If you ever wondered whether or not "AI" is actually "BS"...



  • From the Wikipedia article on Expert Systems:



    "Expert systems are designed to solve complex problems by reasoning about knowledge, like an expert, and not by following the procedure of a developer as is the case in conventional programming."




    Oh, wow, that really sounds novel. My boss will defini... HEY... WHAT'S THAT "IF" DOING THERE IN YOUR CODE??




    "An expert system has a unique structure, different from traditional programs. It is divided into two parts, one fixed, independent of the expert system: the inference engine, and one variable: the knowledge base."




    That's really groundbreaking. A fixed executable program operates on a variable "base" of some sort, which is maintained independently over time. This idea will really improve anything I write in the future, especially if it uses a database or a configuration file.



    "There are generally three individuals having an interaction in an expert system."



    Exactly three? How quaint. It reminds me of square dancing. Let's learn more.




    "In the construction and maintenance of the system there are two other roles: the problem domain expert who builds the system and supplies the knowledge base, and a knowledge engineer who assists the experts in determining the representation of their knowledge, enters this knowledge into an explanation module and who defines the inference technique required to solve the problem. Usually the knowledge engineer will represent the problem solving activity in the form of rules. When these rules are created from domain expertise, the knowledge base stores the rules of the expert system."




    I think I know that guy! And certainly, the existence of an "explanation module" should enable all of us to sleep more soundly. I know I've got a few questions for it.




  • Jesus Christ, is this your first experience with software? Expert systems (or similar systems) are fairly common in complex systems. The entire point is to use a simple, descriptive syntax (or even a domain-specific language) to allow non-programmers to define decision-making processes. The syntax is much more flexible than imperative programming statements and inference engines can usually optimize it to execute much more efficiently than a bunch of if-else statements.



  • @morbiuswilters said:

    Jesus Christ, is this your first experience with software? Expert systems (or similar systems) are fairly common in complex systems. The entire point is to use a simple, descriptive syntax (or even a domain-specific language) to allow non-programmers to define decision-making processes. The syntax is much more flexible than imperative programming statements and inference engines can usually optimize it to execute much more efficiently than a bunch of if-else statements.

    I don't understand how that is "artificial intelligence." It's just normal programming that anyone with a good background should be able to do. Good programmers make domain-specific languages. Joel Spolsky said as much. This may sound troll-worthy, but about 80% of the crap postulated in that article just boils down to fancy applications of Javascript "eval()." AI just had good marketing.



  • Expert systems are, at best, just a subset of artifical inteligence. 


     



  • @esoterik said:

    Expert systems are, at best, just a subset of artifical inteligence. 


     

    But according to a lot of the materials I've looked at while defecating, Expert Systems are actually one of the most successful parts of AI. If this is the good part, what's the bad part? Skynet? Microsoft SharePoint?? The mind recoils.



  • Ever heard about Prolog ?



  • AI is one of these topics with lots and lots of bullshit, like psychology and others (I hope I don't start a flame war). It "looks easy" from a distance, and people just love to speculate about interesting theories and nice-sounding concepts without bothering to get into the "boring mathematical/programming part" (and they love to share their knowledge on Wikipedia too). That doesn't mean, though, that all of the AI is bullshit. Things like playing chess or video games, theorem-proving, optimization, or simple path-finding are AI, even if not the type of AI that appears in Hollywood movies.

     

    If you want to see a real AI crank, google "mentifex".



  • AI = Artificial Idiot



  •  @bridget99 said:

    I don't understand how that is "artificial intelligence." It's just normal programming that anyone with a good background should be able to do. Good programmers make domain-specific languages. Joel Spolsky said as much. This may sound troll-worthy, but about 80% of the crap postulated in that article just boils down to fancy applications of Javascript "eval()." AI just had good marketing.

    There is a big difference between you not understanding and something being BS. A rule based reasoning system is most certainly not "just normal programming". Although by now rule-based programming is known to more developers, it was restricted to AI until the 1990s or so. It is pretty difficult to come up with a good rule-based paradigm. And it is something very, very different from eval(), which, by the way, originates in LISP, a 1950s language, heavily used in AI. An expert system is also not "a domain specific language", and Spolsky, well, let's say he has a big mouth.

    Just let's phrase it this way: if we would "task" a group of "just normal programmers" to build a medical diagnosis system, chances that it'll take them forever, that they will overcharge, and that the system will almost always come up with wrong diagnoses are about 100%.



  • @esoterik said:

    Expert systems are, at best, just a subset of artifical inteligence. 


     

    You make it sound like that's an insult. AI is very wide and nobody claims ES are the whole of AI. Nobody (who knows about AI) claims AI has to be super-cool or smart, either. It's not all neural nets and genetic algorithms, a simple fixed-state-machine is AI, or using the A* algorithm to find a path.

    It's called [b]A[/b]I for a reason.



  • @bridget99 said:

    Good programmers make domain-specific languages.

    As anybody with experience with Gamebryo/Creation Engine and Papyrus can tell you, terrible programmers also make domain-specific languages.



  • @TGV said:

    Just let's phrase it this way: if we would "task" a group of "just normal programmers" to build a medical diagnosis system, chances that it'll take them forever, that they will overcharge, and that the system will almost always come up with wrong diagnoses are about 100%.

     

    So whatever diagnosis it recommends, do the exact opposite and you have a system that gives the right diagnosis 100%.

    Anyone remember Neil Frasers classic AI story ... http://neil.fraser.name/writing/tank/

     

     



  • @daveime said:

    So whatever diagnosis it recommends, do the exact opposite and you have a system that gives the right diagnosis 100%.

    It says to take all his blood out? OK, let's give him a four pint blood transfusion! Oh no, now his heart is going to explode due to too much blood!



  • @daveime said:

    http://neil.fraser.name/writing/tank/

    Heaven help the army if they deploy tanks in formation...

    AI: "Hey! Listen! I calculate a 95% probability that there is a tank in front of us!"

    Tank commander: "STFU, that's a friendly tank!"

    AI: "Hey! Listen!"

    Tank commander: SMASH



  • @ekolis said:

    @daveime said:
    So whatever diagnosis it recommends, do the exact opposite and you have a system that gives the right diagnosis 100%.
    It says to take all his blood out? OK, let's give him a four pint blood transfusion! Oh no, now his heart is going to explode due to too much blood!

    Then you have a new problem, run diagnosis again stating to much blood, and it replies, once again wrong, to add 4 pints of blood.  You do the opposite, removing 4 pints, and you have essentially solved the problem.  See, it can work if you follow through.



  • Expert Systems: Are they good or are they whack?



  • Tank mechanic in a box:



  • @bridget99 said:

    This may sound troll-worthy, but about 80% of the crap postulated in that article just boils down to fancy applications of Javascript "eval()."

    If you're using eval() you're DOING IT WRONG.

     

    An expert system is not like that. It uses its knowledge base to make logical decisions based on the underlying rules. The knowledge is NOT the definition if its behavior. It doesn't determine HOW the logical decisions are made. Here's an example: a text translation engine. The underlying logic of translation is unchangeable, it's fixed, just identify a sentence and, using its knowledge, interpret it and translate that sentence to another language. What changes is the knowledge: the vocabulary, the grammar, the language syntax.



  • "Artifical Intelligence" from a current practical standpoint, is something that is really "Stupid", but complex enough and fast enough that it gives the impression of intelligence. I believe the real trick is to develop a computer program that implements "Artifical Stupidity", in this case you have a computer system that is truely intelligent, self-aware, et. al. and has decided to hide this fact from humanity and give the impression of stupidity.

    Or perhaps this has already happened.... There was a great book titled The Adolesence of P1" written nearly 40 years ago...recommended reading for anyone considering the ramifications of "true" Machine Intelligence.



  • @TheCPUWizard said:

    "Artifical Intelligence" from a current practical standpoint, is something that is really "Stupid", but complex enough and fast enough that it gives the impression of intelligence. I believe the real trick is to develop a computer program that implements "Artifical Stupidity", in this case you have a computer system that is truely intelligent, self-aware, et. al. and has decided to hide this fact from humanity and give the impression of stupidity.

    If only there were some kind of test we could give a computer to determine if it is indistinguishable from a human..



  • @TheCPUWizard said:

    Artifical Intelligence" from a current practical standpoint, is something that is really "Stupid", but complex enough and fast enough that it gives the impression of intelligence.

    So it is exactly like natural intelligence

    @TheCPUWizard said:

    I believe the real trick is to develop a computer program that implements "Artifical Stupidity", in this case you have a computer system that is truely intelligent, self-aware, et. al. and has decided to hide this fact from humanity and give the impression of stupidity.

    Like internet posting?

    @morbiuswilters said:

    If only there were some kind of test we could give a computer to determine if it is indistinguishable from a human.. .

    Beer pong?



  • So many people having an opinion on something they don't understand at all, this must be the interwebs I've read so much about lately. Look up the difference between "strong" and "weak" AI. Expert systems, like computer vision, or machine translation, falls in the latter category. Weak AI studies cognitive strategies and tries to apply them, often in software.



  • @blakeyrat said:

    @bridget99 said:
    Good programmers make domain-specific languages.

    As anybody with experience with Gamebryo/Creation Engine and Papyrus can tell you, terrible programmers also make domain-specific languages.

    +1



  • @bridget99 said:

    @esoterik said:

    Expert systems are, at best, just a subset of artifical inteligence.

    But according to a lot of the materials I've looked at while defecating, Expert Systems are actually one of the most successful parts of AI. If this is the good part, what's the bad part? Skynet? Microsoft SharePoint?? The mind recoils.

     

    Get a Chinese room, you two.

     

     



  • @frits said:

    Expert Systems: Are they good or are they whack?

    Obviously..."it depends"...



  • @daveime said:

    So whatever diagnosis it recommends, do the exact opposite and you have a system that gives the right diagnosis 100%.
    Of course, that presumes the options are binary.  If there are multiple possible diagnoses, you can't just "do the opposite" . . .@daveime said:
    Anyone remember Neil Frasers classic AI story ... http://neil.fraser.name/writing/tank/
    That was quite funny!  I'd never heard that story before.



  • @English Man said:

    AI is very wide and nobody claims ES are the whole of AI
    The OP sure thinks so, as evidenced by the title of this thread!

     



  • @TGV said:

    So many people having an opinion on something they don't understand at all, this must be the interwebs I've read so much about lately. Look up the difference between "strong" and "weak" AI. Expert systems, like computer vision, or machine translation, falls in the latter category. Weak AI studies cognitive strategies and tries to apply them, often in software.

     

    A very simplisic explanation of an expert system (ES) is that you are trying to capture the knowledge and reasoning of an expert in a program. The implementation requires things like a logic inference engine, and knowledge expressed as a set of logical rules (the knowledge base). What something does, and how it does it are two different things, conflating the two causes the confusion which is this thread!

     



  • @TGV said:

    There is a big difference between you not understanding and something being BS.




    I feel like I understand how an Expert System works. I can see a place for such a thing. I just don't agree with all the associated hand-waving and mumbo-jumbo. They're not special, and they're not intelligent.



    @TGV said:
    A rule based reasoning system is most certainly not "just normal programming".

    How different is it, really? More different than writing markup is from writing assembly language? More different than XSLT is from SQL? I don't think so, and the differences I named fall well within the realm of everyday practice among very ordinary programmers.



    @TGV said:
    It is pretty difficult to come up with a good rule-based paradigm.

    So why do things that way?



    @TGV said:
    And it is something very, very different from eval(), which, by the way, originates in LISP, a 1950s language, heavily used in AI.

    I was being a little bit facetious. I realize that being able to dynamically generate and execute code is considered key to certain parts of AI , and that's why I mentioned it. I also mentioned it because just about every knuckle-dragging script kiddie on AOL knows about eval(), and none of them claim (at least not plausibly) to have discovered something that the rest of us missed.



    @TGV said:
    An expert system is also not "a domain specific language"

    The rules engine can be described as a compiler or interpreter for the language in which the rules are written, and it is certainly a domain-specific language. You couldn't write Super Mario Brothers or a printer driver in it.



    @TGV said:
    Just let's phrase it this way: if we would "task" a group of "just normal programmers" to build a medical diagnosis system, chances that it'll take them forever, that they will overcharge, and that the system will almost always come up with wrong diagnoses are about 100%.




    If you were to build a small team of good programmers and physicians and set them to work on such a task, I think they could accomplish this using some relatively simplistic methods. As a patient, I'm not sure I want to rely on a highly flexible or configurable system to diagnose me. Go ahead and just hard-code stuff like "elevated liver enzymes often indicate Paracetamol overdose" into the program. This is not something that should depend on some non-CS person monkeying around with a fancy code or configuration file. And any software project is going to be subject to bugs and delays... do you seriously think that involving a bunch of researchers who think they're truly better than everyone else is going to appreciably speed up anything or enhance quality?



  • When done well, expert systems can do quite well with unexpected input. Watson managed to do better than Ken Jennings at Jeopardy and they plan to use a similar approach to medicine and other areas of decisionmaking.

    Some of the Watson developers attended an airing of one of the Human v. Watson games near me. I had a chance to talk to one of the lead developers afterwards about the limits of approximating human intelligence using an approach like this. I found it interesting that he believes that there is no upper limit in human functions that could be accurately approximated using an approach like this including creativity. I'm not entirely convinced by that, not their explanation that a machine using a buzzer doesn't have an advantage over humans (they say it's fair because the computer could still potentially buzz in incorrectly and lose money...in other words they argue that the buzzer winner has the fastest cognition time and reaction time is secondary)



    [Fixed markup. -ShadowMod]



  • @cmccormick said:

    When done well, expert systems can do quite well with unexpected input. Watson managed to do better than Ken Jennings at Jeopardy and they plan to use a similar approach to medicine and other areas of decisionmaking.

    Some of the Watson developers attended an airing of one of the Human v. Watson games near me. I had a chance to talk to one of the lead developers afterwards about the limits of approximating human intelligence using an approach like this. I found it interesting that he believes that there is no upper limit in human functions that could be accurately approximated using an approach like this including creativity. I'm not entirely convinced by that, not their explanation that a machine using a buzzer doesn't have an advantage over humans (they say it's fair because the computer could still potentially buzz in incorrectly and lose money...in other words they argue that the buzzer winner has the fastest cognition time and reaction time is secondary)



    [Fixed markup. -ShadowMod]

    Yeah, expert systems have a wide range of uses. I'm not sure how far they can be pushed, but they're still impressive. Just ignore bridget, he/she is insane and doesn't know what he/she is talking about. This is the same person who thinks it's impossible to program against an API.



  • @morbiuswilters said:

    This is the same person who thinks it's impossible to program against an API.

    non-pedantic, but still "DW" comment.... Of course it is a bad idea to try to program against an API...you should really try programming in accordance with an API...usually works much better.



  • @bridget99 said:

    do you seriously think that involving a bunch of researchers who think they're truly better than everyone else is going to appreciably speed up anything or enhance quality?

    There we have it. You're either trolling, or you've got some issues feeling inferior to academic researchers. Because the answer to your insinuating and simultaneously self-defending, if not leading, question is a resounding yes.



  • @TGV said:

    @bridget99 said:

    do you seriously think that involving a bunch of researchers who think they're truly better than everyone else is going to appreciably speed up anything or enhance quality?

    There we have it. You're either trolling, or you've got some issues feeling inferior to academic researchers. Because the answer to your insinuating and simultaneously self-defending, if not leading, question is a resounding yes.

    Researchers are good at discovering new things. Most of them would never have claimed to be the sort of experts in practical application development that you tried to make them in your original post. And I do have a problem with people who take conventional techniques and try to create a Wizard of Oz-style mystique around them. I don't know if all proponents of Expert Systems fall into that category, but the ones that contributed to the Wikipedia article I cited clearly have that problem.

    And did it ever occur to you that I might BE an academic researcher (one who does not purport to have discovered "artificial intelligence")?



  • @bridget99 said:

    And did it ever occur to you that I might BE an academic researcher (one who does not purport to have discovered "artificial intelligence")?

    Why would anyone assume that?



  • @pkmnfrk said:

    @bridget99 said:

    And did it ever occur to you that I might BE an academic researcher (one who does not purport to have discovered "artificial intelligence")?

    Why would anyone assume that?

    Well, according to bridget researchers are dunderheads who are befuddled by practical implementation. And bridget is a dunderhead who is befuddled by practical implementation. QED.



  • I have developed a number of expert systems (starting with a DSS back in 1981), and I will attest that they are NOT build using "conventional techniques" that are found in other types of systems. It is a specialized domain, and finding qualified/experienced people is VERY difficult. I have worked with people who are undoubtable top-notch experts with "deterministic" systems, and even with a fair amount of time (months) they typically can not shake the techniques that are well suited to those environments, and develop the radically different mindset required.



  • @TheCPUWizard said:

    I have developed a number of expert systems (starting with a DSS back in 1981), and I will attest that they are NOT build using "conventional techniques" that are found in other types of systems. It is a specialized domain, and finding qualified/experienced people is VERY difficult. I have worked with people who are undoubtable top-notch experts with "deterministic" systems, and even with a fair amount of time (months) they typically can not shake the techniques that are well suited to those environments, and develop the radically different mindset required.

    No way! Nothing in CS is more complicated than programming in VB6! You're just trying to make yourself feel important by talking up expert systems, but bridget and I are on to you!!



  • @bridget99 said:

    And did it ever occur to you that I might BE an academic researcher (one who does not purport to have discovered "artificial intelligence")?
     

    I worked in cogsci, duly studied AI and expert systems, and I've worked in various software companies, and yes,
    the researchers that built those systems, like Mycin, were way, way
    more intelligent and capabable than any non-researcher developer I've
    met. And an expert system is not a "conventional technique", not for ordinary programmers anyway. And your definition of domain specific language seems to include Javascript, since you can't build a printer driver in it. You just don't have a clue, do you?

    But to think that you might be an academic... dear God. I know I had some stupid colleagues back when I was a post-doc, but this dumb? I knew academia was going to end badly, but like this? Although, you could be an ex-physicist. One that didn't quite make it. They are the ones that pretend to know everything and then try to solve linguistics or evolution with a tensor or some other bit of abbreviated linear algebra. "Just ignore the higher order terms, we always do that in physics."

    Anyway, you don't sound like an academic, more like Breivik who claims to know about the Middle Ages after 15,000 hours of study of Wikipedia articles on Assassin's Creed.



  • @TGV said:

    Although, you could be an ex-physicist. One that didn't quite make it. They are the ones that pretend to know everything and then try to solve linguistics or evolution with a tensor or some other bit of abbreviated linear algebra. "Just ignore the higher order terms, we always do that in physics."

    Hey, as someone who's formal education was in Physics and has been working on a next generation Ion Beam system (for etching and deposition) for the past year, I could take a bit of offense at that.



  • @TGV said:

    @bridget99 said:

    And did it ever occur to you that I might BE an academic researcher (one who does not purport to have discovered "artificial intelligence")?
     

    I worked in cogsci, duly studied AI and expert systems, and I've worked in various software companies, and yes,
    the researchers that built those systems, like Mycin, were way, way
    more intelligent and capabable than any non-researcher developer I've
    met. And an expert system is not a "conventional technique", not for ordinary programmers anyway. And your definition of domain specific language seems to include Javascript, since you can't build a printer driver in it. You just don't have a clue, do you?

    But to think that you might be an academic... dear God. I know I had some stupid colleagues back when I was a post-doc, but this dumb? I knew academia was going to end badly, but like this? Although, you could be an ex-physicist. One that didn't quite make it. They are the ones that pretend to know everything and then try to solve linguistics or evolution with a tensor or some other bit of abbreviated linear algebra. "Just ignore the higher order terms, we always do that in physics."

    Anyway, you don't sound like an academic, more like Breivik who claims to know about the Middle Ages after 15,000 hours of study of Wikipedia articles on Assassin's Creed.

    I'm not going to get into school, degree, or company names or anything else that might identify me to the pitchfork-wielding mobs here at WORSETHANFAILURE.COM, but I'll gladly compare GRE scores with you. Or do they not require that at your particular branch of DeVry?

    In any case, you need to realize that just because someone disagrees with you doesn't make that person wrong. For someone who claims to be an academic you have a remarkably poor ability to analyze your own ideas critically. Remember the long, point by point rebuttal I gave your original response here? You never responded to any of my points except the last, and that was with an ad hominem attack. I guess that's the easiest way out for a PAPER PhD: wave your resume around and hope people believe you. Sorry, professor, that doesn't work on me.



  • @TGV said:

    Anyway, you don't sound like an academic, more like Breivik who claims to know about the Middle Ages after 15,000 hours of study of Wikipedia articles on Assassin's Creed.

    That's a pretty cruel thing to say about another human being. We clearly disagree about certain things, and probably even about certain people. Maybe I have an inflated sense of my own expertise. But I've never perpetrated violence on anyone, or even written a manifesto, or even held particularly strong opinions about anything outside of a few relatively narrow engineering topics. If I spend more than five minutes reading Wikipedia articles about something fictional or even just irrelevant to my own goals, I unplug my network and force myself to work. I'm sorry if my snide little post unfairly characterized your particular subdiscipline of computing. There are pretentious people in AI, smart people in AI, and plenty of good people who don't care if you call what they do AI or not. I painted with a broad brush and you were probably quite right in feeling offended. But if you really find reason to compare me to a mass murderer because I thought a Wikipedia article was pretentious, then maybe my apology is misplaced.



  • @bridget99 said:

    @TGV said:
    Anyway, you don't sound like an academic, more like Breivik who claims to know about the Middle Ages after 15,000 hours of study of Wikipedia articles on Assassin's Creed.

    That's a pretty cruel thing to say about another human being.

    Yes, you're right. My apologies.

    What triggered it (this is not an excuse) was that you read an article on Wikipedia, a quite badly written one, actually, jumped to conclusions, and then insisted that your interpretation of expert systems was the right one and that AI, a much broader field, is consequently bull shit. It just isn't like that.

    As I explained before, AI studies cognitive mechanisms, and uses the outcome as a way to solve problems in software. Expert systems are a way to capture knowledge and reasoning steps in a domain independent and implementation independent manner. There are different ways to implement this. Some ES's have ontologies, others don't; some ES's have a blackboard, others use agent communication (agents, BTW, is another AI invention); some use forward inference, others backward inference, or theorem proving, or Bayesian networks; etc. But it keeps things neatly modular, as we all like. In that respect it's nothing like a conventional programming language.

    Yes, I know you can keep  them separated in a conventional language, but the whole point of these languages, let's call them general purpose language, which seems more adequate, is that there is no restriction on your implementation. Consequently, you've got all freedom in the world, but very little support. However, you can write an ES in C, and implement the rules in C, if you like. No need for a DSL. And an ES is not a domain specific language: they can be seen as a rather generic language for representing knowledge and reasoning.

    Now, if you want to know how ES's are different from normal programming, just try one out.

    As a patient, I'm not sure I want to rely on a highly flexible or configurable system to diagnose me. Go ahead and just hard-code stuff like "elevated liver enzymes often indicate Paracetamol overdose" into the program.

    That's exactly the problem. An ES is there to represent such knowledge, but usually as the result of reasoning over known facts. Some are set up in such a way that they require input as they develop hypotheses, others just work on a set of given data. If you hard-code this rule, you might be overlooking other evidence. However, if you've got knowledge about the causes of elevated levels of enzymes in the liver, you can make that corroborate with other data. Then you'll more likely to get a decent ranking of diagnoses. Of course you could implement that too, but then you'll just be hard coding an ES, using techniques that have been developed in ES research. And that's not conventional programming. Perhaps the programmers you know are all smart, razor-sharp intellectuals with extensive knowledge of different computational paradigms, but to me a conventional programmer is someone who copies a Spring configuration, changes two lines, slaps it onto a database, and calls it a web interface.



  • Real AI is when you throw a bunch of data at something to teach it, and then ask it questions based on what it knows. e.g. "here are the spectroscopic signatures of a bunch of satellites and space debris, crossreferenced with what materials they are constructed out of. Now use approximate nonnegative matrix factorization to break that down and, given new spectrograms of unknown space debris objects, determine what materials said space debris is composed of." Or, more commonly, "hey, these users have visited alllll these websites, and they have interacted with some of our advertisements. now try to decide what advertisements to show them in the future!"

    (Other things may also be real AI, but this is some of it. Some century we'll be throwing serious stuff at it, like, principles of philosophy. and expecting insight back. But not today.)

    And hey, even if-statement-level artificial intelligence doesn't need to be smarter than human intelligence to succeed: it can also succeed by being cheaper, or faster, or more readily available and scalable, or capable of surviving in the vacuum of space, or any number of things, depending on the mission.



  • @fennec said:

    Real AI is when you throw a bunch of data at something to teach it, and then ask it questions based on what it knows. e.g. "here are the spectroscopic signatures of a bunch of satellites and space debris, crossreferenced with what materials they are constructed out of. Now use approximate nonnegative matrix factorization to break that down and, given new spectrograms of unknown space debris objects, determine what materials said space debris is composed of." Or, more commonly, "hey, these users have visited alllll these websites, and they have interacted with some of our advertisements. now try to decide what advertisements to show them in the future!"

    (Other things may also be real AI, but this is some of it. Some century we'll be throwing serious stuff at it, like, principles of philosophy. and expecting insight back. But not today.)

    And hey, even if-statement-level artificial intelligence doesn't need to be smarter than human intelligence to succeed: it can also succeed by being cheaper, or faster, or more readily available and scalable, or capable of surviving in the vacuum of space, or any number of things, depending on the mission.

    Your definition of "Real AI" seems very narrow. An AI researcher once told me something I found illuminating: (paraphrasing) "The point of AI isn't to mimic human intelligence. In fact, human intelligence isn't really worth emulating. We can have periods of great insight, but even the smartest person absorbs new information much more slowly than a computer and will need to frequently look things up. The point of AI is to do particular tasks faster, cheaper and more reliably than humans, and that's something that's already been happening for decades."

    And it's true: the fact is, lawyers, doctors, engineers.. outside of a frequently-used base of knowledge they all have to look things up. And they're going to make mistakes whereas a computer intelligence can be refined and its error rate brought down to the point where it is able to diagnose a patient far quicker, cheaper and with fewer errors than a human doctor.


Log in to reply
 

Looks like your connection to What the Daily WTF? was lost, please wait while we try to reconnect.