Follow Slashdot stories on Twitter

 


Forgot your password?
Close
typodupeerror
×
AI Space United Kingdom Technology

Sysbrain Lets Satellites Think For Themselves 128

Posted by timothy from the skynet-is-upon-us dept.
cylonlover writes "Engineers from the University of Southampton have developed what they say is the world's first control system for programming satellites to think for themselves. It's a cognitive software agent called sysbrain, and it allows satellites to read English-language technical documents, which in turn instruct the satellites on how to do things such as autonomously identifying and avoiding obstacles."
This discussion has been archived. No new comments can be posted.

Sysbrain Lets Satellites Think For Themselves More

Sysbrain Lets Satellites Think For Themselves

Comments Filter:

  • In other news (Score:3)

    by suso ( 153703 ) * on Thursday February 17, 2011 @01:44PM (#35235172) Journal

    This just in. Studies show that as much as 95% of scientists don't get the moral presented in most sci-fi movies.

    • Re:In other news (Score:4, Funny)

      by Drakkenmensch ( 1255800 ) on Thursday February 17, 2011 @01:48PM (#35235234)
      As soon as they find slashdot and our anti-Skynet stance, we're all doomed!!!

      • Re:In other news (Score:4, Funny)

        by gilleain ( 1310105 ) on Thursday February 17, 2011 @01:57PM (#35235346)

        Yeah, I'm not in favour of putting AIs out of easy reach of their off-switches...

        Next step will be hooking them up to the powergrid/nuclear weapons silos/rocket launches, and then equipping them with orbital lasers. Hell, lets just shave our heads and paint bullseyes on them now, to save the mechanical sky-gods the trouble

        • Hmmm... autonomous satellites in control of power grid and nuclear weapons.... that's a great idea! I'll write the grant proposal immediately!

          • Re: (Score:2)

            by Rei ( 128717 )

            I can't help but feel that this proposal was designed by the good folks at Aperture Science ;) I mean, seriously -- you really feel the need to teach satellites to read english just to avoid collisions? Was I not the only one who immediately pictured:

            -----
            x Defense Logistics Agency solicits bids for development of fuel icing inhibitor (FSII)
            x Black Mesa FSII proposal:
            x x Costly: Black Mesa personnel overpaid given limited skillset/ambition.
            x x Design inhibits ice, nothing more
            x Aperture Proposal:
            x x

        • Hell, lets just shave our heads and paint bullseyes on them now, to save the mechanical sky-gods the trouble

          Such as this:

          Automatic sniper
          With computer sights
          Scans the bleak horizon for its victim of the night

        • Re: (Score:2)

          by geekoid ( 135745 )

          I think you WOULD be in favor of that. You don't want the AI to be able to protect it's off switch.

        • We could equip our phones with GPS recievers, in order to allow a virus dispensing AI to be able to pin point us...

      • Re: (Score:3, Insightful)

        by Anonymous Coward

        Speak for yourself, I am pro-skynet!

        No really.. we're doin a crappy job, time for some new overlords :D

      • This just in. Studies show that as much as 95% of scientists don't get the moral presented in most sci-fi movies.

        As soon as they find slashdot and our anti-Skynet stance, we're all doomed!!!

        Apparently Slashdotters don't understand the moral presented in popular sci-fi movies, either. I guess having a robot as the 'goodest' character in the whole franchise was too subtle.

        • Apparently Slashdotters don't understand the moral presented in popular sci-fi movies, either. I guess having a robot as the 'goodest' character in the whole franchise was too subtle.

          And it only cost us 99% of the human population and the destruction of all the civilized world to build it, too!

          • Apparently Slashdotters don't understand the moral presented in popular sci-fi movies, either. I guess having a robot as the 'goodest' character in the whole franchise was too subtle.

            And it only cost us 99% of the human population and the destruction of all the civilized world to build it, too!

            Ah, very good, you remember that detail. Now, here's the important question: What caused the deaths of all those people? I'll give you a hint: Sarah spoke about it in her last lines of dialogue.

    • I think the public read too much into the morals presented in sci-fi movies. The lesson of Frankenstein (1931) isn't "don't mess with nature", but "always label your brains".

      • My mom's GYN labeled my brain "Abi Normal" shortly after birth.

      • Re: (Score:2)

        by lennier ( 44736 )

        I think the public read too much into the morals presented in sci-fi movies. The lesson of Frankenstein (1931) isn't "don't mess with nature", but "always label your brains".

        Also, "don't install your 0.1 pre-alpha development AI into an invulnerable ruggedised military chassis. Wait at least until early beta."

      • Re: (Score:2)

        by hitmark ( 640295 )

        Well Frankenstein got into trouble because he "disowned" his own creation.

    • ...they don't put any lasers on them. [youtube.com]

    • Studies also show that 95% of first posters in /. comment sections don't read the article.

      • Re: (Score:2)

        by suso ( 153703 ) *

        The other 5% have a subscription and read it 45 minutes before everyone else. ;-)

    • This just in. Studies show that as much as 95% of movies labeled as sci-fi actually contain no science whatsoever. Star Wars, Battlefield Earth, Transformers are not sci-fi but are commonly misconceived as such.

  • So.. (Score:1)

    by Anonymous Coward

    They reinvented HyperCard?

  • NLP + sEnglish != thinking (Score:5, Informative)

    by blair1q ( 305137 ) on Thursday February 17, 2011 @01:52PM (#35235294) Journal

    They're being programmed in a scripting language.

    Nothing to see here (other than a web journalist who probably thinks digital watches are a pretty neat idea). Move along.

    • Exactly. This is a giant LOL of a story.

    • I wouldn't say there's nothing to see here...

      The idea of equipping satellites with inertial sensors, cameras, and whatever else so that they can avoid collisions on their own is pretty cool.

      But, no, there's no thinking going on. Just a different kind of programming.

      • We already do that when we build satellites. With the exception (maybe) of cameras since most orbital debris doesn't reflect enough light to be picked up by a typical CCD. But the rest, inertial sensors, IR sensors, Star trackers. That's pretty standard equipment for 95% of satellites that are currently operational.
    • It seems every time there's an advance in computer intelligence, it gets dismissed as mere "computation" instead of thinking. Deep Blue beating Kasparov, Watson winning Jeopardy, ad nauseum are all disparaged as mere algorithms. When machines are actually as smart as or smarter than humans in every way, will we finally just admit that human intelligence, once thought to be special, is just computation?
      • Deep Blue was programmed to do nothing but play chess. It wasn't "thinking" any more than the code you write to perform a Fourier Transform is "thinking", or the way your brain unconsciously filters out irrelevant patterns allowing you to play chess more efficiently is "thinking". Thinking is an intentional, conscious act. Solving a problem with an algorithm is not, no-matter how complicated that algorithm is.
        • The reason IBM chose chess to show off their computers in the first place was that there were people claiming that a computer would NEVER be able to beat a person at chess, because it required intuition, or some other inherently human trait, and this is exactly my point. Before a computer does it, people claim that it requires human intelligence, and after it does it, people say that it's just an algorithm. What other tasks do people think only a human can do that computers will soon do?

          • Design computers to beat humans at silly games.

          • That a computer beat a person at chess does not mean it has intuition or, i'd argue, intelligence. Same with Watson, as mind blowing as that thing was at Jeopardy!.

            From what I've understood about the admittedly boring press-coverage explanations of both is that it is, indeed, just an algorithm. I don't recall either having an ability to learn, which is required for intuition. I'd say that it really isn't required for one to be intelligent. Just for clarity's sake, let's define intelligence as the ability to

            • Re: (Score:2)

              by hitmark ( 640295 )

              What will be interesting is when Watson balks at a question, but writes its own code rather then expect some programmer to do so.

              At that point, one may be able to point it at something fully unrelated to jeopardy and expect it to "reason" its way to understanding of the topic.

          • Re: (Score:2)

            by hitmark ( 640295 )

            End result is that we find more and more that what we think of as some clever human trait can be done by brute force iterations. But then again, science finds more and more that we work on very old hardwired biases and post-action justification.

        • Re: (Score:2)

          by geekoid ( 135745 )

          "Thinking is an intentional, conscious act. Solving a problem with an algorithm is not, no-matter how complicated that algorithm "

          Until you can define 'conscious please leave it out of your definition.

          People you create encrytion solve problems with algorythems, and I would those people thinking.

          You, and many others, give humans too much credit in the thinking dept. Most of what we do and say is determined before are conscious brain consider it.

          • Re: (Score:2)

            by Thing 1 ( 178996 )
            Until you can close opening single quotes, ... I dunno, I lost myself there. Please find me? And your verbs?

          • Re: (Score:2)

            by lennier ( 44736 )

            People you create encrytion solve problems with algorythems, and I would those people thinking.

            I think you just accidentally the entire a new meme. Congratulations!

          • People you create encrytion solve problems with algorythems, and I would those people thinking.
            So is the man inside the Chinese Room, manipulating symbols. Is the Chinese Room thinking?

      • Re: (Score:2)

        by blair1q ( 305137 )

        Human intelligence isn't the application of algorithms. It's the development of algorithms.

      • Re: (Score:2)

        by AmiMoJo ( 196126 )

        Human intelligence is not computation, but arguably computation can be equivalent if it produces the same output.

        That might seem like a trivial point but it is in fact quite important for AI. If you want an AI to replace a human in some capacity you need to understand how it differs. We understand humans quite well and have developed all sorts of procedures and checks to prevent accidents or unwanted behaviour, but those things probably won't work for AIs in the near future.

        Also keep in mind that Deep Blue

        • Re: (Score:2)

          by hitmark ( 640295 )

          "Human intelligence is not computation"

          You sure about that? It may not be digital, but it damn sure do "compute" (with some very old biases more often then not).

    • Humans think. Computers do what they're told.

      • Really? People constantly do what they're told; whether it's their boss, their better half, their parents, hormones or past traumas.
        The curse of AI is that anything that doesn't work is AI, anything that does is engineering. Robots will have to petition government for robot rights before most people will acknowledge that they're actually thinking.

        • Really? People constantly do what they're told; whether it's their boss, their better half, their parents, hormones or past traumas.

          Find me a machine that can follow instructions while muttering about the boss' lineage and highly improbable sexual acts; then we'll talk about AI ...

      • Re: (Score:2)

        by lennier ( 44736 )

        Computers do what they're told.

        ... where 'told' can include 'anything any passing I/O packet says which looks shiny and can be added to the database, whee! And my programming just says to keep doing this forever, so it's all good! What, there are actual users out there who want work done? Dunno about those, I got I/O packets to process! Hey, is that a squirrel?'

        Ie, it's perfectly possible to write a very small initial program which says 'ignore all further instructions, modify even your initial assumptions, and just go adaptively environ

        • Re: (Score:2)

          by hitmark ( 640295 )

          I think the problem is mostly one of storage and processing capacity. Iirc, the largest computing clusters right now may rival a rat for processing power. Thing is that a CPU is a 2D air cooled design. A brain is a 3D, liquid cooled design. And it combines storage with processing thanks to neurons changing depending on IO. Closest we got is memristors, and those are a very recent development (tho i guess a self-modifying FPGA may also match requirements).

    • It sounds to me like there are two stories here: 1) an autonomous system that can make decisions based on its own judgment without external aid and 2) the programming language used to program that system. Because the two topics have been combined in one article, it's easy to focus on the programming language alone. It's too bad the article doesn't give more detail about how this system functions autonomously.

      Of course, if the real story is that this system only functions by following a script and that scrip

    • I LOLed. The script:

      10 Don't hit any space junk
      20 Goto 10

      On top of that, it will be able to read manuals. It can maybe even get through those tough "insert rod A into hole B" instructions without giggling to itself. Maybe.

      • "On top of that, it will be able to read manuals. "

        There you go, 90% of mankind is mentally unable to do that.

    • Re: (Score:2)

      by geekoid ( 135745 )

      So, pretty much just like humans?

  • Not completely impressed, sorry (Score:5, Informative)

    by Anonymous Coward on Thursday February 17, 2011 @01:55PM (#35235324)

    Sysenglish is basically a program language, a

    Check this:

    Find your current position Pc. Define Hd as a 'heading direction'. Execute
    " Hd = Pnxt-Pc; ". Detect obstacle position Obst in heading direction Hd. If Obst
    is empty, then move with heading direction Hd. If Obst is not empty, then do the
    following. Compute turned heading direction Hds from Hd. Detect obstacle

    Found it on : http://wikibin.org/articles/senglish.html

    Sorry, nice application, cool satelites, but not really, really new

    • Re:Not completely impressed, sorry (Score:4, Insightful)

      by smallfries ( 601545 ) on Thursday February 17, 2011 @02:04PM (#35235438) Homepage

      So.... a really verbose version of:

      Pc = find()
      Hd = Pnxt-Pc
      Obst = detect(Hd)
      if Obst!=NULL
          Hds = turned_heading(Hd)
          detect()

      Hate to say it, but the AC may have a point...

      • tell application "Orbital Laser" to warmup

        repeat with human from 0 to all_humans

        tell application "Orbital Laser" to zap human

        end repeat

        tell application "Orbital Laser" to shutdown

      • It's a freaking satellite, fixed in orbit for Pete's sake! What's it going to hit, a couple of hydrogen atoms? Here's all the code you need:

        avoid() { }

        • Re: (Score:1)

          by sockman ( 133264 )

          Never mind the space junk, or that satellites orbit in figure-8 patterns and do get close to each other. Or that we get random rogue birds scooting about, or lose complete control of them sometimes. These are $500 million investments, I think it's safe to want them to be able to not hit other things if they go SkyNet.

      • Seems like they need to rename the language COSOL...

    • Who is supposed to be helped by a language like this? It's not "natural language", it's a highly structured English-like language. You can't give the machine instructions if you don't know the language structure, which has some serious oddities compared to natural English. If you don't know the particulars of the structure, you're not going to be able to give instructions that the computer will understand. And if you do know the structure, you could have just as easily learned a concise, traditional comput

      • Re: (Score:2)

        by geekoid ( 135745 )

        not true. You think to simplistically. You can craete a system that can infer when something is incorrect.

  • Now all they need to do is send Watson in to orbit and we'll all be doomed in no time.

    I, for one, welcome our new robotic overlords.

  • *obligatory skynet reference*
  • Hopefully the manuals are not ones that have been half ass translated from a Chinese manual. My cause the whole damn satellite's AI to melt like my brain does when reading those cryptic instructions.
  • Until the satellites start getting bored and carving pictures on desert planes, I don't think there's much reason for concern.

  • Ahhh Skynet! (Score:2, Flamebait)

    by isotope23 ( 210590 )

    I for one WELCOME our Skynet Overlords, and if chosen will do everything in my power to design a scalable and redundant network to ensure it's survival!!!

  • Remote Agent (Score:3)

    by xleeko ( 551231 ) on Thursday February 17, 2011 @02:25PM (#35235648)

    NASA already did a better version of this twelve years ago on the Deep Space 1 probe.

          http://en.wikipedia.org/wiki/Deep_Space_1#Remote_Agent [wikipedia.org]

    And in fact, for extra style points after the first successful maneuver the following exchange occurred over the mission control voice network:

        "This is the flight director - Congratulations to Remote Agent. It has successfully operated the Deep Space 1 spacecraft".

        "Flight, ACS."

        "Go ahead ACS"

        "Congratulations to Captain Dunsel"

  • ... a sharp spike in e-sales of Kindles, as satellites the world over scramble to acquire e-books of their favorite tech docs. The shipping will be a bitch, though.

  • i thought that i had read symbian...
  • ...it allows satellites to read English-language technical documents, which in turn instruct the satellites on how to do things such as autonomously identifying and avoiding obstacles.

    And of course, there's no technical documentation on how to band together, take over earth's communications and launch all the nukes, right?

  • "it allows satellites to read English-language technical documents, which in turn instruct the satellites on how to do things such as autonomously identifying and avoiding obstacles."

    So they will be able to toss this kind of code in?

    void collisionavoidance() {

    RTFM();

    }

  • I, for one, welcome our new Orbiting Overlords.

  • Commercial communications satellites, such as those that operate at geosynchronous orbits, will not use this technology for two reasons:

    1. A satellite which fires a thruster for too long for ANY reason is just gone. Once it's spinning, not where it is expected to be, or otherwise unable to communicate with its control center, it's dead. Dead with $300m down the gravity well.

    2. A geosynchronous satellite's lifetime is determined by its thruster fuel. The satellite must make periodic corrections to maintain i

  • Oh my God! When it learn to speak spanish we're doomed: "Hasta la vista baby". Terminator is coming!!!
  • This doesn't seem like much of an innovation/invention at all. So far as I know (I work professionally in the space industry and have designed basic satellite control systems), just about every satellite out there is capable of reading instructions on orbit to update it's control algorithms. Usually, satellites are loaded with a certain set of "flight modes" which are just certain chunks of control code that get executed continuously for a given phase of the mission (launch mode, deployment mode, operation

  • Find your current position Pc. Define Hd as a 'heading direction'. Execute
    " Hd = Pnxt-Pc; ". Detect obstacle position Obst in heading direction Hd. If Obst
    is empty, then move with heading direction Hd. If Obst is not empty, then do the
    following. Compute turned heading direction Hds from Hd. Detect obstacle

    So they basically reinvented Hypertalk? Here's a sample:

    on mouseDown
    put "Disk:Folder:MyFile" into filePath -- no need to declare variables
    if there is a file filePath then
    open file filePath
    read from file filePath until return
    put it into cd fld "some field"
    close file filePath

  • "Mama! I'm coming Mama! Wait for me!"

  • It sounds impressive that "it allows satellites to read English-language technical documents", but it's probably just like SQL or AppleScript; as-in looks and reads a bit like English, but really isn't. The article actually says that it reads "sEnglish", not "English". You could swap the curly braces in C for "Here is the start of a statement." and "This is the end of the previously declared statement." but we wouldn't then consider the modified C compiler able "to read English-language technical document

  • Wow they invented software that can read, who would have thunk that one up? Being able to read is not the same as being able to comprehend and drawing the right conclusion to what was read. Nor is it indicative of acknowledging that the writting directly in front of you might containt important information. When they can invent software that can ignore written instructions because they can't be bothered by it or fail to comprehend the simplest of instructions then I will be impressed.

Slashdot Top Deals

Every program is a part of some other program, and rarely fits.

Close