Fair enough, but far too many people are going to actually believe it. I see this attitude about our government all the time, where people have diagnosed the problem, but incorrectly identify the biggest cause. It makes me sad to see it perpetuated. That is why I posted.
Your hate blinds you, and works to the advantage to the very policies you despise. Repeat after me: Congress writes the laws; Congress controls the money; The Executive implements the laws. If you don't like laws with bad portions tacked on, blame Congress, not the president, who has no line-item veto (thank Congress for that one too). This is doubly true for anything spending related, as that is Congress' job with its "power of the purse". If you want to change things, fight the battle where it matters, in the congressional elections. Far too many people focus only on the presidential election, losing sight of the true seat of power in our government. The best way a presidental election can be helpful is by electing from the opposite party as the majority in Congress (i.e. voting for gridlock). What you can't expect, however, is a president to veto crappy laws from his own party. Don't blame the messenger.
Think of Clinton in his first two years (Clipper chip, anyone?), versus his last six. When did he do better? Look at who was the majority in Congress during those periods. Educating, isn't it? In other words, a president is at his best when he is a brake on the stupidity of Congress. Of course, for an issue like this, even Clinton/Gore/Kerry are not going to veto some enormous spending bill for some obscure and relatively minor addition. For that kind of thing, you can only blame the ones who created it, which is Congress. That is where you should fight your battle.
Since it seems to be plain USB, I am absolutely going to buy one of these controllers as soon as a Linux driver comes out. I have a Logitech PS2-copy dual-analog controller, and I've used the thing in a lot of my programs (robotics stuff and games), as well as for general gaming. It's always neat to play with new forms of input. I'm still working on a game idea that requires both a dancepad and a dual-analog controller.
My post was meant as a smartass remark, but I think your explanation is probably the right one. Technically, though, it's still only 3 axes (X, Y, Z); "rotational axes" in your comment refers to rotation around X/Y/Z, while "regular axes" refers to linear movement along X/Y/Z.
Well, if one wanted to be a smartass, one could point out that the axes of rotation do not need to lie exactly parallel to the translational axes. Nobody ever claimed all six axes were orthogonal. Really though, you are picking nits. FTFA they say "six degrees of freedom" which is correct. I don't think it's necessarily wrong to call a d.o.f. an axis in colloquial speech.
What's suspicious is the summary says "linear movement" along 6 axes. This could be construed to mean rotation movement is not tracked.
How is that suspicious for a summary written by a random user and posted by Zonk? TFA doesn't say that, which is all that matters. We're lucky if a summary (1) isn't a dupe, (2) does not contain glaring factual errors, and (3) has decent spelling and grammar.
The server is half dead now, so I put up a real mirror of the actual mpg files here. It will not be up permanently, so somebody should post these videos with bittorrent or put it on youtube. You've been warned:)
For radar cross-section, foam and plastic parts basically don't count. So, it's more of a question of how small they can get the engines, electronics, and camera assembly. A lot of the overall dimensions should be taken up by the wing, which could be made of materials that radar will pass through.
I think researchers need to stop this sort of research which is based on hollywood science craption and focus on creating things that are actually real...
What part of Entertainment Technology Center made you think it had any kind of strong AI? Words such as "animatronic character" and "armature" (from puppetry) are used to desctibe Quasi, and give it away immediately. The article doesn't claim it to be more (The summary does, perhaps, but that isn't the project's fault, is it?). It's a robot in the same sense as "robot wars", which IMO kind of diluted the word "robot", but that's what caused the word "autonomous robot" to be coined. This isn't being run out of the CS department nor the Robotics Institute, so I'm not at all suprised at the lack of autonomy.
If you want people at the same university working on modular robots, check out Claytronics and these snake robots. Let's let the different researchers do what they do best, okay?
Well written. You also explain what the heck the article was referring to, because it isn't clear in the summary. I had mod points a few hours ago, but alas they are no more. May you be modded above the junk and the trolls.
You can almost always turn it off, but you won't know its annoying in most games until you have already been playing awhile. This is particularly true for games that decide to have some "earthquake" sequence when a boss comes out. Force feedback would be neat, but the rumble is just simply annoying most of the time, and you have to dig in the menu and turn it off, unless you are playing a multiplayer game and the other people want to leave it on:(
If game designers worked for phone companies, cell phones would vibrate every time you dial a wrong number or someone hangs up on you. I don't think that would add to the experience of owning a phone though:)
Rumble can, and does, enhance some titles. However there are invariably titles which overuse it, ruining an otherwise fine gaming experience. I don't mind getting a thump when I hit a wall in the game. What I can't stand is a constant rumble if an engine is damaged, or during some "The boss is coming out of the ground" sequence that lasts five minutes. Give people's hands a rest, please. It reminds me of the some of the first THX movies, which would abuse the capabilities by overusing them for loud sounds.
So, I for one will not miss rumble. Not for how it could be used, but for how it was too often misused.
I could download an application, run it, have it trash my user folder, add some things to my.profile, etc. The truth is that the current 'security' on just about every system out there is a joke if you consider intentionally running a (secretly) malicious application a security problem.
Well, there's at least one project to do this kind of thing, which got taken up by a popular distribution. The fancy security certified OSes have been doing MAC for a long time. Now it's more a case of getting them distrubted and creating profiles for well behaved apps. It's a big project though, as modelling the 1000s of programs in a normal Linux distribution is harder than the 10s of apps a secure government computer might see.
I depends on what you really mean by balance. FYI QRIO and ASIMO have already demonstrated running capability (though just barely). You are specifically talking about walking though, but obviously a running robot is falling, and thus demonstrating some of the capability that you'd like to see. Their fastest walks are probably still balanced in the ZMP-always-inside-foot sense, although the exact methods are closely gaurded secrets by their respective companies. A robot that maintains the ZMP inside its footprint is still falling in the sense that it cannot simply halt at a given point in its walk cycle without falling over; they are not staticly stable.
- A generation ago is NOT 1994. A generation ago would have been more like 1976, so...WRONG
Good to know you didn't even follow my link. If you looked at the graph, you would see: 1973: 47.7 crimes/1k pop 2005: 21.0 crimes/1k pop
I'd say that's an improvement. A cursory examination of that site would show all sorts of data going back to the 1970s. For some things such as murder, data back to 1900.
Firstly, they DIDN'T include crimes against children UNDER the age of 12, which makes their sources questionable at best with regards to the decline of violent crime against all children.
That's because it was based on interviews. If you would like to argue that the crime rates are independent for those under 12 versus those over 12, please support this with some a reference to a study. Again, if you went to the website, you can reach a crime table split up by age (in two clicks)
The age groups 12-15, 16-19, and 20-24 are all fairly correlated, and show steep drops. If you want to argue that children in the 8-11 range are completely different, it is up to YOU to supply evidence. The null hypothesis says there is no difference unless you demonstrate otherwise (Occam's razor, and all that). Children under 8 should probably not be left unsupervised, so the current level of danger in the environment shouldn't affect choices by the parent on whether to let them go unsupervised.
Secondly, Ohhhhhhhhhh look at that....They interviewed a whole 134,000 people, out of the current population(according to another branch of Minitrue) of 298+ MILLION U.S. Comrades
That's far larger than most medical studies, including ones for treatments where people's lives are at stake. Please go tell the medical community how wrong they are. Also, your focus on the percentage demonstrates a lack of understanding of statistics -- 1% of 100 people is a very biased sample, whereas 1% of a million people can often be quite accurate. Now, it is true that the graph doesn't provide the estimations of accuracy in the study, but I didn't claim that the change was statistically significant either. I'd be suprised if it wasn't, given the sample size however.
Now let's take a look at the source of their "facts":... And who designed and owns the NCVS? The Bureau of Justice Statistics. And what Minitrue arm is the BJS a part of...well looky here, the DOJ.
Yes, I know. It's a conspiracy across administrations, since 1972. I'm well aware of your position.
So, based on the inadvertant admissions of your Minitrue's sites, they summarize an ENTIRE NATION'S crime statistics based on interviews with less than 1% of that same population,
If you would like to claim that the streets are more dangerous, it is up to you to cite a better/bigger study (hint: this is the biggest one in the US). Go ahead and throw this study out. In the absence of proof that it is more dangerous, we default to the null hypothesis that it is the same. That still supports my argument that it is not more dangerous now than a generation ago.
using their own methodologies which cannot be verified by an outside independent source, and you take that at face value.
Actually, that study includes the methodology on the website in quite a bit of detail. You can download PDFs of it. As far as taking the study at face value, I'd rather say that I make decisions on the information available, weighted by my estimation of the accuracy of the source. It is true that the DOJ (or any other source) is not free from bias, but in the absence of other accurate, unbiased sources (the popular press has poor methodology in general), that's what I make my decisions on.
It is pretty clear that your arguments are not based on any evidence, because you don't provide any. Stating that a large port
In other words "I have no real rebuttal of their data collection or methods, so I'll just bring up a 1984 reference." Let me guess, you think it's been an ongoing and successful lie and cover-up of the truth, running since 1972 through 7 administrations and 4 party transitions. Somehow the data are skewed, but not consistently so, since we are a doing a relative comparison. They've managed to hide past data, even though its been published every year and picked up in the press.
That would be an impressive distortion to say the least. In the real world, we have the Washington Post and the New York Times, who make sure no scandal will stay uncovered in the long term. Also, the graph clearly indicates that nearly all of the improvement happened during Clinton's tenure, which is not something the current DOJ leadership would be all that proud of.
I have a 24" monitor at 1920x1200, and I still use all of my 12 virtual desktops on Linux. CTWM has a very simple visual pager, and I've never had to spend more than 5 minutes explaining to someone else how to use it. I don't know why other OS'es don't adopt this, as its no different from RTS games with their map views (even ones by MSFT).
It was also funny to watch OSX users rave over Expose... I can maximize my pager window too, and CTWM has been able to do that since 1993.
nice game... Could you modify them to play on grass?
That's the long term goal. Our robots are in the small-size league, so they are smaller and (relatively) cheaper, which allows us to focus on teamwork and speed. There is a middle-size league with larger robots (0.5 x 0.5 x 0.75 meters), which currently is just on a larger carpet, but may go outside eventually. There's also a humanoid league, but right now it'd be a stretch to call what they play "soccer" just yet -- most of their time is spent walking (slowly) and trying to find out where things are. Someday, humanoids will be on teams that play like the small-size robots outside, but there are a lot of technical hurdles to overcome.
cannot be bought, don't loose morale when 2 goals in advance, don't do drugs and probably can't get killed by drunk
This is actually a pretty interesting part of robot soccer; Score differentials are much larger than in normal sport, becase as you say, they do not let off when ahead. On the other hand, this drops the emotianal link present in so many sports; Our robots don't care if they are ahead or behind, they just play. We had thought of making them play more defensively when ahead, but in reality it makes sense to keep pushing, as the robot aren't getting tired. A few teams have tried some "hail mary" tactics to come from behind, such as pulling the goalie near the end of a match when behind. Such tactics haven't had too much of an impact yet.
In a lot of games, it can actually be a little embarrasing. In many wins, you don't just beat a team, you keep punishing them, so you feel a little bad for the humans who made the robot team. Thus you can feel a little bad for making the other team look so bad. Of course, there are some intense rivalries between some teams, so I don't always feel that way.
poor AI in video games? it's a feature, not a bug: it's much more interesting to plan your game when having idiotic bots in your team than getting your ass kicked by intelligent agents on the other end of the wire.
I think the grail of video games is AIs that play like unskilled humans, rather than automatons. You want the behavior to make sense at some human level, but you don't want it to abuse its true advantages as some sort of super 0ms ping super sniper. It'll also be nice to see games taking advantage of limited sensing in more cases.
FPS's often distribute a graph of waypoints (aka planning mesh, navigation network, etc) with each level file. Unless you have dynamic terrain, you could do the same, right? Of course, FPS's are famous for their exceedingly sparse planning graphs, so maybe if yours was more dense (to make the characters move nicely), I guess they could be too big.
Recent motion planning research has been about how to get away without pre-generating such a graph, and instead building it incrementally on the fly. Two methods in particular that have seen a use are PRM and RRT. While nothing is faster than using a preprocessed grid, using something such as RRT as a "local" planner in order to reach a sparse global graph (from preprocessing) would probably be quite applicable to modern games.
The problem with the state of AI today is not that the algorithms are too processor-intensive, it's that they flat-out suck.
Please don't take what you see in games to be state of the art. Watch this video of my RoboCup team and tell me that AI still completely sucks. You'll see two teams playing 5 on 5 soccer with zero human input, i.e. fully autonomous. Game AIs may suck, but that's because their AI programmers are graphics people trying to do AI. The result looks about as good as me trying to make a state of the art graphics engine.
The only reasonable application of hardware AI acceleration that I can think of would be a massively parallel chip that runs thousands or millions of neural net nodes at once... but this would mainly be a benefit for academic AI research, not for games.
Neural nets died down as a fad in academic circles almost 10 years ago. There's a common saying that "Neural nets are the second best way to do everything."... meaning that if you analyze a problem, some other approach almost always turns out to work better. They are a reasonable approach for unstructured classification problems that aren't fully understood, but after some analysis other approaches almost always take over. This has been the case with things like OCR and face recognition.
I'm pretty sure that most games use simple "heuristic" algorithms for AI, rather than anything complicated like neural nets or Bayesian learning or SVM.
NNs, Naive Bayes, and SVMs are all classifiers (and often slow ones at that). They aren't really directly applicable for defining policies for an agent, so they don't get used much (as well they shouldn't). Most agent decision systems use a combination of heirarchical finite state machines (FSMs), planning on a small set of high level actions, and motion planning.
Games tend toward the absolute simplest of FSMs with only binary inputs, and yield the expected highly rigid behavior in a given situation. For the most part, they don't even use randomness, which is absolutely necessary in any kind of situation where one player is trying to outguess another. I've heard that non-RTS games only budget about 1% of the CPU to AI, and it shows. Rich FSMs, action-based planning, and proper motion planning get swept aside, and that is unfortunate. However the coming multi-core revolution may offer some hope. Game programmers are having trouble splitting up graphics routines, so it might be that AI can get the core or two that it deserves when we hit quad-core CPUs. Due to the many algorithms, AI benefits from general purpose CPUs, and parallelizes quite well.
Whether enough real AI people will ever get hired to do games right remains to be seen. At the moment it seems even primarily systems companies like Google are more interested in AI people than game companies.
We thought about this a fair bit where I worked. We decided that it just wasn't doable.
One thing I could imagine doing for a game is offloading just the path planning. You could make it a fairly dynamic thing where no particular client has responsibility for an NPC. Each client would recieve a source and target position for some NPCs, and the client could plan paths and send back a nearby waypoint for each NPC that will take it partway to its goal. You could assign multiple client computers the same problem and pick from the returned results randomly. This would only work for fairly static worlds, but AFAIK most MMORPGs don't allow dynamic obstacles (one reason for not having players be obstacles).
You do realize that A* if from 1968, and that things have improved a bit since then, right? It is better than algorithms used in current video games maybe, but that's as far as I'd take it.
Slashdot Mirror
Fair enough, but far too many people are going to actually believe it. I see this attitude about our government all the time, where people have diagnosed the problem, but incorrectly identify the biggest cause. It makes me sad to see it perpetuated. That is why I posted.
Your hate blinds you, and works to the advantage to the very policies you despise. Repeat after me: Congress writes the laws; Congress controls the money; The Executive implements the laws. If you don't like laws with bad portions tacked on, blame Congress, not the president, who has no line-item veto (thank Congress for that one too). This is doubly true for anything spending related, as that is Congress' job with its "power of the purse". If you want to change things, fight the battle where it matters, in the congressional elections. Far too many people focus only on the presidential election, losing sight of the true seat of power in our government. The best way a presidental election can be helpful is by electing from the opposite party as the majority in Congress (i.e. voting for gridlock). What you can't expect, however, is a president to veto crappy laws from his own party. Don't blame the messenger.
Think of Clinton in his first two years (Clipper chip, anyone?), versus his last six. When did he do better? Look at who was the majority in Congress during those periods. Educating, isn't it? In other words, a president is at his best when he is a brake on the stupidity of Congress. Of course, for an issue like this, even Clinton/Gore/Kerry are not going to veto some enormous spending bill for some obscure and relatively minor addition. For that kind of thing, you can only blame the ones who created it, which is Congress. That is where you should fight your battle.
Since it seems to be plain USB, I am absolutely going to buy one of these controllers as soon as a Linux driver comes out. I have a Logitech PS2-copy dual-analog controller, and I've used the thing in a lot of my programs (robotics stuff and games), as well as for general gaming. It's always neat to play with new forms of input. I'm still working on a game idea that requires both a dancepad and a dual-analog controller.
Well, if one wanted to be a smartass, one could point out that the axes of rotation do not need to lie exactly parallel to the translational axes. Nobody ever claimed all six axes were orthogonal. Really though, you are picking nits. FTFA they say "six degrees of freedom" which is correct. I don't think it's necessarily wrong to call a d.o.f. an axis in colloquial speech.
How is that suspicious for a summary written by a random user and posted by Zonk? TFA doesn't say that, which is all that matters. We're lucky if a summary (1) isn't a dupe, (2) does not contain glaring factual errors, and (3) has decent spelling and grammar.
Probably from rotation. The PS3 demos showed people rotating the controller to fly an airplane, not translating it.
The server is half dead now, so I put up a real mirror of the actual mpg files here. It will not be up permanently, so somebody should post these videos with bittorrent or put it on youtube. You've been warned :)
For radar cross-section, foam and plastic parts basically don't count. So, it's more of a question of how small they can get the engines, electronics, and camera assembly. A lot of the overall dimensions should be taken up by the wing, which could be made of materials that radar will pass through.
I think researchers need to stop this sort of research which is based on hollywood science craption and focus on creating things that are actually real...
What part of Entertainment Technology Center made you think it had any kind of strong AI? Words such as "animatronic character" and "armature" (from puppetry) are used to desctibe Quasi, and give it away immediately. The article doesn't claim it to be more (The summary does, perhaps, but that isn't the project's fault, is it?). It's a robot in the same sense as "robot wars", which IMO kind of diluted the word "robot", but that's what caused the word "autonomous robot" to be coined. This isn't being run out of the CS department nor the Robotics Institute, so I'm not at all suprised at the lack of autonomy.
If you want people at the same university working on modular robots, check out Claytronics and these snake robots. Let's let the different researchers do what they do best, okay?
Well written. You also explain what the heck the article was referring to, because it isn't clear in the summary. I had mod points a few hours ago, but alas they are no more. May you be modded above the junk and the trolls.
You can almost always turn it off, but you won't know its annoying in most games until you have already been playing awhile. This is particularly true for games that decide to have some "earthquake" sequence when a boss comes out. Force feedback would be neat, but the rumble is just simply annoying most of the time, and you have to dig in the menu and turn it off, unless you are playing a multiplayer game and the other people want to leave it on :(
:)
If game designers worked for phone companies, cell phones would vibrate every time you dial a wrong number or someone hangs up on you. I don't think that would add to the experience of owning a phone though
Rumble can, and does, enhance some titles. However there are invariably titles which overuse it, ruining an otherwise fine gaming experience. I don't mind getting a thump when I hit a wall in the game. What I can't stand is a constant rumble if an engine is damaged, or during some "The boss is coming out of the ground" sequence that lasts five minutes. Give people's hands a rest, please. It reminds me of the some of the first THX movies, which would abuse the capabilities by overusing them for loud sounds.
So, I for one will not miss rumble. Not for how it could be used, but for how it was too often misused.
I could download an application, run it, have it trash my user folder, add some things to my .profile, etc. The truth is that the current 'security' on just about every system out there is a joke if you consider intentionally running a (secretly) malicious application a security problem.
Well, there's at least one project to do this kind of thing, which got taken up by a popular distribution. The fancy security certified OSes have been doing MAC for a long time. Now it's more a case of getting them distrubted and creating profiles for well behaved apps. It's a big project though, as modelling the 1000s of programs in a normal Linux distribution is harder than the 10s of apps a secure government computer might see.
I depends on what you really mean by balance. FYI QRIO and ASIMO have already demonstrated running capability (though just barely). You are specifically talking about walking though, but obviously a running robot is falling, and thus demonstrating some of the capability that you'd like to see. Their fastest walks are probably still balanced in the ZMP-always-inside-foot sense, although the exact methods are closely gaurded secrets by their respective companies. A robot that maintains the ZMP inside its footprint is still falling in the sense that it cannot simply halt at a given point in its walk cycle without falling over; they are not staticly stable.
- A generation ago is NOT 1994. A generation ago would have been more like 1976, so...WRONG
.htm
...
Good to know you didn't even follow my link. If you looked at the graph, you would see:
1973: 47.7 crimes/1k pop
2005: 21.0 crimes/1k pop
I'd say that's an improvement. A cursory examination of that site would show all sorts of data going back to the 1970s. For some things such as murder, data back to 1900.
Firstly, they DIDN'T include crimes against children UNDER the age of 12, which makes their sources questionable at best with regards to the decline of violent crime against all children.
That's because it was based on interviews. If you would like to argue that the crime rates are independent for those under 12 versus those over 12, please support this with some a reference to a study. Again, if you went to the website, you can reach a crime table split up by age (in two clicks)
http://www.ojp.usdoj.gov/bjs/glance/tables/vagetab
The age groups 12-15, 16-19, and 20-24 are all fairly correlated, and show steep drops. If you want to argue that children in the 8-11 range are completely different, it is up to YOU to supply evidence. The null hypothesis says there is no difference unless you demonstrate otherwise (Occam's razor, and all that). Children under 8 should probably not be left unsupervised, so the current level of danger in the environment shouldn't affect choices by the parent on whether to let them go unsupervised.
Secondly, Ohhhhhhhhhh look at that....They interviewed a whole 134,000 people, out of the current population(according to another branch of Minitrue) of 298+ MILLION U.S. Comrades
That's far larger than most medical studies, including ones for treatments where people's lives are at stake. Please go tell the medical community how wrong they are. Also, your focus on the percentage demonstrates a lack of understanding of statistics -- 1% of 100 people is a very biased sample, whereas 1% of a million people can often be quite accurate. Now, it is true that the graph doesn't provide the estimations of accuracy in the study, but I didn't claim that the change was statistically significant either. I'd be suprised if it wasn't, given the sample size however.
Now let's take a look at the source of their "facts":
And who designed and owns the NCVS? The Bureau of Justice Statistics.
And what Minitrue arm is the BJS a part of...well looky here, the DOJ.
Yes, I know. It's a conspiracy across administrations, since 1972. I'm well aware of your position.
So, based on the inadvertant admissions of your Minitrue's sites, they summarize an ENTIRE NATION'S crime statistics based on interviews with less than 1% of that same population,
If you would like to claim that the streets are more dangerous, it is up to you to cite a better/bigger study (hint: this is the biggest one in the US). Go ahead and throw this study out. In the absence of proof that it is more dangerous, we default to the null hypothesis that it is the same. That still supports my argument that it is not more dangerous now than a generation ago.
using their own methodologies which cannot be verified by an outside independent source, and you take that at face value.
Actually, that study includes the methodology on the website in quite a bit of detail. You can download PDFs of it. As far as taking the study at face value, I'd rather say that I make decisions on the information available, weighted by my estimation of the accuracy of the source. It is true that the DOJ (or any other source) is not free from bias, but in the absence of other accurate, unbiased sources (the popular press has poor methodology in general), that's what I make my decisions on.
It is pretty clear that your arguments are not based on any evidence, because you don't provide any. Stating that a large port
Yes, but you have to split it approximately 78,000 ways[1]
[1] http://primes.utm.edu/howmany.shtml
In other words "I have no real rebuttal of their data collection or methods, so I'll just bring up a 1984 reference." Let me guess, you think it's been an ongoing and successful lie and cover-up of the truth, running since 1972 through 7 administrations and 4 party transitions. Somehow the data are skewed, but not consistently so, since we are a doing a relative comparison. They've managed to hide past data, even though its been published every year and picked up in the press.
That would be an impressive distortion to say the least. In the real world, we have the Washington Post and the New York Times, who make sure no scandal will stay uncovered in the long term. Also, the graph clearly indicates that nearly all of the improvement happened during Clinton's tenure, which is not something the current DOJ leadership would be all that proud of.
I declare a subthread for people with prime user id numbers only.
factor 92219: 92219
The media has made you paranoid. The reality is that we are much safer than we were a generation ago:
http://www.ojp.usdoj.gov/bjs/glance/viort.htm
I have a 24" monitor at 1920x1200, and I still use all of my 12 virtual desktops on Linux. CTWM has a very simple visual pager, and I've never had to spend more than 5 minutes explaining to someone else how to use it. I don't know why other OS'es don't adopt this, as its no different from RTS games with their map views (even ones by MSFT).
It was also funny to watch OSX users rave over Expose... I can maximize my pager window too, and CTWM has been able to do that since 1993.
nice game ... Could you modify them to play on grass?
That's the long term goal. Our robots are in the small-size league, so they are smaller and (relatively) cheaper, which allows us to focus on teamwork and speed. There is a middle-size league with larger robots (0.5 x 0.5 x 0.75 meters), which currently is just on a larger carpet, but may go outside eventually. There's also a humanoid league, but right now it'd be a stretch to call what they play "soccer" just yet -- most of their time is spent walking (slowly) and trying to find out where things are. Someday, humanoids will be on teams that play like the small-size robots outside, but there are a lot of technical hurdles to overcome.
cannot be bought, don't loose morale when 2 goals in advance, don't do drugs and probably can't get killed by drunk
This is actually a pretty interesting part of robot soccer; Score differentials are much larger than in normal sport, becase as you say, they do not let off when ahead. On the other hand, this drops the emotianal link present in so many sports; Our robots don't care if they are ahead or behind, they just play. We had thought of making them play more defensively when ahead, but in reality it makes sense to keep pushing, as the robot aren't getting tired. A few teams have tried some "hail mary" tactics to come from behind, such as pulling the goalie near the end of a match when behind. Such tactics haven't had too much of an impact yet.
In a lot of games, it can actually be a little embarrasing. In many wins, you don't just beat a team, you keep punishing them, so you feel a little bad for the humans who made the robot team. Thus you can feel a little bad for making the other team look so bad. Of course, there are some intense rivalries between some teams, so I don't always feel that way.
poor AI in video games? it's a feature, not a bug: it's much more interesting to plan your game when having idiotic bots in your team than getting your ass kicked by intelligent agents on the other end of the wire.
I think the grail of video games is AIs that play like unskilled humans, rather than automatons. You want the behavior to make sense at some human level, but you don't want it to abuse its true advantages as some sort of super 0ms ping super sniper. It'll also be nice to see games taking advantage of limited sensing in more cases.
Maybe he meant queue? Crickets lined up in an orderly circular buffer seems nice to me. Then again, I'm a programmer.
FPS's often distribute a graph of waypoints (aka planning mesh, navigation network, etc) with each level file. Unless you have dynamic terrain, you could do the same, right? Of course, FPS's are famous for their exceedingly sparse planning graphs, so maybe if yours was more dense (to make the characters move nicely), I guess they could be too big.
Recent motion planning research has been about how to get away without pre-generating such a graph, and instead building it incrementally on the fly. Two methods in particular that have seen a use are PRM and RRT. While nothing is faster than using a preprocessed grid, using something such as RRT as a "local" planner in order to reach a sparse global graph (from preprocessing) would probably be quite applicable to modern games.
The problem with the state of AI today is not that the algorithms are too processor-intensive, it's that they flat-out suck.
... meaning that if you analyze a problem, some other approach almost always turns out to work better. They are a reasonable approach for unstructured classification problems that aren't fully understood, but after some analysis other approaches almost always take over. This has been the case with things like OCR and face recognition.
Please don't take what you see in games to be state of the art. Watch this video of my RoboCup team and tell me that AI still completely sucks. You'll see two teams playing 5 on 5 soccer with zero human input, i.e. fully autonomous. Game AIs may suck, but that's because their AI programmers are graphics people trying to do AI. The result looks about as good as me trying to make a state of the art graphics engine.
The only reasonable application of hardware AI acceleration that I can think of would be a massively parallel chip that runs thousands or millions of neural net nodes at once... but this would mainly be a benefit for academic AI research, not for games.
Neural nets died down as a fad in academic circles almost 10 years ago. There's a common saying that "Neural nets are the second best way to do everything."
I'm pretty sure that most games use simple "heuristic" algorithms for AI, rather than anything complicated like neural nets or Bayesian learning or SVM.
NNs, Naive Bayes, and SVMs are all classifiers (and often slow ones at that). They aren't really directly applicable for defining policies for an agent, so they don't get used much (as well they shouldn't). Most agent decision systems use a combination of heirarchical finite state machines (FSMs), planning on a small set of high level actions, and motion planning.
Games tend toward the absolute simplest of FSMs with only binary inputs, and yield the expected highly rigid behavior in a given situation. For the most part, they don't even use randomness, which is absolutely necessary in any kind of situation where one player is trying to outguess another. I've heard that non-RTS games only budget about 1% of the CPU to AI, and it shows. Rich FSMs, action-based planning, and proper motion planning get swept aside, and that is unfortunate. However the coming multi-core revolution may offer some hope. Game programmers are having trouble splitting up graphics routines, so it might be that AI can get the core or two that it deserves when we hit quad-core CPUs. Due to the many algorithms, AI benefits from general purpose CPUs, and parallelizes quite well.
Whether enough real AI people will ever get hired to do games right remains to be seen. At the moment it seems even primarily systems companies like Google are more interested in AI people than game companies.
We thought about this a fair bit where I worked. We decided that it just wasn't doable.
One thing I could imagine doing for a game is offloading just the path planning. You could make it a fairly dynamic thing where no particular client has responsibility for an NPC. Each client would recieve a source and target position for some NPCs, and the client could plan paths and send back a nearby waypoint for each NPC that will take it partway to its goal. You could assign multiple client computers the same problem and pick from the returned results randomly. This would only work for fairly static worlds, but AFAIK most MMORPGs don't allow dynamic obstacles (one reason for not having players be obstacles).
it is better than current algorythms [sic]
You do realize that A* if from 1968, and that things have improved a bit since then, right? It is better than algorithms used in current video games maybe, but that's as far as I'd take it.