Artificial intelligence was used to help develop a drug candidate, rentosertib, for a rare lung condition. Its maker says the drug also seems to reduce the biological hallmarks of age.
Get the latest international news and world events from around the world.
Isaac Newton’s remarkable year plagiarism accusations that followed
1. Newton’s “miracle year” was not quite a miracle.
Newton’s relationships with figures such as Robert Hooke, Gottfried Wilhelm Leibniz and others demonstrate how difficult it can be to determine who “discovered” an idea first.
A deep dive into Newton’s major work, Principia Mathematica, reveals his hate of hypotheses and a habit of courting controversy.
OpenAI’s GPT6 Astra model autonomously completes Portal in 24 hours just $571 in tokens
Astra had to map 3D game spaces and understand the puzzles before it could plan solutions and succeed.
Gene circuits of human immunity
Scientists have unveiled a massive, high-resolution functional map of human immune cells that promises to transform our understanding of how genetics control health and disease.
Published in the journal Cell, the study represents a landmark achievement in immunology and genomics. By systematically stress-testing genes across the genome in 22 million human immune cells, scientists moved beyond mere DNA sequencing to decode the dynamic circuits that govern how these genes actually work in the context of health and disease. This leap from observation to intervention offers a powerful new framework for designing cancer immunotherapies and treating autoimmune conditions, among other things.
“To understand the significance of this study, you have to look at the last three decades of biology,” says a senior author of the study. “First came the Human Genome Project, which gave us the blueprint of our genes. Then, projects like the Human Cell Atlas showed us how different cells read that blueprint. Now, we’re in a grand third wave: discovering what happens to cells when you make targeted changes within the genome. We finally have a way to decode the link between genetic sequence and cell state.”
Atherosclerosis may meet key criteria for autoimmune disease, evidence suggests
In autoimmune diseases, the immune system turns against the body through pathogenic high-affinity B cells or T cells, leading to severe organ damage. Multiple sclerosis (nervous system paralysis), type 1 diabetes (a strong risk factor for atherosclerosis requiring lifelong treatment) and rheumatoid arthritis (debilitating, excruciating joint deformities) are examples of clinically important autoimmune diseases.
Despite decades of international research efforts, no such defining criteria have been documented for atherosclerosis. Research by an international consortium—led by Professor Andreas Habenicht at LMU University Hospital and Professor Changjun Yin at Sun-Yat-Sen University in Guangzhou, China—now reports in Nature Cardiovascular Research that, according to its findings, atherosclerosis satisfies crucial defining criteria of an autoimmune disease.
The universe is finetuned for technology, not just life » IAI TV
The universe seems to be fine-tuned for life. Were nature’s constants even slightly different, life wouldn’t exist. To explain this, scientists reach for bizarre ideas, like an infinite multiverse. But physicist and technologist Jeff Shainline thinks we’ve been missing a crucial clue. The universe isn’t just fine-tuned for life, but also for the creation of technology – humanity itself is just one step in this longer evolutionary process. He uses this observation to build a new case for an overlooked explanation of fine-tuning: Lee Smolin’s proposal that universes reproduce through black holes and, like organisms, evolve across cosmic generations towards fecundity. Life and technology, Shainline argues, are not the point of the universe, but parts of its reproductive strategy.
A deep puzzle has loomed over physics since the 1950s. The world we observe—from sub-atomic quarks to lightyear-spanning galaxies—has the properties it does because a short list of numbers takes certain values. These numbers are often referred to as constants of nature (I will refer to them as parameters), and they include quantities like the speed of light, the masses of fundamental particles, and the strengths of the fundamental forces. The puzzle arises because these numbers have very special values that allow life to exist. This is called fine-tuning. As an example, life depends on carbon, and carbon is made in stars. But the specific process by which carbon is made in stars is extraordinary, relying on a very peculiar coincidence of properties involving electromagnetism, gravity, and the strong nuclear force that causes alpha particles to fuse into carbon nuclei.
Unitree Autonomous Combat: 1 Terrifying Milestone Changing Robotics Forever
Unitree just achieved fully autonomous humanoid combat.
On September 7, 2026, Unitree demonstrated the world’s first fully autonomous humanoid robot combat — powered by its UnifoLM-X2-1.0 world-action model.
No human control. Real-time decision-making and dynamic adaptation.
This is more than a military spectacle. The same world-model technology that enables autonomous combat is exactly what industrial buyers need for robots that can handle unpredictable factory and warehouse environments.
Chinese manufacturers already dominate global humanoid shipments. This milestone shortens deployment timelines for practical industrial use.
Full analysis:
#HumanoidRobots #IndustrialRobotics #PhysicalAI
What’s the deal?
Ineffable Intelligence, a UK-based AI startup led by former Google DeepMind researcher David Silver, has assembled a team of cofounders and hires drawn from DeepMind, InstaDeep, and venture firm Flying Fish. The company aims to build AI superintelligence using reinforcement learning, the training technique with which Silver is closely associated.
Why now? Ineffable was incorporated in November 2025, and Silver became a director in January 2026 after leaving DeepMind. It joins a wave of “neolabs” — frontier AI startups founded largely by veterans of DeepMind, OpenAI, and Anthropic over the past two years.
What’s the endgame? Like its peers, Ineffable’s stated mission is to create AI superintelligence. In reinforcement learning, a model learns by trial and error to maximise a reward set by its trainer — actions that earn rewards get reinforced, while others fade, a process loosely akin to operant conditioning in psychology.
Causal evidence that language models use confidence to drive behaviour
Researchers have long known that AI models generate hidden “confidence signals” when they process information. But a fundamental question remained: Does the AI actually use this internal sense of confidence to decide whether to answer a question or simply say, “I don’t know”? 3. The Proof: To prove this wasn’t just a coincidence, researchers directly intervened in the AI’s internal workings, essentially turning its “confidence volume” up or down. When they artificially boosted the AI’s confidence, it answered more questions. When they suppressed it, the AI abstained more often. This provided direct, causal proof that the model’s confidence level is what drives its decision to speak up or stay quiet. Interestingly, the study found that the AI expresses confidence in two ways: a mathematical one (based on how it calculates the probability of its next word) and a verbal one (when it explicitly states, “I am 80% sure”). The researchers discovered that both of these are actually just simplified glimpses into a much richer, more complex internal understanding of the AI’s own uncertainty. 1. “Saying” it is confident isn’t as accurate as “Being” confident The study makes a crucial distinction between the AI’s internal mathematical confidence and its *verbal* confidence (when the AI literally types out “I am highly confident in this answer”). The researchers found that verbal confidence is a “lossy read-out” and is actually less accurate at predicting whether the AI is right or wrong than its hidden internal probabilities. In short: just because an AI tells you it is certain doesn’t mean its internal mechanics actually back that up. We shouldn’t blindly trust an AI’s explicit claims of certainty. 2. We still can’t see the full “Black Box” Large language models are trained heavily on human feedback (a process called Reinforcement Learning from Human Feedback, or RLHF). During training, human reviewers often reward the AI for saying “I don’t know” when a question is too difficult, rather than making up a false answer (hallucinating). A major ongoing debate in AI research is whether these models have genuinely developed an organic internal sense of uncertainty, or if they are simply executing highly advanced pattern-matching to mimic the abstention behaviors that human trainers rewarded them for during development.
Kumaran et al. show that large language models making decisions on when to answer a question or abstain from answering can be influenced by boosting or suppressing confidence signals in the model.
