If you mean "1's and 0's", I'll submit that the data structure used for computation is irrelevant.
If you mean, human's don't think logically, that might be quite right. Certainly, it takes many ontological assumptions to jump to a classical logic (i.e. a logic that includes the law of the excluded middle). Without them, humans are stuck in the same logic as computers -- so-called "constructive" or "intuitionistic" logics. Every programming language is a constructive logic, by the way.
Humanity exists in part because humans are able to forget...
A computer can be made to forget things.
that we all age evokes compassion.
Fine. And you think that a machine capable of knowing its age, and the ages of others around it, can't as a matter of principle reach the same "feeling"?
Children are special to humans because they cannot be easily reproduced.
Yes, at enormous costs to the tax payer. Of course, my point was that cable and DSL use already existing infrastructure for the last mile, so their complaints have "always" (at least since the year 2000) been null and void. Now the same ISPs are complaining about a lack of bandwidth on their internal networks and border networks, when they have lowered the bandwidth they deliver to me, at the same price. If the last mile had been the bottleneck at 6MB/s, the last mile would continue to be the bottleneck at 5MB/s.
Indeed! The telcos have whined for years that the problem giving people high speed internet access was the "last mile". Suddenly, in the last 10 years, 22MB/s cable and DSL have shown up (i.e., the last mile is faster than ever). And in the last year, my connection dropped from 6MB/s to 5MB/s, at 60$ a month.
I am not thinking of machine intelligence in terms of evolution. I am thinking of machine intelligence in terms of joins on lattices that must be constructed before they can be joined. In particular, that means building the lattices, either programmatically, or through automation (genetic programming).
I estimate that it will be at least 300,000 life cycles before we get anything approaching human intelligence. But we can approach super human intelligence much sooner, by mixing in sub-human intelligence with libraries beyond human comprehension. They are slightly different things.
What about topological computation? Your room ought not be a high dimension simplicial complex. The topological word problem is soluble in two dimensions. (Not three, though)
No, it's not. That is pretty much impossible, unless you stick machine learning systems in machines that actually interact with the world. Our genes have a model of the real world in them. You can't expect a computer system capable of improving itself in "all dimensions" (whatever that means), if it can't even interact with a realistic model of the world. The model will either have to be programmed directly, or programmed by evolution. As far as intelligence is concerned, is no difference between them.
The first part that you propose is COMPLETELY unreasonable
No, it's not. Look up "literate programming". Instead of using LaTeX to typeset mathematics using mathematical symbols, typeset them using interpretable code.
Are you offering to pay for software engineers to do code reviews and audits? I know thousands of scientists who would love a software developer to work with.
I have studied both physics and CS and I can hardly imagine a physics curriculum (especially one in theoretical physics) which doesn't include any course in numerical methods.
Never mind that any physicist worth his salt knows that the real numbers are uncountable, and a computer is finite.
I don't understand what the hell is going on, but it seems to me that if you can send energy someplace faster than the speed of light, that could be used to transmit "information".
You're right. Energy and information are the same thing, looked at from different perspectives.
Why would you think adding antennas would help? They will both be broadcasting into the same spectrum. And if they're not, you're just making a "bigger spectrum". The problem is lack of spectrum.
Slashdot Mirror
Management is scared of the Haskell version: shoot $ huntdown $ head $ developers
Humans don't think in binary
If you mean "1's and 0's", I'll submit that the data structure used for computation is irrelevant.
If you mean, human's don't think logically, that might be quite right. Certainly, it takes many ontological assumptions to jump to a classical logic (i.e. a logic that includes the law of the excluded middle). Without them, humans are stuck in the same logic as computers -- so-called "constructive" or "intuitionistic" logics. Every programming language is a constructive logic, by the way.
Humanity exists in part because humans are able to forget...
A computer can be made to forget things.
that we all age evokes compassion.
Fine. And you think that a machine capable of knowing its age, and the ages of others around it, can't as a matter of principle reach the same "feeling"?
Children are special to humans because they cannot be easily reproduced.
Give it time. It is easier every year.
Yes, at enormous costs to the tax payer. Of course, my point was that cable and DSL use already existing infrastructure for the last mile, so their complaints have "always" (at least since the year 2000) been null and void. Now the same ISPs are complaining about a lack of bandwidth on their internal networks and border networks, when they have lowered the bandwidth they deliver to me, at the same price. If the last mile had been the bottleneck at 6MB/s, the last mile would continue to be the bottleneck at 5MB/s.
Indeed! The telcos have whined for years that the problem giving people high speed internet access was the "last mile". Suddenly, in the last 10 years, 22MB/s cable and DSL have shown up (i.e., the last mile is faster than ever). And in the last year, my connection dropped from 6MB/s to 5MB/s, at 60$ a month.
I am not thinking of machine intelligence in terms of evolution. I am thinking of machine intelligence in terms of joins on lattices that must be constructed before they can be joined. In particular, that means building the lattices, either programmatically, or through automation (genetic programming).
Read:
http://hardware.slashdot.org/comments.pl?sid=1545358&cid=31094828
I estimate that it will be at least 300,000 life cycles before we get anything approaching human intelligence. But we can approach super human intelligence much sooner, by mixing in sub-human intelligence with libraries beyond human comprehension. They are slightly different things.
What about topological computation? Your room ought not be a high dimension simplicial complex. The topological word problem is soluble in two dimensions. (Not three, though)
This works in principle, but the neural net has to know how to quantify over a "rich" domain. See: http://hardware.slashdot.org/comments.pl?sid=1545358&cid=31094828
Word are expensive if you don't know what they mean? A dictionary would make you a millionaire. Or poor. Or something.
No, it's not. That is pretty much impossible, unless you stick machine learning systems in machines that actually interact with the world. Our genes have a model of the real world in them. You can't expect a computer system capable of improving itself in "all dimensions" (whatever that means), if it can't even interact with a realistic model of the world. The model will either have to be programmed directly, or programmed by evolution. As far as intelligence is concerned, is no difference between them.
"I like popups because..."
The first part that you propose is COMPLETELY unreasonable
No, it's not. Look up "literate programming". Instead of using LaTeX to typeset mathematics using mathematical symbols, typeset them using interpretable code.
Are you offering to pay for software engineers to do code reviews and audits? I know thousands of scientists who would love a software developer to work with.
I have studied both physics and CS and I can hardly imagine a physics curriculum (especially one in theoretical physics) which doesn't include any course in numerical methods.
Never mind that any physicist worth his salt knows that the real numbers are uncountable, and a computer is finite.
Don't ask about priorities. Ask what you should work on first.
When did Honda get bailed out??!?!?!?!
Or maybe you just downloaded a trojan...
I've been using TextMate, Macfusion, and GHCi.
Can you tell me more about this "illegitimate" prize in economics? I don't doubt you, as such. I am interested in reading the paper.
Whatever, just get a netbook with an NTP client and sync up to the US Navy, like any fullblooded American scientist already does.
Time is the monad that keeps us from changing what we have already done.
I don't understand what the hell is going on, but it seems to me that if you can send energy someplace faster than the speed of light, that could be used to transmit "information".
You're right. Energy and information are the same thing, looked at from different perspectives.
Argh, as if bash and Unix's reliance on it wasn't bad enough, there's an MS bash now?
This is looking rather promising. The source tarball is 843MB. Talk about gigabloatware.
Why would you think adding antennas would help? They will both be broadcasting into the same spectrum. And if they're not, you're just making a "bigger spectrum". The problem is lack of spectrum.