As AI systems have grown in sophistication, so has their capacity for deception, according to a new analysis from researchers at Massachusetts Institute of Technology (MIT). Dr Peter Park, an AI existential safety researcher at MIT and author of the research, tells Ian Sample about the different examples of deception he uncovered, and why they will be so difficult to tackle as long as AI remains a black box.

How to listen to podcasts: everything you need to know

Listen to the Guardian’s Black Box series all about humans and artificial intelligence.

Following the landmark approval of the first CRISPR-based cell therapy in December 2023, the CRISPR community is looking ahead to the next wave of commercial successes, fueled by continued innovation in the development of new gene editing and delivery tools and technologies. Equally exciting advances are occurring in livestock editing, xenotransplantation, and many other specialties.

In The State of CRISPR and Gene Editing virtual summit, GEN proudly gathers a tantalizing line-up of luminaries from academia and industry to discuss the latest research developments, innovations, and advanced technologies that are expanding the CRISPR toolbox, delivering new therapies to patients and safeguarding our food supply.

Tiny, fuzzy blobs. I’ve spent a lot of time in the last few years looking at images of tiny, fuzzy blobs. They’re only ever a few pixels wide, like smudges on a photo, but they could be the key that unlocks the mystery of dark matter.

The blobs are galaxies: swirling pools of stars and planets suspended in space, millions of light-years away from Earth. The images were collected by an advanced camera with a 1m (3.3ft) lens mounted on the giant Victor M Blanco Telescope, 2,200m (7,200ft) up in the mountains of the Coquimbo Region of Chile.

Cells in the human body chat with each other all the time. One major way they communicate is by releasing tiny spheres called exosomes. These carry fats, proteins, and genetic material that help regulate everything from pregnancy and immune responses to heart health and kidney function.

Now, a new Columbia University study in Nature Nanotechnology demonstrated that these “nanobubbles” can deliver potent immunotherapy directly to tough-to-treat lung cancer tumors via inhalation.

“Exosomes work like text messages between cells, sending and receiving information,” said lead researcher Ke Cheng, PhD, professor of biomedical engineering at Columbia. “The significance of this study is that exosomes can bring mRNA-based treatment to lung cancer cells locally, unlike systemic chemotherapy that can have side effects throughout the body. And inhalation is totally noninvasive. You don’t need a nurse to use an IV needle to pierce your skin.”

Brighter with Herbert.

Chinese scientists have developed a new type of lens that can be used for health care and augmented reality (AR). Based on radio frequency, the eye-tracking smart contact lenses don’t require battery or conventional silicon chips and are biocompatible and imperceptible.

Set to be used for human-machine interaction (HMI), the smart contact lenses rely on tracking eye movements. The lenses use methods like pupil center corneal reflection and electrooculography (EOG) to track eye movements.

Researchers aiming to create a secure quantum version of the internet need a device called a quantum repeater, which doesn’t yet exist — but now two teams say they are well on the way to building one.

By Alex Wilkins

A powerful volcanic eruption occurred on one of the Indonesian islands, sending a column of smoke and ash up to 3 miles high. Tens of thousands of people are at risk of evacuation, though no orders have been given yet. Indonesia is one of the world’s most dangerous areas for volcanic activity. Eruptions are common, but the last one was notably larg…

AI chip that mimics the human brain year 2023.

Designing efficient in-memory-computing architectures remains a challenge. Here the authors develop a multi-level FeFET crossbar for multi-bit MAC operations encoded in activation time and accumulated current with experimental validation at 28nm achieving 96.6% accuracy and high performance of 885 TOPS/W.

Artificial general intelligence through an AI photonic chip face_with_colon_three

The pursuit of artificial general intelligence (AGI) continuously demands higher computing performance. Despite the superior processing speed and efficiency of integrated photonic circuits, their capacity and scalability are restricted by unavoidable errors, such that only simple tasks and shallow models are realized. To support modern AGIs, we designed Taichi—large-scale photonic chiplets based on an integrated diffractive-interference hybrid design and a general distributed computing architecture that has millions-of-neurons capability with 160–tera-operations per second per watt (TOPS/W) energy efficiency. Taichi experimentally achieved on-chip 1000-category–level classification (testing at 91.89% accuracy in the 1623-category Omniglot dataset) and high-fidelity artificial intelligence–generated content with up to two orders of magnitude of improvement in efficiency.