Tag: ai

Gauge CNN

going beyond the Euclidean plane would require them to reimagine one of the basic computational procedures that made neural networks so effective at 2D image recognition in the first place. This procedure, called “convolution,” lets a layer of the neural network perform a mathematical operation on small patches of the input data and then pass the results to the next layer in the network. Gauge CNNs can detect patterns not only in 2D arrays of pixels, but also on spheres and asymmetrically curved objects. “This framework is a fairly definitive answer to this problem of deep learning on curved surfaces”

this has been applied to proteins:

Correia’s system, called MaSIF (short for molecular surface interaction fingerprinting), avoids the inherent complexity of a protein’s 3D shape by ignoring the molecules’ internal structure. Instead, the system scans the protein’s 2D surface for what the researchers call interaction fingerprints: features learned by a neural network that indicate that another protein could bind there. “The idea is that when any 2 molecules come together, what they’re essentially presenting to 1 another is that surface. So that’s all you need,. It’s very, very innovative.”

GPT-2 Chess

What does this imply? I’m not sure (and maybe it will imply more if someone manages to make it actually good). It was already weird to see something with no auditory qualia learn passable poetic meter. It’s even weirder to see something with no concept of space learn to play chess. Is any of this meaningful? I still don’t know.

MuZero

MuZero learns a model that, when applied iteratively, predicts the quantities most directly relevant to planning: the reward, the action-selection policy, and the value function. When evaluated on 57 different Atari games – the canonical video game environment for testing AI techniques, in which model-based planning approaches have historically struggled – our new algorithm achieved a new state of the art. When evaluated on Go, chess and shogi, without any knowledge of the game rules, MuZero matched the superhuman performance of the AlphaZero algorithm that was supplied with the game rules.

Life In Our Phage World

The world of phages is more than a little scary. They have been evolving for billions of years, their numbers are so vast every writer in this anthology resorts to scientific notation, and their generation time is as low as minutes, making for dizzying amounts of selection pressure and optimization – phages seem to have explored every possible way of attacking, subverting bacteria, replicating faster, compacting and making themselves more efficient, and won every arms-race bacteria started with them.

2023-01-12: And the reverse

A species of plankton that populate freshwater worldwide is the world’s first known organism that survives and thrives by dining on viruses alone, an advance that sheds new light on the role of viruses in the global food web. This virus-only diet – “virovory” – is enough to fuel the growth and reproduction of a species of Halteria, a single-celled organism known for the minuscule hairs.

2025-09-18: Viable AI-mutated phages

the researchers mixed all 16 AI-generated phages with ΦX174 and then threw them into a tube with E. coli cells. Because the phages were forced to compete for the same host cells, the variants that reproduced fastest would dominate. By sequencing the phages over time, the researchers could track which phages were gaining ground and which were falling behind. Several of the AI phages consistently outperformed wild ΦX174, with one variant (called Evo-Φ69) increasing to 65x its starting level.

Ultimately, these 16 AI-generated phages were not only viable; in many cases, they were more infectious than wildtype ΦX174 despite carrying major genome alterations that a human would be unlikely to rationally design.

AI applied to Education

The future of online education is adaptive assessment, not for testing, but for learning. Incorrect answers are not random but betray specific assumptions and patterns of thought. Analysis of answers, therefore, can be used to guide students to exactly that lecture that needs to be reviewed and understood to achieve mastery of the material. Computer-adaptive testing will thus become computer-adaptive learning.

More With Less

There’s still a lot of potential to build more efficient and larger scale computing systems, particularly ones tailored for machine learning. And I think the basic research that has been done in the last 5 or 6 years still has a lot of room to be applied in all the ways that it should be. We’ll collaborate with our Google product colleagues to get a lot of these things out into real-world uses.

But we also are looking at what are the next major problems on the horizon, given what we can do today and what we can’t do. We want to build systems that can generalize to a new task. Being able to do things with much less data and with much less computation is going to be interesting and important.

Steppingstone principle

The steppingstone principle goes beyond traditional evolutionary approaches. Instead of optimizing for a specific goal, it embraces creative exploration of all possible solutions. By doing so, it has paid off with groundbreaking results. 1 system based on the steppingstone principle mastered 2 video games that had stumped popular machine learning methods. DeepMind reported success in combining deep learning with the evolution of a diverse population of solutions. The steppingstone’s potential can be seen by analogy with biological evolution. In nature, the tree of life has no overarching goal, and features used for 1 function might find themselves enlisted for something completely different. Feathers likely evolved for insulation and only later became handy for flight. If we want algorithms that can navigate the physical and social world as easily as we can — or better! — we need to imitate nature’s tactics. Instead of hard-coding the rules of reasoning, or having computers learn to score highly on specific performance metrics, we must let a population of solutions blossom. Make them prioritize novelty or interestingness instead of the ability to walk or talk. They may discover an indirect path, a set of steppingstones, and wind up walking and talking better than if they’d sought those skills directly.