Of course, you need a competent interviewer. None of the interviewers in the dozen or so on-line transcript I read qualified. In fact, I didn't even see anybody do a meaningful attempt. They were more than happy to let cleverbot skip from one topic to the next, and never asked any questions that require understanding.
If I were the referee in a boxing match, or the judge in a Olympic gymnastic event, I'd probably give very inappropriate scores as well. The trick is to find competent judges, and then the scores will provide a meaningful estimate of skill. What Turing pointed out that we don't have any objective test for intelligence.
With a Turing test, the challenge is to ask questions that require intelligence and understanding to answer, not some random chit-chat. Some simple example that I tried on cleverbot, and where it fails miserably: "I have two sisters, Jane and Mary. Jane is the oldest, what is the name of the youngest ?".
That is a very simple question, but it requires a (rudimentary) understanding. Simple parroting of canned phrases won't work. A couple of the answers I got:
- That's a strange coincidence. Do you know Jane? - Interesting - No, you ? - This is because it looks pretty.
Now, if you get fooled by that, it shows that you don't qualify as a good interviewer.
Typical questions, such as "what is your name ?" don't attempt to test understanding.
If there was some sort of objective blush test, a la Blade Runner, where a machine was scored on his ability to work through not just human interaction but also moral problems, social problems, perceptual problems, emotional problems, memory problems, language problems - the whole range of human cognition, in other words
There is no reason why a competent interviewer couldn't ask the subject a range of questions designed to show understanding in any of the areas you mentioned.
Why "specific contexts" ? The topics that you can talk about in a Turing test are endless, as well as the duration of the test.
The whole point about the Turing test is that the test itself provides the definition, since we use the same kind of interaction test to judge how well some other human being understands something. If I discuss a problem with a co-worker, I also get a good idea of how intelligent they are, and if they understand the problem. There is no reason why the same approach doesn't work with a computer program.
The only problem is that the Turing test is somewhat limited by the text-only interface. With modern technology, we could propose a Turing 2.0 test, with audio and video interaction, allowing us to make a better and faster judgment. However, I don't really see that as an important issue.
It doesn't have to be just a thought experiment. It's a perfectly fine test to perform in practice, and with a decent interviewer, it should give a good indication of true intelligence and understanding.
Of course, when the interviewers are complete idiots the results of the test are useless.
No, the Turing test is fine. It is just stupid of people to try to pass the Turing test with the current understanding of AI. The bots can do nothing but fail hopelessly (assuming decent interviewers like Turing had in mind).
Compare with the following: a civilization passes the "space test" when they've successfully landed people on the moon. In itself, that's a perfectly fine test. Now, imagine people in the Stone Age trying to pass the test by trying to build taller and taller wooden ladders. That's about what people are trying to do with the Turing test right now.
Indeed, forget about the Turing test right now, and work on AI from different angles. When we are finally ready, the Turing test can be resurrected.
There's nothing wrong with the Turing test, but it requires a good interviewer that knows how to ask questions that require real understanding to answer.
That makes no sense. There is no benefit to useless jobs. All that matters is productivity, and if you can achieve that with less people, it's only better.
The growth rate has been pretty constant so far, somewhere between 1% and 2%. While a growth rate of 0 is possible, it is most likely that this will be result of increased death rates (due to famine, disease and war), rather than a reduced birth rate.
More likely, however, is a positive growth rate, followed by a sharply negative one, rather than a gentle 0.
Maybe GP was talking about growth in absolute sense. Whether growth is 1% or 2% doesn't really matter all that much, since any exponential growth on a bounded earth is going to run into problems at some point in time.
Because "Free as in S(oftware/)peech" = "Free as in Beer", right? Conflating the two arguments is illogical.
It's not really that illogical. If the project is "Free as in speech", and source code can be freely copied, modified and distributed, it is just not practical to charge any more money than the cost of physical distribution, which is basically free. If you tried to charge more, anybody could make a copy, and sell it for less.
Obviously, whether [sic] measurements are not repeatable.
The weather itself is not repeatable, but the measurements around the world to establish the proxy record of that temperature is perfectly repeatable. You can still examine trees, coral, drill holes, and so on. In fact, since Mann's work, it has been repeated several times, confirming his original graph.
Since "the decline" he was talking about is clearly an error in the recent proxy record, hiding it would benefit the truth.
Of course, most of the deniers have no idea what is meant by "the decline", and they assume is has something to with the temperatures in the last decade or so.
The science is much more than the historical proxy temperature reconstruction that Mann is working on. Even excluding all of Mann's work, the science is still settled.
Depends on how detailed the imagination is. He could imagine people as shallow cartoon figures. Just like the little people on your TV screen don't have any thoughts when you watch the movie.
People will just naturally move inland as the water level rises
I live in a low lying river delta, together with a few million other people. It's a very profitable region of the country, and I can assure you it's not going to be moved anytime soon.
What happens instead is that the levee system will have to be upgraded, which means a $ multi billion investment. Or, worse, the threat isn't taken seriously enough, the upgrade is delayed, and the entire area is flooded during a huge storm surge.
Slashdot Mirror
Of course, you need a competent interviewer. None of the interviewers in the dozen or so on-line transcript I read qualified. In fact, I didn't even see anybody do a meaningful attempt. They were more than happy to let cleverbot skip from one topic to the next, and never asked any questions that require understanding.
If I were the referee in a boxing match, or the judge in a Olympic gymnastic event, I'd probably give very inappropriate scores as well. The trick is to find competent judges, and then the scores will provide a meaningful estimate of skill. What Turing pointed out that we don't have any objective test for intelligence.
With a Turing test, the challenge is to ask questions that require intelligence and understanding to answer, not some random chit-chat. Some simple example that I tried on cleverbot, and where it fails miserably: "I have two sisters, Jane and Mary. Jane is the oldest, what is the name of the youngest ?".
That is a very simple question, but it requires a (rudimentary) understanding. Simple parroting of canned phrases won't work. A couple of the answers I got:
- That's a strange coincidence. Do you know Jane?
- Interesting
- No, you ?
- This is because it looks pretty.
Now, if you get fooled by that, it shows that you don't qualify as a good interviewer.
Typical questions, such as "what is your name ?" don't attempt to test understanding.
There is no reason why a competent interviewer couldn't ask the subject a range of questions designed to show understanding in any of the areas you mentioned.
Why "specific contexts" ? The topics that you can talk about in a Turing test are endless, as well as the duration of the test.
The whole point about the Turing test is that the test itself provides the definition, since we use the same kind of interaction test to judge how well some other human being understands something. If I discuss a problem with a co-worker, I also get a good idea of how intelligent they are, and if they understand the problem. There is no reason why the same approach doesn't work with a computer program.
The only problem is that the Turing test is somewhat limited by the text-only interface. With modern technology, we could propose a Turing 2.0 test, with audio and video interaction, allowing us to make a better and faster judgment. However, I don't really see that as an important issue.
It doesn't have to be just a thought experiment. It's a perfectly fine test to perform in practice, and with a decent interviewer, it should give a good indication of true intelligence and understanding.
Of course, when the interviewers are complete idiots the results of the test are useless.
No, the Turing test is fine. It is just stupid of people to try to pass the Turing test with the current understanding of AI. The bots can do nothing but fail hopelessly (assuming decent interviewers like Turing had in mind).
Compare with the following: a civilization passes the "space test" when they've successfully landed people on the moon. In itself, that's a perfectly fine test. Now, imagine people in the Stone Age trying to pass the test by trying to build taller and taller wooden ladders. That's about what people are trying to do with the Turing test right now.
Indeed, forget about the Turing test right now, and work on AI from different angles. When we are finally ready, the Turing test can be resurrected.
There's nothing wrong with the Turing test, but it requires a good interviewer that knows how to ask questions that require real understanding to answer.
That makes no sense. There is no benefit to useless jobs. All that matters is productivity, and if you can achieve that with less people, it's only better.
Obviously, but that doesn't contradict my statement. Both things are true at the same time.
But, ultimately, the problem isn't high prices, but shortages.
The growth rate has been pretty constant so far, somewhere between 1% and 2%. While a growth rate of 0 is possible, it is most likely that this will be result of increased death rates (due to famine, disease and war), rather than a reduced birth rate.
More likely, however, is a positive growth rate, followed by a sharply negative one, rather than a gentle 0.
Worse than rising prices for oil is the fact that despite high prices, production will still be limited.
Maybe GP was talking about growth in absolute sense. Whether growth is 1% or 2% doesn't really matter all that much, since any exponential growth on a bounded earth is going to run into problems at some point in time.
It's not really that illogical. If the project is "Free as in speech", and source code can be freely copied, modified and distributed, it is just not practical to charge any more money than the cost of physical distribution, which is basically free. If you tried to charge more, anybody could make a copy, and sell it for less.
No, I always build my own PC. But I would expect tech support to be available during my own office hours, and I'm not going to worry about it.
99.9% of the people never have a meeting with people on the other side of the world. Changing time zones would bring them only confusion.
The weather itself is not repeatable, but the measurements around the world to establish the proxy record of that temperature is perfectly repeatable. You can still examine trees, coral, drill holes, and so on. In fact, since Mann's work, it has been repeated several times, confirming his original graph.
Since "the decline" he was talking about is clearly an error in the recent proxy record, hiding it would benefit the truth.
Of course, most of the deniers have no idea what is meant by "the decline", and they assume is has something to with the temperatures in the last decade or so.
The science is much more than the historical proxy temperature reconstruction that Mann is working on. Even excluding all of Mann's work, the science is still settled.
Depends on how detailed the imagination is. He could imagine people as shallow cartoon figures. Just like the little people on your TV screen don't have any thoughts when you watch the movie.
Why make people in the first place, when you can just imagine every consequence in your imagination ?
I live in a low lying river delta, together with a few million other people. It's a very profitable region of the country, and I can assure you it's not going to be moved anytime soon.
What happens instead is that the levee system will have to be upgraded, which means a $ multi billion investment. Or, worse, the threat isn't taken seriously enough, the upgrade is delayed, and the entire area is flooded during a huge storm surge.
If they drop prices again below fair market value, just take advantage of it by increasing imports, and use those to build a strategic storage.
No matter how many factors of twenty you reduce the price, it will never get negative.
Thanks. But I don't really think it's valid to extrapolate Manhattan to the rest of the world.
The slight cooling effect will likely be overwhelmed by CO2 warming happening at the same time.
... or maybe it's not such a great idea after all.
Or, more likely, the $1 million+ license fees, and the royalties per core are not a big obstacle for dozens of different licensees.
In return for the license, you get a high quality core for your ASIC, so that's worth it for a lot of customers.