Yes, the Brain Is a Computer | Hacker News Reader

Yes, the Brain Is a Computer

recursed.blogspot.com

32 pointsALee18 comments

The author's analogy at the beginning involving birds, airplanes, and flying machines seems pretty spot on to me, but it's a double edged sword the author apparently hasn't thought much about.

Ask yourself if it's apt to call a bird a flying machine.

You likely see the problem right away. That description is at least accurate within limits, but it's also reductive to the point where it will induce blindness if you don't recognize its limits. A "flying machine" is only one of the things a bird is. And if you were trying to figure out how to make something very much like a bird, you wouldn't get very far by focusing on the idea that a bird is an flying machine like an airplane.

Similarly, it's true enough that the brain is a computer in the sense that it does computation. That description is at least accurate within limits. But brains do things that we don't know how to get computers to do yet (yes, we're getting better at creating systems that have capacity for performing/optimizing tasks with less explicitly pre-encoded domain logic, but we're still doing that, and as far as I can tell nobody has demonstrated any testable or practically applicable idea of what constitutes consciousness). It's reasonable to assume that a computer might be only one of the things a brain is, and it seems likely that you won't get far by focusing on the idea that a brain is like a present-day von neumann digital computer in the same way a bird is like an airplane.

I agree with you completely!

At the same time, it's worth noting that it makes sense to approach the problem in a reductive (using the word loosely, not sure if it's completely accurate) manner, i.e., keep replicating features and capabilities of human brains until we possibly hit a wall in the form of consciousness, although it's quite possible that it turns out to be an emergent property [1], although I suspect it's a bit more complicated than that.

[1] It seems to me that most explanations that consciousness is "emergent" or an "illusion" would never _predict_ the experiences we actually consider consciousness, like "fire is caused by phlogiston": http://lesswrong.com/lw/ip/fake_explanations/. At the same time, I recognize that I couldn't tell the difference between this world and one populated by philosophical zombies (http://lesswrong.com/lw/p7/zombies_zombies/).

Edit: I should clarify regarding my footnote that I don't believe in epiphenomenalism, I believe consciousness is a purely material phenomenon, I like to think of it as an overview process running in the brain, responding to the other stimulus to take appropriate action.

I found myself agreeing at first, but today, you could make much the same argument about computers, by which I mean the term "computer" is so overloaded from the original definition of a "machine to compute numbers/equations".

If you approach a more general definition, something like "a computer is machine capable of interpreting arbitrary instructions in arbitrary ways that as long as those instructions follow arbitrary rules can achieve arbitrary ends", saying a brain is an equivalent machine hardly seems reductive. It mainly seems pointless.

Good article. I also posted my comment suggesting it's a general-purpose, analog computer on his blog. It makes Epsteins claims make so much more sense when you think of it as a distributed, fault-tolerant, reconfigurable, analog machine working on models that approximate stuff. So, it's a computer. Just not a digital one.

https://news.ycombinator.com/item?id=11731697

Considering it runs on electrons, you could even consider it a digital one if you wanted to.

No, you can't. Digital circuits use a few, regular cells in predictable ways to emulate an abstract logic that simulates any program. They often are designed to run clocked where they're only active when pulse hits them. They use converters to deal with analog world's real signals.

Analog circuits, relevant here at least, are primitive components that directly implement mathematical functions on inputs and outputs consisting of real, electrical activity. They... so far... can't implement arbitrary problems like a Turing machine. However, they do run at full-speed, clockless, need no converters, use little power, and can implement things shown to be computers. There's also computer models for them that suggest general-purpose computation is possible.

Both use electrons. One is more like neurons in the brain. Seriously, look at my first link to see how analog computers differed from digital ones.

Isn't a digital signal an analog signal with a particular interpretation? I'm speaking of the static and dynamic disciplines for digital design. A digital logic cell is an analog circuit with a nonlinear filter to bias the signal away from 0.5, ensuring interpretability, keeping noise from amplifying. And there are defined periods after a digital signal settles during which outputs are assumed not to have any valid digital interpretation.

To nitpick: a digital circuit is always running (at the analog level) whether or not a clock pulse is hitting it. It's just that the digital circuit may begin to switch state when the clock signal is interpreted as a pulse. Then the analog transients eventually die out, which means everything has a digital interpretation, before the next pulse.

Of course, none of this applies to an ideal digital circuit with every wire holding a perfect 0 or 1 signal. I think might be what you are talking about when you say "digital."

I think if you used a genetic algorithm to produce digital circuits to model some function, you'd find it would use the logical gates well outside spec, taking advantage of intermediate states that we avoid by principle. In other words, sure it would use logical gates, but it would be an analog circuit.

"Isn't a digital signal an analog signal with a particular interpretation?"

Yes, technically. I love hitting people with that when they say digital "killed" analog. ;) What I'm getting at is that digital computers with digital cells and RTL operate in a way that simulates abstract operations on non-existent structures (bits) usually pretending to be on only periodically and requiring analog for real world interaction. The analog circuits, even Boolean alternatives, are closer to how the brain works as it's primitive components acting on real, physical values whose input and output mapping is more like a mathematical function. It similarly has precision and noise issues.

"a digital circuit is always running (at the analog level) whether or not a clock pulse is hitting it"

It's an abstraction issue. Digital is an abstraction over analog. At the digital level, it's not running since it's not performing the digital function or only partly. At analog level, it may or may not be running due to power management scheme. Clock gating basically.

"I think if you used a genetic algorithm to produce digital circuits to model some function, you'd find it would use the logical gates well outside spec, taking advantage of intermediate states that we avoid by principle."

Did you read that or were you really guessing? If the latter, it's a really good guess. I guessed that too given genetic programming results on antenna design. Our guesses were validated by various optimization techniques being applied to synthesis of analog circuits, including digital gates. Irritated I can't find the specific paper where they evolved cells or something but it used the opposite building blocks of human designers with better characteristics. Did find these, though.

List of GP's evolved analog circuits that are human competitive in some way:

http://www.genetic-programming.com/eh2004.pdf

ATLAS - simpler tool synthesizing individual, analog circuits

http://www.ece.uc.edu/~dasan/iscas08.pdf

Analog framework for fully-automated, standard cells and more

http://www.eecs.berkeley.edu/Pubs/TechRpts/2013/EECS-2013-71...

Note: Sorry I can't find the better stuff. Apparently, bio-tech, post-synthesis optimization, DNA computers, and analog synthesis of digital cells all use the exact, same words in their papers. Floods search results with garbage. Unreal. If you want, send me an email with subject like "synthesis of digital cell libraries" with a reminder to send you the paper if I run across it in future.

Or alternatively, a digital impulse chemical continuum computer.

Features include: * No stack * Mostly type safe * Process complex thoughts known as emotions

Bugs: * ships without a C compiler * Consumes too much coffee * Version upgrades correlated to earth's rotation around the sun.

You're getting closer. Digital computers work in individual, specific operations on a data representation. Analog computers are continuous on actual, raw data. The brain is continuous on raw data with some kind of storage. So, it seems to be a mostly analog system with possibly digital-like features embedded in it. The combo of analog and digital is called mixed-signal in ASIC design, though usually more digital than analog.

Brain seems like a mixed-signal system that's mostly analog.

" ships without a C compiler *"

A feature. Invents & continually improves compilation instead from safer languages to C to safe languages again. ;)

"Version upgrades correlated to earth's rotation around the sun."

That's funny and original.

It doesn't run on electrons. To the extent that the brain runs on electricity, it modulates that electric potential by moving positively charged sodium and potassium ions, not electrons, across cell membranes. Also, the ions flow back and forth across the membrane, not parallel to it. Furthermore, the electrical signal only travels within a single neuron. Transmission from one neuron to another by sending small molecules called neutotransmitters across synapses.

Of course, none of the above precludes the brain being considered a digital computer. However, the brain does not operate on any kind of fixed clock, and the strength of a signal from one neuron to another is measured in neurotransmitter molecules per second, a fundamentally analog quantity.

I was frustrated reading the article to which this one is responding. Nice to read a categorical rebuttal.

There was a time where "computer" used to be a job title. The digital computers we created were basically made in our image. Given this, who can be surprised at the idea that our nervous systems are computers of sorts?

When I read the dialogue at the beginning of this article, I knew exactly what article he was responding to. I had the same annoyances reading that article, good response!

Your brain "is" indeed a computer in the same way that your eye "is" jelly.

Sure it computes, but in case you've forgotten there is a whole lot of other stuff going on as well and we don't have the first clue how.

The brain is a computer, a chemical regulator, and a number of other things. Those are functional descriptions. The eye is an optical sensor, built in part on a jelly and protein substrate, just like the rest of our bodies.

Put another way, the eye is composed of jelly (among other things), but the brain isn't composed of a computer (among other things). It's a difference between function and composition.

Delightful snark AND intellectually interesting. Good one =D

Wow, that guy completely misunderstood the original poster and showcased not only a lack of philosophical depth but also a condescending and bickering attitude.

Bad article, too personal and intellectually frigid.

Reply on news.ycombinator.com