Thing about equilibria is that some of them are stable - no matter how far you push them they'll roll back - wile others are metastable - push them a bit too far and crash. Think of a pencil balanced on end... I know this doesn't affect your point, but a pencil balancing on its end would be an example of an unstable, not metastable equilibrium. Pushing the pencil any amount from vertical, not just past a threshold, will make it collapse which is why in practice you always see them fall over (unless you maybe cheat and flatten the point far down enough).
***
But as for the point you did make, we need to keep in mind that even if we don't feel lucky, we should hedge against that unluckiness in the most efficient way possible (carbon tax or cap-and-trade + sequestration) rather than the least efficient way possible (basically, most of the stuff Gore or Stern recommend).
Without reading the article, I can guess it tracks this format pretty closely:
Q: What would {Obama,McCain} do about $TECH_ISSUE? Obama: Emphasises coming up with solution that works for ALL Americans by making impossible tradeoff. Says soundbite taken from Lawrence Lessig. McCain: Emphasises coming up with solution that works for ALL Americans by making impossible tradeoff. Says soundbite taken from corporate lobbyist.
Well, in fairness, the amount of conscious-being-provided material (spunk) that she can't create herself, is less than that of any known non-biological replicator, so I'd say they're pretty damn close;-)
Make sure one of the requirements on your X-prize is the protocols to make sure those replicators can't engage in undesired, unbounded replication and consumption of their environment.
And that in turn, leads to the Black Raven problem (or "Raven Paradox").
Theory: All ravens are black. Contrapositive: Everything that's not black, is not a raven. Implication: Every time I see something not black, and also not a raven, I support the Contrapositive. Problem: The Contrapositive is logically equivalent to the Theory. But why the hell should seeing a red apple make me think all ravens are black?
Actually, it's not that hard to define optimally intelligent. There are a number of rigorous definitions I can give you, but I'll just pick one:
"An optimally intelligent being is one that, having made the same observations (i.e. taken the same sensory data) as any other kind of being, will form beliefs (assignments of probabilities to future events) such that, if it made bets with any of the other beings (or groups of beings) about future events, it would win more often than it would lose."
Rigorous enough?
Now, there absolutely are rough spots that would arise even if we could program such a being. Specifically: It would necessarily not have the same observations as any biological being that exists today. Why? Because, as counterintuitive as it may sound, part of what feeds into biological beings' beliefs (via the cognitive architecture via the evolved functionality) happened over billions of years. Your brain was selected to perform in *this* environment, performing computations on things you would observe *here*, on earth. For *practical* things that we try to accomplish, our brains have already gotten quite a head start.
Since natural selection was not using the optimal inference algorithm, we know that such a being would figure out all the things we got as a "head start" faster than the length of evolutionary history on earth, but that quite a huge bound to begin with. Faster than a few billion years isn't necessarily fast:-P
So what *is* the optimal inference algorithm? Well, we don't know yet, but I can tell you what researchers have come up with so far. They basically use Bayes's Theorem, because that tells you in what fraction of possible worlds you would observe the data that you have, and therefore how you should update your beliefs. Applying that with a bunch of math gets you Solomonoff induction. Google that and AIXI, but it works basically like this:
Take some programming language. Look at the set of all programs which halt and which assign a probability (even if zero) to all possible data inputs. (Don't worry about the undecidability of halting -- you can make errors and still get values close enough because of how quickly some of them shrink and become insignificant.) For each program, assume its probability of being the generating function for your observations, halves with each additional bit in its length.
So, before observing anything, you have set a "universal prior". THEN, as data comes in, clip out (assign zero probability to) any program inconsistent with gathered sensory data, starting of course with the shortest ones (becaue they impact probability the most). As usual, to get a rough estimate of probability, you don't need to go through the infinite functions because halving probabilities makes them die out quickly. To predict the next data is just a matter of referring to the probability of each of your programs being true, combine with the probability each one assigns to the next bits being whatever.
To take it a step further, you can feed in any structure you know the "observational world" has for the problem you're looking at. i.e., if you're not studying the universe as a system, but just looking at an object (spam, facial, etc.) recognition problem, you can rule out all programs except those that say e.g. "every 1000th byte is purely a function of the preceding 999 bytes" so as to trim out programs you know aren't generating your "sense data".
PHew, okay, now I'm telling you more than you wanted. Hope that clarifies.
Yes, I'm familiar with that device, but it doesn't do the kind of thing that Kurzweil is referring to. It's really just a phrasebook with voice control, where you have to match some very specific templates. They make no pretense that they could pass these out and let English speakers chat with Arabic speakers.
For my part, I've long been planning to develop an easy text translator that basically lets you add xml markup to text sufficient to disambiguate anything that the parser can't get on its own. (Thus, anyone who understands the text, not just the person who wrote it, can provide the necessary input to translate, even if they don't understand the target language.) It would (like IBM's device) provide the reverse translation so you can be sure you got everything right. I already had a freelance programmer develop a primitive (but easy) interface. But, it's way too big a project until I can get a lot more free time.
Interestingly, when I asked some lawyers about a patent search to see if someone's already covered this method, the first three I talked to came back and said they couldn't do it because my idea would create a conflict of interest with their other clients, who included Microsoft. Go fig.
The ultimate goal of course, was not to patent this, but see it one day implemented.
The point is that each incremental change in our evolutionary history did NOT improve fitness. It just didn't hurt it enough Oh, maybe Richard Dawkins could stop perpetuating that misconception then.
it might prove too difficult to reverse-engineer the brain's circuitry because it evolved so haphazardly. "My colleague Francis Crick used to say that God is a hacker, not an engineer," Dr. Ramachandran said. I'd always thought that this constraint would *help* because you know, in advance, that the solution to the problem is constrained by what we know about how evolution works. You have to start with the simplest brain that could have genetic fitness enhancement, and then work up from there, making sure each step performs a useful cognitive function.
Furthermore, looking at the broader picture, I was reading an artificial intelligence textbook that was available online (sorry, can't recall it offhand) which said that current AI researchers have shifted their focus from "how do humans think?" to "how would an optimally intelligent being think?" and therefore "what is the optimal rational inference you can make, given data?" In that paradigm, the brain is only important insofar as it tracks the optimal inference algorithm, and even if we want to understand the brain, that gives us another clue about it: to improve fitness, it must move closer to the what the optimal algorithm would get from that environment.
I'm guessing you've never worked with any type of unmanned vehicle. Any "driverless cars" we could come up with today would be ridiculously expensive to the point that only governments (and not all of them) and people like Gates and Buffet would be able to afford any of them. Really? It's going to be expensive? You mean like... every technology that's ever existed when it was first launched? You're particularly wrong in this case because a) actuators for a car's controls are cheap, and b) the marginal cost of putting the right software in it is cheap after you discover it. It's just going to be like any other innovation that's expensive to solve, and cheap to make more of. Furthermore, there are already computer-driven cars that passed punishing tests that covered Autobahn driving as well as inner-city.
"Special facilities" where people can use drugs to get "inspiration"? Most people who think they have actually received inspiration are people still addicted to the drugs they were taking.... And the money funds terrorists, right? As if the people who get their ideas while smoking weed are going to be really public about it. Even Feynman shut up about his hard drug use in the long-past.
I'm not sure how the tax system "punishes people who are successful in exploiting new, innovative technology" but even if it does, the tax system has nothing to do with stupid juries. Right. That's because -- get this! -- the "stupid jury" and the "tax system" arguments were different complaints! They're not supposed to be directly related except for both being political barriers to innovation. Which was, you know... my whole point.
You "don't know how it punishes people" because you're not capable of abstract reasoning (as shown by your belief that the stupid jury complaint and punishing taxation complaint were part of the same argument). Any time someone brings a successful, useful technology to market, or in fact, engages in ANY voluntary exchange, it is taxed, usually at several stages along the way. That is ultimately a punishment for making a mutually beneficial improvement in two+ people's lives. A more sensible tax system would tax things with negative externalities (or are otherwise undesirable) and/or things don't increase your burden for each individual improvement you make in two people's lives. An example of such a system is a land tax: hard to cheat on, and only snags you at a few specific times, so everyone is free to capture the full gain of each marginal economic interaction they make.
Nuclear power is cheap, and clean, but it isn't a permanent solution any more than any other single solution is. Considering how freaking much energy it can potentially provide, and the fact that it can provide the energy input for drawing CO2 out of the air to make carbon-neutral fuel, yes, it damn well is a permanent solution for several problems: dictators getting enriched by oil revenues, transportation CO2 emissions, energy shocks, and several I probably haven't thought of.
Okay, now remind me real quick where they indicate which of them didn't come true or he had to revise? We're already about done with the "early 2000's" and nowhere close to on-the-fly voice-to-voice translation, for example.
While I don't agree with your general anti-corporatist mentality, I absolutely agree that non-technological factors (specifically, political ones) are by far the biggest impediment to technological growth and the related economic growth (i.e., the good kind, not the bad GDP gain from fixing a hurricane's damage or whatnot).
We've got a cheap carbon-free power source that can provide all our energy needs (nuclear) but political forces (including ignorant voters) keep them from being built and stupid treaties keep us from getting all the energy.
We could have driverless cars (no accident, optimal usage of existing roads), but fears about phony liability damage (omg a GM driverless car caused an accident [at a lower rate than normal cars]! hit them up for millions!!!!)
We could have special facilities in which people can use drugs safely and with supervision so as to get inspiration, but the visceral hate of psychoactive drug use (other than tobacco coffee and alcohol) prevents ANY such facility from being built, so people can only exploit the benefits in secret and dangerously.
We have a tax system that punishes people who are successful in exploiting a new, innovate technology, all while allowing monumentally-stupid juries to assign obscene awards.
If a new pharmaceutical drug relieves pain for hundreds of thousands of people, but just *slightly* increases the risk of heart attack in unhealthy people, we take this valuable drug off the market.
Technological growth can surge, if we just *let it*!!!
I'm more interested in being able to search the 20+ year history of a stock, mutual fund, index (with dividend reinvestment), or futures contract. They only seem to let you go back 10 years, or not see the result of dividend reinvestment. Even google finance only lets you go back a few years in many cases.
Interestingly enough, people on investing forums casually reference these values as if they're easy to get, but I've never seen a free source for that information.
Has there been any serious research into the idea that stories about Jesus (or miracles in other holy books) were the result of someone visiting who had advanced technology? I'm not trying to mock you, I really want to know. It's not that hard to imagine the Gospel as the undevelopeds' side of a Star Trek episode:-P
Stupid question: since LEDs are so insanely energy efficient, why not just:
a) increase their intensity b) slap on a stained glass cover that gets the color and light diffusion just right
Even if you vastly increased the wattage going into an LED, it's still a lot more efficient than a CFL. Yet no one does this, LEDs offered in stores suck, and there are no LED replacements for standard light bulbs.
Hey, I'd love to switch, but can't their makers spare a few seconds of innovation?
Comcast: OMG!!! Outrageous!!! Some HACKERS denied us access to our OWN DOMAIN NAME!!!! Get them!!!! FBI: Why? They didn't take anything that belongs to you. Comcast: What??? Out contract with ICANN gives us unlimited access to the Comcast domain! FBI: Right. And what does unlimited mean? Comcast: Look, it's right here in Websters: "without any..." FBI: No, no, not that one, use your own internal glossary. Comcast: Okay then, "unlimited: "... ah, okay, see your point there.
( ) New York Times ( ) LA Times ( ) Washington Post
is reporting that the government
( ) is censoring scientists with unpleasant news. ( ) is going to cut back program _________. ( ) has been engaging in warrantless wiretapping. ( ) wants to raise taxes. ( ) plans to institute new product regulations
This is just another case of
( ) Big Government ramming itself down our throats! ( ) the War on Science! ( ) how conservatives are killing the safety net! ( ) government punishing anyone productive! ( ) how the country's becoming a totalitarian dictatorship!
HT:
( ) Digg ( ) Slashdot ( ) Technorati ( ) The other blogger who's a carbon copy of me
Okay, "where it should be" might be exaggerating, but this one shows a chart where 20-25% is the acceptable range, meaning that the BMI's claim of "two levels above acceptable" is way off.
8-14% is "recommended" if you take "recommended" to mean "gee, it would be really nice if you pushed yourself this". Of course, I started going to the gym to lose fat, at 24% is where I'm starting from. (I actually thought it would be higher!) But according to the BMI reading, I should have difficulty walking around, and that's just not the case. (Then again, I don't look like a typical 220 lb male.)
No, there is a valid point to be made there. Just for laughs, I found a BMI chart. I'm about 5'9'', 220 lb. According to the chart, my BMI is about 33, well into "obese" territory, which itself is a class above overweight!
Yet just last week I had my bodyfat checked at a gym with some electronic device you're supposed to grip. It put me at 24% bodyfat, right within the range it should be.
Slashdot Mirror
***
But as for the point you did make, we need to keep in mind that even if we don't feel lucky, we should hedge against that unluckiness in the most efficient way possible (carbon tax or cap-and-trade + sequestration) rather than the least efficient way possible (basically, most of the stuff Gore or Stern recommend).
(Paraphrasing what someone said on /. a while back)
It rolls over and...
*scanning for flair*
*16 flair instances detected*
*minimum flair requirement: 16*
*You belong to subset: "Always do the minimum"? y/n*
Goddamnit, who keeps sending self-aware chatbots to access Slashdot?
How about recommending solutions that require LESS than the age of the universe to work?
Without reading the article, I can guess it tracks this format pretty closely:
Q: What would {Obama,McCain} do about $TECH_ISSUE?
Obama: Emphasises coming up with solution that works for ALL Americans by making impossible tradeoff. Says soundbite taken from Lawrence Lessig.
McCain: Emphasises coming up with solution that works for ALL Americans by making impossible tradeoff. Says soundbite taken from corporate lobbyist.
Does that about sum it up?
Well, in fairness, the amount of conscious-being-provided material (spunk) that she can't create herself, is less than that of any known non-biological replicator, so I'd say they're pretty damn close ;-)
Make sure one of the requirements on your X-prize is the protocols to make sure those replicators can't engage in undesired, unbounded replication and consumption of their environment.
And that in turn, leads to the Black Raven problem (or "Raven Paradox").
Theory: All ravens are black.
Contrapositive: Everything that's not black, is not a raven.
Implication: Every time I see something not black, and also not a raven, I support the Contrapositive.
Problem: The Contrapositive is logically equivalent to the Theory. But why the hell should seeing a red apple make me think all ravens are black?
Actually, it's not that hard to define optimally intelligent. There are a number of rigorous definitions I can give you, but I'll just pick one:
:-P
"An optimally intelligent being is one that, having made the same observations (i.e. taken the same sensory data) as any other kind of being, will form beliefs (assignments of probabilities to future events) such that, if it made bets with any of the other beings (or groups of beings) about future events, it would win more often than it would lose."
Rigorous enough?
Now, there absolutely are rough spots that would arise even if we could program such a being. Specifically: It would necessarily not have the same observations as any biological being that exists today. Why? Because, as counterintuitive as it may sound, part of what feeds into biological beings' beliefs (via the cognitive architecture via the evolved functionality) happened over billions of years. Your brain was selected to perform in *this* environment, performing computations on things you would observe *here*, on earth. For *practical* things that we try to accomplish, our brains have already gotten quite a head start.
Since natural selection was not using the optimal inference algorithm, we know that such a being would figure out all the things we got as a "head start" faster than the length of evolutionary history on earth, but that quite a huge bound to begin with. Faster than a few billion years isn't necessarily fast
So what *is* the optimal inference algorithm? Well, we don't know yet, but I can tell you what researchers have come up with so far. They basically use Bayes's Theorem, because that tells you in what fraction of possible worlds you would observe the data that you have, and therefore how you should update your beliefs. Applying that with a bunch of math gets you Solomonoff induction. Google that and AIXI, but it works basically like this:
Take some programming language. Look at the set of all programs which halt and which assign a probability (even if zero) to all possible data inputs. (Don't worry about the undecidability of halting -- you can make errors and still get values close enough because of how quickly some of them shrink and become insignificant.) For each program, assume its probability of being the generating function for your observations, halves with each additional bit in its length.
So, before observing anything, you have set a "universal prior". THEN, as data comes in, clip out (assign zero probability to) any program inconsistent with gathered sensory data, starting of course with the shortest ones (becaue they impact probability the most). As usual, to get a rough estimate of probability, you don't need to go through the infinite functions because halving probabilities makes them die out quickly. To predict the next data is just a matter of referring to the probability of each of your programs being true, combine with the probability each one assigns to the next bits being whatever.
To take it a step further, you can feed in any structure you know the "observational world" has for the problem you're looking at. i.e., if you're not studying the universe as a system, but just looking at an object (spam, facial, etc.) recognition problem, you can rule out all programs except those that say e.g. "every 1000th byte is purely a function of the preceding 999 bytes" so as to trim out programs you know aren't generating your "sense data".
PHew, okay, now I'm telling you more than you wanted. Hope that clarifies.
Yes, I'm familiar with that device, but it doesn't do the kind of thing that Kurzweil is referring to. It's really just a phrasebook with voice control, where you have to match some very specific templates. They make no pretense that they could pass these out and let English speakers chat with Arabic speakers.
For my part, I've long been planning to develop an easy text translator that basically lets you add xml markup to text sufficient to disambiguate anything that the parser can't get on its own. (Thus, anyone who understands the text, not just the person who wrote it, can provide the necessary input to translate, even if they don't understand the target language.) It would (like IBM's device) provide the reverse translation so you can be sure you got everything right. I already had a freelance programmer develop a primitive (but easy) interface. But, it's way too big a project until I can get a lot more free time.
Interestingly, when I asked some lawyers about a patent search to see if someone's already covered this method, the first three I talked to came back and said they couldn't do it because my idea would create a conflict of interest with their other clients, who included Microsoft. Go fig.
The ultimate goal of course, was not to patent this, but see it one day implemented.
Furthermore, looking at the broader picture, I was reading an artificial intelligence textbook that was available online (sorry, can't recall it offhand) which said that current AI researchers have shifted their focus from "how do humans think?" to "how would an optimally intelligent being think?" and therefore "what is the optimal rational inference you can make, given data?" In that paradigm, the brain is only important insofar as it tracks the optimal inference algorithm, and even if we want to understand the brain, that gives us another clue about it: to improve fitness, it must move closer to the what the optimal algorithm would get from that environment.
Their laughter is drowned, the billions of people who previously could not communicate at all but now can communicate.
Not yet, but Google does it with text. HAHAHAHAHAHAHA! Oh, god, you're a riot. We need you at our next party.
You "don't know how it punishes people" because you're not capable of abstract reasoning (as shown by your belief that the stupid jury complaint and punishing taxation complaint were part of the same argument). Any time someone brings a successful, useful technology to market, or in fact, engages in ANY voluntary exchange, it is taxed, usually at several stages along the way. That is ultimately a punishment for making a mutually beneficial improvement in two+ people's lives. A more sensible tax system would tax things with negative externalities (or are otherwise undesirable) and/or things don't increase your burden for each individual improvement you make in two people's lives. An example of such a system is a land tax: hard to cheat on, and only snags you at a few specific times, so everyone is free to capture the full gain of each marginal economic interaction they make. Nuclear power is cheap, and clean, but it isn't a permanent solution any more than any other single solution is. Considering how freaking much energy it can potentially provide, and the fact that it can provide the energy input for drawing CO2 out of the air to make carbon-neutral fuel, yes, it damn well is a permanent solution for several problems: dictators getting enriched by oil revenues, transportation CO2 emissions, energy shocks, and several I probably haven't thought of.
Okay, now remind me real quick where they indicate which of them didn't come true or he had to revise? We're already about done with the "early 2000's" and nowhere close to on-the-fly voice-to-voice translation, for example.
While I don't agree with your general anti-corporatist mentality, I absolutely agree that non-technological factors (specifically, political ones) are by far the biggest impediment to technological growth and the related economic growth (i.e., the good kind, not the bad GDP gain from fixing a hurricane's damage or whatnot).
We've got a cheap carbon-free power source that can provide all our energy needs (nuclear) but political forces (including ignorant voters) keep them from being built and stupid treaties keep us from getting all the energy.
We could have driverless cars (no accident, optimal usage of existing roads), but fears about phony liability damage (omg a GM driverless car caused an accident [at a lower rate than normal cars]! hit them up for millions!!!!)
We could have special facilities in which people can use drugs safely and with supervision so as to get inspiration, but the visceral hate of psychoactive drug use (other than tobacco coffee and alcohol) prevents ANY such facility from being built, so people can only exploit the benefits in secret and dangerously.
We have a tax system that punishes people who are successful in exploiting a new, innovate technology, all while allowing monumentally-stupid juries to assign obscene awards.
If a new pharmaceutical drug relieves pain for hundreds of thousands of people, but just *slightly* increases the risk of heart attack in unhealthy people, we take this valuable drug off the market.
Technological growth can surge, if we just *let it*!!!
I'm more interested in being able to search the 20+ year history of a stock, mutual fund, index (with dividend reinvestment), or futures contract. They only seem to let you go back 10 years, or not see the result of dividend reinvestment. Even google finance only lets you go back a few years in many cases.
Interestingly enough, people on investing forums casually reference these values as if they're easy to get, but I've never seen a free source for that information.
Has there been any serious research into the idea that stories about Jesus (or miracles in other holy books) were the result of someone visiting who had advanced technology? I'm not trying to mock you, I really want to know. It's not that hard to imagine the Gospel as the undevelopeds' side of a Star Trek episode :-P
Stupid question: since LEDs are so insanely energy efficient, why not just:
a) increase their intensity
b) slap on a stained glass cover that gets the color and light diffusion just right
Even if you vastly increased the wattage going into an LED, it's still a lot more efficient than a CFL. Yet no one does this, LEDs offered in stores suck, and there are no LED replacements for standard light bulbs.
Hey, I'd love to switch, but can't their makers spare a few seconds of innovation?
How come no one's made the obvious joke yet?
..." ... ah, okay, see your point there.
Comcast: OMG!!! Outrageous!!! Some HACKERS denied us access to our OWN DOMAIN NAME!!!! Get them!!!!
FBI: Why? They didn't take anything that belongs to you.
Comcast: What??? Out contract with ICANN gives us unlimited access to the Comcast domain!
FBI: Right. And what does unlimited mean?
Comcast: Look, it's right here in Websters: "without any
FBI: No, no, not that one, use your own internal glossary.
Comcast: Okay then, "unlimited: "
Well, folks, it's happening again! The
( ) New York Times
( ) LA Times
( ) Washington Post
is reporting that the government
( ) is censoring scientists with unpleasant news.
( ) is going to cut back program _________.
( ) has been engaging in warrantless wiretapping.
( ) wants to raise taxes.
( ) plans to institute new product regulations
This is just another case of
( ) Big Government ramming itself down our throats!
( ) the War on Science!
( ) how conservatives are killing the safety net!
( ) government punishing anyone productive!
( ) how the country's becoming a totalitarian dictatorship!
HT:
( ) Digg
( ) Slashdot
( ) Technorati
( ) The other blogger who's a carbon copy of me
Okay, "where it should be" might be exaggerating, but this one shows a chart where 20-25% is the acceptable range, meaning that the BMI's claim of "two levels above acceptable" is way off.
8-14% is "recommended" if you take "recommended" to mean "gee, it would be really nice if you pushed yourself this". Of course, I started going to the gym to lose fat, at 24% is where I'm starting from. (I actually thought it would be higher!) But according to the BMI reading, I should have difficulty walking around, and that's just not the case. (Then again, I don't look like a typical 220 lb male.)
No, there is a valid point to be made there. Just for laughs, I found a BMI chart. I'm about 5'9'', 220 lb. According to the chart, my BMI is about 33, well into "obese" territory, which itself is a class above overweight!
Yet just last week I had my bodyfat checked at a gym with some electronic device you're supposed to grip. It put me at 24% bodyfat, right within the range it should be.
I can't speak for anyone else here, but the line of reasoning:
"patents are stupid --> all of IP is stupid --> abolishing IP by itself would immediately bring about world peace"
certainly didn't make my common sense tingle...