There's nothing inherently wrong in publishing in incremental steps. In fact, it can be useful to get peer review at different stages in a project.
The problem here is that the article overstates the achievement, seemingly intentionally misleading the reader into believing more has been done than actually has been.
Yea, it says, "Examples of inputs [to the AI system] include all positional, velocity, and acceleration states, estimated missile range data, visibility of each platform, bogey ratio, and number of shots taken by hostile platforms", which seems to be saying it has direct access to the information in the simulation.
Yep, from TFA, "Examples of inputs [to the AI system] include all positional, velocity, and acceleration states, estimated missile range data, visibility of each platform, bogey ratio, and number of shots taken by hostile platforms..."
Basically, it knows everything about the current state of the dog fight at all times (without having to parse any visual information). If the human pilot had access to this information, s/he would probably perform much better also. Basically, the NPC cheated.
While I'm sure its a nice AI system, its a LONG way from an autonomous fighter pilot.
I remember playing a few different games at school. Off the top of my head: Math Circus, The Incredible Machine and we even did a 5th grade project involving Sim Tower. This was all back in the mid-90's. Video games have been used in schools for decades now.
I agree completely. One hint at where this problem stems from is the fact that scientists often even have trouble communicating with other scientists outside of their own field (and these are the people writing the wikipedia articles). The goal of research is to go very deep, so scientists tend to get entrenched in their own specialized terminology and way of looking at the world. Furthermore, in my experience as a PhD student, being able to communicate scientific ideas with the broader public is not a skill that is trained or even all that valued among most scientists (because it typically doesn't help get grants), so it's not something that many scientists get good at.
"Deep learning" refers a family of machine learning techniques (such as neural-networks, convolutional neural-networks, stacked-autoencoders, etc.) that have a multi-layer architechture, typically allowing the system to learn highly non-linear functions of many variables. Each layer can be thought of as a simple learned function whose output is fed into the next layer. Such systems can often have thousands or millions of parameters to learn and thus require a LOT of training data and a fair bit of computing power/ runtime to train. But if you look at some area (e.g. object reccognition in computer vision), deep networks are currently the top techniques by a fair margin.
This seems more like basic-level stuff...
The devil is in the details. How do you best represent learning mathematically and computationally? What are mistakes and or what are the objectives? How do you encode these and how to you penalize making these mistakes in the future? These are all challenging questions.
That strikes me as the sort of thing that would be "hardwired" in everything from nematodes to primates.
Machine learning approaches have often taken inspiration from biology, however the exact neurological mechanisms of learning are not yet entirely understood. Its difficult to replicate nature. Its even more difficult when you don't yet understand nature.
In this case, it seems like the details matter but when you say "went something like this", its unclear if you've actually got their argument correct. 1) Had there been attacks on women on campus or in the area? 2) Was the closure actually to discourage anyone (including women) from hanging around the labs late at night when the crime rate is higher (before a camera was installed)? 3) Was the workload for the courses that used the labs such that it likely required late night working?
I'm guessing the answer to some or all of those questions is actually yes. You make it sound as if this was unnecessary complaining by some women, but is this the case? Furthermore, have you ever walked around the campus late at night as a woman? (or at least talked to women about this issue)?
In Canada, if you aren't using a cloth/reusable bag, you're just that fucking dude who didn't bring a bag. Almost everywhere asks and charges before handing you one. The only exception, as others have pointed out, is if you just need garbage bags and want them cheap.
What's a fake choice? Do you mean the outcomes are not different enough? Or that you don't care about the stories enough in the first place? Is your problem that you just don't like stories in general? What is your problem?
Maybe your problem is that you haven't played the Mass Effect series? Or maybe you are just curmudgeonly?
there have been a few 'pick your ending' movies before.... That's about the same level of interactivity as most games these days.
Really? What about, you know, the actual game part where you control a person, their actions and solve puzzles and perform tasks? What more do you want?
I hear people bitching without any constructive criticism.
In movies I can't choose if the main character will save the day as a paragon or a renegade, nor can I choose who the main character will pursue a romance with. In a movie I can't explore new places nor spend down time geeking out on stats. And I don't think that my options in these matters have been systematically curtailed since early games. If anything, game freedom and interactivity has progressed as quickly as graphics technology.
While I understand the argument that some AAA games are becoming Hollywoodized, we should all keep in mind that stories with characters are very fundamental to human art and entertainment. It is very natural that artists in a new medium would want to try to tell stories in it. Also, I think that in the last decade, there have been some monumental achievements on this front; games with great mechanics, stories too rich for a film and full of beautiful art, music and writing. Why are we complaining about this?
And, if someone has the next direction for video games in their back pocket, they aren't bitching here, they are making it.
People like him are the reason the middle class no longer exists.
...They all got up and made successful internet startups or car companies or space companies? Is that where the middle class went?
Or do you mean they no longer exist because they all stopped working at North American factories like Tesla or SpaceX?
Or do you mean they went somewhere else because they were no longer were inspired to be entrepreneurial because no-one had the spirit anymore or could prove it was possible to grow a successful business?
Actually, this is nothing new. Most consoles, at launch, would lose on paper to their counter-part mid-range PC at launch. So what?
1.) There's more to hardware than the listed specs. For instance, Xbox360 had lots of architectural features that made it a poor choice for general computing but excellent for a typical gaming load. One example was a cache architecture tuned for streaming data (i.e. reading of a disc or streaming in geometry and textures). Also, consoles typically require less overhead in terms of things like OS footprint, etc. than PCs. Thus, despite the numbers, consoles are typically able to keep up with PCs that out match them in terms of specs alone.
2.) Developers can focus their efforts on a single architecture when developing for consoles. This entails huge performance gains from various optimizations that would not be feasible when developing for the heterogeneous PC architecture landscape. (Certainly there are exceptions to this i.e. Game A gets optimized code for GPU B, but this is certainly not a guarantee when buying the game). Furthermore, developers have much longer to find particular optimizations for the hardware given that the console life-cycle is typically 5 - 10 years. Notice how much better the last titles in a console generation look compared to those at launch.
Finally, $300 to $400 doesn't buy you much in terms of a gaming rig. I know, I just tried to build one on that budget.
How do you define mainstream ideology? Views held by a majority of said population? How would you determine this? With polls? Where are the numbers to back this statement? What is the distribution among nations and other demographics? I can't believe that Slashdot, a community that champions science, would hold this statement up as insightful.
Everyone should be critical of your statement because it is generalizing, not supported with accompanying facts, and potentially damning to hundreds of millions of people. Bold statements require significant, scientifically sound evidence.
Furthermore, this argument is contrary to what I know from my daily life. All of the Muslims I know are very tolerant, peaceful people. Thus, I would need very compelling evidence to convince me that their world view was a threat. (A threat to who, by the way?)
Slashdot Mirror
interests of their country
Who decides the interests of a country? If you think its just the president or just the department of defence then you don't understand democracy.
Many smart people believe its in the best interests of their country (and the human race) that we don't race to weaponize AI.
conventions are dead.
Shall I fact check this for you, or would you like to take a go?
There's nothing inherently wrong in publishing in incremental steps. In fact, it can be useful to get peer review at different stages in a project.
The problem here is that the article overstates the achievement, seemingly intentionally misleading the reader into believing more has been done than actually has been.
I'd upgrade my other machines if my Windows 10 machine didn't BSOD every couple days.
Yea, it says, "Examples of inputs [to the AI system] include all positional, velocity, and acceleration states, estimated missile range data, visibility of each platform, bogey ratio, and number of shots taken by hostile platforms", which seems to be saying it has direct access to the information in the simulation.
Yep, from TFA, "Examples of inputs [to the AI system] include all positional, velocity, and acceleration states, estimated missile range data, visibility of each platform, bogey ratio, and number of shots taken by hostile platforms..."
Basically, it knows everything about the current state of the dog fight at all times (without having to parse any visual information). If the human pilot had access to this information, s/he would probably perform much better also. Basically, the NPC cheated.
While I'm sure its a nice AI system, its a LONG way from an autonomous fighter pilot.
I remember playing a few different games at school. Off the top of my head: Math Circus, The Incredible Machine and we even did a 5th grade project involving Sim Tower. This was all back in the mid-90's. Video games have been used in schools for decades now.
For anyone interested, the "quantized inertia" explanation was proposed by McCulloch, and is here: http://dx.doi.org/10.1209/0295-5075/111/60005
I agree completely. One hint at where this problem stems from is the fact that scientists often even have trouble communicating with other scientists outside of their own field (and these are the people writing the wikipedia articles). The goal of research is to go very deep, so scientists tend to get entrenched in their own specialized terminology and way of looking at the world. Furthermore, in my experience as a PhD student, being able to communicate scientific ideas with the broader public is not a skill that is trained or even all that valued among most scientists (because it typically doesn't help get grants), so it's not something that many scientists get good at.
Note to self, don't start a long-term indie game project on whatever Microsoft's next game dev platform is.
"Deep learning" refers a family of machine learning techniques (such as neural-networks, convolutional neural-networks, stacked-autoencoders, etc.) that have a multi-layer architechture, typically allowing the system to learn highly non-linear functions of many variables. Each layer can be thought of as a simple learned function whose output is fed into the next layer. Such systems can often have thousands or millions of parameters to learn and thus require a LOT of training data and a fair bit of computing power/ runtime to train. But if you look at some area (e.g. object reccognition in computer vision), deep networks are currently the top techniques by a fair margin.
This seems more like basic-level stuff...
The devil is in the details. How do you best represent learning mathematically and computationally? What are mistakes and or what are the objectives? How do you encode these and how to you penalize making these mistakes in the future? These are all challenging questions.
That strikes me as the sort of thing that would be "hardwired" in everything from nematodes to primates.
Machine learning approaches have often taken inspiration from biology, however the exact neurological mechanisms of learning are not yet entirely understood. Its difficult to replicate nature. Its even more difficult when you don't yet understand nature.
In this case, it seems like the details matter but when you say "went something like this", its unclear if you've actually got their argument correct.
1) Had there been attacks on women on campus or in the area?
2) Was the closure actually to discourage anyone (including women) from hanging around the labs late at night when the crime rate is higher (before a camera was installed)?
3) Was the workload for the courses that used the labs such that it likely required late night working?
I'm guessing the answer to some or all of those questions is actually yes. You make it sound as if this was unnecessary complaining by some women, but is this the case? Furthermore, have you ever walked around the campus late at night as a woman? (or at least talked to women about this issue)?
Everyone else sees a blue pill and a black pill, right?
In Canada, if you aren't using a cloth/reusable bag, you're just that fucking dude who didn't bring a bag. Almost everywhere asks and charges before handing you one. The only exception, as others have pointed out, is if you just need garbage bags and want them cheap.
Any particular reason this was tagged with 'global warming'?
What's a fake choice? Do you mean the outcomes are not different enough? Or that you don't care about the stories enough in the first place? Is your problem that you just don't like stories in general? What is your problem?
Maybe your problem is that you haven't played the Mass Effect series? Or maybe you are just curmudgeonly?
there have been a few 'pick your ending' movies before.... That's about the same level of interactivity as most games these days.
Really? What about, you know, the actual game part where you control a person, their actions and solve puzzles and perform tasks? What more do you want?
I hear people bitching without any constructive criticism.
In movies I can't choose if the main character will save the day as a paragon or a renegade, nor can I choose who the main character will pursue a romance with. In a movie I can't explore new places nor spend down time geeking out on stats. And I don't think that my options in these matters have been systematically curtailed since early games. If anything, game freedom and interactivity has progressed as quickly as graphics technology.
While I understand the argument that some AAA games are becoming Hollywoodized, we should all keep in mind that stories with characters are very fundamental to human art and entertainment. It is very natural that artists in a new medium would want to try to tell stories in it. Also, I think that in the last decade, there have been some monumental achievements on this front; games with great mechanics, stories too rich for a film and full of beautiful art, music and writing. Why are we complaining about this?
And, if someone has the next direction for video games in their back pocket, they aren't bitching here, they are making it.
People like him are the reason the middle class no longer exists.
...They all got up and made successful internet startups or car companies or space companies? Is that where the middle class went?
Or do you mean they no longer exist because they all stopped working at North American factories like Tesla or SpaceX?
Or do you mean they went somewhere else because they were no longer were inspired to be entrepreneurial because no-one had the spirit anymore or could prove it was possible to grow a successful business?
Well I thought this was funny, AC.
Actually, this is nothing new. Most consoles, at launch, would lose on paper to their counter-part mid-range PC at launch. So what?
1.) There's more to hardware than the listed specs. For instance, Xbox360 had lots of architectural features that made it a poor choice for general computing but excellent for a typical gaming load. One example was a cache architecture tuned for streaming data (i.e. reading of a disc or streaming in geometry and textures). Also, consoles typically require less overhead in terms of things like OS footprint, etc. than PCs. Thus, despite the numbers, consoles are typically able to keep up with PCs that out match them in terms of specs alone.
2.) Developers can focus their efforts on a single architecture when developing for consoles. This entails huge performance gains from various optimizations that would not be feasible when developing for the heterogeneous PC architecture landscape. (Certainly there are exceptions to this i.e. Game A gets optimized code for GPU B, but this is certainly not a guarantee when buying the game). Furthermore, developers have much longer to find particular optimizations for the hardware given that the console life-cycle is typically 5 - 10 years. Notice how much better the last titles in a console generation look compared to those at launch.
Finally, $300 to $400 doesn't buy you much in terms of a gaming rig. I know, I just tried to build one on that budget.
Mexican criminals have guns because they so easy to obtain in the US.
If it were more difficult to buy, sell and trade guns in the US you can be sure that there would be a lot less gun violence in Mexico.
If you instead had used European countries as examples of the effectiveness of gun control, your argument would be much harder to make.
This is why prediction results are basically meaningless without sensitivity and specificity numbers...
No
Is the era of headings written as questions over?
Unfortunately, also no.
How do you define mainstream ideology? Views held by a majority of said population? How would you determine this? With polls? Where are the numbers to back this statement? What is the distribution among nations and other demographics? I can't believe that Slashdot, a community that champions science, would hold this statement up as insightful.
Everyone should be critical of your statement because it is generalizing, not supported with accompanying facts, and potentially damning to hundreds of millions of people. Bold statements require significant, scientifically sound evidence.
Furthermore, this argument is contrary to what I know from my daily life. All of the Muslims I know are very tolerant, peaceful people. Thus, I would need very compelling evidence to convince me that their world view was a threat. (A threat to who, by the way?)