Category Technology/Electronics

Self-adapting LLMs behave more like students to absorb new knowledge

Self-adapting LLMs behave more like students to absorb new knowledge
Credit: AI-generated image

In an MIT classroom, a professor lectures while students diligently write down notes they will reread later to study and internalize key information ahead of an exam.

Humans know how to learn new information, but large language models can’t do this in the same way. Once a fully trained LLM has been deployed, its “brain” is static and can’t permanently adapt itself to new knowledge.

This means that if a user tells an LLM something important today, it won’t remember that information the next time this person starts a new conversation with the chatbot.

Now, a new approach developed by MIT researchers enables LLMs to update themselves in a way that permanently internalizes new information...

Read More

Wild new “gyromorph” materials could make computers insanely fast

Gyromorphs Boost Light Computing Power
Illustration of a 60-fold gyromorph’s properties. Top row: Structure of the gyromorph. Left: Structure factor. Right: Pair correlation function. Bottom row: Optical properties. Left: Polarized light beam fully reflected by a gyromorph. Right: Density of states depletion in the gyromorph. Credit: The Martiniani lab at NYU

Gyromorphs merge order and disorder to deliver unprecedented light-blocking power for next-generation photonic computers. Researchers engineered “gyromorphs,” a new type of metamaterial that combines liquid-like randomness with large-scale structural patterns to block light from every direction. This innovation solves longstanding limitations in quasicrystal-based designs and could accelerate advances in photonic computing.

Researchers are exploring a new genera...

Read More

Google’s plan for space-based computing

Two CubeSats orbiting around Earth after being deployed from the ISS. Google is looking at ways satellites can be used for data centres of the future (Credit : NASA)
Two CubeSats orbiting around Earth after being deployed from the ISS. Google is looking at ways satellites can be used for data centres of the future (Credit : NASA)

The sun produces more power than 100 trillion times humanity’s entire electricity generation. In orbit, solar panels can be eight times more productive than their Earth-bound counterparts, generating energy almost continuously without the need for heavy battery storage. These facts have led a team of Google researchers to ask what if the best place to scale artificial intelligence isn’t on Earth at all, but in space?

Project Suncatcher, Google’s latest space mission, envisions constellations of solar-powered satellites equipped with processors and connected by laser-based optical links...

Read More

First full simulation of 50 qubit universal quantum computer achieved

New record on JUPITER: Simulating a 50-qubit quantum computer
View between the racks of JUPITER. Credit: Forschungszentrum Jülich / Sascha Kreklau

A research team at the Jülich Supercomputing Center, together with experts from NVIDIA, has set a new record in quantum simulation: for the first time, a universal quantum computer with 50 qubits has been fully simulated—a feat achieved on Europe’s first exascale supercomputer, JUPITER, inaugurated at Forschungszentrum Jülich in September.

The result surpasses the previous world record of 48 qubits, established by Jülich researchers in 2022 on Japan’s K computer. It showcases the immense computational power of JUPITER and opens new horizons for developing and testing quantum algorithms. The research is published on the arXiv preprint server.

Quantum computer simulations are vital for develo...

Read More