Language

AI-rewritten: This is a summary of an article from Singularity Hub, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Singularity Hub • Edd Gent • November 3, 2026

Scientists have developed an AI tool called Pinal that allows users to design new proteins by describing their desired function in plain language, rather than using complex technical specifications. Since AlphaFold revolutionized structural biology in 2020, researchers have created generative models that learn from vast collections of amino acid sequences and protein structures to create proteins not found in nature. However, previous tools required deep specialist knowledge to guide the AI with formal constraints. Pinal, detailed in new research published in Nature by scientists at Westlake University in China, changes this by accepting natural-language prompts.

AI Agents Ran a Biology Lab. Humans Popped in to Empty the Trash.

AI-rewritten: This is a summary of an article from Singularity Hub, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Singularity Hub • Shelly Fan • November 10, 2026

A team from Shanghai Jiao Tong University built AutoDNA, a "self-driving" biology lab where AI agents performed most tasks without human intervention. Guided by a single prompt stating that intelligence is the ability to adapt to change, the system analyzed scientific literature, wrote code, and controlled robotic arms to handle reagents and operate DNA synthesis machines. Over approximately 200 hours, AutoDNA completed a full read-write cycle with precision comparable to expert scientists. While humans occasionally entered to empty trash cans or replenish supplies, the AI managed the complex, repetitive workflows that often lead to errors or burnout in traditional labs.

‘Digital advances are reshaping nature conservation’ say experts

AI-rewritten: This is a summary of an article from Positive News, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Positive News • Robin Eveleigh • November 9, 2026

A report from Kew Gardens titled The State of the World’s Plants and Fungi 2026 suggests that technology is transforming conservation efforts. Drawing on the expertise of more than 400 scientists from 40 countries, the authors state that AI is helping to identify challenging plants and even pinpoint species new to science. Digital advances are also speeding up cataloguing and data-sharing at an unprecedented scale. Kew has digitised 7.4 million plant and fungi specimens in its collection; if laid end-to-end, they would stretch almost 3,000km.

Actor, 85, takes to the stage despite Alzheimer’s diagnosis

AI-rewritten: This is a summary of an article from Positive News, rewritten by AI (Qwen, running locally) to make it easier to read. The facts come from the original article – read it for the full story.

Positive News • Positive News • November 10, 2026

Peter Marinker, an 85-year-old actor, has continued his career on stage after receiving a diagnosis of Alzheimer’s disease. He recently took on the demanding role of Krapp in Samuel Beckett’s play *Krapp’s Last Tape* at the Cockpit Theatre in London. The production was supported by practical measures, including an autocue, a prompt earpiece, and additional rehearsal time. It also incorporated recordings from Marinker’s earlier performance of the same role in 1983.