Tag: science

Mapping the Ocean Floor

One of the most amazing things to find unexpectedly is often a wreck. There are tons of those around, we don’t know where they are, and we often find them unexpectedly. We’ve found seamounts—4000-meter-high seamounts that we find when we think nothing is there, and these have impacts on biodiversity and circulation. We find all kinds of interesting structures on the seafloor that we just don’t understand yet. And again we find it looking at it now through a new lens of high resolution.

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.

Against File systems

File systems unfit as distributed storage backends

Breaking the assumption that a distributed storage backend should clearly be layered on top of a local file system allowed Ceph to introduce a new storage backend called BlueStore with much better performance and predictability, and the ability to support the changing storage hardware landscape.

Cosmological Bootstrap

There’s no “time” variable anywhere in the new bootstrapped equation. Yet it predicts cosmological triangles, rectangles and other shapes of all sizes that tell a sensible story of quantum particles arising and evolving at the beginning of time. This suggests that the temporal version of the cosmological origin story may be an illusion. Time can be seen as an “emergent” dimension, a kind of hologram springing from the universe’s spatial correlations, which themselves seem to come from basic symmetries. The approach has the potential to help explain why time began, and why it might end. “The thing that we’re bootstrapping is time itself.”

Magnetars

For each chunk, the team extracted information about the elemental composition of the gases there, creating a more accurate picture of the supernova remnant’s composition than if they had averaged over the whole. They estimated that the supernovas came from stars between 10 and 20 times the mass of the sun — which means they were less massive than what’s needed to give birth to dynamo-powered magnetars.

Interpretable models

Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead

2023-03-01: Math approaches can help with interpretation

Let’s take the set of all cat images and the set of all images that aren’t cats. We’re going to view them as topological shapes, or manifolds. One is the manifold of cats and the other is the manifold of non-cats. These are going to be intertwined in some complicated way. Why? Because there are certain things that look very much like cats that are not a cat. Mountain lions sometimes get mistaken for cats. Replicas. The big thing is, 2 manifolds are intertwined in some very complex manner.
I measure the shape of the manifold as it passes through the layers of a neural network. Ultimately, I can show that it reduces to the simplest possible form. You can view a neural network as a device for simplifying the topology of the manifolds under study.

Cross-organelle response

The problem with serial repitching has beer makers and cell biologists scratching their heads. Clearly, something about the fermentation process is making new generations of yeast less able to ferment. But what? Heat shock response is a complex process because cells are capable of making 1000s of different proteins. But only a subset of these proteins are expressed at any one instant, and the subsets differ inside each organelle in the cell. So maintaining the function of these proteins—proteostasis—requires a complex signaling mechanism that switches on the relevant genes in each organelle. This switching process must be coordinated across the cell, since organelles depend on each other. The process of communication and coordination is called cross-organelle response, or CORE, and it is poorly understood. But this is an important emerging area of cell biology: biologists are beginning to realize that CORE plays a crucial role not just in heat shock response but in metabolism in general, and even in processes such as aging.