Human brain has more switches than all computers on Earth | Hacker News Reader

Human brain has more switches than all computers on Earth

news.cnet.com

82 pointscwan31 comments

The title and article are not clearly written IMHO. What they mean is that the brain has more molecular switches than the number of transistors (switches) of all computers. Some numbers:

  * 1.25e14 synapses in a brain
  * they discovered each synapse has 1000 molecular switches
  * so 1.25e17 molecular switches in a brain
  * a post-2005 CPU has 1e8 transistors or so (2010 CPUs barely hit 1e9 transistors)
  * I estimate 1e9 computers on earth
  * so 1e17 transistors in all CPUs
  * 1.25e17 > 1e17 so the statement is right

There are 1.25e14 synapses in the cerebral cortex.

The total number is higher still (the cerebellum is heavilly wired too), but I don't know by how much.

According to Wikipedia[1], citing [2], the human brain has 'as many as' a quadrillion (10e15) synapses. I presume this includes the entire central nervous system (including the cerebral cortex, cerebellum, and subcortical structures).

[1] http://en.wikipedia.org/wiki/Human_brain

[2] http://www.ncbi.nlm.nih.gov/pubmed/8527499

Edit: I just skimmed that paper, and it doesn't look like a reliable source. It's an estimate based on a mathematical analysis, and the authors say "We emphasize that these conclusions are preliminary, based as they are on an incomplete database and simplified models of the brain". I also don't see where the Wikipedia article gets the 10e15 number.

While comparing apples and oranges, they neglected to mention the speed of the "molecular switches." The fastest neurons fire at around 200Hz. Most fire at around 20Hz. Transistors are around 100,000,000 times faster.

Oh and circuits conduct signals at 0.5c while neurons are lucky to go at 0.000001c (300 meters/sec).

This really puts the book 'True Names' in perspective.

May be the mailman could be a fast thinker.

True, but our brains absorb massive amounts of information and process it in parallel whereas it is much more serial for computers.

All parallel tasks can be efficiently serialized on a fast CPU but not all serial tasks can be efficiently parallelized on many slow CPUs.

From http://lesswrong.com/lw/k5/cached_thoughts/ :

Can you imagine having to program using 100Hz CPUs, no matter how many of them you had? You'd also need a hundred billion processors just to get anything done in realtime.

I think the number of neurons should not be missed in those calculations.

This deserves a few more upvotes. There's some ongoing research to determine the role glial cells have in that wonderful computation we call "thought." See, for example this interview: http://www.scientificamerican.com/article.cfm?id=the-root-of... .

Having read the original paper I can tell you a couple things:

1. The technique used to visualize synapses is awesome. 2. This quote about molecular switches comes out of nowhere. It's related but not really part of this study.

More importantly though the idea that these are switches and that there are way more of them than transistors in the world misses two important points.

1. Are the molecular switches relevant to the computation performed by the brain?

Only in so much as to build a transistor you need materials with certain specific properties, and to build a calcium channel you need proteins with certain specific responses to the environment. Comparing the computational unit of one to the building blocks of another isn't quite accurate.

2. Even if they were computationally relevant are they comprable in terms of key metrics like performance?

Two important things to remember about the brain are that its slow and very very efficient. Silicon logic on the other hand is very very fast and inefficient. Even if there were a thousand fold increase in the number of computational units assigned to the brain the processing speed of a modern transistor decimates synaptic level computation.

"...about the brain are that its slow and very very efficient. Silicon logic on the other hand is very very fast and inefficient..."

Can you elaborate? In what way is the brain efficient in a sense that a transistor isn't?

In the energetic efficiency meaning of the word.

The energy consumption of the brain is around 20W in an adult human.

Compare that to the multi hundred Watt consumption of the meager machine you're reading this on...

Oh I see. But like the rest of your comment indicated, it's hard to compare until we understand more of the brain, right? Depending on the speed of the brain, the 'energy per thought' or 'energy per operation' can go either way when we have better measurements and understanding of the things brains are good at and not so good at?

I don't think the analogy was pulled out of ... hrm, /dev/random. afaik, the brain researchers have data for opinions.

E.g. the functions of the first levels of the visual cortex is relatively well understood -- both from mapping the nerve connections and from copying the mechanism in computers.

There is a long time evolutionary pressure to conserve energy, all the way back to the evolution of nerves. (The brain use quite a lot of your total energy use, unless you're a non-mechanized lumberjack...)

Edit: The visual cortex might work differently than other parts of the brain (ask a researcher) because of speed demands, which otoh supports the point about energy efficiency.

Wow, a mouse genetic engineered just to make certain bigger neurons (but not all of them) to emit a green glow. One can do that!?

Low-level neurological concepts like synapses do not translate well into (low-level) computer terms if you do it by force instead of by reasonable consideration. A synapse is not a switch. A neuron is neither a simple weighted node nor is it an 800 MHz signal processing device with gigabytes per second of throughput. By going arbitrarily deep into the biochemistry, people could (and probably do) come up with ridiculous numbers. For example, maybe somebody wants to count and represent all the macromolecules in a neuron that modulate its function. It's easy to come up with an arbitrarily large number and land a nice gig on CNet for it.

When I was working on AI research in the late 80s and 90s (on the symbolic reasoning side of things) I got the distinct impression that Artificial Neural Networks were really just a nice statistical technique that had a marketing breakthrough - they really appeared to have very little to do with exploiting the behaviors found in actual brain cells.

Artificial neural networks do work well as a model of behavior of biological neural nets, up to a point. They are definitely a piece of the puzzle, a module to duplicate a certain type of information that is represented in biological brains. When they became suddenly famous in the 80s and 90s, the mistake was to assume we could build everything, including an artificial mind, if we just had a large enough neural net. That was essentially the IBM approach: just throw enough resources at it and it will become intelligent. Turns out, more is needed for building actual intelligence.

I firmly believe that neural nets and other techniques are still essential components needed for implementing artificial minds. We now know that processing power and storage space are alone are not enough, a brain needs actual software that tells it what to do with information and how to organize itself. That's essentially how I became very skeptical of the kind of brute force AI research that is being conducted today. For instance, modeling a synapse chemically down to the atomic level is nice for basic research, but it's definitely not the way to implement AI. For this, we need larger abstractions that are functionally equivalent and translate well into efficient computer code, and we need to figure out how to make these pieces of code interact with each other in a meaningful way. My wild guess would be that today we're not even constrained by computing power or storage needs, we just lack the correct design.

My own suspicion is that real "general" AI probably will be developed by reverse engineering the human brain and working backwards to the key processes and structures that provide general intelligence.

Of course, this is assuming that there isn't something deeply spooky going on driving human consciousness - which is a possibility I used to regard as terribly silly but some of the concepts alluded to (in all places) Neal Stephenson's Anathem have got be wondering about such things again.

Thank you for mentioning Stephenson. I haven't read him other than Snow Crash, and I don't think I finished it.

Googling Anathem brought me to this blog review of it:

http://neopythonic.blogspot.com/2008/10/thoughts-after-readi...

which reminded me of this NIH neuro anatomist who studied her own stroke, including during her multi-year recovery.

Her TED talk: http://www.ted.com/talks/jill_bolte_taylor_s_powerful_stroke...

YouTube of same: http://www.youtube.com/watch?v=UyyjU8fzEYU

She sees the right brain hemisphere as being our "consciousness" wetware connecting us to others.

Parts of the video are esoteric, but it's fascinating to hear this first-person account from a brain researcher, especially of the morning of her stroke when her left hemishphere was damaged by a spontaneous brain hemorrhage.

Edit: couple of typos

Cool - thanks, that looks pretty interesting.

There is, however, now a company called Numenta that's working on AI using structures they call Hierarchical Temporal Memory that are, in fact, based on trying to model the sort of structures actually found in the brain.

http://www.numenta.com/for-developers/education/general-over...

I was also working in AI in the late 80s and 90s, but in a lab that loved the neural network side of things. I decided pretty quickly that the biological inspiration for NNs was more marketing than useful. The more interesting developments were being done with Bayesian stuff (at least there was a mathematical theory behind the performance).

FWIW, I ended up doing my PhD on Genetic Algorithms / Genetic Programming : where the biological inspiration (and understanding) works both ways (IMHO).

I think the REAL question we're all wondering is when we will have access to a super computer with the capacity of the brain IF we follow Moore's Law.

I don't think this question is well defined enough to be meaningful. In terms of how quickly it can carry out complex calculations, low end modern laptops leave your brain in the dust. What few functions the brain can still claim any advantage on are the result of dedicated wetware and our ability to learn and thereby specialize software. Computers will get there, but it's not really a Moore's Law problem.

The title of this article should not be misconstrued to mean that the brain is a more powerful computation device, only that it is more complicated.

I don't think this question is well defined enough to be meaningful.

You can certainly get closer than asserting that, e.g., a hand calculator is more powerful than a human brain because it can do 7 digit long division almost instantly.

At some point, we'll have computers powerful enough to run a working simulation of a human brain. Shortly before that, there will be a point when computers of that sort are "as powerful as a human brain", though we might not realize it when that happens if the software lags.

So that's an upper bound on charlesju's question.

Computational Neural Networks are getting further and further from the reality of their inspirational biological models. It'd be interesting to see some of this physiology research transition into revised computational models.

There's more than one way to skin a cat.

By design Evolution is not perfect, which suggests there are better ways to create AI. To me it seems that trying to replicate how our human brain works is a wrong approach.

On the other hand, there is no reason why we cannot replicate human brain design if we wanted to[1]. Evolution had millions of years and we are just getting our hands wet.

[1]http://www.bbc.co.uk/news/technology-11734909

AI is also not always about building a human out of silicon, but about building something that is in some specific ways better (albeit much worse in other ways). For example, we ideally want to understand what it's doing and why, something we so far understand fairly imperfectly for humans. We also want to be able to control it fairly directly and reliably--- a society of AIs that you have to herd like cats and coax into doing what you want isn't quite what most engineers looking to plug in intelligent modules into their systems are looking for. We also want it to be very reliable and fast at tasks that humans do poorly or get bored doing, like scanning huge quantities of data to find patterns.

At least, that's how it looks from an engineering side. If the end result is that all you get is a human, well, we already have humans; just hire them instead. From a philosophical and technical side creating artificial humans does still remain quite fascinating.

Take that, Mr. Kurzweil.

Hehehe, and 83% of brains are only used to watch porn...

Reply on news.ycombinator.com