Thursday, July 10, 2014

Big Brain Boondoggle?

Last year the European Commission launched a 2.7 billion dollar Human Brain Project. Its incredibly ambitious goal is to create a super-computer model of the human brain. But now some scientists are criticizing the project, saying it is doomed to failure. “The main apparent goal of building the capacity to construct a larger-scale simulation of the human brain is radically premature,” says expert Peter Dayan.

Such criticism must be a downer for Singularity enthusiasts, some of whom may have been hoping to cast aside this fleshy existence in the not-too-distant future, after a “mind upload” that transferred their consciousness to an endless virtual playground hosted by a supercomputer. But such criticism will not come as a surprise to those who have pondered how different the human brain is from any computer, and the severe difficulty (or perhaps impossibility) of creating computer functionality based on studying brain functionality.

Below are some of the main reasons why it may never be possible to create a supercomputer that in any very substantial sense mimics or models the human brain.

Reason #1: Neither the brain nor the mind is digital

All modern computers are digital, in the sense that they store software and data as a stream of digits, almost always binary digits such as 1010100010110000111010101. But neither the brain nor the mind works that way. There is no low-level part of the brain that functions as a Yes/No or “1/0” toggle. As for the mind, consider almost any concept in your mind, and ask yourself: is there any way I could digitize that? The answer is almost invariably: no. For example, could you possibly digitize your feelings towards your mother, putting them down as a stream of numbers? Of course not.

Given that neither the brain nor the mind is digital, what hope do we have of creating a computer that models the human brain?

Reason #2: Software is one of the two main components of a computer, but we don't know whether a brain uses something like software, and have no idea exactly where such software is if it exists.

A computer is built from two things: hardware and software. It is certainly possible that a computer might be built from some physical design resembling the physical arrangement of the human brain. But in order to have anything like a computer that really mimics the human brain, you would also have to have the software in the computer mimic the brain. However, there's a huge problem involved with that. First, we don't know whether the brain uses anything like software. Secondly, if such software exists in the brain, we don't know where it is located.

One can contrast this situation with what we know about the “software” of life. The software of life is the information stored in DNA. We know exactly where that software is stored: it is stored in the chromosomes of each of our cells. We also know exactly how the information in that software is represented, through sequences of nucleotide pairs.

But we don't have any similar information like this in regard to a software of the mind or a software or the brain. We don't even know whether such software exists. If it does exist, we don't where it exists. Is some software of the brain stored in chemical arrangements in the brain? Or is that software stored by electrical states in the brain? Or is the software stored by physical arrangements of neurons or their connections? We don't know, nor do we even know whether the brain or mind even works through anything like software.

Reason #3: Even if we were to discover some software of the human brain or mind, it would probably take centuries to unravel it.

Let us imagine that we did discover some software of the human brain. In all likelihood, it would be something that would take centuries to unravel. It would not be high-level easily readable lines of programming code. It would instead be some low-level thing that would be indecipherable for a long, long time.

Consider the software stored on your computer. When it is written, it exists on a programmer's computer in the form of relatively easy-to-read high-level instructions, such as “if (income < 1) give_warning_message()”. But when the software is distributed and ends up on your computer, it takes the form of low-level binary instructions such as 10001011001010011100100100. If we were ever to discover some software of the human brain or mind, it would in all likelihood be something as indecipherable as such a string of binary instructions, and it would be like gobbledygook that would take us centuries or more to figure out. Whether you believe that God or evolution is behind the marvels of the human brain, there is no particular reason to think that either one was interested in allowing us humans to be able to reverse-engineer the brain by easily reading its software.

But perhaps I'm wrong about such an assumption. Perhaps God specifically designed the human brain in a way that will make it easy for us humans to reverse-engineer it. Perhaps this is because God wants us to easily find a way to upload our minds into robot bodies and computers. Perhaps mind uploads are the next step in the divine master plan for cosmic evolution. I have my doubts about this notion, but maybe some spiritually-minded geek can develop such an idea into the next silicon-scented religion. 

computer religion
 Digitally divine: a high priest of holy heuristics
  
Postscript: The day after I wrote this, the New York Times published an opinion piece making some of the same points, particularly the suggestion that neuroscience won't be ready for the huge breakthroughs that are hoped for until there is a "Rosetta stone" kind of discovery similar to the discovery of DNA.  

No comments:

Post a Comment