For the same reason the people that make $70,000 a month part time with their home business have infomercials: because what they're really making money from is convincing people to buy into some scheme of theirs.
It's invariably the case that even when these systems really do work (a few of them, particularly the real estate redevelopment models, actually do work,) selling them to millions of people is significantly more profitable than actually running them.
It's called franchising. Pizza Hut doesn't own all its stores; neither do McDonalds, Wendys, and so on. Why? The stores are immensely profitable.
The answer is simple: reaping margin on people investing their own money is far more profitable than doing it all yourself, because the growth rate is far larger, and you step back from all of the personal risk. If they're profitable, they break you off a proper chunk. If they fail, well, it's just not your problem.
Be honest: would you rather own five fast food resteraunts, or reap a percentage on two hundred? The franchise model is so successful that better granchises, like fast food, often have smaller companies subfranchising. Next time you go to a Burger King or whatever, take a look at the drive through window, where they have the complaint phone number. Surprisingly often, it'll list a smaller company you've never heard of, but which owns a few hundred local fast food places from a variety of name brands. In Pittsburgh, most of the fast food in the college district, regardless of who the sign reads as, is owned by either of two companies (to the tune of some 100 or so storefronts covering a good two dozen brands.)
In short, selling a scalable business model is usually more profitable than using it.
There was just an article several days ago about a standardized CAPTCHA defeater toolkit. However, whereas the principle of turing testing may be sound, it's not enough.
Consider the case of the virtual money sweatshops for MMORPGs. They have people whose job is to sit and do boring repetitive things to exploit games to generate things which can be traded for real money.
Now, consider the scenario where poker bots become significantly profitable. You will quickly see sweatshops where it's someone's job to sit there answering CAPTCHA all day to allow the bots to continue to reap.
In most cases, I'd say that "cheating" is doing anything that is against the rules of the game that gives you an unfair advantage. But what here is the bot doing?
Communicating its hand to second instances of the bot. Would you let your poker buddies share hands during play?
Besides, whereas cheating may not be the best word for the practice of putting machines into human tournaments, it's still not exactly fair; you wouldn't let me into the olympic several mile race riding a motorcycle.
But I predict that that will be a self-correcting problem.
Poker has been running for 400 years, and the problem has gotten worse, not better.
Since after all, it won't be long before someone makes an alternate version of the BOT that feeds incorrect data to the other BOTs so that you're more likely to win money from them.
Er, that's a bit naive. Presume for a moment that I'm a cheater. I'm running three bots which are colluding. What makes you think:
That I would let you know who my bots were
That I would accept data from anyone I don't control
That I wouldn't immediately realize why someone was trying to cheat for me
It's not like there's a mesh of arbitrary cheaters out there trusting each other to cheat for the greater good of cheaters everywhere. They're intensely self focussed.
Anyway, as long as the BOTs aren't actually hacking the system, or forcing other peoples' clients to crash, etc, then I think you could make a good argument that it's not really cheating.
Oh sure, breaking the rules by sharing hands, that's not cheating at all. (cough) I'd love to see the described good argument.
And no, I'm not a bot user or apologist.
Well, maybe you're at least not a bot user, but you're attempting to make the active claim that sharing information which by the rules of the game is not public is not cheating, so that second half is a bit more than just debatable.
statistically speaking... about 90% of them are chumps, yes. The other 10% win.
It's more like 97.2% vs 2.8%.
In conclusion: stop talkin about things you don't understand, kiddo. I have no doubt that you lose at online poker, but the problem isn't that you're getting cheated.
You seem knowledgeable; it's a pity you've left no method to be contacted. Given that you claim to be making real money playing poker, I'd like to talk to you about game design. (Besides, we have shockingly similar fan and foe lists; that generally means similar interests.)
Do you really believe that the operators of these on-line "casinos" are above playing poker against you while they can watch your hands, or when they can tell the computer what to deal next?
Yes.
Casino operators have the fundamental quality of mathematicians: they understand that 1% gets very big when run over very big numbers. The amount of money they would make playing personally is essentially paltry compared to the amount of money their site makes. The risk of being discovered cheating, which isn't terribly unlikely considering the US Government's habit of raiding online casinos, is immense; even if they didn't get shut down, their play population would dwindle to nearly nothing overnight.
In short, it would be foolish to risk a machine making a million dollars a week, in order to score $200 a month against some asshats. I just don't think any of the system operators are that stupid. Even if they're compulsive gamblers, they're playing on other people's systems, just to be safe.
And while dealing themselves the good cards too often might be caught by statistical analysis of the decks
You would do well to read about the cryptanalysts in World War 2, who almost wouldn't act on the broken Purple (a Japanese navy code,) because they knew how to detect the enemy acting on information they shouldn't have, and therefore knew the enemy could almost certainly do the same thing.
It's not that hard to detect a cheater, even if it's only knowledge cheating, and the more successful the cheater is, the easier that detection gets.
Another form of cheating that I know is going on (because I know someone who admits to doing it) is to play multiple hands in the same game against another player and share information about your hands.
That would be the team play issue being discussed of the bots.
I find it unfortunate that the person with the clue to bring something like that to bear posted anonymously; I would like to be in contact with someone like that. To exploit too-perfect play is a hell of an interesting concept.
That said, this still makes the tenuous presupposition that the bot author is unaware of said approach. Whereas one might consider that in terms of the avant garde - similar in many ways to virus authors being just barely ahead of viruses - it is important to remember that agent behavior has certain tools available to it that most other settings do not, especially due to the wildly constrained option tree and the clearly known ruleset.
Simulated Annealing can figure shit out on its own, to be plain. Never forget that annealing can beat you up and steal your lunch money without anyone teaching it a damned thing.
Actually applying what Sklansky writes takes a lot of knowledge of the game. You have to be able to recognize betting patterns, calculate pot odds on the fly, accurately estimate your implied odds, put your opponents on ranges of hands, and many other things. All in real-time.
Some of those are things computers are good at. Many of them are not.:)
In my estimation none of those things are particularly difficult for people with a good knowledge of agent behavior techniques. In fact, again in my estimation, the only thing in that list that's even remotely difficult is to recognize betting patterns, and to recognize patternality trends is an extremely well documented and explored field in agent action.
First of all, I'll be amazed when they ever come up with the technology to play no limit hold 'em. That would be a miracle program.
Why? TNLH is not terribly statistically complex, essentially no different than normal poker, and the amount of randomness involved is significantly higher than typical Poker, which is an advantage against most average players.
The reason bots are so slow on the uptake is that people with real agent experience are still playing Chess, Go, Shogi and Han. Poker would be easy to exploit with a screen scraper; a friend of mine with a serious online addiction tried to talk me into it several years ago. (Whether I'm ethical or a pussy is a matter for debate, but for one of those reasons I never did it.)
Chess bots were a mild problem back when I worked at USCF/GamesParlor. Yes, the stakes are higher here, primarily because they, um, exist. And yes, poker bots will eventually get smart. This generation isn't. Say what you will; the only reason I'm not rattling off a laundry list of potential improvements is that I find the practice disgusting and don't want to fuel it.
Just as important are reading your opponent, making bets that damage others pot odds, and playing your position in relation to the blinds.
The first of those three claims is the only of those three that's difficult to code for, and there are many poker afficionados who I've heard say "play your hand, not your opponents." I'm no poker expert; maybe I'm listening to the wrong people. That said, it makes good sense to someone like me that no matter what fidgeting or blinking you see in chat or in people pausing, that the rules are hinged solely on the cards, and that the theoretically perfect model would not benefit from any attempt to read one's opponents. (Besides, how often have you hung out with your poker buddy online, and seen them pause because they dropped something, because their dog started barking, or because they were trying to psych their opponents out?)
Plus, there's just a certain amount of feel needed in the game. Even Doyle Brunson claims ESP is important in Super System.
There are people that claim the same thing of chess and go. Chess fell; Go's due to fall in something like six years (yeah yeah, I know, people make bad claims all the time; I stand by my estimation.)
ESP, "feel," estimation, knowledge, those are all well and good, potentially critical to human modes of play. Just because you rely on them doesn't mean there aren't alternative approaches. Just because the alternative seems like it'd be too hard for a human doesn't mean a gigantic calculator with enough RAM to wish your mom a happy birthday couldn't pull it off.
Yes, to watch ESPN2, there's a lot of reading and poker facing that goes on during poker; the prevalence of that phrase bespeaks its importance deeply. That said, card games are typically mathematically very simple, and approximations can often constrain them successfully (people's "systems.") Modern computers do not need to approximate. Mathematically speaking, computers can easily handle five ply realtime, and I don't know of a poker game that reaches deeper than five ply.
To be clear, computers are now so powerful that they can look ahead to the degree to choose whether or not to draw then fold on a losing hand based solely on how one less card in the deck changes card statistics for the following hand.
Limit ring games are a different ballgame, and a bot does have some chance of success.
It would be kind of you to explain what a limit ring is, for those of us who don't actively follow poker terminology.
Any mid to high stakes game has players who will quickly figure out the way a bot plays, and milk it for all it's worth.
You direly underestimate the level of complexity a bot can bring to play. I've written several game playing architectures (I'm a game designer,) many of whom know how to bluff and how to keep track of the way "they've bee
Oh, I'm certainly not defending the problem; you're quite right to suggest what an issue it is. I just tend to believe that the problem comes from the programmers, rather than the language.
Except that for years PHP and MySQL, tools marketed to beginners, didn't have the "correct" option available.
And beginners won't know to ask about it.
Eh. There's no way to give amateur developers the ability to produce professional code. If you want a secure system, don't hire a novice. It's not PHP's fault that a developer is a failure, to be quite blunt.
And, once again, there is NO excuse for building a network-aware technology that allows for setting variables from the URI query string. None. Even PHP's predecessors had better sense.
Actually, in PHP and PHP/FI days, that was common practice, and in some languages still is.
I don't have to. That we lack them makes my point, no matter how you look at it.
Fallacy: logical error.
Bzzt. Read it again.
And your link is useless, to boot.
Perhaps, if you spend so little time reading it (or, I suspect, none at all) that you can still make the mistake above.
Religious nut: no, actually, that's poisoning the well, by attributing an evil purpose to my method in order to attack the method.
Straw man by fallacy. I did no such thing. Show me a quote if you want to make an accusation like that; it did not happen. Besides, there's no "method" at hand here; it's a question of disagreeing with what you said, and explaining by allegory. That you want to cast it into something it isn't is a clear exposition of desperation. This is a none too subtle form of ad Hominem.
Reversal: I specifically deny perfection, so your notion that statistics trump perfection really only serves my point.
Saying you deny it carries none of the weight of actually making your case. You've yet to do that. Slandering someone for pointing out errors in your original statement does nothing to obviate those errors. This is textbook red herring.
A friend of mine calls this "rephrase and riposte"
I'd call this ad Hominem, except your friend is neither an authority nor in fact particularly interesting. Why are you so desperate to avoid answering the questions I asked you?
Evidence: no, actually, we don't have evidence of this happening hundreds of thousands of times.
Ignorance of data is not absence of data.
Finding theories that explain a lot, seem plausible, fit the data, and aren't ugly hacks is all well and good -- but the data should not be equated with the theory.
Yes yes, you keep handwaving about these pseudoideals of argument. Nonetheless, the errors I pointed out in your arguments still exist, and you've now through two replies managed to get angrier and angrier without actually taking time to address the questions asked you.
Dodge it all you want; your argument is still flawed, and ignoring the holes doesn't make them go away.
Citations: this one's funny. It's your first citation of any project name, and you're whining that I'm not providing citations?
Wait, let me get this straight. You make an unfounded argument. I make a founded reply. You proceed to complain that I don't defend my statements? Tu quoque, except of course not actually tu quoque. I've backed up several points, not one as you claim. You've backed up none, and dodged every single concern addressed you.
This is nothing more than a thin veneer and an attempt at spin. If you won't answer the questions asked you, stop replying. This is meant to be a discussion board, not a soapbox.
Ooh, and a straw man, attributing to me the accusation that these methods are incapable of producing useful results, too.
That's not what a straw man is, and I said no such thing. Please stop making accusations without quotes. You've made nothing but groundless personal attacks. If you want to be taken seriously, don't say "oh there's a straw man," point it out with a quote, and take the time to explain it.
You'll find the reason that you've managed to avoid that three messages in a row, despite that I've already noted the behavior, and the reason that you'll almost certainly do it in the upcoming response, is that you don't actually have specific concerns to address. It's a bunch of hot air. You like throwing around the names of fallacies you don't understand because it makes you feel smart.
In the future, if you want to pretend there's a fallacy, have the decency to quote it.
Yes, actually, I have read some of Darwin's work
I don't believe you.
no, not every word of it.
I asked you about one specific work, not "all of it." Don't pretend th
You're a little late, but you needn't make up for it with rudeness.
I find it quite cathartic to speak down to people who are speaking down to others from platforms of ignorance. If you dislike that, you have three options: put me in my place, admit what I believe about your comprehension, or don't reply.
No. Theory only predicts that there is likely to be a 'selection' if there is a means for the selection (variation) and a reason for the selection (preference). An organ may be useless, but if there's no variation, or if its cost is minimal compared to other selection factors, then selection isn't expected to occur.
In fact, that was exactly my point, hinged on the fairly obvious contention that maintaining a five pound slab of meat is very expensive in terms of food and immunological cost.
No. We don't lack gills simply because they take energy to grow and therefore children without gills somehow survive more easily.
Oh really? Because we used to have them. Where'd they go? And why, if not because maintaining two oxygen blood exchange systems is fantastically expensive? (Hint: after the brain, the skin and the liver, the lungs are the most expensive organs to maintain. Lungs and lung-analogs will be the fourth to go.)
You fall prey to the same fallacy as everyone else, assuming that because an organ is useless in your eyes, it will be useless in the virtual eye of natural selection.
2) That's one of the critical underpinnings of natural selection - maintenance cost as an environmental stress. You brought natural selection up, not me. Now you want to say it's off of the point?
If it did come with guarantees, that might almost be proof of a Creator, who would have designed the process with a goal in mind.
Yes yes, trying to make me out to seem like a religious nut (that's called a Straw Man, by the way) is cute and all, but I never said or even implied anything about perfection. I was quite clear to couch my predictions in terms of statistical significance. If you don't know statistics, you may not understand what that means, but it's quite poison to this notion of perfection of yours. It's a big dice game, nothing more. That said, over the course of the fossil record, and recently in life, we have evidence of this sort of thing happening hundreds of thousands of times.
If you evolve AIs in a lab, and then have them meet real-world gamers, you're introducing an entirely new, foreign constraint to them.
Look, I don't know where you're getting this stuff. A human can't give a chess game any different inputs than another chess game can. It's chess rules, pure and simple. I'm sure you're trying to make some point that the way a human plays chess is somehow significantly different an input, but that's just baloney. That's like suggesting that the vicar's wife method is significantly different than the bingo cage method for picking random numbers. It's malarky.
You can go ahead and wave your hands all you want, but projects like Blondie24 have shown several times that annealing generates admirable opponents even when trained against its own. You can insist there are subtle qualitative differences, but experience suggests otherwise, so I'll wait for your examples until I give any credence to your appeals.
Give me citations, not assertions.
Furthermore, if you try to hurry things up in the lab by keeping only the top z% of the agents (cream of the cream), it's entirely possible that your population will skew in a random direction -- butterfly effect, essentially.
Only in the complete absence of an evolutionary pressure. Have you ever even done one of these systems? And why didn't you answer when I asked you if you'd read Darwin's books?
Without continuous outside testing and some way of rectifying these anomalies, you may in the end have AIs that are
No, dear. Let me hold your hand through what I thought was a relatively simple example.
For a 3x3 grid, the travelling salesman problem can be solved quite quickly through brute force; for 4x4, still so but less so. The output from brute force will be correct. However, you don't want to brute force a 500x500 TS problem, even though the output will be correct. Perhaps now you'll understand the bit about there being more to it than output? Time matters.
True artificial intelligence is only ten years away - and has been for the past three decades.
You know, it's funny how often people make smarmy comments like that. What respectable AI researcher have you ever, even once, seen make the claim that AI is only ten years away?
Someone said that to Bert Dreyfuss at a conference at the Univ. Pittsburgh once. He had them thrown out of the room. Why? Because AI is only a "disappointment" to people that invent goals that the researchers never claimed, then watch those invented goals fail, then claim that AI is useless.
Did you know that your car has more than a dozen components modelled on techniques that derive from modern AI research? Those antilock brakes, for example. If you have a Ford, the interior shape of your combustion engine was determined with a genetic algorithm. If you have a GM, it was determined with simulated annealing.
Just because you're ignorant of the goals and progress of a field doesn't mean there aren't goals or progress. The failure is not AI as a field, but you as a journalist.
Most 'AI' problems that have been solved have been solved via brute force combined with the advance of Moore's law.
This is simple ignorance. No amount of processing horsepower has any impact on any aspect of AI research, nor has it since the 70s. An Apple ][ is enough for every technique in modern use. Stop pretending to know things you don't actually know.
From what I've seen of game 'AI'
By context I suspect that's virtually nothing. Chances are that you'd fail a context-limited Turing test with Smaug. AI tools are in common use identifying 1-800 numbers and giving directions to movie theaters, booking tickets, finding routes through traffic for Acura customers, designing industrial tools and vehicles, finding new plastic composites for your bike, all sorts of crap.
This is a little like saying electric motors could never be used to make a vehicle fleet, why look at that HotWheels. Video game simulated agents are about ten miles behind the rear guard in AI techniques. Q*Bert does not make you a researcher.
That's my 2 cents - flame away.
My, my, how wise you must be, to slander anyone who replies to you before they've even replied. I wonder if that's because you know you're full of crap, or if you're just so universally unable to restrict your own statements to knowledge that you actually believe people react to one another the way they react to you?
A Neural Network stops learning at some point, but it's still AI right?
No. It never was and it never will be. Neural Networks are special-case massively parallel adaptive filters. They're interesting research tools with broad and powerful implications.
This is why I get so annoyed when scientists (and creationists) ask "what is this organ useful for?" expecting that every animal's every organ is entirely well-suited to its environment -- because either evolution or the hand of god made it perfect.
There's a difference between observing that something is imperfect and being aware that ancilla have a cost. The human appendix is expensive, and unless it kicks in and learns to do something, it'll be gone in 10,000 natural generations (that is, after the collapse of modern medicine, so we can get back to natural selection.)
That's not what the theory(!) of evolution predicts.
Why is it that people try to play semantic games like this? There's no such thing as the theory of evolution. It's called the theory of natural selection, and yes, it most certainly predicts that over a statistically significant amount of time, useless organs will be discarded. That's why we don't have gills, tails or fur anymore: they take energy and resources to make, so children that don't have them get by on less.
Just because we haven't finished discarding the appendix yet doesn't mean we won't. Try to take the long view.
Weird results from genetic algorithms are even more likely in small-population scenarios like games.
Oh, read a book. Genetic algorithms are a design step, not a runtime step. I use simulated annealing, a similar competing technique, to generate players; I've run through almost two and a half million networks for my chess game, and I expect that training batch is around one third done. "Small-population scenarios" are the definition of niche ecologies, which are well understood and recorded to be the statistical genesis of most new species. We're even watching it happen in a butterfly right now.
Have you even read Darwin's work, or are you just arguing something you believe you've absorbed by contact from the rest of the well meaning but ill-educated masses?
Metamoderate all modders to parent down. You're only supposed to use insightful when you understand what the post is talking about and can verify its veracity. This guy thinks genetic algorithms get tested one at a time against humans.
This distinction is one of several reasons it's so frustrating that people refuse to differentiate between AI and simulated agent behavior. There isn't any AI on Earth anywhere (except in deep dark government lockdown vaults maybe.) What drives your video game is called a simulated agent.
Man, this would be a great argument if the brain was analog, instead of time-pulse domain like it actually is.
Reminds me of a guy who used to tell athletes to not drink water before an event, because water was very heavy, so it'd weigh you down and you'd get tired faster. It's very easy to make superficially compelling arguments when you don't have any idea what you're talking about.
Uh, Microsoft Foundation Classes, Smalltalk, OWL/VCL/CLX and REXX are also model view controller. The "promise" of MVC has been well understood for decades.
Wait, you're lambasting someone who thinks the most important feature of a language is how it denotes its blocks for not grokking a language? Isn't that like berating your dog for insufficient calculus skills?
Another point: Your "(((((this)))))" example is a straw man argument. Lisp programs use white space and indentation to make the structure clear, just like any other programming language.
The phrase you're probably looking for is "red herring," though in fact it's also not a red herring. A straw man argument is where you argue against what you want a person to be saying rather than what they're actually saying. For example, consider a debate in a fictional parent teacher association, where John advocates private security and Dan advocates real police officers:
John: I think we just don't need officers trained to deal with murderers and rapists in our school system.
Dan: I can't believe you want to leave our children unprotected!
This is a straw man because John does not actually want to leave the children unprotected; he simply disagrees with Dan's choice of form of protection.
The thing I believe you're looking for, a red herring, is what people generally say when they mean "an argument which is not germane." That's not actually what a Red Herring is, though the difference is subtle.
A red herring is an argument which is not germane, which was injected specifically for the purpose of diverting attention from the issue at hand. It's not a red herring if it's just unimportant; it has to be specifically used intentionally to derail the real argument. Arguments which end in "and isn't that what we all wanted anyway" are frequently red herrings.
So the ways Javascript differs from C are due to being LISP like? I find it hard to imagine a more damning indictment of a language.
Then you've never used Intercal, which has the expressiveness of cobol, the readability of k, the write-once-read-many of perl cubed, the speed of logo and the linkage of apple basic. And then there are Ook, befunge, kvakkiutl, and that damn bear-oriented language I can't find anymore.
That said, you're not the first person I've heard say that all languages are, ultimately, Lisp machines, and there's some merit to that.
Pfft. Same merit there is in looking at them as Turing Machines, Babbage Machines, piles of assembly or a physics project involved in the correspondance of capacitance.
Just because something can be expressed in terms of an X doesn't make it an X. That's why the universe isn't a turing machine, no matter how many times you watch The Matrix - first, you have to find evidence of the tape. You try writing your C programs inside a LISP machine, and you're in for a dry shock regarding logical equivalence and program equivalence.
There's a lot more to a program than the results it outputs. That's why nobody pays attention to, say, iterative Travelling Salesman machines, even though there are several running; they're never going to finish in your lifetime.
Similarly, if you write C programs in a LISP machine, they're not going to finish in your lifetime. So who cares?
Is it about "*aweful*" syntax, or awful spelling? Or are you full of awe?
Well as long as you're playing dress-up with your daddy's grammar cop outfit, let me show you how the safety on that gun works. Awful means "filled with awe." God is awful. Awful does not mean bad, and awe does not mean good. "God is awful" means "omg god is scary."
That's why when you say "huhu i hate when people say awful smart" educated people within earshot sigh and shake their heads.
Lisp syntax is excellent because it's simple and consistent, and that's the reason Lisp macros are so powerful. Perl syntax is absolutely awful, and that's the reason Perl will never have macros like Lisp.
Yeah, and that's why LISP killed Perl.
Don't bother calling me a Perl fanatic; I barely use it. I'm just a pragmatist. You guys can swear by a syntax you're just used to until you're blue in the face; LISP has a bad syntax whether you like it or not. It's simple, consistent, bulky, hard to read, and so famously a source of parse bugs based on a visual hanoi's tower of brackets that there's in fact an entire class of jokes revolving around its baditude.
Next, tell me how simple and elegant your double bucky keyboard is, or how you didn't have to go through X-Ray machines to use the ornithopters growing up. Or how lunch cost a nickel and how walking uphill both ways in bare feet builds character. If LISP is so simple, elegant and powerful, and since it's one of the very first programming languages, since it's had a free implementation since about two days before it was invented, then riddle me this, quiz master: why is it so dead?
There's nothing worse than listening to velociraptors opine the importance of the 140-foot-tall predator. Move on, Jurassic Jim. The rest of the world is tired of calling your museum number.
And you know, if being able to identify the trailing end of a block out of context was a common job in programming, that might be a serious issue. Similarly, you might well claim that MathML is a better programming language than C because it provides better control of vertical positioning of complex algorithms in flow than C's flat text, and what if the screen was too small to see the whole alg?
Back in the real world, the answer to your question is simple.
Now tell me what it means. Specifically, tell me what expression it ends.
The one that opened the parenthesis 20 or 30 letters ago.
Let me show you how XML fails it given your criterion of design, from the perspective of an application developer. (People will tell you LISP is a data language, because it's for list processing, but that's nonsense: I've never seen a document that uses LISP as markup. LISP is a data-oriented language, sure, but it's application only. Nobody uses LISP for markup.)
It's not nessecarily about markup tags or enclosure or whatever; PHP isn't nearly so bulky, and PHP lives inside of processing instructions. It's that XML is too data-oriented; it specifies everything in numbing detail, and it's fantastically verbose.
And, don't tell me this is unrealistic. This is actually tame. Take a look at XSLT, or the boilerplate code C# throws down. In C# Managed, if you use a wizard to generate an app, Hello World is 124 lines.
It's about programmer usability. Examples of the other direction include Erlang, Haskell, K and Forth - they're extremely expressive in a very dense space, and as a result even though they require a lot of learning and effort to use, they're very heavily used by their followers, and there's compelling reason to switch. Compare DTD and RelaxNG some time. They solve the same problem domain with the same technologies, and RelaxNG can do everything DTD can (and more,) yet RelaxNG is far far less painful than DTD.
XML is markup. You wouldn't want to use it for programming. That XML has simple parsing rules regarding nesting isn't material; most languages are relatively easy to lex, and lexers and parsers are automatically generatable if you have a grammar (say, an LALR or an SLR grammar, or an exotic.)
A language should not be chosen for the difficulty involved in parsing it; we can almost parse English ffs, and there's nothing less parseable than English. You can get signal from radioactive decay faster than you can parse English. Parsing just isn't a germane concern.
Worry about expressiveness and ease of threat. That's how you'll end up with CLEAN and K and s-expressions instead of XSLT and PL/1 and Schema.
Slashdot Mirror
For the same reason the people that make $70,000 a month part time with their home business have infomercials: because what they're really making money from is convincing people to buy into some scheme of theirs.
It's invariably the case that even when these systems really do work (a few of them, particularly the real estate redevelopment models, actually do work,) selling them to millions of people is significantly more profitable than actually running them.
It's called franchising. Pizza Hut doesn't own all its stores; neither do McDonalds, Wendys, and so on. Why? The stores are immensely profitable.
The answer is simple: reaping margin on people investing their own money is far more profitable than doing it all yourself, because the growth rate is far larger, and you step back from all of the personal risk. If they're profitable, they break you off a proper chunk. If they fail, well, it's just not your problem.
Be honest: would you rather own five fast food resteraunts, or reap a percentage on two hundred? The franchise model is so successful that better granchises, like fast food, often have smaller companies subfranchising. Next time you go to a Burger King or whatever, take a look at the drive through window, where they have the complaint phone number. Surprisingly often, it'll list a smaller company you've never heard of, but which owns a few hundred local fast food places from a variety of name brands. In Pittsburgh, most of the fast food in the college district, regardless of who the sign reads as, is owned by either of two companies (to the tune of some 100 or so storefronts covering a good two dozen brands.)
In short, selling a scalable business model is usually more profitable than using it.
There was just an article several days ago about a standardized CAPTCHA defeater toolkit. However, whereas the principle of turing testing may be sound, it's not enough.
Consider the case of the virtual money sweatshops for MMORPGs. They have people whose job is to sit and do boring repetitive things to exploit games to generate things which can be traded for real money.
Now, consider the scenario where poker bots become significantly profitable. You will quickly see sweatshops where it's someone's job to sit there answering CAPTCHA all day to allow the bots to continue to reap.
Communicating its hand to second instances of the bot. Would you let your poker buddies share hands during play?
Besides, whereas cheating may not be the best word for the practice of putting machines into human tournaments, it's still not exactly fair; you wouldn't let me into the olympic several mile race riding a motorcycle.
But I predict that that will be a self-correcting problem.
Poker has been running for 400 years, and the problem has gotten worse, not better.
Since after all, it won't be long before someone makes an alternate version of the BOT that feeds incorrect data to the other BOTs so that you're more likely to win money from them.
Er, that's a bit naive. Presume for a moment that I'm a cheater. I'm running three bots which are colluding. What makes you think:
It's not like there's a mesh of arbitrary cheaters out there trusting each other to cheat for the greater good of cheaters everywhere. They're intensely self focussed.
Anyway, as long as the BOTs aren't actually hacking the system, or forcing other peoples' clients to crash, etc, then I think you could make a good argument that it's not really cheating.
Oh sure, breaking the rules by sharing hands, that's not cheating at all. (cough) I'd love to see the described good argument.
And no, I'm not a bot user or apologist.
Well, maybe you're at least not a bot user, but you're attempting to make the active claim that sharing information which by the rules of the game is not public is not cheating, so that second half is a bit more than just debatable.
statistically speaking... about 90% of them are chumps, yes. The other 10% win.
It's more like 97.2% vs 2.8%.
In conclusion: stop talkin about things you don't understand, kiddo. I have no doubt that you lose at online poker, but the problem isn't that you're getting cheated.
You seem knowledgeable; it's a pity you've left no method to be contacted. Given that you claim to be making real money playing poker, I'd like to talk to you about game design. (Besides, we have shockingly similar fan and foe lists; that generally means similar interests.)
Do you really believe that the operators of these on-line "casinos" are above playing poker against you while they can watch your hands, or when they can tell the computer what to deal next?
Yes.
Casino operators have the fundamental quality of mathematicians: they understand that 1% gets very big when run over very big numbers. The amount of money they would make playing personally is essentially paltry compared to the amount of money their site makes. The risk of being discovered cheating, which isn't terribly unlikely considering the US Government's habit of raiding online casinos, is immense; even if they didn't get shut down, their play population would dwindle to nearly nothing overnight.
In short, it would be foolish to risk a machine making a million dollars a week, in order to score $200 a month against some asshats. I just don't think any of the system operators are that stupid. Even if they're compulsive gamblers, they're playing on other people's systems, just to be safe.
And while dealing themselves the good cards too often might be caught by statistical analysis of the decks
You would do well to read about the cryptanalysts in World War 2, who almost wouldn't act on the broken Purple (a Japanese navy code,) because they knew how to detect the enemy acting on information they shouldn't have, and therefore knew the enemy could almost certainly do the same thing.
It's not that hard to detect a cheater, even if it's only knowledge cheating, and the more successful the cheater is, the easier that detection gets.
Another form of cheating that I know is going on (because I know someone who admits to doing it) is to play multiple hands in the same game against another player and share information about your hands.
That would be the team play issue being discussed of the bots.
I find it unfortunate that the person with the clue to bring something like that to bear posted anonymously; I would like to be in contact with someone like that. To exploit too-perfect play is a hell of an interesting concept.
That said, this still makes the tenuous presupposition that the bot author is unaware of said approach. Whereas one might consider that in terms of the avant garde - similar in many ways to virus authors being just barely ahead of viruses - it is important to remember that agent behavior has certain tools available to it that most other settings do not, especially due to the wildly constrained option tree and the clearly known ruleset.
Simulated Annealing can figure shit out on its own, to be plain. Never forget that annealing can beat you up and steal your lunch money without anyone teaching it a damned thing.
Actually applying what Sklansky writes takes a lot of knowledge of the game. You have to be able to recognize betting patterns, calculate pot odds on the fly, accurately estimate your implied odds, put your opponents on ranges of hands, and many other things. All in real-time.
:)
Some of those are things computers are good at. Many of them are not.
In my estimation none of those things are particularly difficult for people with a good knowledge of agent behavior techniques. In fact, again in my estimation, the only thing in that list that's even remotely difficult is to recognize betting patterns, and to recognize patternality trends is an extremely well documented and explored field in agent action.
First of all, I'll be amazed when they ever come up with the technology to play no limit hold 'em. That would be a miracle program.
Why? TNLH is not terribly statistically complex, essentially no different than normal poker, and the amount of randomness involved is significantly higher than typical Poker, which is an advantage against most average players.
The reason bots are so slow on the uptake is that people with real agent experience are still playing Chess, Go, Shogi and Han. Poker would be easy to exploit with a screen scraper; a friend of mine with a serious online addiction tried to talk me into it several years ago. (Whether I'm ethical or a pussy is a matter for debate, but for one of those reasons I never did it.)
Chess bots were a mild problem back when I worked at USCF/GamesParlor. Yes, the stakes are higher here, primarily because they, um, exist. And yes, poker bots will eventually get smart. This generation isn't. Say what you will; the only reason I'm not rattling off a laundry list of potential improvements is that I find the practice disgusting and don't want to fuel it.
Just as important are reading your opponent, making bets that damage others pot odds, and playing your position in relation to the blinds.
The first of those three claims is the only of those three that's difficult to code for, and there are many poker afficionados who I've heard say "play your hand, not your opponents." I'm no poker expert; maybe I'm listening to the wrong people. That said, it makes good sense to someone like me that no matter what fidgeting or blinking you see in chat or in people pausing, that the rules are hinged solely on the cards, and that the theoretically perfect model would not benefit from any attempt to read one's opponents. (Besides, how often have you hung out with your poker buddy online, and seen them pause because they dropped something, because their dog started barking, or because they were trying to psych their opponents out?)
Plus, there's just a certain amount of feel needed in the game. Even Doyle Brunson claims ESP is important in Super System.
There are people that claim the same thing of chess and go. Chess fell; Go's due to fall in something like six years (yeah yeah, I know, people make bad claims all the time; I stand by my estimation.)
ESP, "feel," estimation, knowledge, those are all well and good, potentially critical to human modes of play. Just because you rely on them doesn't mean there aren't alternative approaches. Just because the alternative seems like it'd be too hard for a human doesn't mean a gigantic calculator with enough RAM to wish your mom a happy birthday couldn't pull it off.
Yes, to watch ESPN2, there's a lot of reading and poker facing that goes on during poker; the prevalence of that phrase bespeaks its importance deeply. That said, card games are typically mathematically very simple, and approximations can often constrain them successfully (people's "systems.") Modern computers do not need to approximate. Mathematically speaking, computers can easily handle five ply realtime, and I don't know of a poker game that reaches deeper than five ply.
To be clear, computers are now so powerful that they can look ahead to the degree to choose whether or not to draw then fold on a losing hand based solely on how one less card in the deck changes card statistics for the following hand.
Limit ring games are a different ballgame, and a bot does have some chance of success.
It would be kind of you to explain what a limit ring is, for those of us who don't actively follow poker terminology.
Any mid to high stakes game has players who will quickly figure out the way a bot plays, and milk it for all it's worth.
You direly underestimate the level of complexity a bot can bring to play. I've written several game playing architectures (I'm a game designer,) many of whom know how to bluff and how to keep track of the way "they've bee
Oh, I'm certainly not defending the problem; you're quite right to suggest what an issue it is. I just tend to believe that the problem comes from the programmers, rather than the language.
Except that for years PHP and MySQL, tools marketed to beginners, didn't have the "correct" option available.
And beginners won't know to ask about it.
Eh. There's no way to give amateur developers the ability to produce professional code. If you want a secure system, don't hire a novice. It's not PHP's fault that a developer is a failure, to be quite blunt.
And, once again, there is NO excuse for building a network-aware technology that allows for setting variables from the URI query string. None. Even PHP's predecessors had better sense.
Actually, in PHP and PHP/FI days, that was common practice, and in some languages still is.
Gills: you never proved you knew why we lack them
I don't have to. That we lack them makes my point, no matter how you look at it.
Fallacy: logical error.
Bzzt. Read it again.
And your link is useless, to boot.
Perhaps, if you spend so little time reading it (or, I suspect, none at all) that you can still make the mistake above.
Religious nut: no, actually, that's poisoning the well, by attributing an evil purpose to my method in order to attack the method.
Straw man by fallacy. I did no such thing. Show me a quote if you want to make an accusation like that; it did not happen. Besides, there's no "method" at hand here; it's a question of disagreeing with what you said, and explaining by allegory. That you want to cast it into something it isn't is a clear exposition of desperation. This is a none too subtle form of ad Hominem.
Reversal: I specifically deny perfection, so your notion that statistics trump perfection really only serves my point.
Saying you deny it carries none of the weight of actually making your case. You've yet to do that. Slandering someone for pointing out errors in your original statement does nothing to obviate those errors. This is textbook red herring.
A friend of mine calls this "rephrase and riposte"
I'd call this ad Hominem, except your friend is neither an authority nor in fact particularly interesting. Why are you so desperate to avoid answering the questions I asked you?
Evidence: no, actually, we don't have evidence of this happening hundreds of thousands of times.
Ignorance of data is not absence of data.
Finding theories that explain a lot, seem plausible, fit the data, and aren't ugly hacks is all well and good -- but the data should not be equated with the theory.
Yes yes, you keep handwaving about these pseudoideals of argument. Nonetheless, the errors I pointed out in your arguments still exist, and you've now through two replies managed to get angrier and angrier without actually taking time to address the questions asked you.
Dodge it all you want; your argument is still flawed, and ignoring the holes doesn't make them go away.
Citations: this one's funny. It's your first citation of any project name, and you're whining that I'm not providing citations?
Wait, let me get this straight. You make an unfounded argument. I make a founded reply. You proceed to complain that I don't defend my statements? Tu quoque, except of course not actually tu quoque. I've backed up several points, not one as you claim. You've backed up none, and dodged every single concern addressed you.
This is nothing more than a thin veneer and an attempt at spin. If you won't answer the questions asked you, stop replying. This is meant to be a discussion board, not a soapbox.
Ooh, and a straw man, attributing to me the accusation that these methods are incapable of producing useful results, too.
That's not what a straw man is, and I said no such thing. Please stop making accusations without quotes. You've made nothing but groundless personal attacks. If you want to be taken seriously, don't say "oh there's a straw man," point it out with a quote, and take the time to explain it.
You'll find the reason that you've managed to avoid that three messages in a row, despite that I've already noted the behavior, and the reason that you'll almost certainly do it in the upcoming response, is that you don't actually have specific concerns to address. It's a bunch of hot air. You like throwing around the names of fallacies you don't understand because it makes you feel smart.
In the future, if you want to pretend there's a fallacy, have the decency to quote it.
Yes, actually, I have read some of Darwin's work
I don't believe you.
no, not every word of it.
I asked you about one specific work, not "all of it." Don't pretend th
You're a little late, but you needn't make up for it with rudeness.
I find it quite cathartic to speak down to people who are speaking down to others from platforms of ignorance. If you dislike that, you have three options: put me in my place, admit what I believe about your comprehension, or don't reply.
No. Theory only predicts that there is likely to be a 'selection' if there is a means for the selection (variation) and a reason for the selection (preference). An organ may be useless, but if there's no variation, or if its cost is minimal compared to other selection factors, then selection isn't expected to occur.
In fact, that was exactly my point, hinged on the fairly obvious contention that maintaining a five pound slab of meat is very expensive in terms of food and immunological cost.
No. We don't lack gills simply because they take energy to grow and therefore children without gills somehow survive more easily.
Oh really? Because we used to have them. Where'd they go? And why, if not because maintaining two oxygen blood exchange systems is fantastically expensive? (Hint: after the brain, the skin and the liver, the lungs are the most expensive organs to maintain. Lungs and lung-analogs will be the fourth to go.)
You fall prey to the same fallacy as everyone else, assuming that because an organ is useless in your eyes, it will be useless in the virtual eye of natural selection.
1) That's not what a fallacy is.
2) That's one of the critical underpinnings of natural selection - maintenance cost as an environmental stress. You brought natural selection up, not me. Now you want to say it's off of the point?
If it did come with guarantees, that might almost be proof of a Creator, who would have designed the process with a goal in mind.
Yes yes, trying to make me out to seem like a religious nut (that's called a Straw Man, by the way) is cute and all, but I never said or even implied anything about perfection. I was quite clear to couch my predictions in terms of statistical significance. If you don't know statistics, you may not understand what that means, but it's quite poison to this notion of perfection of yours. It's a big dice game, nothing more. That said, over the course of the fossil record, and recently in life, we have evidence of this sort of thing happening hundreds of thousands of times.
If you evolve AIs in a lab, and then have them meet real-world gamers, you're introducing an entirely new, foreign constraint to them.
Look, I don't know where you're getting this stuff. A human can't give a chess game any different inputs than another chess game can. It's chess rules, pure and simple. I'm sure you're trying to make some point that the way a human plays chess is somehow significantly different an input, but that's just baloney. That's like suggesting that the vicar's wife method is significantly different than the bingo cage method for picking random numbers. It's malarky.
You can go ahead and wave your hands all you want, but projects like Blondie24 have shown several times that annealing generates admirable opponents even when trained against its own. You can insist there are subtle qualitative differences, but experience suggests otherwise, so I'll wait for your examples until I give any credence to your appeals.
Give me citations, not assertions.
Furthermore, if you try to hurry things up in the lab by keeping only the top z% of the agents (cream of the cream), it's entirely possible that your population will skew in a random direction -- butterfly effect, essentially.
Only in the complete absence of an evolutionary pressure. Have you ever even done one of these systems? And why didn't you answer when I asked you if you'd read Darwin's books?
Without continuous outside testing and some way of rectifying these anomalies, you may in the end have AIs that are
For the same reason that spark plugs are taught in Mechanic school, even though spark plugs aren't cars.
Though I will accept the adage that "Once an AI problem is solved it's no longer AI."
I wouldn't.
No, dear. Let me hold your hand through what I thought was a relatively simple example.
For a 3x3 grid, the travelling salesman problem can be solved quite quickly through brute force; for 4x4, still so but less so. The output from brute force will be correct. However, you don't want to brute force a 500x500 TS problem, even though the output will be correct. Perhaps now you'll understand the bit about there being more to it than output? Time matters.
True artificial intelligence is only ten years away - and has been for the past three decades.
You know, it's funny how often people make smarmy comments like that. What respectable AI researcher have you ever, even once, seen make the claim that AI is only ten years away?
Someone said that to Bert Dreyfuss at a conference at the Univ. Pittsburgh once. He had them thrown out of the room. Why? Because AI is only a "disappointment" to people that invent goals that the researchers never claimed, then watch those invented goals fail, then claim that AI is useless.
Did you know that your car has more than a dozen components modelled on techniques that derive from modern AI research? Those antilock brakes, for example. If you have a Ford, the interior shape of your combustion engine was determined with a genetic algorithm. If you have a GM, it was determined with simulated annealing.
Just because you're ignorant of the goals and progress of a field doesn't mean there aren't goals or progress. The failure is not AI as a field, but you as a journalist.
Most 'AI' problems that have been solved have been solved via brute force combined with the advance of Moore's law.
This is simple ignorance. No amount of processing horsepower has any impact on any aspect of AI research, nor has it since the 70s. An Apple ][ is enough for every technique in modern use. Stop pretending to know things you don't actually know.
From what I've seen of game 'AI'
By context I suspect that's virtually nothing. Chances are that you'd fail a context-limited Turing test with Smaug. AI tools are in common use identifying 1-800 numbers and giving directions to movie theaters, booking tickets, finding routes through traffic for Acura customers, designing industrial tools and vehicles, finding new plastic composites for your bike, all sorts of crap.
This is a little like saying electric motors could never be used to make a vehicle fleet, why look at that HotWheels. Video game simulated agents are about ten miles behind the rear guard in AI techniques. Q*Bert does not make you a researcher.
That's my 2 cents - flame away.
My, my, how wise you must be, to slander anyone who replies to you before they've even replied. I wonder if that's because you know you're full of crap, or if you're just so universally unable to restrict your own statements to knowledge that you actually believe people react to one another the way they react to you?
A Neural Network stops learning at some point, but it's still AI right?
No. It never was and it never will be. Neural Networks are special-case massively parallel adaptive filters. They're interesting research tools with broad and powerful implications.
They are also not self-aware.
This is why I get so annoyed when scientists (and creationists) ask "what is this organ useful for?" expecting that every animal's every organ is entirely well-suited to its environment -- because either evolution or the hand of god made it perfect.
There's a difference between observing that something is imperfect and being aware that ancilla have a cost. The human appendix is expensive, and unless it kicks in and learns to do something, it'll be gone in 10,000 natural generations (that is, after the collapse of modern medicine, so we can get back to natural selection.)
That's not what the theory(!) of evolution predicts.
Why is it that people try to play semantic games like this? There's no such thing as the theory of evolution. It's called the theory of natural selection, and yes, it most certainly predicts that over a statistically significant amount of time, useless organs will be discarded. That's why we don't have gills, tails or fur anymore: they take energy and resources to make, so children that don't have them get by on less.
Just because we haven't finished discarding the appendix yet doesn't mean we won't. Try to take the long view.
Weird results from genetic algorithms are even more likely in small-population scenarios like games.
Oh, read a book. Genetic algorithms are a design step, not a runtime step. I use simulated annealing, a similar competing technique, to generate players; I've run through almost two and a half million networks for my chess game, and I expect that training batch is around one third done. "Small-population scenarios" are the definition of niche ecologies, which are well understood and recorded to be the statistical genesis of most new species. We're even watching it happen in a butterfly right now.
Have you even read Darwin's work, or are you just arguing something you believe you've absorbed by contact from the rest of the well meaning but ill-educated masses?
Metamoderate all modders to parent down. You're only supposed to use insightful when you understand what the post is talking about and can verify its veracity. This guy thinks genetic algorithms get tested one at a time against humans.
This distinction is one of several reasons it's so frustrating that people refuse to differentiate between AI and simulated agent behavior. There isn't any AI on Earth anywhere (except in deep dark government lockdown vaults maybe.) What drives your video game is called a simulated agent.
Man, this would be a great argument if the brain was analog, instead of time-pulse domain like it actually is.
Reminds me of a guy who used to tell athletes to not drink water before an event, because water was very heavy, so it'd weigh you down and you'd get tired faster. It's very easy to make superficially compelling arguments when you don't have any idea what you're talking about.
In summation: read a book.
Uh, Microsoft Foundation Classes, Smalltalk, OWL/VCL/CLX and REXX are also model view controller. The "promise" of MVC has been well understood for decades.
Another point: Your "(((((this)))))" example is a straw man argument. Lisp programs use white space and indentation to make the structure clear, just like any other programming language.
The phrase you're probably looking for is "red herring," though in fact it's also not a red herring. A straw man argument is where you argue against what you want a person to be saying rather than what they're actually saying. For example, consider a debate in a fictional parent teacher association, where John advocates private security and Dan advocates real police officers:
This is a straw man because John does not actually want to leave the children unprotected; he simply disagrees with Dan's choice of form of protection.
The thing I believe you're looking for, a red herring, is what people generally say when they mean "an argument which is not germane." That's not actually what a Red Herring is, though the difference is subtle.
A red herring is an argument which is not germane, which was injected specifically for the purpose of diverting attention from the issue at hand. It's not a red herring if it's just unimportant; it has to be specifically used intentionally to derail the real argument. Arguments which end in "and isn't that what we all wanted anyway" are frequently red herrings.
So the ways Javascript differs from C are due to being LISP like? I find it hard to imagine a more damning indictment of a language.
Then you've never used Intercal, which has the expressiveness of cobol, the readability of k, the write-once-read-many of perl cubed, the speed of logo and the linkage of apple basic. And then there are Ook, befunge, kvakkiutl, and that damn bear-oriented language I can't find anymore.
That said, you're not the first person I've heard say that all languages are, ultimately, Lisp machines, and there's some merit to that.
Pfft. Same merit there is in looking at them as Turing Machines, Babbage Machines, piles of assembly or a physics project involved in the correspondance of capacitance.
Just because something can be expressed in terms of an X doesn't make it an X. That's why the universe isn't a turing machine, no matter how many times you watch The Matrix - first, you have to find evidence of the tape. You try writing your C programs inside a LISP machine, and you're in for a dry shock regarding logical equivalence and program equivalence.
There's a lot more to a program than the results it outputs. That's why nobody pays attention to, say, iterative Travelling Salesman machines, even though there are several running; they're never going to finish in your lifetime.
Similarly, if you write C programs in a LISP machine, they're not going to finish in your lifetime. So who cares?
Is it about "*aweful*" syntax, or awful spelling? Or are you full of awe?
Well as long as you're playing dress-up with your daddy's grammar cop outfit, let me show you how the safety on that gun works. Awful means "filled with awe." God is awful. Awful does not mean bad, and awe does not mean good. "God is awful" means "omg god is scary."
That's why when you say "huhu i hate when people say awful smart" educated people within earshot sigh and shake their heads.
Lisp syntax is excellent because it's simple and consistent, and that's the reason Lisp macros are so powerful. Perl syntax is absolutely awful, and that's the reason Perl will never have macros like Lisp.
Yeah, and that's why LISP killed Perl.
Don't bother calling me a Perl fanatic; I barely use it. I'm just a pragmatist. You guys can swear by a syntax you're just used to until you're blue in the face; LISP has a bad syntax whether you like it or not. It's simple, consistent, bulky, hard to read, and so famously a source of parse bugs based on a visual hanoi's tower of brackets that there's in fact an entire class of jokes revolving around its baditude.
Next, tell me how simple and elegant your double bucky keyboard is, or how you didn't have to go through X-Ray machines to use the ornithopters growing up. Or how lunch cost a nickel and how walking uphill both ways in bare feet builds character. If LISP is so simple, elegant and powerful, and since it's one of the very first programming languages, since it's had a free implementation since about two days before it was invented, then riddle me this, quiz master: why is it so dead?
There's nothing worse than listening to velociraptors opine the importance of the 140-foot-tall predator. Move on, Jurassic Jim. The rest of the world is tired of calling your museum number.
And you know, if being able to identify the trailing end of a block out of context was a common job in programming, that might be a serious issue. Similarly, you might well claim that MathML is a better programming language than C because it provides better control of vertical positioning of complex algorithms in flow than C's flat text, and what if the screen was too small to see the whole alg?
Back in the real world, the answer to your question is simple.
Now tell me what it means. Specifically, tell me what expression it ends.
The one that opened the parenthesis 20 or 30 letters ago.
Let me show you how XML fails it given your criterion of design, from the perspective of an application developer. (People will tell you LISP is a data language, because it's for list processing, but that's nonsense: I've never seen a document that uses LISP as markup. LISP is a data-oriented language, sure, but it's application only. Nobody uses LISP for markup.)
<program type="x-source/c++" encoding="utf-8" lang="en" xml:lang="en">
<main>
<svc name="std::cout" type="stream" encoding="utf-8" method="xsputn" href="crt@@cout">Hello, World!<svc>
</main>
</program>
It's not nessecarily about markup tags or enclosure or whatever; PHP isn't nearly so bulky, and PHP lives inside of processing instructions. It's that XML is too data-oriented; it specifies everything in numbing detail, and it's fantastically verbose.
And, don't tell me this is unrealistic. This is actually tame. Take a look at XSLT, or the boilerplate code C# throws down. In C# Managed, if you use a wizard to generate an app, Hello World is 124 lines.
It's about programmer usability. Examples of the other direction include Erlang, Haskell, K and Forth - they're extremely expressive in a very dense space, and as a result even though they require a lot of learning and effort to use, they're very heavily used by their followers, and there's compelling reason to switch. Compare DTD and RelaxNG some time. They solve the same problem domain with the same technologies, and RelaxNG can do everything DTD can (and more,) yet RelaxNG is far far less painful than DTD.
XML is markup. You wouldn't want to use it for programming. That XML has simple parsing rules regarding nesting isn't material; most languages are relatively easy to lex, and lexers and parsers are automatically generatable if you have a grammar (say, an LALR or an SLR grammar, or an exotic.)
A language should not be chosen for the difficulty involved in parsing it; we can almost parse English ffs, and there's nothing less parseable than English. You can get signal from radioactive decay faster than you can parse English. Parsing just isn't a germane concern.
Worry about expressiveness and ease of threat. That's how you'll end up with CLEAN and K and s-expressions instead of XSLT and PL/1 and Schema.