My point is just that this talk about simulating a human brain is meaningless until we first have proved that we can simulate a flea brain. Hell, can we even exactly simulate a single neuron that precisely matches the behavior of a neuron in a petri dish? Solve the easy stuff first.
Can you explain to me how USING a website constitutes SIGNING a contract? How can a contract be valid when I never even lifted a pen? Where's the signature?
Maybe because we want google to win because we like what they do, but we want microsoft to lose because we don't like what they do? In wartime you don't "fight fair", why would you want to grant your enemies the same benefits as your friends?
It is evil for office to use a secret windows API because it gives microsoft an unfair advantage against the competing office suites for windows. It is not evil for google to use a secret youtube api because... why would it be? Google has a youtube server, and they make a youtube client app. They generously provide a public API so anybody can write a youtube app. There is no reason in law or common sense why they are obligated to provide any API at all, or why they should allow anybody else to make youtube apps that compete with their own. They are going above and beyond by ever allowing anybody to make such an app.
This would be a fair comparison to the microsoft office scenario if windows was JUST a server for running office. If microsoft had an office server, would they really be out of line for wanting to restrict what clients connect to it? Go make your own open source office server if you want to own it and set the rules.
Evil depends on your point of view. If you side with google, then google is doing good. If you side with microsoft, google is doing evil. If you side with the consumer, the question is more complicated and I don't know the answer. Probably it is good for the consumer to have access to youtube from whatever and whichever device they own, but if microsoft is evil then doing anything and everything to put them out of business might be good, even if it temporarily pisses off people who own windows phones (temporary until they get a phone from a non-evil company).
You fall into the common trap of thinking that treating everyone equal is somehow the definition of good, but in reality good means treating good people/companies well and treating evil people/companies with the evil they deserve. The trick is in judging who is good and who is evil so that we know who to punish and who to praise.
> There is no way you can prove that something is objectively bad (that I see).
You might be able to prove that something is subjectively bad, however. Don't forget that something being subjective doesn't make it any less real, dangerous, or painful. For example, I don't know if there is a medical device that can objectively tell me if you are in pain. However, if you are hurting, you know beyond a shadow of a doubt that you are subjectively feeling pain. How would you like it if a doctor refused to treat you even though you had horrendous pain, just because you had a type of pain that couldn't be objectively proven to be real? Migraines might fall into this category.
Note: I know that I took just one sentence that you said and removed the context of the conversation.
Huh? The existence of a soul at all, eternal or not, is a matter of religion, i.e. personal belief. This is an article about science and technology, i.e. about things we can prove in a lab and things we can build in a factory. I'm not an atheist but I don't get what you are going on about.
True, however, you will need to store at least one byte in memory for each neuron (probably a lot more), and you will need to execute at least one instruction per time-step (probably a lot more than one instruction). This would mean that to simulate the vast number of neurons and vaster number of connections, you would need a REALLY big supercomputer.
And consider that if we really want something that behaves like a live brain, we probably can't just simulate a neuron as a few numbers, like in the old-school neural network algorithms. It might be necessary to simulate a dense tetrehedral finite element mesh in the shape of a neuron, where each node stores chemical, physical, and electrical properties. Even simulating one neuron to the requisite level of accuracy could require a very powerful computer. The brain has a lot of cells, and each cell is made out of a lot of molecules. What if we have to simulate all of the molecules one-by-one? I admit this is pessimistic and we might have clever models that don't require such detail, my point is just that simulating a human brain most definitely requires an absurdly powerful computer.
Considering we don't know everything about the human brain, you may be right. But for all we know, the brain and sentience might be fully explainable using ordinary chemistry and physics. I am just curious, what leads you to suspect otherwise?
> My qualm with cloning is that it involves wiping the genetic from the egg (thus destroying that potential individual) and populating it with the desired genetic code.
Your qualm is that the genetic material from an egg is destroyed? But doesn't this happen every month in a woman who isn't pregnant?
I learned a bit about the Chomsky language hierarchy in the undergrad theory class. Stuff like context free grammars and pushdown automatas. I don't think our course covered any languages beyond the scope of a turing machine, but that sounds interesting. Intuitively it seems possible that we could define a language that a turing machine can't recognize, but aside from possibly the human brain, are there any known machines that CAN recognize those languages? Or is this just a way of saying that some problems are so hard that there is no way to solve them?
Regarding poetry, I don't understand how this is a valid example. First of all, can humans even agree on what constitutes poetry? Wouldn't there be some passages that some experts consider poetry, and other prose? That would make the problem unsolvable even to humans. Second, has it actually been proven that computers can't recognize poetry when they see it? It doesn't seem that hard to write a program to recognize certain specific kinds of poems. Something simple like a haiku? The program could check for the right number of syllables, and check to see if some of the words are in the category "nature", if we strictly abide by the rule that Haikus must be about nature.
I'd say the line is when you commit bribery by handing money over. If you want to spend your own money to broadcast your own speech over the tv in an ad, that should be legal. But if you pay the politician a donation so they can ostensibly run those same ads but in reality pocket some of the cash, that is bribery and should not be considered speech.
> I don't think he's talking specifically about digital per say, rather it's been proven that some problems can't be solved by Turing machines.
Are you talking about interaction machines? I heard about those recently, but I haven't had the chance to follow up to see what they can do that a turing machine can't. But on the surface it doesn't make sense. An interaction machine is a network of communicating turing machines. Why couldn't this network be simulated on a turing machine and get the same result albeit more slowly?
> some people theorize that our brain is a machine that is superior to computers, and can solve problems that computers never can.
It sounds like you are talking about Penrose's shadows of the mind. He speculated (not theorized, just guessed) that maybe the brain uses some unknown "spooky" aspect of quantum mechanics to do things that a deterministic computing machine can never do. I don't think anybody has ever given this notion much credence, considering there is no evidence and no mechanism, but until we know the brain a lot better I guess we can't be sure.
This sounds a lot like super audio cd. Rather than use a given number of bits to encode a number that represents the height of the sound wave at a point in time, and then use a sequence of numbers to approximate a sound wave, the signal is represented as a continuous stream of individual bits. When you want to know the height of the sound wave at a point in time, you take an average of the bit stream centered around the point you are interested in.
I once heard it said that AI is defined as any task that a computer can't yet do. Once we learn how to write a program to perform that task, it is no longer considered AI. Chess is a good example. We once thought that it required intelligence to beat a grandmaster at chess. Now we know it just requires an algorithm, no intelligence required. If AI is defined this way, then we will indeed never achieve it.
> A lot of the reason people complain college is useless is that it doesn't teach you things that can only be taught with actual experience in industry.
Indeed this is true for many things such as the examples you give. Although actually I did take a college class that did simulate the clueless pointy-haired boss. The class was in our senior year, and we had to form small teams and design and implement a software product according to a customer's requirements. Us undergrads did the coding, and the teams were run by graduate student "managers". The professor was the "CEO" and had final say. The undergrads did all the work, but the graduate student said this: "My grade depends on me running the meetings. So you till me what we are meeting about, and I will then repeat your words and thus earn my grade by leading the meeting". Part of the customer requirements was that our application be distributed across a network. We were aware of CORBA, but choose to use a simpler, cheaper, more appropriate RPC system. The professor insisted that we use a full-blown Borland CORBA product, so that she would have an excuse to buy it for her research team and bill it as a classroom expense. Sounds like something right out of Dilbert if you ask me.
I still think there are a lot of useful things that could be taught in school, but aren't. What programmers need is experience writing programs. Not just theoretical knowledge of how to find the big-O of an algorithm, but how to actually design and implement a substantial amount of code. A lot of what it takes to create software is tedious, obnoxious practical stuff like figuring out compiler flags, selecting appropriate libraries, learning how to use those libraries, and figuring out unintended interactions between components that lead to bugs. Programming assignments in school are usually of the form "Here is a framework where everything is architected and coded except for one algorithm, go code that algorithm". This is fine for teaching the algorithm, but it misses out on all those other things I just mentioned.
When I was a TA for a graphics class, a big part of my job was handing out the programming assignments. I was given a fair amount of leeway, but I roughly stuck with what was done the previous year. An early assignment was to write a polygon rasterizer. We had a framework that allowed the students to just write the rasterizer and nothing else; they were given code to take input from the mouse to describe the vertices, and an output framework in the form of a setpixel function. The framework displayed the pixels as large blocks so you could see gaps between polygons that should have been adjacent (in case your implementation was flawed), and used color to indicate overlap (in case your implementation was flawed).
I thought this made the task too easy, so later when it was time to write a raytracer, I just gave them a set of requirements and suggested they use libPNG to write their output. Everybody succeeded in making a raytracer, and they learned how to think through the task of setting up the whole program. A more traditional approach would have just asked them to write ray-object intersection code, losing sight of the big picture.
I went to an awful lot of school to learn computer science. My shiny fancy degree got me an interview for a job. But guess what? None of the questions in the interview were answerable based on what I learned in school. I knew the answers because I wrote code for fun as a hobby, starting at the age of 15. But I knew the answers and got the job. And guess what? I have not used anything from school in the job. My job is all about the useful programming skills that I had to pick up to write fun little toy programs as a hobby.
While it is possible to be extremely intelligent and yet not know simple facts (such as if the person was raised by wolves), I think that truly smart people who are given the chance to learn will choose to learn a lot of facts. Therefore there is some correlation between knowing raw facts and being intelligent, even if the correlation is not always true one hundred percent of the time.
If somebody doesn't even know how to write without making errors, it may be that they don't care about writing but have chosen to become good at other things, and so in that case it is a mistake to assume that they are not credible on the basis of writing alone. However, in the real world (not the theoretical hypothetical world), smart capable people are usually good at the basics, so generally speaking it is a good idea to assume that lack of basic skills means lack of advanced skills.
I myself am a good counterexample to my own general claim. I can't remember my multiplication tables or do long division to save my life. But I got straight A's in graduate level linear algebra and differential equation classes. Still, if you saw how bad I was at basic math, you would not be insane for thinking I am probably bad at math in general, the burden would be on me to prove otherwise.
Why is it that schools are grouped into either the comparatively prestigious category of 4 year academic B.S. degrees, or the lowly 2 year trade school degree?
If you want to learn theory and go on to do basic research or become a professor, then the B.S. degree is ideal. But if you want to get a really good education of the type that would prepare you to work a skilled job, where is there to turn to? A trade school is geared towards career training, but these are not prestigious and are considered lowly and are typically just 2 year programs. Why can't there be a four year program from a high quality school that has the emphasis on teaching the skills you need for a career in industry? People complain that college is useless, and maybe it is in practice, but couldn't college be very useful if it taught the right things? Why not teach what students need to know to succeed?
When the code is good and the reviews just say "looks good", doesn't that take very little time? Even when no issues are found, the code review lets the team see the changes before they are checked in, so everybody is kept up to date about all changes. This is not a waste of time.
Seriously? How is this even legal? If you are working for your employer, whether in an actively productive role or in training, it is part of the job and should be considered as such. Now if you are on a salary they might have the legal right to ask to work more than 40 hours a week to make this happen, but then it isn't really on your own time, they just want your workday to be longer. And what do you mean by "at your own expense"? Can't these kinds of skills be learned for free from any computer with an Internet connection?
Why can't his manager simply order him to learn the necessary skills? Learning those skills would become part of his job. If he doesn't learn, he isn't doing his job, and could get fired. Why put up with someone who refuses to do his job?
Slashdot Mirror
My point is just that this talk about simulating a human brain is meaningless until we first have proved that we can simulate a flea brain.
Hell, can we even exactly simulate a single neuron that precisely matches the behavior of a neuron in a petri dish?
Solve the easy stuff first.
Can you explain to me how USING a website constitutes SIGNING a contract?
How can a contract be valid when I never even lifted a pen?
Where's the signature?
Maybe because we want google to win because we like what they do, but we want microsoft to lose because we don't like what they do?
In wartime you don't "fight fair", why would you want to grant your enemies the same benefits as your friends?
It is evil for office to use a secret windows API because it gives microsoft an unfair advantage against the competing office suites for windows.
It is not evil for google to use a secret youtube api because... why would it be? Google has a youtube server, and they make a youtube client app. They generously provide a public API so anybody can write a youtube app. There is no reason in law or common sense why they are obligated to provide any API at all, or why they should allow anybody else to make youtube apps that compete with their own. They are going above and beyond by ever allowing anybody to make such an app.
This would be a fair comparison to the microsoft office scenario if windows was JUST a server for running office. If microsoft had an office server, would they really be out of line for wanting to restrict what clients connect to it? Go make your own open source office server if you want to own it and set the rules.
Evil depends on your point of view. If you side with google, then google is doing good. If you side with microsoft, google is doing evil. If you side with the consumer, the question is more complicated and I don't know the answer. Probably it is good for the consumer to have access to youtube from whatever and whichever device they own, but if microsoft is evil then doing anything and everything to put them out of business might be good, even if it temporarily pisses off people who own windows phones (temporary until they get a phone from a non-evil company).
You fall into the common trap of thinking that treating everyone equal is somehow the definition of good, but in reality good means treating good people/companies well and treating evil people/companies with the evil they deserve. The trick is in judging who is good and who is evil so that we know who to punish and who to praise.
> There is no way you can prove that something is objectively bad (that I see).
You might be able to prove that something is subjectively bad, however.
Don't forget that something being subjective doesn't make it any less real, dangerous, or painful.
For example, I don't know if there is a medical device that can objectively tell me if you are in pain. However, if you are hurting, you know beyond a shadow of a doubt that you are subjectively feeling pain. How would you like it if a doctor refused to treat you even though you had horrendous pain, just because you had a type of pain that couldn't be objectively proven to be real? Migraines might fall into this category.
Note: I know that I took just one sentence that you said and removed the context of the conversation.
Huh? The existence of a soul at all, eternal or not, is a matter of religion, i.e. personal belief. This is an article about science and technology, i.e. about things we can prove in a lab and things we can build in a factory. I'm not an atheist but I don't get what you are going on about.
True, however, you will need to store at least one byte in memory for each neuron (probably a lot more), and you will need to execute at least one instruction per time-step (probably a lot more than one instruction). This would mean that to simulate the vast number of neurons and vaster number of connections, you would need a REALLY big supercomputer.
And consider that if we really want something that behaves like a live brain, we probably can't just simulate a neuron as a few numbers, like in the old-school neural network algorithms. It might be necessary to simulate a dense tetrehedral finite element mesh in the shape of a neuron, where each node stores chemical, physical, and electrical properties. Even simulating one neuron to the requisite level of accuracy could require a very powerful computer. The brain has a lot of cells, and each cell is made out of a lot of molecules. What if we have to simulate all of the molecules one-by-one? I admit this is pessimistic and we might have clever models that don't require such detail, my point is just that simulating a human brain most definitely requires an absurdly powerful computer.
Considering we don't know everything about the human brain, you may be right. But for all we know, the brain and sentience might be fully explainable using ordinary chemistry and physics. I am just curious, what leads you to suspect otherwise?
> My qualm with cloning is that it involves wiping the genetic from the egg (thus destroying that potential individual) and populating it with the desired genetic code.
Your qualm is that the genetic material from an egg is destroyed?
But doesn't this happen every month in a woman who isn't pregnant?
Since when does IT mean software developer? You are thinking "programmer". I always thought IT meant sysadmin.
I learned a bit about the Chomsky language hierarchy in the undergrad theory class. Stuff like context free grammars and pushdown automatas.
I don't think our course covered any languages beyond the scope of a turing machine, but that sounds interesting. Intuitively it seems possible that we could define a language that a turing machine can't recognize, but aside from possibly the human brain, are there any known machines that CAN recognize those languages? Or is this just a way of saying that some problems are so hard that there is no way to solve them?
Regarding poetry, I don't understand how this is a valid example. First of all, can humans even agree on what constitutes poetry? Wouldn't there be some passages that some experts consider poetry, and other prose? That would make the problem unsolvable even to humans. Second, has it actually been proven that computers can't recognize poetry when they see it? It doesn't seem that hard to write a program to recognize certain specific kinds of poems. Something simple like a haiku? The program could check for the right number of syllables, and check to see if some of the words are in the category "nature", if we strictly abide by the rule that Haikus must be about nature.
I'd say the line is when you commit bribery by handing money over.
If you want to spend your own money to broadcast your own speech over the tv in an ad, that should be legal.
But if you pay the politician a donation so they can ostensibly run those same ads but in reality pocket some of the cash, that is bribery and should not be considered speech.
> I don't think he's talking specifically about digital per say, rather it's been proven that some problems can't be solved by Turing machines.
Are you talking about interaction machines? I heard about those recently, but I haven't had the chance to follow up to see what they can do that a turing machine can't. But on the surface it doesn't make sense. An interaction machine is a network of communicating turing machines. Why couldn't this network be simulated on a turing machine and get the same result albeit more slowly?
> some people theorize that our brain is a machine that is superior to computers, and can solve problems that computers never can.
It sounds like you are talking about Penrose's shadows of the mind. He speculated (not theorized, just guessed) that maybe the brain uses some unknown "spooky" aspect of quantum mechanics to do things that a deterministic computing machine can never do. I don't think anybody has ever given this notion much credence, considering there is no evidence and no mechanism, but until we know the brain a lot better I guess we can't be sure.
This sounds a lot like super audio cd. Rather than use a given number of bits to encode a number that represents the height of the sound wave at a point in time, and then use a sequence of numbers to approximate a sound wave, the signal is represented as a continuous stream of individual bits. When you want to know the height of the sound wave at a point in time, you take an average of the bit stream centered around the point you are interested in.
> The one true measure of truth is that it's spoken by "earned authority" after all.
So you think people who know nothing but have a hunch are just as likely to be right as someone who has spent years studying and learning?
I once heard it said that AI is defined as any task that a computer can't yet do. Once we learn how to write a program to perform that task, it is no longer considered AI. Chess is a good example. We once thought that it required intelligence to beat a grandmaster at chess. Now we know it just requires an algorithm, no intelligence required. If AI is defined this way, then we will indeed never achieve it.
> A lot of the reason people complain college is useless is that it doesn't teach you things that can only be taught with actual experience in industry.
Indeed this is true for many things such as the examples you give.
Although actually I did take a college class that did simulate the clueless pointy-haired boss.
The class was in our senior year, and we had to form small teams and design and implement a software product according to a customer's requirements.
Us undergrads did the coding, and the teams were run by graduate student "managers". The professor was the "CEO" and had final say.
The undergrads did all the work, but the graduate student said this: "My grade depends on me running the meetings. So you till me what we are meeting about, and I will then repeat your words and thus earn my grade by leading the meeting". Part of the customer requirements was that our application be distributed across a network. We were aware of CORBA, but choose to use a simpler, cheaper, more appropriate RPC system. The professor insisted that we use a full-blown Borland CORBA product, so that she would have an excuse to buy it for her research team and bill it as a classroom expense. Sounds like something right out of Dilbert if you ask me.
I still think there are a lot of useful things that could be taught in school, but aren't. What programmers need is experience writing programs. Not just theoretical knowledge of how to find the big-O of an algorithm, but how to actually design and implement a substantial amount of code. A lot of what it takes to create software is tedious, obnoxious practical stuff like figuring out compiler flags, selecting appropriate libraries, learning how to use those libraries, and figuring out unintended interactions between components that lead to bugs. Programming assignments in school are usually of the form "Here is a framework where everything is architected and coded except for one algorithm, go code that algorithm". This is fine for teaching the algorithm, but it misses out on all those other things I just mentioned.
When I was a TA for a graphics class, a big part of my job was handing out the programming assignments. I was given a fair amount of leeway, but I roughly stuck with what was done the previous year. An early assignment was to write a polygon rasterizer. We had a framework that allowed the students to just write the rasterizer and nothing else; they were given code to take input from the mouse to describe the vertices, and an output framework in the form of a setpixel function. The framework displayed the pixels as large blocks so you could see gaps between polygons that should have been adjacent (in case your implementation was flawed), and used color to indicate overlap (in case your implementation was flawed).
I thought this made the task too easy, so later when it was time to write a raytracer, I just gave them a set of requirements and suggested they use libPNG to write their output. Everybody succeeded in making a raytracer, and they learned how to think through the task of setting up the whole program. A more traditional approach would have just asked them to write ray-object intersection code, losing sight of the big picture.
I went to an awful lot of school to learn computer science. My shiny fancy degree got me an interview for a job. But guess what? None of the questions in the interview were answerable based on what I learned in school. I knew the answers because I wrote code for fun as a hobby, starting at the age of 15. But I knew the answers and got the job. And guess what? I have not used anything from school in the job. My job is all about the useful programming skills that I had to pick up to write fun little toy programs as a hobby.
While it is possible to be extremely intelligent and yet not know simple facts (such as if the person was raised by wolves), I think that truly smart people who are given the chance to learn will choose to learn a lot of facts. Therefore there is some correlation between knowing raw facts and being intelligent, even if the correlation is not always true one hundred percent of the time.
If somebody doesn't even know how to write without making errors, it may be that they don't care about writing but have chosen to become good at other things, and so in that case it is a mistake to assume that they are not credible on the basis of writing alone. However, in the real world (not the theoretical hypothetical world), smart capable people are usually good at the basics, so generally speaking it is a good idea to assume that lack of basic skills means lack of advanced skills.
I myself am a good counterexample to my own general claim. I can't remember my multiplication tables or do long division to save my life. But I got straight A's in graduate level linear algebra and differential equation classes. Still, if you saw how bad I was at basic math, you would not be insane for thinking I am probably bad at math in general, the burden would be on me to prove otherwise.
Why is it that schools are grouped into either the comparatively prestigious category of 4 year academic B.S. degrees, or the lowly 2 year trade school degree?
If you want to learn theory and go on to do basic research or become a professor, then the B.S. degree is ideal. But if you want to get a really good education of the type that would prepare you to work a skilled job, where is there to turn to? A trade school is geared towards career training, but these are not prestigious and are considered lowly and are typically just 2 year programs. Why can't there be a four year program from a high quality school that has the emphasis on teaching the skills you need for a career in industry? People complain that college is useless, and maybe it is in practice, but couldn't college be very useful if it taught the right things? Why not teach what students need to know to succeed?
When the code is good and the reviews just say "looks good", doesn't that take very little time? Even when no issues are found, the code review lets the team see the changes before they are checked in, so everybody is kept up to date about all changes. This is not a waste of time.
What does being good at jeopardy have to do with having basic communication skills like knowing that "have" and "of" are completely different words?
Seriously? How is this even legal? If you are working for your employer, whether in an actively productive role or in training, it is part of the job and should be considered as such. Now if you are on a salary they might have the legal right to ask to work more than 40 hours a week to make this happen, but then it isn't really on your own time, they just want your workday to be longer. And what do you mean by "at your own expense"? Can't these kinds of skills be learned for free from any computer with an Internet connection?
Why can't his manager simply order him to learn the necessary skills? Learning those skills would become part of his job. If he doesn't learn, he isn't doing his job, and could get fired. Why put up with someone who refuses to do his job?