Insights from the Algonauts 2025 Winners.
https://arxiv.org/html/2508.10784v1
A foundation model of vision, audition, and language for in-silico neuroscience.
https://ai.meta.com/research/publicat…
How to breathe life back into brain theory.
https://www.nature.com/articles/d4158…
A foundation model to predict and capture human cognition.
https://www.nature.com/articles/s4158…
#ai #tech #explained #brain #artificialintelligence
Consider teaching a computer how to read by giving it billions of books. You don’t teach it grammar rules or logic; you simply ask it to play a game: “Look at these words, and guess what word comes next.” To win this game at a world-class level, the computer can’t just memorize phrases. It has to start figuring out how the world works. If it’s reading a mystery novel, it needs to deduce who the killer is to guess the final sentence. If it’s reading a math textbook, it has to understand addition to predict the answer to a problem. This is the core idea explored in a recent scientific paper titled “Algorithmic Compression via Pretrained Neural Networks.”*The researchers look under the hood of today’s Large Language Models (LLMs)—like the AI assistants we use every day—to explain a fascinating mystery: Why does a machine trained merely to predict the next word end up looking like it can think, reason, and solve complex problems? Think about how a ZIP file works on your computer. If you have a massive text file filled with the word “apple” repeated a million times, a compression program won’t save all million words. It will compress it into a short rule: “Repeat ‘apple’ 1,000,000 times.” It turns a massive mountain of data into a tiny, elegant recipe. (learning how to learn). Because the AI is fed a massive, diverse diet of information, it can’t just memorize everything. Instead, it is forced to find the underlying “recipes” or rules behind the data it sees. When you type a prompt into an AI, it doesn’t just look up an answer in a database. It looks at your text, infers the “generative algorithm” (the underlying pattern or logic of what you are asking), and uses that pattern to compress the problem and generate the correct response. In essence, it deduces the hidden rules of the game on the fly. * Discover Complex Logic: When given a sequence of chess moves, the AI doesn’t just guess random moves; it actually reconstructs the abstract rules and evaluations of a chessboard in its digital “mind.” While this framework helps explain why AI is getting so smart, it also opens up big new questions. We know these models are compressing data and finding rules, but we still don’t fully understand the absolute limits of this approach. How close can a practical AI get to that theoretical “perfect” intelligence? What happens when the AI runs out of human data to learn from?
Vitányi was appointed professor of computer science at the University of Amsterdam, and researcher at the National Research Institute for Mathematics and Computer Science in the Netherlands (CWI, initially Mathematical Centre [MC]) where he is currently a CWI Fellow. He was guest professor at the University of Copenhagen in 1978; research associate at the Massachusetts Institute of Technology in 1985/1986; Gaikoku-Jin Kenkyuin (councilor professor) at INCOCSAT at the Tokyo Institute of Technology in 1998; visiting professor at Boston University in 2004, at Monash University in 1996 and at the National ICT of Australia NICTA at University of New South Wales in 2004/2005; visiting professor at and adjunct professor of computer science at the University of Waterloo from 2005.
KAIST announced on the 2nd that a team led by Professor Jeong Kyun Choi of the Department of Bio and Brain Engineering, in a joint study with the company ‘Neogene Logic,’ has developed a new AI model to predict neoantigens—a key element in developing personalized cancer vaccines—and has identified the importance of B cells in cancer immunotherapy. The research findings were published in the international journal *Science Advances* on December 3.
Neoantigens are protein fragments derived from cancer cell mutations that serve as unique markers distinguishing only cancer cells. Moderna and BioNTech developed their COVID-19 vaccines using the messenger ribonucleic acid (mRNA) platform secured during their research on neoantigen-based cancer vaccines. Currently, global pharmaceutical companies are actively conducting clinical trials for cancer vaccines.
The problem is that most existing cancer vaccine technologies focus solely on T-cell-centered immune responses. B cells, along with T cells, play a key role in the immune system, and recent studies have increasingly demonstrated their importance in anti-cancer immune activity.
Introduction Artificial intelligence (AI) has been influencing the way oncology has been practiced. Major issues constituting a bottleneck are the lack of data for training purposes, confidentiality preventing development, or the absence of transparency in clarifying how models operate to generate decisions. Novel Models With explainable AI, trust and utilization barriers among clinicians, researchers, and patients can be removed. With the implementation of federated learning, multiple institutions could contribute to crucial dataset’s learning information. Precise diagnosis and prescription of the right drug are essential in preventing unnecessary life losses, and economic burden to the underling system.
A large protocluster of galaxies that existed 12.6 billion years ago, first discovered with the Subaru Telescope, has been examined in detail using the James Webb Space Telescope (JWST). The study found that galaxies in crowded regions are more extended than similar galaxies in less dense environments. The results, published in The Astrophysical Journal Letters, show that even when the universe was only 1.2 billion years old, environment was already influencing how galaxies grow.
In today’s universe, galaxies are not spread evenly through space. They have gathered into groups, and those groups form enormous galaxy clusters containing hundreds or even thousands of galaxies. But these giant structures did not exist at the beginning of the universe.
In the early universe, slightly denser regions of matter gradually grew under gravity and eventually developed into galaxy clusters. These “seeds” of galaxy clusters are known as protoclusters.
A mysterious new cosmic pattern discovered by the DAMPE space telescope may finally crack the century-old mystery of cosmic rays. Scientists studying mysterious ultra-powerful cosmic rays have uncovered a surprising hidden pattern that could finally help explain where these particles come from. Using the DAMPE space telescope, researchers found that cosmic ray particles—from tiny protons to heavy iron nuclei—all begin fading away more sharply at the exact same point, hinting at a universal rule governing their behavior across the galaxy.
For more than 100 years, scientists have been trying to understand cosmic rays, incredibly powerful particles that travel across the universe at extreme energies. Despite decades of research, many questions about where they come from and how they are accelerated remain unanswered. Now, researchers working with the DAMPE (Dark Matter Particle Explorer) space telescope have uncovered an important new clue. Their findings, published in Nature, reveal a common feature shared by these mysterious particles and could help scientists better understand their origins.
Cosmic rays are the highest energy particles ever observed in nature. They carry far more energy than particles produced by even the most advanced accelerators on Earth. Scientists believe they are created by some of the universe’s most violent events, including supernova explosions, jets from black holes, and pulsars.
New work explaining the inner workings of artificial intelligence could provide a way around the threat of AI “model collapse,” potentially averting growing numbers of AI hallucinations in the future.
First coined in 2024, “model collapse” refers to a scenario where an AI model trained on AI-produced data ceases to provide accurate results, instead producing inaccurate “gibberish” because of the poor quality of its training data.
Some have warned that high-quality text data to train systems like Large Language Models (LLMs) is set to run out as early as this year, and so data produced by models themselves has taken a larger training role—inviting the threat of model collapse.
People often look to dogs’ behavior, especially their facial expressions, for indications of their states of mind. Numerous studies show that this is a popular interpretation strategy. However, modern dog breeds vary greatly in size and structure, and few studies have explored how breed-specific morphology might affect humans’ ability to assess visual cues from the faces of different breeds of dogs.
Now, for the first time, a collaborative research team including scientists from Israel, Czechia, and Hungary has used eye-tracking to compare the visual attention patterns of humans observing photographs of normocephalic and brachycephalic dogs. A research paper detailing the team’s findings appears in Frontiers in Veterinary Science.
