Are processing speed and reaction times related?

[Tabaristiio]

Is there any relationship between processing speed and reaction time (speed at which something reacts to external stimuli) of information processors such as the brain and computers?

So if a computer had higher processing speed, would it be able to react quicker to external stimuli, For example, if one computer had a 1 terahertz processor and the other had a 1 gigahertz processor, would the 1 terahertz processor have the potential to react faster to external stimuli provided it was programmed to do so.

Suppose we have a robotic body which has a computer programmed to catch things and dodge things. If it has a higher processing speed like a 1 terahertz processor, would it be able to catch and dodge faster moving objects than a slower 1 gigahertz processor? If so, how much processing speed is required for that robot's computer to catch / dodge a bullet or something traveling at the speed of an average bullet?

Will that also mean that a autonomous car with high processor computers can travel at a much faster speed reacting and dealing with obstacles at a much faster speed to humans without causing accidents due to being able process the data fast enough.?

 

[DEvens]

Maybe. Maybe not.

Reaction time is not only a function of processing the response. It is also a function of transmission up the equivalent of nerve sensors. For example, in your own body, you have several reflexes. These are responses that do not require involvement of the brain. A signal impinges on a never ending. The signal goes as far as a nerve node, the nerve node "makes the decision", and the signal goes back down to the muscle. And your hand jerks away from the hot thing. This happens faster than could ever happen due to you getting the info "hot" and thinking "should I do something about it?" There are a variety of "tricks" of this sort available to applications such as autonomous cars.

In addition, a large part of reacting to driving conditions consists of projecting what is about to happen. My driving instructor trained me to look forward on the road and figure out what is going to happen in the next 20 seconds. That is because, at highway speeds, 20 seconds is a reasonable non-panic stopping time. Once you get "adequate" processing to project 20 seconds ahead, then more does not make a lot of difference.

In even more addition, a large part of driving is physical law. You simply cannot react any quicker than the vehicle supports and road conditions support. Stopping times and distances are constrained. Turning is constrained. If an obstacle appears (that "dear little dog" that next door has) then physics will limit the possible responses.

So the only possible difference that being faster could make is those few situations where reaction time is the difference. It means that following distances could be slightly shorter at any given speed, for example. But not much. And safe highway speeds are determined much more by road conditions and the typical condition of the typical vehicle. Your typical transport truck can't go around that corner any faster without tipping over just because the driver has micro-second response time. So speed limits would be unlikely to change much.

The main hope for autonomous cars is that they don't lose focus, their attention does not wander./they don't mess with the radio or make cell phone calls when they should be driving. And they don't get emotional if somebody cuts them off. And they don't drink and drive.

 

[pilofrogs]

Computers not only have to react, they have to analyse the stimulus. For you and me, the time it takes to recognize that a dog is in the road and the time it takes to push the brakes are in the same basic ballpark, in terms of the time they take. A computer is going to have a hard time recognizing the dog but an easy time pushing the brakes. The computer can go from "I need to push the brakes" to "pushing the brakes" in essentially zero time. How long it takes the computer to recognize that brake pushing might be required is going to vary hugely depending on the task.

If your ghz computer and your thz computer are competing in a race where the stimulus is really easy for a computer to detect, say, are the lights on? The thz computer won't have a huge advantage. It will probably win, but you won't be able to tell. On the other hand, if they're in a race to categorize paintings by artist, your thz computer is going to win every time by a fair margin.

 

[newjerseyrunner]

High reaction time computing is usually more about how information in processed, rather than the processing speed. We program systems that are security or safety critical to be able to drop large amounts of data in order to focus on more important data. Multitasking is usually done by constantly switching what the CPU is working on, and it's possible for the programmer to specify that a particular task is high priority, and the OS will starve other processes in order to get critical ones information that it needs.

For example, if your car detects that you are about to be in an accident, it will stop scanning the road for potholes, it'll stop checking on the GPS, it'll stop monitoring the temperature of your engine. It'll devote 100% of it's resources into figuring out where the passengers are and making sure the correct air bags deploy at the correct time. (Realistically these are actually different systems that don't have to divert resources, but that's the idea.)

Of course being able to process information faster makes reaction times better, but algorithms are more important. For example, you don't need image recognition. The car doesn't need to know that a dog is in it's path, it only needs to know that something is in it's path. If it has time to analyze what it is and attempt to predict it's future behavior, that's just gravy at this point in time.

The processing power of the human mind far exceeds that of any computers. Our reaction time, however, is quite slow.

 

[GTOM]

The processing power of the human mind far exceeds that of any computers. Our reaction time, however, is quite slow.

Is the slow chemical signalling is a main bottleneck of human reaction time?

 

[newjerseyrunner]

Is the slow chemical signalling is a main bottleneck of human reaction time?

First off, yes and no. Compared to a computer? Yes, chemical signaling is WAY slower, no biological life form will ever think as fast as a theoretical AI. Is it not possible to react faster because of it? No, there are advanced species that can react to things much faster than we can: cats for example.

There are probably many factors affecting reaction time of species:
Small prey would probably react extremely quickly, predators would eat slower ones: rabbits, flies, squirrels...
Large/armored/poisonous prey with minimal predation would probably react extremely slowly, they would have no evolutionary pressure to keep their minds fast: hippos, turtles, snakes,
Ambush predators probably have very high reaction time, they only have one shot and need to adapt to the prey's tactics in real time: cats, crocodiles, raptors...
Pack hunters will probably be a little slower, overwhelming numbers means that there is more room for mistake: dogs, humans, dolphins
Social creatures probably think slower too, it's beneficial to stop and think before reacting in social groups: chimps, merecats, elephants

 

[jack476]

That would depend more on architecture than anything else.

The 1 GHz CPU could, for instance, have better thread management and more interrupt vectors, where the hypothetical 1 THz CPU could just as well be limited to a single thread and have less sophisticated interrupt handling routines and instruction sets. And it might be implemented in such as a way as to be optimized for quickly responding. For instance, an ATMega128 microcontroller with a maximum CPU clock rate of 16 MHz (and usually scaled down significantly to save on battery) set up for a real-time application could easily respond to an interrupt far more quickly than a personal computer (which would have a CPU orders of magnitude faster in terms of clock rate) that would have to organize the response through the OS and respond by USB.

Likewise, if the task is image recognition, there are architecture considerations. For image processing tasks, you need hardware that is optimized for certain kinds of vector operations. That is, you need a GPU rather than a CPU. And as anyone who's ever suffered the unending Hell that is integrated Intel "graphics" can tell you, a GPU with a clock rated at a fraction of the frequency of your CPU will vastly outperform your CPU in terms of image processing tasks every single time.

And this is something that I always point out when the subject of CPU speeds gets brought up: architecture > thread management > cache > frequency.

As for how that translates to reaction time in terms of the human brain, they really don't compare with each other. The brain is a computer, but nothing at all like the kind of computer you have on your desk, and meaningful comparisons to concepts like clock frequency aren't really possible. Now, if your brain is working faster, such as with intensive focus and mindfulness training or with some kinds of stimulants, reaction time will improve, but that's still going to be limited by how slowly (compared to an electronic computer) a signal can be carried from your brain to your muscles.

 

[meBigGuy]

1. In the real world a faster processor would mean a programmer (or product manager, actually) would add features until the system was slow again

2. If you took something that sped up the chemical processes of nerve transmission and neuron communication then reaction times would be faster (in my opinion).

AND:

I object to calling the mind a computer. That word connotes things that do not apply.
I'll accept calling the mind an information processor, but the mind is not "computer-like".

There is a wide gap between what people think their brain is doing (and what they think they are experiencing) and what is actually going on in that mass of neural goo.
The relationship between reality and our sense of it is WAY more tenuous that most suspect.

We see/operate on a sophisticated model of reality in our brain that is delayed from external reality by sense and processing speeds and limited sensory data. There are paths that we are conscious of, and paths that we are not. The incredible high level neural processing that builds our comfortable view of reality from sparse sense data is mind boggling (so to speak).

This "model" theory of internal reality is excellently described by David Eagleton in a recent NPR show:
http://video.pbs.org/video/2365580655/

 

[Tabaristiio]

OK. Firstly, my goal was to find out what the requirements are for a fast reaction time. I intuitively thought that fast processing speed will lead to faster reaction times in any information processing systems, such as the brain and the computer. Due to the fact that with faster processing speeds, you can process more information per second. Which made me conclude that a brain or a computer would react quicker to a fast moving object if it has a faster processing system than a slower processing system due to being able to process the moving object with more detail and quicker. However, reading some of the responses, it seems to be a little more complex than that.

So I have a few follow up response questions in regards to reaction times:

1) What are some ways a human can truly quicken general reaction times (Visual reactions to all external stimuli) with very little to no harmful side effects?
- By reaction time, I mean mental speed only. Not counting physical movement speed. I.E. mentally processing the external stimuli without being overwhelmed by the speed of it. I mean to some animals, I believe 1 second feels like a longer time relative to us because they process more information in 1 second than we do in multiple seconds. Thus their internal clock is quicker due having a faster brain. So relatively speaking, suppose an alien processes information in 1 second which will take us 10 seconds to process the same information. That means they should be 10 times or so faster than us reaction time wise. A moving object that is too fast for us to see let alone react to, should be normal for that being. Or a moving object that is normal for us, should appear in slow motion to that alien. So given that I've specified my goal, is there a way to become like that alien or animal where I am able to perceive one second as much longer due to being able to process more information in a single second thus giving me quicker reaction time? If so, how? Some possible methods I've heard of are electrical stimulation into the brain, taking specific stimulants / supplements, adrenaline in fight / flight mode usually causes it etc. What other ways are there? If someone knows of any specific supplements that can be taken, please give me the names of it.

2) What would one have to do to improve reaction times of a computer if just increasing raw processing speed is not enough?
- Someone mentioned CPU architecture and GPU, can you please explain what you mean by them, especially CPU architecture? I just like to give examples in which the
computer would apply its reaction time. For instance, if I wanted to program a computer and insert it into a robot so that the robot would be able to dodge, block or parry the bullet, how would I program that computer? What would I have to do? This question ignore all other factors related to physics such as momentum for example. This is purely to with reaction time. Or for instance, suppose I wanted to program a computer so that it is able to drive a plane on autopilot through tight areas at 900 mph. Again, ignoring all other factors related physics, how would I program that computer so that it'll be able to process and react to obstacles and make constant changes in its movement? A final example would be a simpler one, if I wanted to program a computer so that it can keep track of an enemy aircraft in the air traveling at supersonic speeds, how would I program and build that computer? For example, say an enemy aircraft is traveling at supersonic speeds and is constantly changing directions in the air, which makes it extremely difficult for a human to keep track of that aircraft manually, missing the eluding target every time. This is where the computer comes in, it should be able to keep constant track of the enemy aircraft even whilst its moving and changing directions fast. The computer should be able to adjust in millionth of seconds to the aircraft changing directions, thus instantly keeping with the aircraft. How would I build such a computer system?

3) What is the difference between frames per seconds (FPS) in cameras and frequencies such as hertz, megahertz and gigahertz used in CPU's and is FPS more
important in fast reaction times than CPU frequency?
- So if I built a robot with a camera vision with an extremely an extremely high FPS, such as 1 billion, would that play a bigger role in how fast it reacts to external stimuli
than its CPU frequency?