I don't see if it's all that "simple". They just found that we use the same information flow for doing a complicated task, but that applies pretty much to a computer; when it carries out a difficult task, its execution loop is still the same. The simplicity of structure does not necessarily mean the simplicity of the method.
Agreed, and they're having to use extensive analogies just to communicate the concept in a coherent way. If it was simple then you would think it could just be explained as it is in itself.
This doesn't seem simple to me. A fully connect external coordinator, helping to determine which preexisting central weightings can be re-used for the specific task at hand, which tasks need to be redirected to less well-trained networks, and which need novel training, with the final prediction being an amalgamation of all results. Just the idea of trying to design a DL network with similar properties blows my mind!
Further to this picture, in recent years the processing potential of extensive microtubule networks which feature in all cells, and can have special characteristics in nerve cells is beginning to be recognised:
> Microtubules are major architectural elements without which the neuron could not achieve or maintain its exaggerated shape. In addition to serving as structural elements, microtubules are railways along which molecular motor proteins convey cargo. Microtubule arrays in axons, dendrites, growth cones, and migratory neurons are tightly organized with respect to the intrinsic polarity of the microtubule, which is relevant to both its assembly and transport properties. Vibrant research is being conducted on the mechanisms by which microtubules are organized in different compartments of the neuron, how microtubule dynamics and stability are regulated, and the orchestration of microtubule-based transport of organelles and proteins. While all of this is surely enough to cause one to marvel, we cannot avoid pondering - what other work might microtubules do for neurons?
The notion of manifolds as mechanism for behavioral or cognitive control is not new; much work has been done in the field of movement science e.g. juggling and other forms of dynamic pattern generation. See for example JAS Scott Kelso's Dynamic Patterns or Spivey's Continuity of Mind or the newish field of Radical Embodied Cognition.
This is fairly systems neuroscience oriented approach much welcome, but not ground breaking.
That's what it means here, too. The idea is that if you take the N-dimensional state-space of neural activity --- each dimension being the activity of a neuron or electrode or voxel or whatever --- and look at how the state evolves through time, you discover that it "lives" in a manifold, k, of much smaller dimension than the full space, N.
This is, I think, simply a restatement of the fact that neurons exhibit correlations in their activity patterns. That is, not all vectors of activity are "allowed". Still, it has implications for everything from learning to brain-machine interfaces.
A meta comment: in academia, for an outsider a specific paper's contribution is usually hard to decipher. Many papers will sound fairly similar. Popular articles like TFA will essentially have to resort to describing a whole branch of research or even a whole subfield. Authors of the academic paper often don't even recognize their work from just reading the PR summary produced by the marketing department of the university (although this one is written by the original authors).
Hardly any work is ground breaking in science nowadays. There are incremental advances most of the time.
Manifold in the sense of a low dimensional locally flat object, in a high dime sional space. An autoencoder has a manifold at the bottleneck layer, for example.
The post is weirdly similar to the American Scientist article from the July 2019 issue entitled "How the Mind Emerges from the Brain's Complex Networks"
In the natural world, ants of some species (initially) wander randomly, and upon finding food return to their colony while laying down pheromone trails. If other ants find such a path, they are likely not to keep travelling at random, but instead to follow the trail, returning and reinforcing it if they eventually find food
