What does it mean to reverse engineer the brain?
PZ Myers takes Ray Kruzweil to task for claiming that we might be able to reverse engineer the “program” that our brain runs, and indeed make a computer-based “port” of it, sometime in the next 10 years. He doesn’t just refute Kruzweil’s claim, he seems to be using it as evidence for his argument that Kruzweil is a moron.
I can’t (and don’t particularly care) to comment on the intelligence of Kruzweil, I’ll leave the defense of his honor to the ready and willing. Instead I want to share an alternative interpretation of what Kruzweil meant. Or at least some food for thought about what seems to me a fascinating subject.
Myers makes a lot of great points about why it would be difficult (within 10 years) to simulate all the detailed biological interactions between cells in the brain. He also focuses a lot on a claim Kurzweil apparently made about the potentially small size of the computer program in question, based on the assertion that Kurzweil was only talking about “data” in the genome and that this is not sufficient for building a simulator of brain biology.
But when I read the Kurzweil quote, a simulator of brain biology is not at all what came to mind. Indeed, reading Myers’ argument had me scratching my head, since it took me a moment or two to realize what he was on about. You see, I read Kurzweil’s quote from the perspective of a software engineer, whereas Myers’ interpreted it as a biologist would see fit to do. This is a reasonable thing given that he is indeed a biologist. What seems less reasonable to me is how vigorously he attacks Kruzweil’s claim without giving the acknowledgement that he’s made an awful lot of potentially incorrect assumptions about what Kurzweil actually meant.
So what is my software engineers’ perspective?
I don’t see why reverse engineering the brain would require any ability to simulate protein and brain cell interactions. Doing so would be a brilliant achievement, but it seems orthogonal to the idea as I understood it. Myers says:
To simplify it so a computer science guy can get it, Kurzweil has everything completely wrong. The genome is not the program; it’s the data. The program is the ontogeny of the organism, which is an emergent property of interactions between the regulatory components of the genome and the environment, which uses that data to build species-specific properties of the organism. He doesn’t even comprehend the nature of the problem, and here he is pontificating on magic solutions completely free of facts and reason.
In my opinion this is not only wrong, but incredibly harsh.
It’s wrong because I think of biology itself as the programming language, and perhaps also the runtime. If a C++ developer were asked to reverse-engineer a program written in Pascal, they would not have to learn Pascal (or worse yet, reverse-engineer the compiler and runtime) in order to create a functionally identical C++ program. So I believe it could be with the biology and the functionality of the brain.
I understood Kurzweil’s statement to be about the creation of a computer program (in some existing, modern programming language) which implements at least a rudimentary version of the human brain’s algorithm. For example, its ability to continously observe input and recognize patterns, resulting in emergent properties that make up at least part of what we call consciousness. That’s why I think “reverse engineering” seems like an appropriate description, while “simulation” does not.
I won’t claim that this is what Kurzweil meant. He may have meant exactly what Myers claims he did. But I have the feeling Myers didn’t give this subject due consideration before lambasting Kurzweil over the supposed ridiculousness of his claim.
Bickering aside, what do you think of this subject? Can we reverse engineer intelligence/consciousness or at least some fundamental components? Will we have true “brain simulators” any time soon? I’m not remotely qualified to answer such questions, but I’m inclined to share Kurzweil’s optimizism and ambition about at least the former.
Comments are closed.
I agree with you. Its like a gravity model, where you don’t need a real object to reproduce the gravitational effects between objects (the sun and the earth, for instance).
Brandon,
If I understand your argument you’re saying reverse-engineering is creating a program that to the best of your knowledge creates the same results. The problem though isn’t that different from the one Myers mentions: we don’t really know what happens under what conditions. In his case, he says we can’t use the genome to create a brain because the genome is just a set of data that leads to different code paths. In your case you say we don’t need the genome or the exact ways functions of each cell and cell type use the genome, we just need to end up overall with the same results.
Think about what these results are: emotions, thoughts, problem-solving. None of those are objectively quantifiable. We don’t have an objective unit of measurement for sadness, it’s reported by each individual and they subjectively rate their sadness. Additionally, it’s difficult to experimentally create the situations necessary for even a single person to find patterns.
The advantage of what Myers is talking about is that it uses objective, repeatable measurements. Abstracting from the lowest level is of course important, we can’t model the brain on the atomic or sub-atomic level but the abstraction needs to be at a point where all subjective results are still being directly created by interaction with objective results.