A long long time ago, when Clinton was president:
I hear a Macintosh computer start up, (that glorious stretched out C -Major chord) and I get a chill down my spine. I log into my (now out of date) AOL e-mail and get a "Welcome Bryan" (I thought it was awesome that I could program my AOL e-mail to say my name as it started up...oh 7th grade) at the start up. I then get bored, and start to mess with text to speech and had my computer reassuring me that everything would be all right, and that my sister didn't really mean to say the things she said to me...I then proceed to log off and go to bed to start the whole process of being a teen all over again.
I'll stop with the stories, and get to Hofstadter and the Eliza Effect. I brought up this tale of me way back when to connect with what Hofstadter brings up on page 157 when he is talking about how an ATM is grateful that it received a deposit slip and very thoughtfully printing out "THANK YOU" on its screen. Now, I agree with Hofstadter that the ATM has no lifelike structure on the inside, and that people just mistakenly think that something like this is all right to say. A mistake yes, reality no. A person could defend the issue, saying that since it reacts to what a human does it has intelligence, but no sane human would ever think an ATM can think on its own. It needs some type on user input before it can get to the end result of "THANK YOU" .
This is what Hofstadter was getting to when he was talking about the ELIZA project. That it is not necessarily the case that the computer it "helping" a person through a hard time, but merely reacting to what the user is inputting. This is no different than the AIM bot such as smarterchild.
A side note, as I am typing a red line will appear below a misspelled word. Does the computer "know" that it is misspelled and wants you to correct it? No, it is just linked up to a dictionary and when an unknown word appears or a word that is close to a word in the dictionary appears it will throw the red line on the page which tells the user to correct it.
-Bryan
Thursday, October 22, 2009
Thursday, October 8, 2009
A Little Numbo Never Hurt Anyone...
For this blog we had to read pages 138 to 154. This section in the book discussed how Defays set up the architecture of Numbo, a sample run of Numbo, and how Numbo compares to other computer models. He also discusses how strict comparisons between humans and Numbo are not possible, and this is what caught my eye (p. 151).
There are three reasons why Defays does not think that comparing Numbo and humans is a great possibility. The first one states that Numbo's knowledge base is impoverished and that major aspects of adults mathematical background is lacking in the Pnet. The second is that the way that humans tend to approach a problem have been ignored. He gave the example of the way the bricks are layed out in a linear fashion, which causes humans to read (left to right in this case) in one way, causing some subconscious desire to favor earlier bricks rather than latter ones. Finally, the third comparison was that of ad hoc solutions added to the architecture of Numbo.
This interested me because in figure III-6 Defays shows two different protocols happening side by side. Now, if a person were to go up to the protocols and look at them, I don't think that they could tell the differences from the human and machine protocols, on more complex problems. But, I do believe that on problems that only require a person to recall what the answer is, rather than try to figure it out, would easily be spotted by another person. He gives us the target 6, and bricks 3 3 17 11 22 at one point. Any human (that reads right to left) would get 6 using 3 + 3, whereas Numbo might use 17 - 11 or something else. As soon as we (as humans) have to start doing major computations is when it is harder to tell the difference between human and Numbo.
(Would this be the case with something like language? What about Jumbo? Would we be able to tell the difference with longer versus shorter words? What about word phrases?)
-Bryan
There are three reasons why Defays does not think that comparing Numbo and humans is a great possibility. The first one states that Numbo's knowledge base is impoverished and that major aspects of adults mathematical background is lacking in the Pnet. The second is that the way that humans tend to approach a problem have been ignored. He gave the example of the way the bricks are layed out in a linear fashion, which causes humans to read (left to right in this case) in one way, causing some subconscious desire to favor earlier bricks rather than latter ones. Finally, the third comparison was that of ad hoc solutions added to the architecture of Numbo.
This interested me because in figure III-6 Defays shows two different protocols happening side by side. Now, if a person were to go up to the protocols and look at them, I don't think that they could tell the differences from the human and machine protocols, on more complex problems. But, I do believe that on problems that only require a person to recall what the answer is, rather than try to figure it out, would easily be spotted by another person. He gives us the target 6, and bricks 3 3 17 11 22 at one point. Any human (that reads right to left) would get 6 using 3 + 3, whereas Numbo might use 17 - 11 or something else. As soon as we (as humans) have to start doing major computations is when it is harder to tell the difference between human and Numbo.
(Would this be the case with something like language? What about Jumbo? Would we be able to tell the difference with longer versus shorter words? What about word phrases?)
-Bryan
Monday, October 5, 2009
Non-Deterministic Determinism
As soon as I saw that the section that we had to read for this post dealt with non-determinism the Koch Snowflake immediately jumped into my consciousness. If you don't know what the Koch Snowflake is I highly suggest taking a look at it: Koch Snowflake. I won't go into a lot of detail about it, but it deals with taking something that seems very complex and non-deterministic (random), and is able to make it more deterministic. Now, with real snowflakes this isn't the case, since a snowflake is like a human and every human is different, but has the same "structure."
Now that I got that out of the way, we can go to Defays' project called Numble. On page 132, he gives a math problem, consisting of a 'target' and five 'bricks.' The idea is a lot like out initial Crypto assignment when we were given five numbers (like bricks) and were expected to find the solution (like a target). The biggest difference between the two is that in Numble, a person can use one or all of the bricks to get the target, whereas in Crypto we had to utilize all of our 'bricks' to get our 'target.' Other than that they hold the same concepts. Which is trying to model how we as (as intelligent beings) can take symbols and create a solution from the given solutions. (I use the word 'symbols' because we don't always have to use numbers, or even letters to get a solution. It could be music, where a person is able to start writing a song, and can get a solution by finishing the song, as an example).
Defays' goes on to almost structure a (neural) network on page 136 with his Pnet. (If you don't like neural network, you can just say a network, but since we're dealing with AI, I thought it to be appropriate to connect another aspect of cogsci). Structure is important with how one is to model, and if something has a weak structure (like a horribly designed bridge) it will fall apart. But, if one is able to build something complex out of a simple structure, then you can go from determinism to non-determinism, such as the Koch Snowflake.
(This all deals with a person using "top-down" or "bottom-up" processing that is. A person will seem more deterministic if they use "bottom-up" rather than "top-down" where they seem more stochastic or non-deterministic. But that's my opinion only, some may not see it this way).
-Bryan
Now that I got that out of the way, we can go to Defays' project called Numble. On page 132, he gives a math problem, consisting of a 'target' and five 'bricks.' The idea is a lot like out initial Crypto assignment when we were given five numbers (like bricks) and were expected to find the solution (like a target). The biggest difference between the two is that in Numble, a person can use one or all of the bricks to get the target, whereas in Crypto we had to utilize all of our 'bricks' to get our 'target.' Other than that they hold the same concepts. Which is trying to model how we as (as intelligent beings) can take symbols and create a solution from the given solutions. (I use the word 'symbols' because we don't always have to use numbers, or even letters to get a solution. It could be music, where a person is able to start writing a song, and can get a solution by finishing the song, as an example).
Defays' goes on to almost structure a (neural) network on page 136 with his Pnet. (If you don't like neural network, you can just say a network, but since we're dealing with AI, I thought it to be appropriate to connect another aspect of cogsci). Structure is important with how one is to model, and if something has a weak structure (like a horribly designed bridge) it will fall apart. But, if one is able to build something complex out of a simple structure, then you can go from determinism to non-determinism, such as the Koch Snowflake.
(This all deals with a person using "top-down" or "bottom-up" processing that is. A person will seem more deterministic if they use "bottom-up" rather than "top-down" where they seem more stochastic or non-deterministic. But that's my opinion only, some may not see it this way).
-Bryan
Thursday, October 1, 2009
More Than Meets The Eye...
When I was reading this section of the book (111-126) it dealt a lot with how Jumbo takes words and transforms them into other formations of the word. It gets back to my post on anagrams, and how "hot shots" can be perceived as "hots hots." Hofstadter talks about two different transformations in Jumbo. The first one is "entropy-preserving" and the other one is "entropy-increasing." He goes on to define "entropy" as "perceived disorder." So, working by analogy, a person is able to take the latter part of the phrase and make sense of it. For e-preserving, we can assume, without reading ahead that we are trying to keep the perceived disorder in check, and for e-increasing we can say that we want to increase or raise the perceived disorder.
"Cognition equals recognition" (119). This is the thesis that Hofstadter said we are trying to prove, and he goes on to say that there are infinite amounts of ways to write the letter 'A' but we can all recognize it as the letter 'A'. He also goes on to say that there has to be flexibility with Platonic abstraction and a mental representation. In this sense we have the ability to recognize things that we have seen before, and at some level we learned what it was. I always think of the story of the sailor on a foreign island trying to tell the residents of the island that a ship is coming towards the shore. But, since they have never seen a ship, they do not recognize it and think the sailor is crazy.
A side note, this section reminded me of the T.V. shows from my childhood, and how almost everything that I watched had something transform. Such as Transformers (shouldn't have to explain here) to Superman when Clark Kent would run into a phone booth and "transform" into Superman. I just thought that virtually everything that we are exposed to can have an aspect of transformation.
-Bryan
"Cognition equals recognition" (119). This is the thesis that Hofstadter said we are trying to prove, and he goes on to say that there are infinite amounts of ways to write the letter 'A' but we can all recognize it as the letter 'A'. He also goes on to say that there has to be flexibility with Platonic abstraction and a mental representation. In this sense we have the ability to recognize things that we have seen before, and at some level we learned what it was. I always think of the story of the sailor on a foreign island trying to tell the residents of the island that a ship is coming towards the shore. But, since they have never seen a ship, they do not recognize it and think the sailor is crazy.
A side note, this section reminded me of the T.V. shows from my childhood, and how almost everything that I watched had something transform. Such as Transformers (shouldn't have to explain here) to Superman when Clark Kent would run into a phone booth and "transform" into Superman. I just thought that virtually everything that we are exposed to can have an aspect of transformation.
-Bryan
Subscribe to:
Posts (Atom)
