I think we, the AI community, are making actual progress. The problem is that the problem is much harder than people thought it would be back when it first emerged.
Statistical models have done wonders for a lot of things. Classification, mentioned above, is one of the most obvious successes. Natural language processing is another surprising success of statistical methods. The use of hidden markov models has solved a number of problems that were difficult using symbolic approaches (mostly dealing with syntax). The natural language understanding is still a long ways away of course.
Partially observable markov decision processes have also been used a lot in learning in uncertain environments with good success -- another technique from stats.
The problem with AI as a whole is that there is so much knowledge. It is really incredible how much we know. Not even in an academic sense, you know things will fall, how to balance, and all sorts of "common sense" knowledge. Modeling this in a symbolic way is very difficult because of the large amount of information. It is also hard to express. Formalisms such as first order predicate calculus are often used, but they have limitations.
Statistical models are appealing because we do not have to manually write down knowledge. The machine can learn by itself (to some extent). This is probably why machine learning is one of the hottest topics right now.
So keep faith in the smart people trying to work on AI -- just don't expect true intelligent machines for some time yet. Advances are constantly being made in smaller domain-specific areas though.
He just spoke at the University of Illinois at Chicago on Friday and the first thing did was tell people they should buy some GNU/GPL/FSF merchandise that he brought with him. Then he milled around for a while so people could before he began his talk.
When public key crypto came out the government had a fit. Actually, the whole history of cryptography and the NSA is interesting. They fought it tooth and nail, but eventually had to come to terms with the public having strong crypto. Check out "Crypto, How the code rebels beat the government - saving privavy in teh digital age," by Steven Levy for some history in an interesting storylike format.
It may be cliche, but I would mod this up if I had the points. If they would drop all 16bit support it would help a ton. Hell, if there were actually enough 64bit software out there it would be even better if they released a new completely 64bit OS. That can't happen until more vendors recompile their apps, so I figure we will always have at least two generations being supported.
200GB is already on the small end for new drives, so by your estimates it will affect everything from now on. I have heard estimates that 1TB drives will be reasonably priced by 2007, by then it should be standard.
They are, it is just that things occur on different scales. Graphics processing is hugely parallel, most other code isn't unfortunately. There are fundemental differences that must be dealt with,
a general purpose CPU is much more difficult to design than a dedicated GPU.
You are probably right, even as a drunk AC. $35 is cheap though, even in the smaller town I worked in high school we charged $65/hr for repair work. Making Sony pay for actual damages (whatever the actual cost is) would be more effective than having them pay $7.50 and a CD.
2 Hours of PC repair at $100/hr. per computer affected. It seems reasonable. The average user doesn't have the tools/knowledge to un-root their system, so lets assume they had to pay someone to do it. Time is money anyway, having to spend an afternoon to fix it is worth something.
That would be a painful settlement. How many thousands/millions? of PCs were hit?
Thanks for pointing that out. Still, most people aren't gamers. Most use only a tiny fraction of the available power they have. The difference between using 5% and 15% of the CPU will be undetectable to them. The casual computer user is also more apt to be affected by rootkits & viruses as well, making the problem worse.
I should probably do more research on systems stuff in the future, it really isn't my focus.
It can be very very close though. We are not talking about emulation (running ppc code on x86 or similar) here, it is virtulization. The instructions are still native, just being passed through another transparent layer.
I think your last sentence hit it. There are groups producing huge amounts of data that needs to be stored then processed. What is the point in having 10,000 CPUs crunching numbers only to have the system I/O bound by the hard disk? Memory is still a couple orders of magnitude behind hard drives in size so they have to cache data on the disk at some point.
I'm looking into it. It is a rather small project, possibly with huge ramifications if it works, so a simple solution is best. And I got XP SP2 & Visual Studio 2005 for free with the right academic license
.
I agree. I just started in C# because a colleague had a large base of code we needed in the language already. It is actually quite a nice language. The most painful part was installing Windows for the first time in years.
Good point. In those cases local exploits would be a real problem. I would also assume that there are some heavy restrictions on what users can run. On OS X all users by default have the ability to run & install anything so long as it doesn't modify a system file/setting, which requires the root password.
Being on a cable modem puts your machine on the same network as others, but it does not give them an account on your machine. This hack would be no different than the person sitting down in front of the actual mini and logging in there. He already had some permissions on the machine and used a local exploit.
It would be like asking the Pentagon for a username on their server, because hey, it isn't root, you can't do any damage. No admin in their right mind would do it.
I don't know anyone that uses the AOL client on OS X. Most use iChat, and some use AdiumX. I stick to iChat since I don't use any advanced features and it integrates into the address book and email apps nicely.
Probably the same reason my dad got me a BB gun, model rockets, chemistry kits, and a bunch of other stuff. He had fun with them as a kid, he had a kid, why not have fun with you kid?
He didn't get into video games with me much, but it was him who started me on computers. We did play a lot of Sim City and similar games back in the day though. He also got me a GameCube for Christmas a year ago, just what a college senior needed. You never need to grow up all the way.
I plug my iPod, laptop, whatever, into the aux in on my stereo. A mini-jack to RCA cable costs about $10 or $15 and works great. If I needed to charge my iPod while playing I suppose I could move an extra powered USB hub next to my stereo also. It may not look as cool, but it saves me $300 and probably (definitely) sounds better anyway.
Having a laptop can be very usefull. Requiring students to purchase one, and more importantly a specific brand, is a bit questionable. I can see requiring a computer of some sort. They should be free to use whatever brand they want though. The school won't provide support for it anyway (even if they say they will, they won't).
I made it through my undergrad just fine without a laptop. I always lived near/on campus so my computer was close, and there are machines all over campus anyway. A laptop would have been nice, that I will grant. Now I commute a good 40 minutes to campus for grad school. Having a laptop now is great. I always have all of my files, the same software, etc. with me. Since I never know how long I will be somewhere it helps to have the flexibility. I could still make do without it though.
Slashdot Mirror
I think we, the AI community, are making actual progress. The problem is that the problem is much harder than people thought it would be back when it first emerged.
Statistical models have done wonders for a lot of things. Classification, mentioned above, is one of the most obvious successes. Natural language processing is another surprising success of statistical methods. The use of hidden markov models has solved a number of problems that were difficult using symbolic approaches (mostly dealing with syntax). The natural language understanding is still a long ways away of course.
Partially observable markov decision processes have also been used a lot in learning in uncertain environments with good success -- another technique from stats.
The problem with AI as a whole is that there is so much knowledge. It is really incredible how much we know. Not even in an academic sense, you know things will fall, how to balance, and all sorts of "common sense" knowledge. Modeling this in a symbolic way is very difficult because of the large amount of information. It is also hard to express. Formalisms such as first order predicate calculus are often used, but they have limitations.
Statistical models are appealing because we do not have to manually write down knowledge. The machine can learn by itself (to some extent). This is probably why machine learning is one of the hottest topics right now.
So keep faith in the smart people trying to work on AI -- just don't expect true intelligent machines for some time yet. Advances are constantly being made in smaller domain-specific areas though.
He just spoke at the University of Illinois at Chicago on Friday and the first thing did was tell people they should buy some GNU/GPL/FSF merchandise that he brought with him. Then he milled around for a while so people could before he began his talk.
When public key crypto came out the government had a fit. Actually, the whole history of cryptography and the NSA is interesting. They fought it tooth and nail, but eventually had to come to terms with the public having strong crypto. Check out "Crypto, How the code rebels beat the government - saving privavy in teh digital age," by Steven Levy for some history in an interesting storylike format.
BZip2 uses Huffman Encoding, Move to Front Encoding, and Burrow-Wheelers Transformations. So Huffman does get a lot of credit.
It may be cliche, but I would mod this up if I had the points. If they would drop all 16bit support it would help a ton. Hell, if there were actually enough 64bit software out there it would be even better if they released a new completely 64bit OS. That can't happen until more vendors recompile their apps, so I figure we will always have at least two generations being supported.
200GB is already on the small end for new drives, so by your estimates it will affect everything from now on. I have heard estimates that 1TB drives will be reasonably priced by 2007, by then it should be standard.
Yet Mario64 was 3D and not a side scroller. It was also a great game. Simple to pick up and lots of fun.
They are, it is just that things occur on different scales. Graphics processing is hugely parallel, most other code isn't unfortunately. There are fundemental differences that must be dealt with, a general purpose CPU is much more difficult to design than a dedicated GPU.
I pretty sure he was joking.
You are probably right, even as a drunk AC. $35 is cheap though, even in the smaller town I worked in high school we charged $65/hr for repair work. Making Sony pay for actual damages (whatever the actual cost is) would be more effective than having them pay $7.50 and a CD.
2 Hours of PC repair at $100/hr. per computer affected. It seems reasonable. The average user doesn't have the tools/knowledge to un-root their system, so lets assume they had to pay someone to do it. Time is money anyway, having to spend an afternoon to fix it is worth something.
That would be a painful settlement. How many thousands/millions? of PCs were hit?
Thanks for pointing that out. Still, most people aren't gamers. Most use only a tiny fraction of the available power they have. The difference between using 5% and 15% of the CPU will be undetectable to them. The casual computer user is also more apt to be affected by rootkits & viruses as well, making the problem worse.
I should probably do more research on systems stuff in the future, it really isn't my focus.
It can be very very close though. We are not talking about emulation (running ppc code on x86 or similar) here, it is virtulization. The instructions are still native, just being passed through another transparent layer.
I think your last sentence hit it. There are groups producing huge amounts of data that needs to be stored then processed. What is the point in having 10,000 CPUs crunching numbers only to have the system I/O bound by the hard disk? Memory is still a couple orders of magnitude behind hard drives in size so they have to cache data on the disk at some point.
Just google for Richard Geere and Gerbil.
I'm looking into it. It is a rather small project, possibly with huge ramifications if it works, so a simple solution is best. And I got XP SP2 & Visual Studio 2005 for free with the right academic license .
I agree. I just started in C# because a colleague had a large base of code we needed in the language already. It is actually quite a nice language. The most painful part was installing Windows for the first time in years.
Good point. In those cases local exploits would be a real problem. I would also assume that there are some heavy restrictions on what users can run. On OS X all users by default have the ability to run & install anything so long as it doesn't modify a system file/setting, which requires the root password.
It would be like asking the Pentagon for a username on their server, because hey, it isn't root, you can't do any damage. No admin in their right mind would do it.
I don't know anyone that uses the AOL client on OS X. Most use iChat, and some use AdiumX. I stick to iChat since I don't use any advanced features and it integrates into the address book and email apps nicely.
He didn't get into video games with me much, but it was him who started me on computers. We did play a lot of Sim City and similar games back in the day though. He also got me a GameCube for Christmas a year ago, just what a college senior needed. You never need to grow up all the way.
I plug my iPod, laptop, whatever, into the aux in on my stereo. A mini-jack to RCA cable costs about $10 or $15 and works great. If I needed to charge my iPod while playing I suppose I could move an extra powered USB hub next to my stereo also. It may not look as cool, but it saves me $300 and probably (definitely) sounds better anyway.
In all fairness, the next two results are for Jimmy Carter and Micheal Moore. Then it is page after page of story about the results.
Having a laptop can be very usefull. Requiring students to purchase one, and more importantly a specific brand, is a bit questionable. I can see requiring a computer of some sort. They should be free to use whatever brand they want though. The school won't provide support for it anyway (even if they say they will, they won't).
I made it through my undergrad just fine without a laptop. I always lived near/on campus so my computer was close, and there are machines all over campus anyway. A laptop would have been nice, that I will grant. Now I commute a good 40 minutes to campus for grad school. Having a laptop now is great. I always have all of my files, the same software, etc. with me. Since I never know how long I will be somewhere it helps to have the flexibility. I could still make do without it though.
There is research going on in this field, with some success even. It is a hard problem though.