Slashdot Mirror
Neuroscience Can't Explain How a Microprocessor Works (economist.com)
mspohr writes: The Economist has an interesting story about two neuroscientists/engineers -- Eric Jonas of the University of California, Berkeley, and Konrad Kording of Northwestern University, in Chicago -- who decided to test the methods of neuroscience using a 6502 processor. Their results are published in the PLOS Computational Biology journal. Neuroscientists explore how the brain works by looking at damaged brains and monitoring inputs and outputs to try to infer intermediate processing. They did the same with the 6502 processor which was used in early Atari, Apple and Commodore computers. What they discovered was that these methods were sorely lacking in that they often pointed in the wrong direction and missed important processing steps.
Anyone with even an elementary education in cognitive science will tell you that attempting to model thought processes is always done according to the dominant technology of the time in question. First it was machinery, then it was circuits, then it was computers.
This does not mean the model is accurate or even useful.
In order to understand the DNA of an Orange, we "scientists" dissected an alarm clock. This _proved_ that our methods of studing oranges, and fruit in general, have been wrong for centuries.
I know I'll catch hell for my religious beliefs, but...
I think that the 6502 was not the result of evolution, but rather it had a Creator and was the product of Intelligent Design. There are just so many subtle clues that suggest features that were deliberately put in there. Could natural selection really explain how it had two different indirect access modes, one that selects a direct index from an offset, and the other adds the offset to the index?
These researchers may be trying to apply the wrong methods to a device that is almost certainly the product of a higher power.
A dingo ate my sig...
We once had electronic circuits designed to perform calculations (example: Enigma). It seemed natural to try to model the brain as a complex electronic device.
We now routinely use silicon integrated circuits to perform calculations (example: the IBM PC-XT). It seems natural now to try to model the brain as a complex general computing device.
The take-away point I get from this is that we may need another revolutionary technology or two (fully three-dimensional integrated circuits? IC's based on carbon instead of silicon?) before we can model the sentient mind as similar to an artificially created device. Such advances may also be required before we can create (invent?) a true "artificial intelligence".
Neurons aren't digital processors. A set of connected neurons isn't either. Neuroscience already knows that it's really difficult to learn about the structure and function of the brain from the available tools. What was more interesting was that they were able to pick up anything. They found that the chip had a master clock, for example.
There are people already challenging the use of viewing the brains as a computer (signal ins and outs) in terms of really understanding how brains organize and function. So, given all this, it's not surprising that the methods didn't fare well. The neuroscientists already knew they had a very tough task, it's those in CS and AI that are assuming that understanding the brain is the same as understanding a collection of digital circuits.
It's not brain surgery.
...a 6502 is not a brain.
The issue is, the 6502 is several orders of magnitude less complex than a brain. It could be likened to a massively parallel computer that is running thousands of programs all at once. So it is completely reasonable, on the scale of the brain, to suggest that damage to an area in a dozen people that affects their hearing to draw the conclusion that that part of the brain is responsible for hearing. Damaging a couple transistors in a 6502, a single processor, is akin to damaging a few neurons in each of a million different areas of the brain at once. The researchers correctly determined that a particular damage which caused Donkey Kong not to boot did not mean the part of the CPU which is responsible for Donkey Kong was damaged. That does not, however, say anything about the methods used to study the brain. It is, unfortunately, yet another reason why researchers need to stop making silly comparisons between computers and brains, and also need to stop "playing for the crowd", which really was what this seems to have been.
It's like asking my boss to explain the technical details of what I do. Whenever he asks me to explain the details, I know it's going to be a really short conversation/meeting. About 3 sentences in, he waves his hands in the air and says "I don't need to know the details!"
Because he really shouldn't try to explain anything either.
They probably would've found the workings of 286 16-bit protected mode very similar to a human being who has spent the last 10 years high on opium drugs, and has just undergone a lobotomy.
"Neuroscience Can't Explain How a Microprocessor Works" is like saying "Herpetology Can't Explain How a Bicycle Works."
I think the biggest flaw in this paper is that perturbing an analog system is nothing like perturbing a digital system. To be clear, if the brain is anything comparable to a computer, it is a computer built from millions of parallel analog processors. Perturbing an analog system can be informative in ways that perturbing an digital system would not be -- analog systems can reveal half answers and shades of grey even when severely disrupted. A microprocessor will throw a fit if a single bit gets flipped unexpectedly.
As if anyone could doubt this: neuroscientists have developed methods for the system they actually study. A microprocessor would require a completely different set of methodologies to "study" a priori. If neuroscientists were studying a microprocessor, I guarantee you they would shift to methods that work for digital systems almost instantly.
I suspect the authors know this, so I can only conclude that they are more willing to troll the discipline than engage in serious work (i.e. work that is not proposing straw-man arguments).
Also SHOT FIRED: Maybe these authors should put their money where their mouth is! The paper literally asks "can one neuroscientist figure out one microprocessor"? Why would any serious conclusion depend on the work of one neuroscientist (with clearly biased motivation)? Where is the competition website that models a simple microprocessor to any desired perturbation, system output and wiring diagram included. After all, this is the information available to a modern neuroscientist. Put the reward at say, $500,000 minimum (grant equivalency), to figure out what is going on? Please?
I know. It's really not that difficult and does not take any math beyond simple logic.
Before we can answer the question, "Is there any reason for brains to look like each other if it was magic'd into existence by divine fiat?" it might be helpful to look into the question, "Do microprocessors look like each other?" Alternatively we can ponder the statement, "If brains were magic'd into existence by divine fiat, there is no reason for brains to either look or not look like others." However if information is self perpetuating, one of the forms it might take is the form of something that has been referred to as the divine, but most of the purported attributes of such a thing seem to be tautologies at best. God is good and perfect merely when he has been defined as good and perfect from the outset.
We don't know everything about how the brain works but we know a lot. How much is left to be known remains to be seen.
This is such a good idea.
Learning how something works by incrementally breaking it?
I'm sure glad no elected official uses that approach...
The microprocessor is the result of decades of research and this experiment is an effort to short-circuit (pun intentional) much of that research. What would be a more interesting experiment would be to start with neurological model of Boolean logic and then present it progressively more challenging problems to solve. It would be very interesting to see if those solutions follow the evolution of Von Neumann machines, Harvard architecture or something entirely different.
We've been replacing "transistor infrastructures" with ever smaller "transistor infrastructures". They will eventually find a way to make cost effective "memristor infrastructures" and then they will in short order be built... and then improvements will be made and the same cycles will be seen with memristors, and light based designs, and quantum designs... and as an outlier maybe even biological ones. But biologic-nonbiologic interfaces which may employ some of the above technologies will abound as well.
I'm not sure how much the brain has been viewed as a bunch of digital circuits, but what those who make the claim that it has been and therefore our understanding of the brain is flawed as a result, don't seem to know is that artificial analog circuits have been around longer than digital ones.
And that makes it a strawman, how? The general thrust of his argument is that the detractors of AI are so hung up on the concept of intelligence that they never address whether we will be able to artificially replicate the functions of the brain or how close we are, or what is in the works to get us closer, or anything in that area really.
Once we eliminate all the posturing around the concept of intelligence by changing the term to machine learning, all the arguments collapse.
The claim that complexity implies intelligent design is one thing.
The claim that a set of ancient writings of dubious origin gives us useful information about this intelligent designer, is quite another.
lets engineer human brains..oh, wait, you can't do that with humans or primates..or even in drosophilia are techniques are still primitive, optogenetics nonwithstanding
idiots: science is about the practical; you go with the experiment you can do
This article is really a rehash of a beautiful classic paper advocating for the field of systems biology, not that that takes away from the message, its just funny that there is nothing new under the sun:
Can a Biologist Fix a Radio?
http://math.arizona.edu/~jwatkins/canabiologistfixaradio.pdf
Microprocessors are system with a defined instruction set and storage.
Now let's talk about how well the instruction set is defined in the human mind--oops; I don't think we quite have that model yet.
It was fairly obvious that it wouldn't work.
But perhaps the real goal was to see if some parts which should work did work, or if others which weren't expected did.
This is why I don't get a nephrologist to fix my car.
It's true I tell you, feller at work's next door neighbour read it in the paper.
Haven't read the article yet but just the summary makes me think "brilliant!" what a fantastic way to check your own field's methods. I wonder if this has been applied to other fields. Clearly the other fields won't use cpus.
Viewing the brain as digital or analogue is a gross oversimplification, at different scales and in different areas it can look like either (or both).
For example neurons will commonly spike, if you were to plot the activity across the neuron you may see a period of no activity followed by a short period of high activity, then a return to no activity. The processes that cause a spike can be highly non linear and force the output into either a very high or very low state. In this respect you could say the neuron is digital, it's either firing or not and there is no real middle ground. However the various factors that determine whether a neuron will spike such as charge across the membrane, or different hormone levels, are continuous and in that respect you could say the actual computation
being done by the neuron is analogue.
In my opinion however the biggest problem with comparing the brain to any type of circuit is that (as far as I know) pretty much every circuit we use is deterministic. One thing we do know about the brain is that neurons are highly stochastic, and treating them as a deterministic system is doomed to failure. Luckily this is something that the computational neuroscience community at least is well aware of.
https://slashdot.org/submissio...
There a fundamental flaw.
Brain are extremely parallel and highly distributed processing units.
Some region are more specialised in some tasks, but as a whole, no part of the brain absolutely needs another part for the brain to keep working.
From that perspective, CPU are a small single function device. They either work, or not. It's hard to have a *half functionning" CPU (unless you very specifically manage to burn a peculier par of the silicon that isn't core to the functionning. I don't see how that would be possible on a 6502 - except maybe burning a part of the microcode that is seldom used. Maybe on modern processors it would be possible to burn some acceleration core while keeping the main function intact).
If they wanted to apply fault analysis to analyse computers, the best situation would be approximated by randomly pulling *daughter boards* and see whcih function go missing and/or cause the boot process to hang.
(e.g.: remove the graphics adapter. Computer still boots but produces no video output, thus correctly confirming that these daughter board was the CGA).
Or you could reason at the scale of a cluster, by remove nodes.
(But that won't be much interesting. In a cluster, usually most nodes are entirely interchangeable. It would be as much informative as applying the method to analyse sponges).
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
How do they work?
I do not want your cheap brainburning drugs. They are useless for work. And I am a working man today.
The take-away from this article is the debugger neuro-scientists use is buggy. The algorithms they use to examine brain structure and connections to tell them what does what - when applied to a simple known system (a test-problem) - fails to find those connections. This implies they need to work on their algorithms.
can't explain how a steam engine works. So?
Oh dear, do we really that stuff here these days?
That's because a processor isn't a brain, it's a false and misleading equivalency. Engineers understand perfectly well how processors function. Have I said lately that I hate astroturfing in the guise of science?
Computer people commenting on neuroscience like they're experts. Yikes. Move along nothing to see here other than a profound lack of knowledge a great ignorance.
That these researchers were able to obtain *any* information about the underlying hardware is remarkable. Models can never be completely right; the map will never, ever be the territory. Empirical adequacy is the best we can hope for. Looking at failure states to infer causal connections is exactly what I did as a sysadmin back in the day. These researchers are doing the same thing. It worked for me as a sysadmin, and it works in neuroscience as well, though with one caveat. Ethically, you can't just switch off a patient's cognitive apparatus at will and see how the patient responds; you have to wait until an automobile or industrial accident, combat wound, or disease does it for you. The map you get won't be perfect, but that doesn't invalidate the methodology. In this particular case, the researchers couldn't completely identify the target, but they did get at least one key subsystem, the clock, right. That is way cool...
The patterns were a mishmash of unrelated structures that were as misleading as they were illuminating.
This pretty much describes the state of every branch of science after a major influx of new data. Just look at the maps of the world produced after Europe became aware of North America. Early maps sometimes show California as an island; and it's not because the cartographer is stupid; he just put the data at his disposal together into what was at the time a plausible conjecture. And in fact the problem might not even have been that he was ignorant. He may have misinterpreted some of the (at that stage) imprecise data he had to work with.
New information confounds. The detection and resolution of conflicts in data is arguably what science is.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Microprocessor designers can't explain how the brain works, either.
yfbds
I think the real problem here is that the results are open to interpretation with nothing to compare against.
We only see a test of how these techniques worked. There is no competing set of techniques to compare against, so all conclusions about what this really means are incomplete. How SHOULD these techniques have fared? What techniques produce better results?
Drawing any conclusion from this is kind of silly, all we can really say is we would expect some results to be misleading. Yet, that is always the case and always to be expected.
But that logic only makes sense if microprocessors and brains were similar enough that comparable methods could be used to attempt to understand them. But that isn't true.
Yes, there's no guarantee that the methods that work on the brain also have to work on microprocessors. However, there's also no guarantee that they won't work in both cases. There are many methods/observations that are so generally useful that they apply to a huge range of problems. This is important because there's no guarantee that methods that work on one part of the brain also work on another part of the brain. Maybe the part of the brain responsible for facial recognition and the part of the brain that controls your muscles are as different as horses and tractors?
However, the real issue is that we don't even know which methods work on any part of the brain! As far as I can tell, the methods used to study the brain are basically just shots in the dark. So, we have chosen methods that we hope to be generally useful. Sure, they kind of make sense to apply to brains, but I know of no proof they are really useful. They certainly haven't explained how the brain works yet! Furthermore, before this paper was published, these general methods appeared to make about as much sense for microprocessors as they do on the brain.
So the point of this article is this: the methods we have to study the brain are just shots in the dark; we don't know if they work. So let's try another shot in the dark to see if these methods work on a known system. Since we are assuming that these methods are useful in a wide range of situations, the fact that the methods are total failures on a 6502 should indicate that the methods aren't as generally useful as we hoped.
We already have flash, which performs the same system function as memristors. I don't see any dramatic advantage coming from the incorporation of memristors in commercial CPUs.
Contribute to civilization: ari.aynrand.org/donate
is there a good way to figure out the architecture features of a simple CPU without documentation? and then can we apply THOSE methods on the brain?
They should've funded a brain-scanning gadget for Apple IIs.
https://hardware.slashdot.org/...
Yup, have never had experience coding for 6502. (Only from 8088/8086 up)
Just noticed now that it lacks multiplication/division instruction (and thus probably lacks microcode to do them as a series of addition/substraction and shifts).
Thank for correcting me.
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]