Wholesale biochemical re-engineering - Not too far off now

[Izawwlgood]

Right, and I said that I'm rather okay with that. There are no extra parts there, it's reliable, and it does the job efficiently.

How do you figure? The 'extra parts' don't go away. The machinery to duplicate or such is still there. They just don't get prioritized. You're saying turning the headlights and A/C off in your car makes it drive faster.

All parts of your computer act to serve its function of absorbing data, computing it, and outputting it as specified by the hardware designers, programmers, and users. There isn't a part of the computer that, for example, creates duplicates of itself. If there were, I'd want that removed because it's just another part of the computer that can break.

You're simply fixating on processes you don't understand and calling them 'extra' and objecting to it. I personally don't know why a number of processes happen on my computer, but they don't interfere with the computer serving as a tool for the things I want. Similarly, bacteria will do a number of things you may not understand, but are all required for the survival and maintenance of the bacteria. So just like various antivirus and security protocols are happening on my computer to prevent horrible things from happening to it, so to does a bacteria run 'extra processes' to keep itself alive. You're objecting to these processes, but you haven't actually really stated which process, or why you object to them.

Also;

Actually, I kind of think humans are also probably more complicated than the simplest possible mechanism to store, execute, and express human thought.

I think you're incredibly wrong about this.

[sourmìlk]

Right, and I said that I'm rather okay with that. There are no extra parts there, it's reliable, and it does the job efficiently.

How do you figure? The 'extra parts' don't go away. The machinery to duplicate or such is still there. They just don't get prioritized. You're saying turning the headlights and A/C off in your car makes it drive faster.

No, I'm saying turning off the portion of your car that copies itself makes the car less complicated. The extra parts are just dead there: they don't add to the complexity because they don't function and thus can't fail.

All parts of your computer act to serve its function of absorbing data, computing it, and outputting it as specified by the hardware designers, programmers, and users. There isn't a part of the computer that, for example, creates duplicates of itself. If there were, I'd want that removed because it's just another part of the computer that can break.

You're simply fixating on processes you don't understand and calling them 'extra' and objecting to it. I personally don't know why a number of processes happen on my computer, but they don't interfere with the computer serving as a tool for the things I want. Similarly, bacteria will do a number of things you may not understand, but are all required for the survival and maintenance of the bacteria. So just like various antivirus and security protocols are happening on my computer to prevent horrible things from happening to it, so to does a bacteria run 'extra processes' to keep itself alive. You're objecting to these processes, but you haven't actually really stated which process, or why you object to them.

I singled out the reproduction processes, actually. I'm fine with processes that keep the manufacturer from collapsing, but on philosophical grounds I'm uncomfortable with ones that don't affect manufacturing until they fail.

Actually, I kind of think humans are also probably more complicated than the simplest possible mechanism to store, execute, and express human thought.

I think you're incredibly wrong about this.

Really? We can theoretically make a computer that can simulate human thought. No matter how complex this computer is going to get, it's almost certainly not going to involve more parts in more complexity than quadrillions upon quadrillions of proteins folding and interacting with each other.

[Izawwlgood]

I singled out the reproduction processes, actually. I'm fine with processes that keep the manufacturer from collapsing, but on philosophical grounds I'm uncomfortable with ones that don't affect manufacturing until they fail.

You really don't understand this process!
When trying to purify a protein for pharmaceutical use, you go about this by making a small amount of bacteria take up a DNA plasmid that contains instructions for that protein, and some selectable marker, say, antibiotic resistance. You then select for bacteria that successfully took up the plasmid. So now you've got a small amount of bacteria that have a plasmid in it that you want. To get a bunch of protein, you grow up more of the bacteria. You do this in media with the selection agent. Now you've gone from, say, 50 microliters of media with good density bacteria, to, say, 5 liters of media with good density bacteria. You then add a chemical agent that induces the bacteria to start synthesizing the protein on the plasmid, and wait a few hours. You then spin the bacteria down into a pellet, and lyse the bacteria, and purify the protein.

The 'extra process' you are saying is problematic is bacterial replication. It is vital to the process of making protein, because without the ability of bacteria to replicate, you require not just a hundred nanograms of DNA or so, but a few hundred milligrams of DNA. Think of the scales involved. It's even larger for big pharma when they produce, say, insulin. And it all starts from a seed culture of bacteria that are allowed to replicate.

No, I'm saying turning off the portion of your car that copies itself makes the car less complicated. The extra parts are just dead there: they don't add to the complexity because they don't function and thus can't fail.

How do you think they are affecting protein synthesis?

Really? We can theoretically make a computer that can simulate human thought. No matter how complex this computer is going to get, it's almost certainly not going to involve more parts in more complexity than quadrillions upon quadrillions of proteins folding and interacting with each other.

What on earth are you basing this claim on? Show me the computer that replicates a human mind, and explain to me how the various programs and subprograms and circuits and whatnot are 'perfectly simplified'.

[sourmìlk]

I suppose if bacterial replication is a significant part of the protein manufacturing process then bacteria don't really have many processes irrelevant to protein synthesis, and are, in fact, simple. So yay, I'm satisfied with that.

As for the latter one: let's say we go about simulating the human mind in a very inefficient way: by describing the rules governing how neurons react and interact, and then creating a virtual network of neurons that communicate with each other according to those rules. Unless it is necessary to simulate every molecule of the neuron to decide how it interacts with other neurons (and admittedly it might be, although the behaviour of neurons (from what I understand) is governed by specific patters and rules), the computer simulation is less complex than the brain itself. And the actual hardware that the simulation runs on is pretty simple. Electrical engineering isn't that complicated, relative to gaining a functionally equivalent understanding of molecular biology.

[Izawwlgood]

As for the latter one: let's say we go about simulating the human mind in a very inefficient way: by describing the rules governing how neurons react and interact, and then creating a virtual network of neurons that communicate with each other according to those rules.

I think this is vastly more complex than you realize. We are still in our infancy of our understanding of how the brain is wired. Saying 'just simulate a bunch of interconnected neurons' is on par with saying 'just simulate the universe down to the microsatellite', with our currently observational power.

And the actual hardware that the simulation runs on is pretty simple.

Out of curiosity sourmilk, how many connections do you think there are in the human brain?

[sourmìlk]

As for the latter one: let's say we go about simulating the human mind in a very inefficient way: by describing the rules governing how neurons react and interact, and then creating a virtual network of neurons that communicate with each other according to those rules.

I think this is vastly more complex than you realize. We are still in our infancy of our understanding of how the brain is wired. Saying 'just simulate a bunch of interconnected neurons' is on par with saying 'just simulate the universe down to the microsatellite', with our currently observational power.

I get that it can be absurdly complicated. My point is that, for an arbitrarily complicated system in reality, if the simulation doesn't have to solve the schrodinger equation for each particle, then it is less complicated than what the universe does.

And the actual hardware that the simulation runs on is pretty simple.

Out of curiosity sourmilk, how many connections do you think there are in the human brain?

I would have guessed a few trillion. Wikipedia says 100 - 500 trillion.

[Izawwlgood]

I would have guessed a few trillion. Wikipedia says 100 - 500 trillion.

Go ahead and simulate that. You should be aware that our ability to computationally simulate neuronal circuits is limited to, I dunno, 10-20 neurons, each forming at a maximum, perhaps 5 connections? Maybe someone with a stronger neuroscience background can chime in on that, that's just what I recall from a couple of our neurobio's computational neurology labs presentations. I might actually be overestimating the number of neurons we're capable of simulating...

[sourmìlk]

At the moment I recognize that we are not able to simulate a human brain.

[gmalivuk]

Unless it is necessary to simulate every molecule of the neuron to decide how it interacts with other neurons..., the computer simulation is less complex than the brain itself.

Sure, but only so long as you conveniently ignore the hardware requirements of said computer simulation.

[sourmìlk]

Considering that the behaviour of an entire processor is less complicated than that of a single protein (in the sense that we can perfectly predict how a processor will behave given certain instructions), I don't really think the hardware factors into it to any significant degree. 100 trillion of the most complex computers is still conceptually less complex than a single neuron.

[Izawwlgood]

100 trillion of the most complex computers is still conceptually less complex than a single neuron.

I think you just underlined why this;

Actually, I kind of think humans are also probably more complicated than the simplest possible mechanism to store, execute, and express human thought.

Is a grossly fantastical, unsupported position.

[sourmìlk]

Only if you think that the rules governing a neuron's behaviour cannot be expressed in anything less than the entirety of the neuron or a complete simulation of it.

[morriswalters]

I know zip about biology, but I have to ask. Isn't the behavior of a single neuron and everything else in the body an expression of the genome. That seems pretty efficient to me.

[sourmìlk]

Well yeah it's pretty efficient, but it's extremely complicated.

[The Mighty Thesaurus]

Considering that the behaviour of an entire processor is less complicated than that of a single protein (in the sense that we can perfectly predict how a processor will behave given certain instructions), I don't really think the hardware factors into it to any significant degree. 100 trillion of the most complex computers is still conceptually less complex than a single neuron.

Our ignorance says nothing about the complexity of the underlying mechanics.

[sourmìlk]

Doesn't it? They're so complex we haven't been able to figure them out so far. Whereas processors have been thoroughly understood for decades.

[morriswalters]

Worrying about complexity is a trap. A processor requires a whole infrastructure to support its manufacture and development. It is just not the chip, it's everything required to make it. You just substitute one type of complexity for another. Using bacteria to assist in manufacturing any biological strikes me as elegant. You get the process for free, the bacteria does the work for you.

[sourmìlk]

My understanding is that chips are simple (albeit not easy) to make. It's pretty much just lithography, right? Light lithography is an elegant process.

[Izawwlgood]

Sourmilk, you're not listening. We can understand a great deal about a single neurons behavior. We can understand a small amount about how neurons connect or how regions of the brain are associated with various behavior or brain function. You are suggesting that it is feasible to fully simulate a human brain. I am telling you that that is science fiction at this point. Science fiction on par with 'simulate the entire universe'.

I have given you the scale of why this is not feasible (500 TRILLION connections) and I have pointed out to you contradictions within your own understanding of this subject. This;

Only if you think that the rules governing a neuron's behaviour cannot be expressed in anything less than the entirety of the neuron or a complete simulation of it.

is a fine point to make, that yes, we can simulate a SINGLE neurons behavior, even perhaps, a handful of neurons behavior. I don't know how many connections are in a microprocessor, but I wager it's a lot smaller than the number of connections in a single brain. Couple this with the fact that neurons are singular in/out switches; they receive input from a number of neurons, and can output to a number of neurons (even themselves!).

I know zip about biology, but I have to ask. Isn't the behavior of a single neuron and everything else in the body an expression of the genome. That seems pretty efficient to me.

Sort of. A decent analogy, I suppose, is that an office building may be the expression of the blueprints that went into making it, but that's not a really good way to describe what sort of business happens on the fourth floor. But yes. Close enough.

And sourmilk, go ahead and make a chip. Do it now. Lets address how many 'points of failure' there are in this process.

[Ghostbear]

My understanding is that chips are simple (albeit not easy) to make. It's pretty much just lithography, right? Light lithography is an elegant process.

Haha. This is comedy gold right here. "Just lithography". Getting lithography to make device widths that are less than the wavelength of light being used is, by itself, beyond non-trivial. Then you have to deal with all of the different process styles -- gate first, gate last? High-K, Low-K? How much (and how?) are you doping the whole thing? What's your error checking process to find chips that got a defect? How much money did you spend making your fab have a solid enough foundation as to be unaffected by external vibrations? What about making the whole lithography process happen in a clean room, because if there's enough contaminants your whole process was just turned into a really expensive sand buffer?

What about actually designing the chip itself? Making all billion+ transistors work together to produce the end result consistently and without error? What about all of that error checking you need to build into the whole system because, well, errors are going to happen anyway (as one example -- we like to imagine the signals as being nice, clean, square waves, but they can actually be hideously noise ridden)? Oh and those billion transistors? If you fuck up just one of them, you can make your entire chip garbage. You'll also need to make the chip actually fast at it's job, instead of just really slow but technically a processor. And after all of that designing, you need to find ways to translate that design to something that can then be passed to the fabrication process -- and if you switch to a different process (same node but different company, different node, switching some gate properties, etc.) you now need to modify your design to make it work for that

Chips are an insanely complicated thing to get working in a practical sense, no matter how simple any of the individual steps are on a basic scale. It doesn't matter that boolean logic or basic lithography are simple, or that you can represent something with some high level programming language elegantly, when the entire foundation all of it is built upon to make things work is insanely complicated.

[sourmìlk]

Ghostbear, I think we may be confusing definitions of 'complicated' and 'easy'. I never said that making chips was easy, because I'm pretty damn sure it's not. But I don't think what you've said there makes the process particularly complicated in the sense that it doesn't add that many more steps. Certainly not anywhere near as many steps as an actual neuron takes to do one brain-equivalent-of-a-calculation.

You are suggesting that it is feasible to fully simulate a human brain.

No, no, I am absolutely not suggesting this at all. In any way. I'm saying that, for an arbitrarily complex system, if its simulation doesn't need to emulate every aspect of the system to mimic its behaviour, then the simulation is less complex than the base system, even if the simulation is still very complex. Extending that principle: if, for a brain full of neurons, its simulation doesn't need to simulate every single molecule at every single point as it happens inside the neuron to mimic its behaviour, then the simulation is less complex than the brain.

And sourmilk, go ahead and make a chip. Do it now. Lets address how many 'points of failure' there are in this process.

There are a lot. I don't think there are as many as there are proteins in a neuron.

[Izawwlgood]

There are a lot. I don't think there are as many as there are proteins in a neuron.

Proteins in a neuron is like comparing atoms in a transistor though; it's not the same as comparing 'points of failure' in the process of printing microprocessors.

Dude, with due respect, you're doing it again; you're moving the goal posts and redefining your original statement ad nauseum rather than just admitting you were wrong in your initial assertions.

[sourmìlk]

Okay, so if a single protein is off in a neuron, it won't affect the neuron's overall function? And also, how many single-base-pair mutations are there that impair or disable a neuron? Is this comparable in number to the number of points of failure in chip manufacturing? I'm pretty sure it isn't, though I'm mostly hazing a guess. Admittedly it is kind of difficult to describe biological and technological complexity in similar terms.

[morriswalters]

Is the point of this that it would be easier to simulate a bacteria then to use the the bacteria itself? He is correct in that a simulation doesn't have to simulate biological processes required for a biological organism like a neuron. But that's a moot point isn't it? The complexity you lose in the simulation doesn't go away, it just moves somewhere else. You have to do exactly those things that the organism does, you just do them differently, not to mention orders of magnitude slower. For instance a simulation of a neuron doesn't have to deal with the heat generated by the process, but entropy will require that energy to be somewhere. Or am I mistaken?

[sourmìlk]

We're actually talking about brains now. Assuming that it's the interneural activity that defines consciousness, it should be substantially similar to just simulate a neuron's behaviour based on a set of rules about how it behaves rather than by simulating all its internal goings-on.

[morriswalters]

Never the less you have to pay thermodynamics to play. The energy budget of the human body has been made efficient by evolution or so I'm told. While you could certainly lose the biological parts in a simulation then you have to trade something for dumping that efficiency, and in the case of a digital computer that would come down to power and its buddy heat Just look at a modern data center. Good luck.

[sourmìlk]

I'm not sure that's entirely right, that simplifying requires more energy. For example, if I have a circuit that involves 20 gates that's logically equivalent to one with 5 gates, I've thrown out a lot of gates but I end up using less energy.

[Ghostbear]

Ghostbear, I think we may be confusing definitions of 'complicated' and 'easy'. I never said that making chips was easy, because I'm pretty damn sure it's not. But I don't think what you've said there makes the process particularly complicated in the sense that it doesn't add that many more steps. Certainly not anywhere near as many steps as an actual neuron takes to do one brain-equivalent-of-a-calculation.

The process is complicated because you're not just making a boolean logic machine. You're making a boolean logic machine with error checking, heat dissipation, I/O protocols, internal microcode, it's own assembly language, etc. Then you're taking that 1,001 devices-in-one device and tying together a dozen other processes -- lithography, design rules, physical fab requirements (etc.) -- to make that device. Quite frequently, that process will fail -- you'll have a design error (such as the example I linked to) that blows the whole thing up, or you'll be dealing with the fact that 100% yields just plain don't exist for anything big enough (aka anything that is considered a "computer" part) or something else will go wrong. Look at how common computer crashes were back a decade or two ago -- those are the manifestations of how many points of failure we have to deal with. The only reason it appears better today is because we've made computers more able to smoothly handle those errors, not because we've eliminated them*.

You aren't making the simple device that you think it is, because there's a million other things that need to be handled to make it functional. What you're doing is looking at the high level view of it, which is why I highlighted that just because you can do something "simply" with a high level programming language or abstract it to a (relatively) simple logic system does not mean that it is a simple, elegant process. It just means that the programmer has less work to do.

* As an example of those -- Legend of Grimrock was causing frequent graphics driver crashes for me. Five years ago, that would have meant constant BSODs. Today it means that the screen goes blank for a moment, then I get a popup from windows telling me that the shit hit the fan but it already cleaned up the mess.

[Izawwlgood]

Okay, so if a single protein is off in a neuron, it won't affect the neuron's overall function? And also, how many single-base-pair mutations are there that impair or disable a neuron? Is this comparable in number to the number of points of failure in chip manufacturing? I'm pretty sure it isn't, though I'm mostly hazing a guess. Admittedly it is kind of difficult to describe biological and technological complexity in similar terms.

I mean, yes, small errors in protein structure CAN be catastrophic. There might also be thousands of 'errors' in protein structure right now all over your brain that have no affect whatsoever.

It really sounds like you have zero idea about the scales of things involved here. Tell me, how many proteins do you think are in a single cell? Remember! Neurons are probably the largest cell in the body. Do you think a single problem protein in an entire neuron is going to impair or disable it? The analogy here is, say, a single atom in a microprocessor being a heavier isotope.

Often it's quite easy to discuss biological and technological complexity in similar terms, actually. Your previous analogy;

There are a lot. I don't think there are as many [transistors in a microprocessor] as there are proteins in a neuron.

was an incorrect one to make if you were interested in 'points of failure' in a manufacturing or simulation process.

[sourmìlk]

Oh, I see the confusion. I was unclear in what I was referring to. Yeah, there are probably nowhere near as many transistors in a microprocessor as there are proteins in a neuron. I meant to say that there aren't as many points of failure in the manufacturing of a CPU as there are proteins in a neuron. Or even as there are fatal potential genetic mutations in a neuron.

[nitePhyyre]

Five years ago, that would have meant constant BSODs. Today it means that the screen goes blank for a moment, then I get a popup from windows telling me that the shit hit the fan but it already cleaned up the mess.

And how sweet it is.

[sourmìlk]

Five years ago, that would have meant constant BSODs. Today it means that the screen goes blank for a moment, then I get a popup from windows telling me that the shit hit the fan but it already cleaned up the mess.

And how sweet it is.

Yeah, until Windows keeps going black and then eventually BSODs on you (which is what was happening to one of my desktops). Was just an overheating problem though.

[Izawwlgood]

Oh, I see the confusion. I was unclear in what I was referring to. Yeah, there are probably nowhere near as many transistors in a microprocessor as there are proteins in a neuron. I meant to say that there aren't as many points of failure in the manufacturing of a CPU as there are proteins in a neuron. Or even as there are fatal potential genetic mutations in a neuron.

You're still not being clear; you think there aren't as many points of failure in the manufacturing of a CPU, as there are number of proteins in the neuron, or as there are in the MANUFACTURING of a protein in a neuron?

If the former, you think there are less points of failure from raw ore to finished microprocessor than there are from amino acid to protein? Because humans do it better than 4 billion years of evolution?

If the latter, that's why I told you the comparison is useless; it's comparing two entirely different things. Number of ways a process can fail vs number of things present in a given unit. You're also comparing the wrong scale of organization here; CPU:transistor as brain:neuron.

[Zamfir]

the sourmilk show is now over