Parallel computing, AIs and the cloud

In summary, the setting strongly require rather unimpressive processors and not high enough demand to create multiple processors. On the other hand there would be clearly a big need for computation power. The solution that I can think of is going simple. In case of devices requiring little computation power using single processor and using the same processors in hundreds for parallel computing for devices requiring huge computation power (AIs).Both problems and workarounds are possible. The high suicide rate among coders forced to write programs for parallel computing could be due to the huge amount of pressure and
  • #1
Czcibor
288
132
The setting strongly require rather unimpressive processors and not high enough demand to create multiple processors. On the other hand there would be clearly a big need for computation power. The solution that I can think of is going simple. In case of devices requiring little computation power using single processor and using the same processors in hundreds for parallel computing for devices requiring huge computation power (AIs).

Which in story consequences do you see? Both problems and workarounds.

General:
-High suicide rate among coders forced to write programs for parallel computing? ;)

AIs:
-I for sure understand the problem of critical path, but not sure about direct consequences, visible for an user. The real life consequence that I see is poor reaction time of AI in case of accidents and sudden attack in combat.
-If there is insane number of processors, it may be possible to disperse them within the robot a bit. In combat it would mean that robot would not be vulnerable to any equivalent of "head shot" with relation to CPU. (sure, engine, fuel tank, etc. would still present a vulnerability)

Partial workarounds:
-I would expect an AI, to start a process calculating the best way of destroying a potential target, just to have the whole process aborted to prevent a friendly fire. I mean that system would have to waste computation power just to analyse potential scenarios to be able react on time.
-overlocking a few processors dealing with the critical path, maybe even change them from time to time to prevent overheating

Cloud computing:
-having so huge computation power but in tiny chunks there would be a temptation to run hundreds of tiny applications in the same time
-computer games could easily have nice graphics and quite developed multi agent system

Any interesting limitations that you see? Or maybe interesting consequences?
 
Physics news on Phys.org
  • #2
Czcibor said:
The setting ...
What setting? This sounds like jumping into the middle of a conversation without having heard any of the first part.
 
  • #3
phinds said:
What setting? This sounds like jumping into the middle of a conversation without having heard any of the first part.

The OP has a habit of presuming people have read their other threads:
https://www.physicsforums.com/threads/future-political-system.818290/
https://www.physicsforums.com/threa...nt-units-for-a-tidal-locked-exoplanet.826060/
https://www.physicsforums.com/threa...efence-for-a-loosely-populated-planet.823855/
https://www.physicsforums.com/threads/consequences-of-denser-atmosphere-on-earth-like-planet.800041/

TL;DR the setting is a tidaly locked planet with a small population of humans (around 10 million) that are cut off from Earth and have 21st century level technology (yet somehow are exceedingly rich despite the low population and lack of trading partners). Their world can be accessed by portals and at some point they're invaded.

Czcibor what do you mean by "AI"? Do you mean Strong AI, something as generally intelligent as a human being? If so it's going to be hugely expensive given your requirement for unimpressive technology using one type of processor. The human brain project estimates it will require an exascale computer to model a brain at the neural level. Assuming that's all we need (and the jury is still out) to create a human level intelligence then you would require 30x more processing power than the world's current most powerful super computer: Tianhe-2 . Some googling reveals that the footprint for this system is 720 square meters so you'd need 21,600 square meters just to have enough computing power to get to the current estimate of the minimum for a human level digital brain. Probably more because your colony doesn't produce good computing technology at all.
 
  • #4
Ryan_m_b said:
The OP has a habit of presuming people have read their other threads:
Jeez, and I thought *I* was arrogant. Czcibor, that's a very bad habit. You should do the opposite and assume that people have NOT read your other posts. At the very least, provide a link if you are continuing something from another thread.
 
  • #5
AI is a bit confusing, since it focuses on battlerobots.
(Otherwise i glad to read that estimations about human level AI, so it looks like i am not terribly conservative to estimate a big server room for it in my setting.)

Well, parallel processing and cloud computing, I'm not the best programmer, but i try to give as many accurate info as possible.

So multi-thread programs should be syncronized if they have access to the same resources (memory, hard drive, control system etc).

The bottlenecks for autonomous combatbots are the following : fuzzy image processing, to determine something is a target or not. Multiple processors can take different parts of the image, but i guess a decision making will still take significant time with lower grade processors.
(Well, soldiers in Vietnam attacked monkeys multiple times, if tech level isn't that high, i guess that robots will make even more mistake.)
Analyze combat situation : whether to charge, flank, suppress fire, retreat.
Industry and material for semiconductors, sensors etc.

Cloud computing and remote control, in this case they don't have to rely on own intelligence, but take the experience of their operators, huge servers and databanks with all the uploaded experiences of operators and robots.
The bottlenecks are bandwidth (with encode and decode information), and enemy countermeasures like anti radiation missiles, jammers etc. (Possibly delay time of light if it has to go through on many relays.) (And in PA setting, possibly the server park itself...)
Shorter wavelengths scatter less in vacuum, but more in air, and they can be reflected from dust, water, so enemy with right technology can detect it, and after detecting enough signals, estimate coordinates and send a missile to track them.

No headshot with scattered processors, enemy heat seekers will target the engine. Or if the robot exposes a camera and a gun to see and fire, they will target that part, and if it hit, the robot won't be able to fight anymore, unless it has multiple of theese things, but still, the enemy learned its position, they can send a grenade, or a missile there.
 
  • #6
Ryan_m_b said:
The OP has a habit of presuming people have read their other threads:
https://www.physicsforums.com/threads/future-political-system.818290/
https://www.physicsforums.com/threa...nt-units-for-a-tidal-locked-exoplanet.826060/
https://www.physicsforums.com/threa...efence-for-a-loosely-populated-planet.823855/
https://www.physicsforums.com/threads/consequences-of-denser-atmosphere-on-earth-like-planet.800041/

TL;DR the setting is a tidaly locked planet with a small population of humans (around 10 million) that are cut off from Earth and have 21st century level technology (yet somehow are exceedingly rich despite the low population and lack of trading partners). Their world can be accessed by portals and at some point they're invaded.

Czcibor what do you mean by "AI"? Do you mean Strong AI, something as generally intelligent as a human being? If so it's going to be hugely expensive given your requirement for unimpressive technology using one type of processor. The human brain project estimates it will require an exascale computer to model a brain at the neural level. Assuming that's all we need (and the jury is still out) to create a human level intelligence then you would require 30x more processing power than the world's current most powerful super computer: Tianhe-2 . Some googling reveals that the footprint for this system is 720 square meters so you'd need 21,600 square meters just to have enough computing power to get to the current estimate of the minimum for a human level digital brain. Probably more because your colony doesn't produce good computing technology at all.

No, I thought about something a bit better than google car. Pattern recognition, avoiding obvious obstacles on the way. In combat - shooting to everything that moves and fails FoF signal.

GTOM said:
AI is a bit confusing, since it focuses on battlerobots.
(Otherwise i glad to read that estimations about human level AI, so it looks like i am not terribly conservative to estimate a big server room for it in my setting.)

Well, parallel processing and cloud computing, I'm not the best programmer, but i try to give as many accurate info as possible.

So multi-thread programs should be syncronized if they have access to the same resources (memory, hard drive, control system etc).

The bottlenecks for autonomous combatbots are the following : fuzzy image processing, to determine something is a target or not. Multiple processors can take different parts of the image, but i guess a decision making will still take significant time with lower grade processors.
(Well, soldiers in Vietnam attacked monkeys multiple times, if tech level isn't that high, i guess that robots will make even more mistake.)
Good idea about such system shooting all medium size animals on the way.
Analyze combat situation : whether to charge, flank, suppress fire, retreat.
Industry and material for semiconductors, sensors etc. Cloud computing and remote control, in this case they don't have to rely on own intelligence, but take the experience of their operators, huge servers and databanks with all the uploaded experiences of operators and robots.
The bottlenecks are bandwidth (with encode and decode information), and enemy countermeasures like anti radiation missiles, jammers etc. (Possibly delay time of light if it has to go through on many relays.) (And in PA setting, possibly the server park itself...)
Shorter wavelengths scatter less in vacuum, but more in air, and they can be reflected from dust, water, so enemy with right technology can detect it, and after detecting enough signals, estimate coordinates and send a missile to track them.
I mostly thought about a system that in case of loosing connection with HQ follows some simple, pre-programmed strategy. Otherwise operators may try something more advanced
However, I have mixed feelings about jamming communication. Frequency hopping is not a rocket science. But targeting signal - sure. I thought about using some cheap transmitters as the loudest part of any robot team, to draw away attention from more valuable targets.

No headshot with scattered processors, enemy heat seekers will target the engine. Or if the robot exposes a camera and a gun to see and fire, they will target that part, and if it hit, the robot won't be able to fight anymore, unless it has multiple of theese things, but still, the enemy learned its position, they can send a grenade, or a missile there.
I thought about multitude of cheap cameras and combining the picture. But sure, next serious missile would make the robot destroyed anyway.
 
  • #7
Combining the picture, i think that requires quite a computing force, i don't have the artice (and it was hungarian) but it talked about making lots of aerial photos and combine them to a super resolution image... with months or at least weeks of computing.

They already have HARM missiles that tracks radars (with passive radar and an active mm wavelength radar, well a radar dish has quite a radar cross-secrion)
If they use laser comm, i guess it will be IR, since shorter wavelengths swallowed and scatter even more, so a heat seeker can detect it, although if they turn it off in time, the missile will be less able to follow it. (I don't know whether frequency agile technology also works so fine with IR or no?)
Well, it is also good to confuse anti-rad missiles with having multiple sources in a formation and random turn on/offs according to Heinlein's Moon is a harsh mistress; and that one :
https://books.google.hu/books?id=ANEM6nI3tosC&pg=PA196&lpg=PA196&hl=hu#v=onepage&q&f=falseJamming, Iran did quite a fine job with jamming that spy drone, i think the russians or chinese didnt sell their best equipment to them.

Shooting medium sized animals, and possibly empty soldier helmets, and uniforms if they move it with a rope. :P
 
Last edited:
  • #8
Czcibor said:
General:
-High suicide rate among coders forced to write programs for parallel computing? ;)
I write parallel code all the time, it's not hard.
 
  • #9
GTOM said:
Combining the picture, i think that requires quite a computing force, i don't have the artice (and it was hungarian) but it talked about making lots of aerial photos and combine them to a super resolution image... with months or at least weeks of computing.
So just use redundancy and don't try to combine it?

They already have HARM missiles that tracks radars (with passive radar and an active mm wavelength radar, well a radar dish has quite a radar cross-secrion)
If they use laser comm, i guess it will be IR, since shorter wavelengths swallowed and scatter even more, so a heat seeker can detect it, although if they turn it off in time, the missile will be less able to follow it. (I don't know whether frequency agile technology also works so fine with IR or no?)
I think that laser don't give significant heat signature.
The main limitation of laser communication is range in atmosphere - a few kilometres. Under such denser one getting more than one km could be an ambitious aim.

Shooting medium sized animals, and possibly empty soldier helmets, and uniforms if they move it with a rope. :P

Luckily the benchmark hit probability is not so high:
"US forces have fired so many bullets in Iraq and Afghanistan - an estimated 250,000 for every insurgent killed - that American ammunition-makers cannot keep up with demand." :D
http://www.belfasttelegraph.co.uk/n...e-250000-for-every-rebel-killed-28580666.html
 
  • #10
I think most of the bullets were fired for training and suppress fire, well, professional snipers don't waste ammunition much.
IMHO an IR laser will create a heat sign if it hits the missile's detector.

Redundancy, we are fine with parity, probably the robots should also have two main computers, and two independent high-res cameras and guns.
(More probably means that too much things get destroyed if they hit the engine, or damage the outer Faraday cage and send EMP.)
 
  • #11
http://www.bbc.com/news/science-environment-33867941

Interesting, altough i'd like to know, how it actually learns? I could simply write a program, that iterates through a set of possibilities, till it finds the best solution.
The problem, that it won't be effective, if possibilities become polinomial or exponential.

But theese things still give me a headache, what can be the role of humans, that they can't solve with electronic slaves, in a not so near future? :(
 
  • #12
GTOM said:
I think most of the bullets were fired for training and suppress fire, well, professional snipers don't waste ammunition much.
IMHO an IR laser will create a heat sign if it hits the missile's detector.
From that what my friend who served in Iraq told me - they used spray and pray heavily.

Redundancy, we are fine with parity, probably the robots should also have two main computers, and two independent high-res cameras and guns.
(More probably means that too much things get destroyed if they hit the engine, or damage the outer Faraday cage and send EMP.)
Maybe not think "keep a duplicate" but "have excess computing power unless severely damaged"?

GTOM said:
http://www.bbc.com/news/science-environment-33867941

Interesting, altough i'd like to know, how it actually learns? I could simply write a program, that iterates through a set of possibilities, till it finds the best solution.
The problem, that it won't be effective, if possibilities become polinomial or exponential.

But theese things still give me a headache, what can be the role of humans, that they can't solve with electronic slaves, in a not so near future? :(
Engine from full throttle to null, rudder, elevators, communicating or staying silent, launching rocket or not... After a few seconds the number of combinations would choke supercomputers.

Other approach. Modules responsible for processing inputs. Module responsible for keeping tab on already detected objects. Module for keeping tabs on known friendly units nearby. General tactics module. General flight module with a few variants, like nap-of-the-earth. Plus a few scripts, like when to use a damaged aircraft as kamikaze weapon.

The funny part starts with communication. Such devices could with coordination beat human reaction time. For example suddenly ordering 100 drones to divide known targets and make a salvo.
 
  • #13
Czcibor rather than just making things up why not look to drones in existence? I can't see anything you've posted here that modern machines couldn't do, for example:
https://en.wikipedia.org/wiki/Guardium

Whilst it might be difficult to get proper information on the technology involved looking into these sorts of examples will give you an idea of what's possible with modern technology. From there you can reasonably speculate a bit given that your setting has less advanced computers but better programming technology.
 
  • #14
It is an interesting question, whether real world tactics and strategy could be more similar to chess. or GO?
(Later is still human dominated)
Difficulty of environment ranges from space to urban warfare.
(I guess the first one barely requires human operators, if any... but be a bit epic :P)
 
  • #15
phinds said:
What setting? This sounds like jumping into the middle of a conversation without having heard any of the first part

I've always assumed the Czcibor translates his posts into English from another language (Czech I think--but I could be wrong). The abrupt nature of his sentences may be the grammar getting lost in translation.
 
  • #16
Ryan_m_b said:
Czcibor rather than just making things up why not look to drones in existence? I can't see anything you've posted here that modern machines couldn't do, for example:
https://en.wikipedia.org/wiki/Guardium

Whilst it might be difficult to get proper information on the technology involved looking into these sorts of examples will give you an idea of what's possible with modern technology. From there you can reasonably speculate a bit given that your setting has less advanced computers but better programming technology.
I've already made such basic research, my question was mostly concerning a speculative AI part.

GTOM said:
It is an interesting question, whether real world tactics and strategy could be more similar to chess. or GO?
(Later is still human dominated)
Difficulty of environment ranges from space to urban warfare.
(I guess the first one barely requires human operators, if any... but be a bit epic :P)
Problem with comparing it to any classical game:
- fog of war;
- quite often unclear aims;
- at best you deal with probabilities, while in chess and go you deal with determinism.
 
  • #17
GTOM said:
It is an interesting question, whether real world tactics and strategy could be more similar to chess. or GO?
(Later is still human dominated)

I'm sure the answer is neither, beyond some very basic fundamentals. Real life conflicts feature a wealth of concerns that aren't represented in the mechanics of games, even quite advanced video games. In virtually every game you have things like: direct line of sight, perfect communication with units, no chance that units will disobey orders, perfect uniformity amongst units, few logistics to consider (small exception to video games with resource mechanics), no non-military considerations such as political/social consequences etcetera etcetera.

GTOM said:
Difficulty of environment ranges from space to urban warfare.
(I guess the first one barely requires human operators, if any... but be a bit epic :P)

Would it be epic? I imagine most forms of space warfare would just consist of trying to dodge incoming fire from incredibly distant enemies until you are destroyed in the blink of an eye, or they are, or both.
 
  • #18
Czcibor said:
I've already made such basic research, my question was mostly concerning a speculative AI part.

And what did it tell you? Its very unclear what exactly you want. You ask for speculation but provide basically no information for us to go on like AI capabilities, theatre, technology level etcetera.
 
  • #19
Khatti said:
I've always assumed the Czcibor translates his posts into English from another language (Czech I think--but I could be wrong). The abrupt nature of his sentences may be the grammar getting lost in translation.
Quite close, I'm a Polish native speaker.

Theoretically I write my posts directly in English. But its quite possible that my syntax of English is influenced by my Polish.
 
  • #20
Czcibor said:
Quite close, I'm a Polish native speaker.

Theoretically I write my posts directly in English. But its quite possible that my syntax of English is influenced by my Polish.
My comment had nothing to do with your English, or your syntax, or your thread content. I objected, and continue to object, to your assumption that you can just jump right into the middle of a discussion under the assumption that people who happen across one of your threads will have read your other threads. I think that's a rude assumption. If you are going to continue or extend a discussion that started in another thread REFERENCE the thread!
 
  • #21
Ryan_m_b said:
I'm sure the answer is neither, beyond some very basic fundamentals. Real life conflicts feature a wealth of concerns that aren't represented in the mechanics of games, even quite advanced video games. In virtually every game you have things like: direct line of sight, perfect communication with units, no chance that units will disobey orders, perfect uniformity amongst units, few logistics to consider (small exception to video games with resource mechanics), no non-military considerations such as political/social consequences etcetera etcetera.

Mostly i use the other, non-military considerations to counter "good guys" rely too much on robotics. Of course human soldiers will be also rather superhuman thanks to advancements in neural interfaces, drugs, bio-mods, powered battle armors etc.

Would it be epic? I imagine most forms of space warfare would just consist of trying to dodge incoming fire from incredibly distant enemies until you are destroyed in the blink of an eye, or they are, or both.

Due to laser scattering and broadband reflective armor, i rather consider few hundred kms for lasers and supervelocity weapons.
Few thousands for missiles and fighters. Yes the later will be unmanned most times.
 
  • #22
Czcibor said:
Quite close, I'm a Polish native speaker.

I was basing my assumption on the arrangement of consonants in your online name. We have a small but visible community of Czech-Americans where I live. One of the common surnames among these folks is Krzmarzick.
 
  • #23
GTOM said:
http://www.bbc.com/news/science-environment-33867941

Interesting, altough i'd like to know, how it actually learns? I could simply write a program, that iterates through a set of possibilities, till it finds the best solution.
The problem, that it won't be effective, if possibilities become polinomial or exponential.

But theese things still give me a headache, what can be the role of humans, that they can't solve with electronic slaves, in a not so near future? :(

https://en.wikipedia.org/wiki/Deep_Blue_(chess_computer)

Seems that such brute force approach is being used. However, when you get more expertise you improve the program and concentrate on testing good paths, thus reducing overall required computation power.
 
  • #24
Edit by mentor: deleted link removed

Personally i don't really believe in this naturally emerged AI thing... although i found it interesting, that in Endymion, they said that actually AIs evolved from computer viruses :P

Otherwise sorry for OFF, but back to hit ratio : I read a book about the criminal world of Hungary, the main character started in the war in ex-Yugo. Where his merc mate told him : 100 basic payment, 50 for hits, big money for special missions and selling bullets and grenades.

Isnt is possible, that is why so much ammunition is wasted in Iraq as well?
 
Last edited by a moderator:
  • #26
Parallel programming is useful for problems with very repetitious but independent calculations. Like predicting weather, where the computations are pretty much the same everywhere, though they may have different results. Or other modeling of uniform physical systems.
 

FAQ: Parallel computing, AIs and the cloud

1. What is parallel computing?

Parallel computing is a type of computing architecture where multiple processors or computers work together to solve a problem. This allows for faster and more efficient processing of data and tasks.

2. What are the benefits of parallel computing?

The main benefits of parallel computing include increased speed and efficiency, the ability to handle large and complex data sets, and the potential for cost savings by using multiple processors instead of one high-powered processor.

3. What is an AI?

AI, or artificial intelligence, is a branch of computer science that focuses on creating intelligent machines that can perform tasks that typically require human intelligence. This includes tasks such as problem-solving, decision making, and pattern recognition.

4. How does the cloud relate to parallel computing and AI?

The cloud refers to a network of remote servers that are used to store, manage, and process data over the internet. Parallel computing and AI can both benefit from the use of cloud computing, as it allows for faster and more efficient processing of data and tasks, as well as access to vast amounts of data.

5. What are some real-world applications of parallel computing, AIs, and the cloud?

Some examples of real-world applications include data analysis and processing in industries such as finance and healthcare, image and speech recognition, natural language processing, and virtual assistants.

Similar threads

Writing: Input Wanted Number of Androids on Spaceships
Replies
21
Views
2K
Replies
1
Views
1K
Writing: Input Wanted Captain's choices on colony ships
Replies
4
Views
2K
Replies
6
Views
2K
Replies
5
Views
2K
Replies
62
Views
7K
Replies
7
Views
2K
Back
Top