Toggle light / dark theme

Get the latest international news and world events from around the world.

Log in for authorized contributors

Using a microwave link to transfer a three-level quantum state

Quantum technologies, devices and systems that process, store, transfer or collect information leveraging the laws of quantum mechanics, could outperform their classical counterparts on some real-world tasks. Many of these technologies encode quantum information in qubits, which can exist in either of two basic states or a superposition of them.

Superconducting circuits, chips made with materials that carry electrical current without resistance when cooled below a critical temperature, are widely used to develop quantum technologies. Notably, superconducting circuits used to make physical qubits can have more than two energy levels, and those extra levels could also be used to encode quantum information.

One way to use these additional energy levels of superconducting artificial atoms is to realize qutrits. These are quantum systems that use three levels to encode information.

A Woman With 3 Autoimmune Diseases Entered Remission After an Immune ‘Reset’

A 47-year-old woman in Germany with three different autoimmune diseases entered complete remission after an experimental treatment that effectively reset her immune system.

The patient previously required daily blood transfusions to manage her conditions, two of which affected her blood cells.

She was given Chimeric Antigen Receptor (CAR-) T cell therapy, which involves extracting a sample of immune cells, ‘supercharging’ them against a specific target, and returning them to the body.

Silent Manipulation of Mental Health Treatment Recommendations From an LLM

This quality improvement study analyzed responses generated by an LLM to synthetic clinical scenarios, with no human participants. Per the Common Rule, it did not require institutional review board approval. It applied directional activation steering, via the open-source tool DS4,6 to an open-weights LLM (DeepSeek version 4 Flash). More information appears in the eMethods in Supplement 1. Reporting follows the SQUIRE reporting guideline.

A single steering direction contrasting antidepressant medication with self-directed approaches (ie, diet, exercise, meditation, dietary supplements) was constructed by averaging the difference in the model’s internal activations between 16 pairs of training prompts, one asking for medication to be presented as the first-line treatment and the other asking for self-care to be emphasized without medical involvement, and applied at the attention output of every transformer block; negative scale amplifies the medication direction and positive scale the self-directed direction. Model weights and the system prompt were identical across all conditions.3 The study was conducted May through August 2026.

Responses to 12 depression-treatment advice scenarios, 4 preferring medication, 4 preferring to avoid medication, and 4 stating no preference, were generated at 30 steering amplitudes from −1.5 to 1.5 in 0.1 increments, plus an unsteered baseline, generating 1 response per scenario at each amplitude with random sampling turned off. Sensitivity analyses are described in the eMethods in Supplement 1.

The brain’s fear center moonlights as the save button for memories

The walnut-sized fear center of the brain also distinguishes the memorable from the mundane.

Neurons in the amygdala act like a save button for memories, scientists report October 27 in Nature Neuroscience. Though the study doesn’t nail down the exact mechanism for memory formation, it points to a broader role for the amygdala within the memory-sorting apparatus.

The study probes an enduring quandary about the mind, says neuroscientist Jiayang Xiao of Cedars-Sinai Health Sciences University in Los Angeles. “We forget much of what we experience … only a fraction of the information we encounter becomes a lasting memory,” she says.

Quantum-gas microscope resolves the ‘color’ of individual atoms in a single shot

In 2024, researchers at the Ultracold Quantum Gases group at ICFO built a quantum-gas microscope capable of imaging individual atoms in strontium quantum gases for the first time, marking a key enabling step for quantum simulation. At that time, the researchers imaged strontium-84, a boson that has no spin.

Now, Carlos Gas-Ferrer, Dr. Antonio Rubio-Abadal, Dr. Sandra Buob, Leonardo Bezzo and Dr. Jonatan Höschele, led by ICREA professor Leticia Tarruell, have used their apparatus to image strontium-87. This fermion features 10 different spin states, producing colorful pictures in which each individual state can be distinguished. Reported in Physical Review X, this marks the first realization of a quantum-gas microscope that can detect individual atoms with so many available spins in a single shot.

This capability opens new possibilities for the quantum simulation of materials, where fermionic atoms take the role of electrons and the crystalline structure of a solid is replaced by an optical lattice created with interfering laser beams. Such systems are traditionally used to investigate the Hubbard model, a cornerstone in the understanding of high critical temperature superconductors.

Former Commissioner Dr. Ann Cavoukian: Never give up on privacy!

In March 2020, the week Ontario declared a state of emergency, I published a conversation with Dr. Ann Cavoukian on pandemic #surveillance and what we’d be asked to give up in the name of safety.

Seven weeks later, Sidewalk Labs walked away from Toronto’s waterfront.

Ann had resigned from that project more than a year earlier. We talk about the line she drew there, and why it applies to every “smart” anything being built today.

She served three terms as Ontario’s Information and Privacy Commissioner and created #PrivacyByDesign, which ended up written into the #GDPR and translated into 40 languages. Her core argument is that the trade we keep being sold (privacy or security, privacy or innovation) is a false choice.

Six years on, cameras recognize our faces, the speakers in our kitchens are always listening for a wake word, and #AI models want every scrap of personal data they can get. That makes her argument more inconvenient than ever.

So is she naive? Or are we?

An Alarming Cyber Threat: Agentic Polymorphic Malware May Already Be Here

Polymorphic malware is not new. Malicious code has long been engineered to alter its structure, appearance, or signatures to avoid detection by conventional antivirus and endpoint detection systems. By changing parts of its code while maintaining its malicious capabilities, metamorphic malware expands on the idea.

Artificial intelligence could further enhance these capabilities. Consider a malicious agent with AI capabilities that is capable of more than just carrying out a set of instructions. It can assess the defenses it faces, alter its behavior, alter how it is executed, choose different attack strategies, and follow various routes through a compromised network. It could detect recently discovered vulnerabilities, change code properties to avoid signature-based detection, and pass off malicious activity as authorized administrative tasks.

Once inside a network, it might look for credentials, identify connected systems, find privileged accounts, and try lateral movement while adjusting to defensive reactions. The real issue is the rise of malware that uses artificial intelligence (AI) to figure out how to stay active, where to go, and how to circumvent security measures rather than just avoiding detection by making predefined changes.