Biologist and skeptic PZ Myers has ripped into Ray Kurzweil for his recent claim that the human brain will be completely modeled by 2020 (Note: Not that it's particularly important, but Kurzweil did say it'll take two decades at the recent Singularity Summit, not one). In a rather sweeping and insulting article titled, "Ray Kurzweil does not understand the brain," Myers takes the position that the genome cannot possibly serve as an effective blueprint in our efforts to reverse engineer the human brain. In regards to he claim that the design of the brain is in the genome, he writes,
Kurzweil knows nothing about how the brain works. It's [sic] design is not encoded in the genome: what's in the genome is a collection of molecular tools wrapped up in bits of conditional logic, the regulatory part of the genome, that makes cells responsive to interactions with a complex environment. The brain unfolds during development, by means of essential cell:cell interactions, of which we understand only a tiny fraction. The end result is a brain that is much, much more than simply the sum of the nucleotides that encode a few thousand proteins. He has to simulate all of development from his codebase in order to generate a brain simulator, and he isn't even aware of the magnitude of that problem.Myers continues:
We cannot derive the brain from the protein sequences underlying it; the sequences are insufficient, as well, because the nature of their expression is dependent on the environment and the history of a few hundred billion cells, each plugging along interdependently. We haven't even solved the sequence-to-protein-folding problem, which is an essential first step to executing Kurzweil's clueless algorithm. And we have absolutely no way to calculate in principle all the possible interactions and functions of a single protein with the tens of thousands of other proteins in the cell!
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.Okay, while I agree that Kurzweil's timeline is ridiculously optimistic (I'm thinking we'll achieve a modeled human brain sometime between 2075 and 2100), Myers's claim that Kurzweil "knows nothing" about the brain is as incorrect as it is disingenuous. Say what you will about Kurzweil, but the man does his homework. While I wouldn't make the claim that he does seminal work in the neurosciences, I will say that his efforts at describing the brain along computationally functionalist terms is important. The way he has described the brain's redundancy and massively repeating arrays is as fascinating as it is revealing.
Moreover, Myers's claim that the human genome cannot inform our efforts at reverse engineering the brain is equally unfair and ridiculous. While I agree that the genome is not the brain, it undeniably contains the information required to construct a brain from scratch. This is irrefutable and Myers can stamp his feet in protest all he wants. We may be unable to properly read this data as yet, or even execute the exact programming required to set the process in motion, but that doesn't mean the problem is intractable. It's still early days. In addition, we have an existing model, the brain, to constantly juxtapose against the data embedded in our DNA (e.g. cognitive mapping).
Again, it just seems excruciatingly intuitive and obvious to think that our best efforts at emulating an entire brain will be informed to a considerable extent by pre-existing data, namely our own DNA and its millions upon millions of years of evolutionary success.
Oh, and Myers: Let's lose the ad hominem.
7 comments:
Kurzweil may be religious, but that doesn't mean he's completely and absolutely wrong.
And I suspect the date error is because it hasn't sunk in how far we are in the future yet.
While I agree that the genome is not the brain, it undeniably contains the information required to construct a brain from scratch.
I think Myers would counter, arguing the environment containing the brain also contains some amount of the information necessary to construct a brain from scratch. Without a complex environment populated by other life forms (a life mesh) a genome may well not be able to produce a functional brain. For instance, we already know that mitochondrial DNA and RNA play a larger role in genetics than was initially expected. This distinction about development and interaction w/ the environment may end up being trivial as we reverse engineer the brain, but it also may end up being totally critical as we may find the environment serves as fundamental, more-or-less intelligent scaffolding for the creation of a functional brain.
Kurzweil himself has recently opened the door to the effects of development, even Evo Devo, on the brain, with comments like "But even a perfect simulation of the human brain or cortex won’t do anything unless it is infused with knowledge and trained" - link here - http://bit.ly/9yXCye
That said, I wholeheartedly agree with your main point that the genome will very probably help us to reverse-engineer the brain.
Overall, it's fun to watch the debate about brain emulation/simulation move from the IF realm (whether or not it'd ever be possible to create a brain) over to the WHAT IF or HOW realm (not like this, it'd have to be more like this). Thanks for being a leading voice on this!
Hey Alvis: I don't disagree; I believe that evo devo, epigenics and a myriad number of other external factors are responsible for the development of a fully functional and healthy brain. And I'm not suggesting that the first fully emulated brain is something we will create from scratch. It'll likely be something akin to what Venter did with artificial life. We'll riff off an existing biological precedent. This first 'emulated brain' will have to be fostered in a suitable environment (whether real or virtual) so that it can properly develop. As for second and third generation brain emulation, that's still an open question and will necessarily involve headier concepts like uploading/mind transfer.
This first 'emulated brain' will have to be fostered in a suitable environment (whether real or virtual) so that it can properly develop.
True dat! I do wonder though what is / will be the minimum suitable environment for brain dev emulation? To what extent will a brain dev simulation need to mimic our biosphere/memosphere/noosphere? Will it require a simulated body, spinal cord, nerves, input arrays? Or not - the brain has been shown to be plastic in certain circumstances. Obv, the greater the dependence on the environment, the more complex and robust the entire brain simulation mechanism will need to be, which adds years and pushes us closer to your predictions and much further from Ray's.
Also, will it be possible to accelerate the learning of the first simulated brains or might such efforts be slowed by the necessity of developmental windows? Acceleration of artificial brain learning seems much more likely (during the initial stages) if it can all happen in sim space. But will that be the path?
It'd be very interesting to poll AI, tech and acceleration theorists on what they see as the critical bottlenecks to whole brain emulation, just to have a list that could then be checked off as breakthroughs occur and as we move forward.
Alvis, you beat me to the punch. PZ's argument was that the claims that 1) 800 MB of code will be sufficient to emulate a brain, and 2) we will accomplish this in 10 years, are patently ridiculous, and they are.
One problem that PZ points out is that we can't even do protein folding correctly yet, and that's just the first level of biological organization where information outside of the genome matters, because the genetic code alone doesn't specify the chemistry of amino acids that leads to accurate folding. Some proteins require chaperones to overcome energy barriers and achieve their native state, and the identification of such chaperones can only be done empirically. Identifying the sequences of chaperones in the genome won't tell us which chemical reactions they are involved in.
And again, protein folding is just the lowest level of biological organization. Then there are protein-protein interactions, other chemical reactions, cell-cell interactions, tissues, organs, systems, and the interaction of everything with the environment.
Have you heard of Gini, the girl who was chained in a bathroom from age 3 until 13? She never achieved full cognitive function. Neither genes, proteins, nor neurons alone build normally functioning brains. The environment is a huge part of it.
I suppose I misread Kurzweil's intent, but my interpretation was not that he thinks someone will copy 25MM lines of code from the genome into a computer and produce a brain, but that an equivalent amount of coding should be necessary to create a synthetic brain.
For example, linux, Microsoft Windows, and Apple's OS are all of equivalent size and produce equivalent complexity, but they are not copies of each other.
Using the genome code itself and simulating such things as protein folding would require a brain simulation at the atomic level. Blue Brain is working at the neuron level now, and is still many times slower than real time. An atomic level simulation might be possible in the 2100s I suppose.
However, I don't think it's necessary to simulate at that level to produce an artificial brain (although it might be needed for mind uploading, should that ever happen).
Finally, Kurzweil's claims are in line with Prof. Henry Markram of the Blue Brain Project who has said we could see an artificial, conscious brain in the 2020s with appropriate funding. It seems likely that the funding will happen considering the payoff to the financial backers.
Singularity Hub has an article up today quoting Venter from Der Spiegel as saying that "We have learned nothing from the genome" after 10 years.
This shouldn't come as a surprise when you realize that the genome is nothing more than a parts list and merely the beginning of the complex interactions (for which we need extra information about) that produce humans.
It's also why DTC genetic testing is useless at best and a scam at worst.
doubtertom: I think Kurzweil's meaning is clear. He took the number of bits needed to describe the human genome, cut it in half to describe the 50% of genes expressed in the brain, and said an equivalent amount of information in lines of code could reproduce the brain. He didn't mention that any information outside the genome would need to be reproduced in the code to make a functional brain. He basically said that the information contained in the genome is sufficient.
That's absurdly reductionist.
Post a Comment