Do you also find it odd that deadlocking a computer doesn't cause the cooling fans to stop and RAM refresh cycle to fail?
That is, what is the basis of "should"?
We have computer systems with a hardware watchdog. If the keep-alive heartbeat fails - like when the CPU deadlocks - it resets the computer. https://en.wikipedia.org/wiki/Watchdog_timer
Computers have multiple systems to help prevent total system shutdown. Why can't humans also have such?
The autonomic system doesnt have a global observer doing that job. No part of the brain has watchdogs to make sure it keeps doing what it's supposed to be doing. You instantly hit the homunculus problem assuming the brain had another brain watching it because then it would need another brain to watch that, ad infinitum.
That's all fine and good, but you didn't understand my argument.
I was objecting to your assertion that a "deadlock" would necessarily cause the autonomous nervous system to fail.
Since computers don't do that, why do you think that evolved systems - which have billions of years to build up robustness - must do that?
Your objection to the "homunculus problem" is not relevant. A watchdog does not need to be a "brain" in any but the most poetic sense. It can be switch on a timer, where the timer is reset by the system.
Computer can have deadlocks on many levels. I could write a multi-threaded program which deadlocks. Even though that's frozen, the rest of the computer could act like normal.
Since you didn't specify which sort of deadlock you were looking for, I pointed to ones which occur at the most recently evolved parts of the brain.
Shrug That wasn't the point. You said that if computers act like X then why don't humans act like X. I pointed out some cases where humans acted a bit like X.
You then asked why those cases weren't also Y. I pointed out that computers aren't always also Y, so why did you think that humans with X must also be Y?
I am not implying a global watchdog, for the simple reason that distributed computing systems can have local watchdogs which reset a node when the node is not active, even without global intervention.
The mechanisms can be different - animals swim while submarines don't, but both move underwater. My point was that I disagreed with your intuition and conclusion, and since I can find one counter-example to show that X doesn't imply Y, you can't conclude that X must imply Y. Nor can you conclude my counter-example is the only possible mechanism.
You still haven't described the basis for your intuition.
Distributed systems also deadlock in ways that local watchdogs can't account for. You still imply the global watchdog of the brain of the systems designer to navigate those problems.
All of your examples require a human homunculus to steer a system away from deadlocking.
The brain doesn't have an architectural homunculus to do that for it, and yet, it does not deadlock in any way resembling the deadlocks we see in computer science.
My question is how did the brain do this without a system architect? How can natural selection have a bias that prevents deadlocking in neural compositions?
While you haven't expressed the source of your intuition, based on your answers, you see the architecture of human-designed systems, which are engineered based on economic considerations, as being predictive of what natural systems are like.
This is not, generally speaking, true.
Even when built on standard computing architecture, we see that designs influenced by evolutionary design can result in "bizarre, mysterious, and unconventional" designs. I'm quoting from Thompson and Layzell's "Analysis of Unconventional Evolved Electronics" (1999). In that paper, note that the evolve system had a different response to temperature change than the "brittle failure" of normal digital circuits. System crashes like what you describe are "brittle".
Evolution, by human design senses, is incredibly wasteful. We are not going to build computers or distributed computing systems that way, hence any intuition based on those design principles are likely not applicable.
You asked "How can natural selection have a bias that prevents deadlocking in neural compositions?"
The "how" is easy - organisms with hard crashing deadlocks don't reproduce.
You want to know what that mechanism is, which is a different question.
If you really want to understand this, you'll have to talk to biologists who focus on this sort of topic, not a "hacking news" sites.
It's more like massive amount of reactive asynchronous communicating processes, where deadlocks cannot happen. Even if two processes accidentally enter a loop where they react by sending messages to each other, they still will be able to react to messages from other processes.
14 comments
[ 2.8 ms ] story [ 25.1 ms ] threadDeadlocking the brain should appear in the underlying autonomous nervous system, but it doesn't. That's very odd.
That is, what is the basis of "should"?
We have computer systems with a hardware watchdog. If the keep-alive heartbeat fails - like when the CPU deadlocks - it resets the computer. https://en.wikipedia.org/wiki/Watchdog_timer
Computers have multiple systems to help prevent total system shutdown. Why can't humans also have such?
I was objecting to your assertion that a "deadlock" would necessarily cause the autonomous nervous system to fail.
Since computers don't do that, why do you think that evolved systems - which have billions of years to build up robustness - must do that?
Your objection to the "homunculus problem" is not relevant. A watchdog does not need to be a "brain" in any but the most poetic sense. It can be switch on a timer, where the timer is reset by the system.
Computer can have deadlocks on many levels. I could write a multi-threaded program which deadlocks. Even though that's frozen, the rest of the computer could act like normal.
Since you didn't specify which sort of deadlock you were looking for, I pointed to ones which occur at the most recently evolved parts of the brain.
Find a global watchdog behaving in the manner you describe within any neural system for any animal at any period in the history of Earth's biosphere.
Find any "neural system resetting" anywhere, in fact.
You then asked why those cases weren't also Y. I pointed out that computers aren't always also Y, so why did you think that humans with X must also be Y?
I am not implying a global watchdog, for the simple reason that distributed computing systems can have local watchdogs which reset a node when the node is not active, even without global intervention.
The mechanisms can be different - animals swim while submarines don't, but both move underwater. My point was that I disagreed with your intuition and conclusion, and since I can find one counter-example to show that X doesn't imply Y, you can't conclude that X must imply Y. Nor can you conclude my counter-example is the only possible mechanism.
You still haven't described the basis for your intuition.
All of your examples require a human homunculus to steer a system away from deadlocking.
The brain doesn't have an architectural homunculus to do that for it, and yet, it does not deadlock in any way resembling the deadlocks we see in computer science.
My question is how did the brain do this without a system architect? How can natural selection have a bias that prevents deadlocking in neural compositions?
This is not, generally speaking, true.
Even when built on standard computing architecture, we see that designs influenced by evolutionary design can result in "bizarre, mysterious, and unconventional" designs. I'm quoting from Thompson and Layzell's "Analysis of Unconventional Evolved Electronics" (1999). In that paper, note that the evolve system had a different response to temperature change than the "brittle failure" of normal digital circuits. System crashes like what you describe are "brittle".
Evolution, by human design senses, is incredibly wasteful. We are not going to build computers or distributed computing systems that way, hence any intuition based on those design principles are likely not applicable.
You asked "How can natural selection have a bias that prevents deadlocking in neural compositions?"
The "how" is easy - organisms with hard crashing deadlocks don't reproduce.
You want to know what that mechanism is, which is a different question.
If you really want to understand this, you'll have to talk to biologists who focus on this sort of topic, not a "hacking news" sites.
Or maybe epilepsy?
-- Robot Santa on "Futurama"