News:

PD.com: Ten minutes of your life that you can never get back.

Main Menu

Kurzweil hilarity

Started by Triple Zero, August 18, 2010, 10:33:53 AM

Previous topic - Next topic

Triple Zero

(via reddit)

http://scienceblogs.com/pharyngula/2010/08/ray_kurzweil_does_not_understa.php

quoting:


* [It is] Kurzweil's assessment that about a million lines of code may be enough to simulate the human brain.

* Here's how that math works, Kurzweil explains: The design of the brain is in the genome.

* The human genome has three billion base pairs or six billion bits, which is about 800 million bytes before compression, he says.

* Eliminating redundancies and applying loss-less compression, that information can be compressed into about 50 million bytes, according to Kurzweil.

* About half of that is the brain, which comes down to 25 million bytes, or a million lines of code.



uhuh. Yep. That's exactly how it works. Yup.

Before you all rush in to point out how he's utterly wrong from a biological point of view, let me say that he is ALSO wrong from a programming/software engineering point of view. 25 megabytes of procedural data, optimized against redundancies, does NOT equal a million lines of code. Unless you're talking about a million "oneliners" in Perl or C code.

A "oneliner" is a single line of code that does way more than a line of code should do by any sensible standard of code readability. it often has the functionality of a simple program or powerful procedure. Looking for examples, I think this Twitter feed makes the case rather clear: http://twitter.com/perloneliner

The point is, yeah, you can have a million of those, and it would (maybe) have the same functionality as 25MB of procedural data, except that it wouldn't make you any wiser, cause even a seasoned Perl programmer can't read oneliners fluently (without puzzling) and once they start interacting with eachother (as they would, in brain software), it becomes rather impossible to grasp.

Oh, there's another good analogy between "brain software" and these Perl oneliners. As some of you will undoubtedly realize, the data and software of the DNA is useless without the organism's hardware to run it on. Just the DNA, the basepairs, it's not enough, you need to figure out how they encode proteins and then what these proteins will do in the context of the living organism. It's bootstrapping. Like Perl code, the statements refer to a huge collection of libraries with Operating System calls, specialized algorithms, built-in stuff like the regular expression parser. Without that it's just a fancy calculator.
Ex-Soviet Bloc Sexual Attack Swede of Tomorrow™
e-prime disclaimer: let it seem fairly unclear I understand the apparent subjectivity of the above statements. maybe.

INFORMATION SO POWERFUL, YOU ACTUALLY NEED LESS.

Requia ☣

Why is he including compression, compression doesn't effect the number of lines of code, just the amount of space you need to store it.
Inflatable dolls are not recognized flotation devices.

Triple Zero

Quote from: Requia ☣ on August 18, 2010, 03:41:47 PM
Why is he including compression, compression doesn't effect the number of lines of code, just the amount of space you need to store it.

I assume it's a way to get around to the computational complexity of DNA, given all the "junk" DNA and such. Also to make his statement more bold, perhaps?
Ex-Soviet Bloc Sexual Attack Swede of Tomorrow™
e-prime disclaimer: let it seem fairly unclear I understand the apparent subjectivity of the above statements. maybe.

INFORMATION SO POWERFUL, YOU ACTUALLY NEED LESS.

Bebek Sincap Ratatosk

Transhumanists are teh funnay

:lulz: :lulz: :lulz: :lulz: :lulz:
- I don't see race. I just see cars going around in a circle.

"Back in my day, crazy meant something. Now everyone is crazy" - Charlie Manson

Doktor Howl

A team of two can make a perfectly functioning human brain in about 9 months.
Molon Lube

Triple Zero

Quote from: Doktor Howl on August 18, 2010, 08:04:40 PM
A team of two can make a perfectly functioning human brain in about 9 months.
:spit2:
Ex-Soviet Bloc Sexual Attack Swede of Tomorrow™
e-prime disclaimer: let it seem fairly unclear I understand the apparent subjectivity of the above statements. maybe.

INFORMATION SO POWERFUL, YOU ACTUALLY NEED LESS.

Kai

And from the biological angle....

Let's talk proteins.


Consider this. You have a length of DNA sequence, double stranded, and that sequence length, a string of ATCGs, corresponds to a fully functional protein designed by natural selection for some particular molecular task.

HOWEVER,

That length of sequence is only a template. And at the moment, it's tightly bound up. So we unwind it. This takes a few other proteins, already functional, just to get the DNA to a state where we can even do anything with it. Then we copy. This requires several more proteins, including RNA polymerase.

Then theres a copy, a transcript, but it's not the TRUE mRNA yet, it still has to be processed, removing the promotor region, adding a polyadenine tail, splicing out the introns.

Then, yes, it's a messenger RNA, which can serve as a full template for a protein sequence.

But it doesn't stop there. No, no. The mRNA needs a few other intermediates, ribosomes, tRNA (to line up the code with the correct amino acids) and then the assembly line process transcribes itself down the chain and hooks together the amino acids.

Think it's finished now? Not quite. If the above seemed amazingly complex, then the final step is even more so. A protein is only functional when it's folded in the right configuration. We can easily model and predict transcription and translation in the laboratory. We are only able to crudely model protein folding. There is a massive amount of information there, so complex that some entreupeneurial molecular biologists have made a game out of it, called
Foldit. Turns out the human mind is able to twist and compress strings of amino acids into the correct and functional shape much better than any current computer.

Even the most basic of biological functions is phenomenally complex. Some biologists believed it wouldn't be long before a DNA sequence could be entered into a computer and out would pop the organism. How naive. The whole of an organism is epigenetic, it is determined not only by genes, but by environment, by spacial and temporal gradients, by maternal factors, by interactomes of such breadth we are only on the cusp of discovering how they really work.

So, there is no way a million base pairs could simulate a human brain. It doesn't even work that way in nature, though many confuse it that way.
If there is magic on this planet, it is contained in water. --Loren Eisley, The Immense Journey

Her Royal Majesty's Chief of Insect Genitalia Dissection
Grand Visser of the Six Legged Class
Chanticleer of the Holometabola Clade Church, Diptera Parish

Bebek Sincap Ratatosk

Quote from: Kai on August 23, 2010, 09:48:20 PM
And from the biological angle....

Let's talk proteins.


Consider this. You have a length of DNA sequence, double stranded, and that sequence length, a string of ATCGs, corresponds to a fully functional protein designed by natural selection for some particular molecular task.

HOWEVER,

That length of sequence is only a template. And at the moment, it's tightly bound up. So we unwind it. This takes a few other proteins, already functional, just to get the DNA to a state where we can even do anything with it. Then we copy. This requires several more proteins, including RNA polymerase.

Then theres a copy, a transcript, but it's not the TRUE mRNA yet, it still has to be processed, removing the promotor region, adding a polyadenine tail, splicing out the introns.

Then, yes, it's a messenger RNA, which can serve as a full template for a protein sequence.

But it doesn't stop there. No, no. The mRNA needs a few other intermediates, ribosomes, tRNA (to line up the code with the correct amino acids) and then the assembly line process transcribes itself down the chain and hooks together the amino acids.

Think it's finished now? Not quite. If the above seemed amazingly complex, then the final step is even more so. A protein is only functional when it's folded in the right configuration. We can easily model and predict transcription and translation in the laboratory. We are only able to crudely model protein folding. There is a massive amount of information there, so complex that some entreupeneurial molecular biologists have made a game out of it, called
Foldit. Turns out the human mind is able to twist and compress strings of amino acids into the correct and functional shape much better than any current computer.

Even the most basic of biological functions is phenomenally complex. Some biologists believed it wouldn't be long before a DNA sequence could be entered into a computer and out would pop the organism. How naive. The whole of an organism is epigenetic, it is determined not only by genes, but by environment, by spacial and temporal gradients, by maternal factors, by interactomes of such breadth we are only on the cusp of discovering how they really work.

So, there is no way a million base pairs could simulate a human brain. It doesn't even work that way in nature, though many confuse it that way.

:mittens:

Kai, Bringin the SCIENCE
- I don't see race. I just see cars going around in a circle.

"Back in my day, crazy meant something. Now everyone is crazy" - Charlie Manson

Jasper

There are approximately 0.15 quadrillion synapses in the cortex.

Can 1M lines of code account for that?

(I don't know the answer, no really.)

Kai

Quote from: Sigmatic on August 23, 2010, 09:56:46 PM
There are approximately 0.15 quadrillion synapses in the cortex.

Can 1M lines of code account for that?

(I don't know the answer, no really.)

Also, the number of synapses can grow over a lifetime, and change, etc.

Sure, you could have 150 trillion bits on a computer representing the synapses in the brain. But they just sit there, unless there is some OTHER code to tell them how to "interact". How are they receiving information, what sort of imput? How are they outputting information? Not to mention compilation. Because as much as it's nice to think that I can throw a whole bunch of neurons together randomly and something intelligent will emerge, the structure that permits our higher functions is deeper than that, there is a regionality of neurons by function, a very developed structure spanning the whole of vertebrate evolution.

Can we simulate a fish brain? No. That alone tells how distant we are from Turing AI.

I'd like to add that Zero's answer was as much science as mine.
If there is magic on this planet, it is contained in water. --Loren Eisley, The Immense Journey

Her Royal Majesty's Chief of Insect Genitalia Dissection
Grand Visser of the Six Legged Class
Chanticleer of the Holometabola Clade Church, Diptera Parish

Kai

Quote from: Triple Zero on August 18, 2010, 04:03:41 PM
Quote from: Requia ☣ on August 18, 2010, 03:41:47 PM
Why is he including compression, compression doesn't effect the number of lines of code, just the amount of space you need to store it.

I assume it's a way to get around to the computational complexity of DNA, given all the "junk" DNA and such. Also to make his statement more bold, perhaps?

[Aside] Yeah, junk DNA,  :lulz: More research suggests that a good amount of that junk DNA isn't junk, but rather promoters, enhancers, and introns that lead to mutations which all happen to be non coding regions. Just like dark matter isn't "dark", just non-luminous.[/Aside]
If there is magic on this planet, it is contained in water. --Loren Eisley, The Immense Journey

Her Royal Majesty's Chief of Insect Genitalia Dissection
Grand Visser of the Six Legged Class
Chanticleer of the Holometabola Clade Church, Diptera Parish

Requia ☣

Quote from: Sigmatic on August 23, 2010, 09:56:46 PM
There are approximately 0.15 quadrillion synapses in the cortex.

Can 1M lines of code account for that?

(I don't know the answer, no really.)

Easily, the code doesn't have to cover every synapse, rather it needs to govern the way the synapses grow and interact and so forth.

In terms of simulating cells (which is a giant chunk of what that DNA does) its even more complicated than Kai points out, because the computer doesn't even have rules built in for how atoms chain together, DNA doesn't need to simulate chemistry, it can use the real thing. 

From the opposite perspective, why bother simulating a whole cell in the first place?  Makes more sense to treat cells and proteins as black boxes and not worry about how they get the results as long as the simulation gets the same results.  Looking at it that way there's probably a tiny tiny amount of DNA responsible for the brain when compared to the whole, so comparisons of amount of code to DNA don't work in either direction.
Inflatable dolls are not recognized flotation devices.

President Television

#12
Quote from: Doktor Howl on August 18, 2010, 08:04:40 PM
A team of two can make a perfectly functioning human brain in about 9 months.

Perfectly functioning?

Quote from: Requia ☣ on August 23, 2010, 11:50:06 PM
Quote from: Sigmatic on August 23, 2010, 09:56:46 PM
There are approximately 0.15 quadrillion synapses in the cortex.

Can 1M lines of code account for that?

(I don't know the answer, no really.)

Easily, the code doesn't have to cover every synapse, rather it needs to govern the way the synapses grow and interact and so forth.

In terms of simulating cells (which is a giant chunk of what that DNA does) its even more complicated than Kai points out, because the computer doesn't even have rules built in for how atoms chain together, DNA doesn't need to simulate chemistry, it can use the real thing. 

From the opposite perspective, why bother simulating a whole cell in the first place?  Makes more sense to treat cells and proteins as black boxes and not worry about how they get the results as long as the simulation gets the same results.  Looking at it that way there's probably a tiny tiny amount of DNA responsible for the brain when compared to the whole, so comparisons of amount of code to DNA don't work in either direction.

But that doesn't sound like it would be very accurate at all. You wouldn't have a working human brain, you'd have a representation of a brain. At best.
Well, that's just my gut feeling, anyway, coming from someone who doesn't know anything about computer science.
My shit list: Stephen Harper, anarchists that complain about taxes instead of institutionalized torture, those people walking, anyone who lets a single aspect of themselves define their entire personality, salesmen that don't smoke pipes, Fredericton New Brunswick, bigots, philosophy majors, my nemesis, pirates that don't do anything, criminals without class, sociopaths, narcissists, furries, juggalos, foes.