Hacker News Reader

Comment by ianai | Hacker News Reader

Key to understanding the thesis:

“They found that groups of neurons connect into 'cliques', and that the number of neurons in a clique would lead to its size as a high-dimensional geometric object (a mathematical dimensional concept, not a space-time one).”

Edit-more to the point;

“"Networks are often analysed in terms of groups of nodes that are all-to-all connected, known as cliques. The number of neurons in a clique determines its size, or more formally, its dimension," the researchers explained in the 2017 paper.”

View on HN

Right. In other words, it has nothing to do with the human brain creating structures in 11 dimensions. It's a total misreading of the science.

This particular claim has been making the rounds for quite some time. It's total rubbish.

> In other words, it has nothing to do with the human brain creating structures in 11 dimensions.

I think you misread that. It means that the title is technically correct but relies on a different definition of the term 'dimension', not that the title is wrong.

I don't agree. The brain does not create structures in up to 11 dimensions, full stop. This is a popular science article. To use that word from standard vocabulary to describe something that just happens to use the same word as some utterly esoteric mathematical definition is misleading at best. And then to imply that the brain creates such structures is just compounding the problem.

What about, "The amazing connectivity of the human brain", or "Researchers unravel the mathematics of connections in the human brain"?

There's absolutely no need to confuse readers by conflating unrelated concepts.

I could make up my own definition of dimension, e.g. the dimension of a manifold needed to embed some bizarre variety with some cleverly defined homological correspondence to slices of brain material I looked at under the microscope, thought of as simplices. I might then get 276 dimensions. Do I then get a press release for this? It's utterly meaningless. "Mathematician discovers that our brains live in a 276 dimensional universe".

> The brain does not create structures in up to 11 dimensions, full stop.

Except it clearly does, using the definition of the mathematics used to analyse connections.

> I could make up my own definition of dimension, e.g. the dimension of a manifold needed to embed some bizarre variety with some cleverly defined homological correspondence to slices of brain material I looked at under the microscope, thought of as simplices. I might then get 276 dimensions. Do I then get a press release for this? It's utterly meaningless

Just because we can make up words, all words that have different meanings in different contexts, are meaningless? I agree that perhaps not enough context was provided, but to say that the words itself do not have a certain meaning because you misunderstood the context in which it's used is ludicrous.

The world does not revolve around you.

Words are used for communication. The definition of a word derives from the mutually understood definition of the word between the author and the reader.

>The world does not revolve around you.

The GP was quite careful to speak from the standpoint of a reader from a general audience. A work of writing does in fact revolve around its audience.

> The GP was quite careful to speak from the standpoint of a reader from a general audience. A work of writing does in fact revolve around its audience.

So was I, otherwise I wouldn't have said:

> I agree that perhaps not enough context was provided, but to say that the words itself do not have a certain meaning because you misunderstood the context in which it's used is ludicrous.

Regardless,

> The definition of a word derives from the mutually understood definition of the word between the author and the reader.

Of course, and the word is defined in the article. So if the reader had actually read the article given, rather than just commenting on the title, then they wouldn't have been confused.

> To use that word from standard vocabulary to describe something that just happens to use the same word as some utterly esoteric mathematical definition is misleading at best.

So are we trying to raise or lower the quality of public discourse? If a journalist uses a technical term in a headline, and then goes on to explain what the term means in the article, how is that not fine? It sounds like you only want journalists talking at a eighth grade level.

Why don't you try to re-write the findings yourself then? I think you'll find it difficult to say the least. Your first obstacle might be finding a better word to describe the mathematical shape of the structures other than 'dimension'. That's what mathematicians use, why can't journalists use it too?

I don’t know about re-writing the findings, but if I was to do so, I’d start with this statement from early in the article:

“We're used to thinking of the world from a 3-D perspective,”

Here 3-D clearly refers to our three spatial dimensions, and the article clearly abuses this definition of the term to contrast with the totally different meaning of “dimension” that is to come.

It’s fine to use non-intuitive mathematical definitions in a work for a non-expert audience, but you must clearly define them from the start — and not lead the audience astray by deliberately referencing the most common definition to form a contrast at the start of your piece.

Journalists of all people should understand the connotation of words.

The title was likely picked to be confusing on purpose as click-bait. Misleading headlines lower the quality of public discourse.

> Misleading headlines lower the quality of public discourse.

Misleading headlines only affect people who have not read the article. People who have not read the article, or can't bother to read the article, are essentially incapable of giving a proper high-quality response to the article.

I don't see how you can lower the quality of discourse any further than someone commenting on an article that they didn't read.

Time is a valuable commodity, it is disingenuous to trick people into reading articles through the use of clickbait headlines.

The article itself also continues to conflate the definitions of dimension by referring to 3-D. When little timmy draws a star on his paper we don't say "congrats on your 5 dimensional drawing".

> Time is a valuable commodity,

Right, so don't waste other people's time by commenting on an article that you haven't properly read.

> If a journalist uses a technical term in a headline, and then goes on to explain what the term means in the article, how is that not fine?

I don't wee where the term is ever very clearly explained, the best I see is:

>> the number of neurons in a clique would lead to its size as a high-dimensional geometric object (a mathematical dimensional concept, not a space-time one)

It is a "network topology" concept, the distinction between "mathematical" and "space-time" is nonsensical.

> That's what mathematicians use, why can't journalists use it too?

Not only is the term poorly explained, but it also appears to be used incorrectly. I don't think it is correct to talk about these structures being created in 11 dimensions, but that as "with 11 dimensions", "having a dimension of 11" or "being 11-dimensional" (which is how the term is used in the paper).

> It sounds like you only want journalists talking at a eighth grade level.

The problem is that the article is talking at about a 8th grade level using collegiate terms that it doesn't take the time to explain at at an 8th grade level. This doesn't help to elevate public discourse, it just sows misunderstanding.

> Why don't you try to re-write the findings yourself then? I think you'll find it difficult to say the least.

This seems needlessly combative.

In terms of accurate and responsible reporting of this science, I think the article that the author cites and paraphrases heavily is much more well written: https://www.eurekalert.org/pub_releases/2017-06/f-bbt060617....

> If a journalist uses a technical term in a headline, and then goes on to explain what the term means in the article, how is that not fine?

If a marketing team uses a financial term in a commercial with an asterisk, and then goes on to explain what the term with an asterisk means in the actual 117 pages long financial contract this commercial is trying to sell, how is that not fine?

It does create 11 dimensional structures, in the same sense programs can have 11 dimensional arrays. Not at the quantum reality level.

I think you probably object because "11 dimensions" sounds a lot like the 11 temporal dimensions of in certain space-time related theories. But this is a completely unrelated set of dimensions.

The article describes what are most definitely dimensions, but ... of the array sort, and not of the String Theory sort.

It's right up there with Lyme Disease bioweapon conspiracy theories and flat-earthers.

It's hard to tell from the post what the terms are being used to describe. First of all, I would consider a dimension either space-time or mathematical to be an orthogonal/independent concern. Now where this gets distractingly blurry is that the paper itself says "The number of neurons in a clique determines its size, or more formally, its dimension." which uses dimension to mean size, like a length.

If I'm generous I could maybe believe that larger cliques form more complex connection patterns and can perform higher dimensional processing. But even this fails as the paper then goes on to describe the cliques and cavities as being higher-dimensional.

All that aside, I do think there's something to the paper, just using some poor choice of terms. It seems what's being described as complex is both the physical non-uniform pattern of connection and the subgraphs of activation which may be able to consider a high number of dimensions for a given input stimulus. I couldn't follow the Simplices/Complexes/Betti numbers to see how it relates to a measure of dimension.

Why would the space-time and mathematical concept of dimensions be orthogonal? Don't we use the mathematical concept of dimension when dealing with space-time dimensions? Software developers in particular should have intuitive understanding of this connection, because implementing space-time in code is usually done through methods that closely resemble mathematical dimensions. Eg if you were to implement a point in 3 dimensional space time you'd give the point xyz and t as variables.

I don't mean the math one orthogonal to the space-time one. I mean that dimensions are orthogonal to each other whether in math or space-time (except special relativities).

May well turn out later on that AI is best modeled for speed and detail at 11 bit resolution - after all - evolution has done a lot of work that we equate as machine learning.

Also interesting that 11 is also a prime number. But equally, may just be one of life's coincidences and the human drive to see patterns and understand exactly what is going on - a trait that still advances humanity today.

Evolution also determined a giraffe's lyrngeal nerve should detour down the neck and back.

Apparently the signal arrived too early so a delay loop was spliced in. Score one for ID.

When the giraffe class was first released it was too OP with its tall neck, allowing it to access food previously unreachable to many other classes while also being able to keep an eye on predators lurking at distance. Upon next evolutionary tick, it had to be nerfed.

For anyone who doesn't get the reference: Watch TierZoo [1]. Hilarious and informative, especially for gamers.

[1]: https://www.youtube.com/watch?v=6xbQ2WbTp0E

> May well turn out later on that AI is best modeled for speed and detail at 11 bit resolution - after all - evolution has done a lot of work that we equate as machine learning.

I don't see gradient descent or back propagation in biology.

Maybe not yet, but for example gradient descent - https://journals.plos.org/plosone/article?id=10.1371/journal...

As for back propagation, I'm sure there are aspects in nature in which translate - maybe the nerves system of some species.

Way to see it is that evolutionary change of any form could in theory be modeled in machine learning. There may be aspects of evolution and how it operates that help refine how we do machine learning.

Well for gradient ascent you have this.

https://en.wikipedia.org/wiki/Chemotaxis

iirc, back propagation sounds similar in inspection to some of the processes seen in sleeping brains. Ie what they see in MRIs looks like what back propagation might look like from outside. But I’m really not comfortable enough with these terms to speak about them.

The jury is still out on back-propagation (i.e. reverse accumulation automatic differentiation) within the brain (retrograde signalling from the post-synaptic neuron to the pre-synaptic neuron may still be discovered and turn out to approximate RM AD)

But regarding evolution implementing back-propagation, I agree that is just bonkers, as it would require cells to back-propagate in time information about death or procreation! If our biology implements information time machines, biology would be a lot weirder than it is (and it would imply that we could use that by sequencing an organisms genome (or ensemble of them) to read from the future, and keeping it fed without allowing reproduction, and later when we want to write information back either kill or let it reproduce).

That's just plain nonsense. There can't be back-propagation in evolution without biological time-travel of information.

And machine learning without back-propagation is basically a genetic algorithm, so the claim that "evolution has done a lot of work that we equate as machine learning" would then reduce to a widely accepted fact that nobody disputes and is thus uninformative: "evolution has done a lot of work that we equate as genetic algorithms"

"There can't be back-propagation in evolution without biological time-travel of information." You still can't do such time-travel in computers, sure you could simulate it but on a chronological scale in reality, your not time traveling.

To equate this into evolution, if you viewed a species as the model, each generation adapts to the previous generations experience, hence it feeds back into the next generation. That evolution may prevail or that whole lineage dies off. It gets down to the perspective of how you measure that time aspect and define the start point of your time perspective. In a simulation you are arbitrary resetting that time baseline on each model, so the equivalent would be to view the human as a baseline of each birth. yes it gets confusing, but to dismiss it just because you have no actual time-travel from a cosmological perspective and compare to an arbitrary time perspective that runs in a linear temporal realm of the universe is unfair.

But many things are hard to equate, you could say it is like comparing apples to pears, but then there are many perspectives in which you can compare them and find them equal. But you can't take one single perspective and transpose that upon another perspective and presume a conclusion.

if natural selection were to use back-propagation, then killing or allowing procreation would be measurable not only in the genome right after birth of an IVF organism, but even be measurable in the half genomes of the sperm and egg cells before they met (assuming we reach some non-destructe genome sequencing technology). That would simply make actual natural selection useless. Moreover, if all cells had this capacity, and our neurons are cells too, why don't I have this skill inately?

I stand by my opinion that natural selection does not employ timetravel, and hence also does not employ backpropagation

EDIT: adding, one does not need timetravel for backpropagation in the brain or in a computer, just enough memory or delay bandwidth to store the intermediate values for backpropagation. i.e. the feedback is nearly instantaneous, or low bandwidth from example to scoring, but scoring in natural selection depends on how much offspring one makes before one dies, how can this inform the choice mutations??

EDIT: upon rereading you seem to believe I claim computing backpropagation on computers is time travel?

Oh ok, so they actually mean that they found 11-dimensional directed simplices in neuronal connectivity, which then implies the barycentric coordinates describing relative activity levels in the simplex will have 11 dimensions. Simple.

The correct title to this article is that the topology of the Human brain corresponds to a set of connections in an eleven dimensional space.

Still too long to post it here..

Reply on news.ycombinator.com