DeepScientist is a goal-oriented, fully autonomous scientific discovery system. It uses Bayesian Optimization and a hierarchical 'hypothesize, verify, and analyze' process with a Findings Memory to balance exploration and exploitation. It generated and validated thousands of scientific ideas, surpassing human SOTA on three AI tasks.
Alan Turing and John von Neumann saw it early: the logic of life and the logic of code may be one and the same. This article explores the idea that life, at its core, might be computational, drawing parallels between DNA, computation, and the work of Turing and von Neumann.
All for One: LLMs Solve Mental Math at the Last Token With Information Transferred From Other TokensThis paper investigates how large language models (LLMs) solve mental math problems. It proposes that meaningful computation occurs late in the network (in terms of layer depth) and primarily at the last token, receiving information from other tokens in specific middle layers. The authors introduce techniques (CAMA and ABP) to identify an 'All-for-One' subgraph responsible for this behavior, demonstrating its sufficiency and necessity for high performance across various models and input styles.
Physicists and computer scientists are using stochastic thermodynamics to understand the energy costs of computation, with implications for designing more energy-efficient devices.
The article discusses a phenomenon known as the "Dynamic Quantum Cheshire Cat Effect", which is a type of quantum effect that allows physical properties to be separated from the objects to which they belong. The authors show that this effect can be generalized to dynamical settings, where the property that is separated from the particle can propagate in space and lead to a flux of conserved quantity.
In this essay, Lance Fortnow, a computer scientist, argues that by embracing the computations that surround us, we can begin to understand and tame our seemingly random world. He discusses how even seemingly random events, like a coin flip or the mailing of a letter, can be seen as computational processes. The essay also touches on the progress made in artificial intelligence and machine learning, and how they are helping us manage randomness and complexity in our world.
