Not necessarily what you were looking for, but a number of years ago I was taking Zoloft (anti-depressant -- SSRI like Prozac and Paxil, though all three have different properties). Normally I don't really remember a dream but once a week, maybe. Starting on about the third night I was taking it, I started remembering several dreams a night, often as many as five, and I remembered a lot of detail.
After several weeks of this, I realized something: "nightmares" are not some special category of dreams. In fact, most if not all dreams, even so called "good" ones, are just plain weird. Really weird, to the point of disturbing. You decide quickly that these people who want to remember more of their dreams might change their minds if they knew. And those who are successful in remembering more of their dreams and still enjoy it are not likely remembering as much detail as they could. To really remember that kind of disturbing detail took a psychoactive drug.
Anyway, I wasn't into lucid dreaming at the time and I went back off the SSRI after only a couple months, so I don't know if this helps lucid dreaming. But I thought you might enjoy the story even it was slightly tangential.
> Also, don't forget that an LCD display last also about 10,000-15,000 hours, after which the backlight has to be replaced (usually about as, if not more expensive than buying a new display).
That may be true for smaller (computer) displays, but not for HDTVs. RP LCD TVs themselves cost about $3000 for a 50" and the lightbulbs are well under $500.
Ah, yes, things have probably progressed pretty quickly in this arena. The vertex shaders are not nearly as flexible and powerful as the pixel shaders. A common technique is to draw a single quadilateral across the entire framebuffer, and with the right mapping every pixel will be visited once in the fragment (pixel) program. This fragment program is where you write the raytracer.
(Simplified concept, of course, but you get the point.)
Open up their specs so you can write a real-time raytracer? Why can't you use Cg or HLSL like others have done? Why do you need to write to the video card directly? You have full access to the programmability of the GPU through these languages. If not, program the damned thing in their version of assembler through the DirectX or OpenGL APIs. Unless by "tweaking OpenGL or DirectX" you mean "programming the GPU", your statement seems flat-out wrong.
I don't think the fat tree died with Thinking Machines. For example, MCR at LLNL uses a Quadrics fat tree. I imagine many sizeable clusters (way more than 64 nodes) use one. There's one link here, and the MCR link here but you can probably google for quadrics and fat tree to find some more. I'd be surprised if fat trees didn't show up in Myrinet / other interconnects, but you typically need to have a sizeable cluster before there's any point in calling it a fat tree.
(Oh, and if you meant something else entirely by fat tree, I apologize. I'm not too familiar with the particulars of the CM5 fat tree, so the Quadrics one is the only usage I'm aware of.)
You may need to explain that more. By using the words "brute force trial and error", you imply that I suggested the iteration was done randomly. I did no such thing.
A random walk might get you better results, but it is much more likely not to. I don't think anyone would use a purely random search function in their learning procedure and expect good results. Typical learning iterations involve directed search based on the observation of how the predicted results match the test results. Heck even look at a simple Newton iteration step -- it takes its current state, predicts based on some observations of the tangent at its currect point where the zero of the function lies, and tries to guess there. In fact, many learning algorithms are guaranteed to converge.
Now -- what you said was "When I learn, it is a sophisticated process of observation and inference." Certainly you can make that claim, but it only holds true at the macroscopic level. When you look at any individual neuron in your brain, it learns using a fairly deterministic set of mathematical rules based on the strengths of its incident connections, their firing rates, neurotransmitter levels, and so on.
This "observation and inference" you are doing is at a much higher level. In fact, if you look at what an entire neural network (or SVM, or whatever) is doing, you might see it behaving in such a way as a whole.
usually the neural network is just a very simple, possibly linear, adaptive filter which means that really contains no more than a few matrix multiplications...
The simplicity of the calculation does not mean it is not a learning algorithm. Real neural networks are quite simple, as each "neuron" is simply a weighted average of the inputs passed through a sigmoid or step function. However, en masse they perform better than most other algorithms at handwriting recognition. They take a training set and operate on it repeatedly, updating their parameters, until some sort of convergence is reached. Their performance on a test set is a measure of how well they have learned. This is a learning algorithm.
Even linear regression is a learning algorithm. You give it a bunch of training data as input (i.e. x,y pairs), iterate on that data until it converges, and is then used to predict new data. There happens to be an analytic solution to the iteration, but this does not make it any less of a learning algorithm.
I think maybe your definition of "learning" is unnecessarily strict. The simplicity of the computation is not what defines this category of algorithms.
Also, you get less of the "screen door effect" with a DLP than with an LCD projection because of the better fill percentage. This may let you get a bigger DLP screen for the same viewing distance.
So in many ways I think DLP has the potential to be a better technology. It is lacking right now (in my opinion) because I think most screens are a single chip solution with a color wheel, and that gave them perceptually less color depth despite the better contrast ratio. That and the cost was significantly higher than my LCD RP.
It makes sense, but I still don't think it's a big deal. First, we're talking about 90% brightness vs 80% brightness given the same wattage bulb. That's less than a 12% relative difference.
Second, don't LCD flat panels suffer the same amount as LCD rear projection? Obviously there are a ton of LCD monitors out there, so it doesn't seem to be impacting their perception as a viable technology for computer monitors.
You're right -- DLPs do tend to have less blank space. Your numbers seem high, but not unbelievable. However, from a proper viewing distance neither one should be noticable, even if you are using it as a monitor instead of a television. If they are, you're sitting too close.
Ah, yes. The T221. Formerly knows as the "Big Bertha" display. I saw a prototype of this thing running Quake 3. At that time, it required four separate computers with specialized software and a custom, physically very large piece of hardware.
Yes, that's right, I saw $100,000 worth of equipment used to play Quake at 3840x2400. I cried tears of joy that day, not just for the beauty of the sight, but for the thought that dozens of people, in triumphant togetherness, were able to work so hard for so long so that I might behold the wonder of 9 megapixel Quake.
> Why? I use everyday to work 1280x720 (using it right now)... and I'm curious....
It depends entirely on (a) what kind of work you are doing, and (b) your style.
I do coding full-time, and I typically have 6 half-screen-height shells and 3 full-height editor windows open at a time, or maybe 4 editors and 4 shells. This gives me 600 lines of text on the screen at once. I'm running a dual-monitor setup with both monitors at 1600x1200. And yet, I could probably make good use of three monitors.
Got to agree with you there. You'll note I also called that resolution "not really adequate for full-time use as a computer monitor". At the same time, for some applications bigger is better. Particularly, this resolution is quite adequate for most 3D games, but you better turn on the best damn antialiasing your card can handle.
Keep in mind that HDTV does not even mean a full 1080 lines of resolution necessarily.
Specifically, I have an HDTV LCD rear projection 50". Its native resolution is 1280x720, but with a little overscan you have to cut that down to about 1200x680 (roughly). I believe this resolution is typically the same for DLP rear projections and LCOS. I suspect that LCD flat panels are the same. Some DLP TVs appeared to me to have a limited color depth and too much dithering was apparent. I don't think this is an inherent problem with the technology, however, as DLP projectors work quite well hooked up to computers.
An "EDTV" plasma flat-panel TV is (IIRC) 768x480. That is clearly inadequate for use as a computer monitor. I think even the HDTV plasmas are commonly only 720 vertical lines. The few TVs that actually have 1080 lines of resolution are mostly CRT tubes (e.g. CRT RP).
The most important question is what the native resolution of these LCD flat panels is, and whether or not there is a computer-compatible connector that makes full use of it. For example, my TV I specifically got because it has both RGB (HD15) and DVI inputs, and I can get a resolution that maps directly to the pixels on the screen.
Unfortunately, this resolution (again, 1280x720) is not really adequate for full-time use as a computer monitor. It's great for the occasional web surfing, but I wouldn't want to do any real work on it.
In summary: If you can deal with the resolution, and there is a good connector on the TV (DVI is ideal, VGA is acceptable), then you will be fine. There's nothing particularly wrong with the attributes of these LCD TVs for use as computer monitors, in general, including color depth and pixel response times. (Once you start looking at other technologies like CRT RP, DLP, and Plasma, these other issues may become problematic.)
I'm with you on that one.... cell phone was definitely a bad example.:)
However, you could easily argue that (1) having a camcellcorderphone pointed at the screen during the previews indicates an intent to record the movie, although it might not be as rude, or else that (2) merely opening and examining a cell phone might result in the lens being pointed at the screen, without the user having to talk on it.
> > The law has been written with future technologies in mind and can equally apply to any type of recorder, including a mobile phone. So in California at least it is soon going to be illegal to take your phone into the cinema. > Again, only if you're intent on copying the film. Don't aim your phone at the screen and hit record and you'll be fine. Besides, does anybody have a camera phone with two to three hours of memory?
Not that I disagree with you, but your reasoning in this point is a little off. The parent said it can apply equally well to any type of recorder, including a mobile phone. You asserted no one had 2-3 hours of memory in their phone as a way to invalidate this line of reasoning.
There are two flaws with this. First, in a couple years, everyone's mobile phone will double as a high resolution digital camera and camcorder with hours and hours of capacity. So just because it you couldn't record the whole movie today doesn't mean you can't tomorrow.
Second, there is no provision that you need to intend to record the *whole* movie. Do you think they waited to make sure he caught the whole 2 hour movie before they decided he was breaking the law? Of course not.
So the original argument was quite on the mark, in fact, overly paranoid or not. If your cellphone in 2007 has a built in camcorder, and you pull it out to talk, and by the position of your hand the lens is pointed at the screen, does this mean you should be arrested and sent to jail for a year?
> One major issue, every time I booted it would alternate between thinking my Orinoco gold pcmcia wireless card would be not present, or thing a new one had been put in and try to reinstall it
Weird. I have that issue with the previous Mandrake. (9.2, right?) I "solved" it (loose usage, I know) by just ejecting the Orinoco card during bootup and re-inserting after it passed the eth0/1 initialization. Kinda stupid, but it stopped complaining.
It does something similar with the floppy drive, though. I haven't put the floppy drive in the bay since before I ever installed Mandrake on the thing, and yet every other reboot it tries to install it or remove it.
> SDL is a reasonable answer to portability while still accomplishing the integration that MS has achieved, but SDL isn't really as mainstream as OpenGL is.
SDL and OpenGL are not mutually exclusive. I have very successfully used SDL to handle joystick input, window creation, and sound output, with OpenGL for 3D. SDL in fact is designed to work this way, since SDL will create OpenGL rendering contexts for you when you create windows, and it handles the fullscreen video modes far easier than any other method of creating OpenGL contexts that I have used.
And just for comparison: I ported a DirectX/3D game from windows to SDL/OpenGL on Linux, and cut out about 1000 lines of code in the process because the DirectX API is so ugly. (I realize that might imply DirectX is more powerful, and I also realize my abhorrence of the MS DX API may be simply personal preference, and to boot, this was a DX5 program, so the API may be simpler now. I'm just relating the facts as I found them.)
Oh, and wasn't SDL used by Loki quite often for porting games to Linux?
But to make things better, my top four slashboxes are: 1. Mine, with links to commonly visited sites like dictionary.com (would the editors please add one for this!) and gamespot. 2. Google! 3. AllMusic 4. Pricewatch
Therefore, I have google and slashdot, as well as some other sites, as my homepage.
I would not call that disinformation at all. I assume you are attempting to make a subtle distinction between management of the labs, and what one typically means by "belong to", not that you are attempting to troll or anything.
You fell victim to one of the classic blunders! The first is never get involved in a land war in Asia. The second, only slightly less well known, is this: never go up against a nuclear weapons laboratory when death is on the line!
things like decent refraction, area lights, global illumination, caustics, raytracing. We can expect to see at least some of those implemented in hardware somewhere in the future.
Good points, but I'm not sure I agree with you on everything. First, you can implement refraction and reflection with current generation (as least current for the PC world) texture mapping techniques. See sphere maps for an example. What we might see is more native support for them.
Raytracing is a technique, not an effect, and while there are implementations of hardware raytracers using vertex/pixel shader programs, I don't see any need to use raytracing directly as an algorithm in real time for quite a while. Most effects can be handled without having to resort to it.
There are several techniques for caustics, e.g. photon maps, but I have yet to see those used dynamically. Same for global illumination, where radiosity has been used for quite a while for static lighting. But these are phenomenally expensive, and while I expect that we will see them some day, there probably won't be any real-time implementations of them in the next couple years in time for the next generation of consoles.
Coincidentally, what I think of as junebugs are apparently also called "maybeetles". (e.g. see the filename for the JPEG I linked to.) This is the first I ever heard of it, though.
But I agree; those little dragonflies in the original post I'm pretty sure are most often known by the name mayflies.
Slashdot Mirror
Not necessarily what you were looking for, but a number of years ago I was taking Zoloft (anti-depressant -- SSRI like Prozac and Paxil, though all three have different properties). Normally I don't really remember a dream but once a week, maybe. Starting on about the third night I was taking it, I started remembering several dreams a night, often as many as five, and I remembered a lot of detail.
After several weeks of this, I realized something: "nightmares" are not some special category of dreams. In fact, most if not all dreams, even so called "good" ones, are just plain weird. Really weird, to the point of disturbing. You decide quickly that these people who want to remember more of their dreams might change their minds if they knew. And those who are successful in remembering more of their dreams and still enjoy it are not likely remembering as much detail as they could. To really remember that kind of disturbing detail took a psychoactive drug.
Anyway, I wasn't into lucid dreaming at the time and I went back off the SSRI after only a couple months, so I don't know if this helps lucid dreaming. But I thought you might enjoy the story even it was slightly tangential.
> Also, don't forget that an LCD display last also about 10,000-15,000 hours, after which the backlight has to be replaced (usually about as, if not more expensive than buying a new display).
That may be true for smaller (computer) displays, but not for HDTVs. RP LCD TVs themselves cost about $3000 for a 50" and the lightbulbs are well under $500.
Ah, yes, things have probably progressed pretty quickly in this arena. The vertex shaders are not nearly as flexible and powerful as the pixel shaders. A common technique is to draw a single quadilateral across the entire framebuffer, and with the right mapping every pixel will be visited once in the fragment (pixel) program. This fragment program is where you write the raytracer.
(Simplified concept, of course, but you get the point.)
Open up their specs so you can write a real-time raytracer? Why can't you use Cg or HLSL like others have done? Why do you need to write to the video card directly? You have full access to the programmability of the GPU through these languages. If not, program the damned thing in their version of assembler through the DirectX or OpenGL APIs. Unless by "tweaking OpenGL or DirectX" you mean "programming the GPU", your statement seems flat-out wrong.
Don't believe you can do it? Here's a link some projects that do real-time raytacing, radiosity, photon mapping, and subsurface scattering, all on GPUs. These GPUs are programmable without them opening up their specs.
(The desire for them to open up their specs is for other reasons, not because they are hiding some functionality from you.)
I don't think the fat tree died with Thinking Machines. For example, MCR at LLNL uses a Quadrics fat tree. I imagine many sizeable clusters (way more than 64 nodes) use one. There's one link here, and the MCR link here but you can probably google for quadrics and fat tree to find some more. I'd be surprised if fat trees didn't show up in Myrinet / other interconnects, but you typically need to have a sizeable cluster before there's any point in calling it a fat tree.
(Oh, and if you meant something else entirely by fat tree, I apologize. I'm not too familiar with the particulars of the CM5 fat tree, so the Quadrics one is the only usage I'm aware of.)
So you learn by brute force trial and error?
You may need to explain that more. By using the words "brute force trial and error", you imply that I suggested the iteration was done randomly. I did no such thing.
A random walk might get you better results, but it is much more likely not to. I don't think anyone would use a purely random search function in their learning procedure and expect good results. Typical learning iterations involve directed search based on the observation of how the predicted results match the test results. Heck even look at a simple Newton iteration step -- it takes its current state, predicts based on some observations of the tangent at its currect point where the zero of the function lies, and tries to guess there. In fact, many learning algorithms are guaranteed to converge.
Now -- what you said was "When I learn, it is a sophisticated process of observation and inference." Certainly you can make that claim, but it only holds true at the macroscopic level. When you look at any individual neuron in your brain, it learns using a fairly deterministic set of mathematical rules based on the strengths of its incident connections, their firing rates, neurotransmitter levels, and so on.
This "observation and inference" you are doing is at a much higher level. In fact, if you look at what an entire neural network (or SVM, or whatever) is doing, you might see it behaving in such a way as a whole.
usually the neural network is just a very simple, possibly linear, adaptive filter which means that really contains no more than a few matrix multiplications ...
The simplicity of the calculation does not mean it is not a learning algorithm. Real neural networks are quite simple, as each "neuron" is simply a weighted average of the inputs passed through a sigmoid or step function. However, en masse they perform better than most other algorithms at handwriting recognition. They take a training set and operate on it repeatedly, updating their parameters, until some sort of convergence is reached. Their performance on a test set is a measure of how well they have learned. This is a learning algorithm.
Even linear regression is a learning algorithm. You give it a bunch of training data as input (i.e. x,y pairs), iterate on that data until it converges, and is then used to predict new data. There happens to be an analytic solution to the iteration, but this does not make it any less of a learning algorithm.
I think maybe your definition of "learning" is unnecessarily strict. The simplicity of the computation is not what defines this category of algorithms.
(OT a little)
Also, you get less of the "screen door effect" with a DLP than with an LCD projection because of the better fill percentage. This may let you get a bigger DLP screen for the same viewing distance.
So in many ways I think DLP has the potential to be a better technology. It is lacking right now (in my opinion) because I think most screens are a single chip solution with a color wheel, and that gave them perceptually less color depth despite the better contrast ratio. That and the cost was significantly higher than my LCD RP.
It makes sense, but I still don't think it's a big deal. First, we're talking about 90% brightness vs 80% brightness given the same wattage bulb. That's less than a 12% relative difference.
Second, don't LCD flat panels suffer the same amount as LCD rear projection? Obviously there are a ton of LCD monitors out there, so it doesn't seem to be impacting their perception as a viable technology for computer monitors.
You're right -- DLPs do tend to have less blank space. Your numbers seem high, but not unbelievable. However, from a proper viewing distance neither one should be noticable, even if you are using it as a monitor instead of a television. If they are, you're sitting too close.
Ah, yes. The T221. Formerly knows as the "Big Bertha" display. I saw a prototype of this thing running Quake 3. At that time, it required four separate computers with specialized software and a custom, physically very large piece of hardware.
Yes, that's right, I saw $100,000 worth of equipment used to play Quake at 3840x2400. I cried tears of joy that day, not just for the beauty of the sight, but for the thought that dozens of people, in triumphant togetherness, were able to work so hard for so long so that I might behold the wonder of 9 megapixel Quake.
> Why? I use everyday to work 1280x720 (using it right now)... and I'm curious....
It depends entirely on (a) what kind of work you are doing, and (b) your style.
I do coding full-time, and I typically have 6 half-screen-height shells and 3 full-height editor windows open at a time, or maybe 4 editors and 4 shells. This gives me 600 lines of text on the screen at once. I'm running a dual-monitor setup with both monitors at 1600x1200. And yet, I could probably make good use of three monitors.
> > if you can deal with the resolution...
> But who would want to?
Got to agree with you there. You'll note I also called that resolution "not really adequate for full-time use as a computer monitor". At the same time, for some applications bigger is better. Particularly, this resolution is quite adequate for most 3D games, but you better turn on the best damn antialiasing your card can handle.
holy cow! Let's fix that link....
Keep in mind that HDTV does not even mean a full 1080 lines of resolution necessarily.
Specifically, I have an HDTV LCD rear projection 50". Its native resolution is 1280x720, but with a little overscan you have to cut that down to about 1200x680 (roughly). I believe this resolution is typically the same for DLP rear projections and LCOS. I suspect that LCD flat panels are the same. Some DLP TVs appeared to me to have a limited color depth and too much dithering was apparent. I don't think this is an inherent problem with the technology, however, as DLP projectors work quite well hooked up to computers.
An "EDTV" plasma flat-panel TV is (IIRC) 768x480. That is clearly inadequate for use as a computer monitor. I think even the HDTV plasmas are commonly only 720 vertical lines. The few TVs that actually have 1080 lines of resolution are mostly CRT tubes (e.g. CRT RP).
The most important question is what the native resolution of these LCD flat panels is, and whether or not there is a computer-compatible connector that makes full use of it. For example, my TV I specifically got because it has both RGB (HD15) and DVI inputs, and I can get a resolution that maps directly to the pixels on the screen.
Unfortunately, this resolution (again, 1280x720) is not really adequate for full-time use as a computer monitor. It's great for the occasional web surfing, but I wouldn't want to do any real work on it.
In summary: If you can deal with the resolution, and there is a good connector on the TV (DVI is ideal, VGA is acceptable), then you will be fine. There's nothing particularly wrong with the attributes of these LCD TVs for use as computer monitors, in general, including color depth and pixel response times. (Once you start looking at other technologies like CRT RP, DLP, and Plasma, these other issues may become problematic.)
I'm with you on that one.... cell phone was definitely a bad example. :)
However, you could easily argue that (1) having a camcellcorderphone pointed at the screen during the previews indicates an intent to record the movie, although it might not be as rude, or else that (2) merely opening and examining a cell phone might result in the lens being pointed at the screen, without the user having to talk on it.
> > The law has been written with future technologies in mind and can equally apply to any type of recorder, including a mobile phone. So in California at least it is soon going to be illegal to take your phone into the cinema.
> Again, only if you're intent on copying the film. Don't aim your phone at the screen and hit record and you'll be fine. Besides, does anybody have a camera phone with two to three hours of memory?
Not that I disagree with you, but your reasoning in this point is a little off. The parent said it can apply equally well to any type of recorder, including a mobile phone. You asserted no one had 2-3 hours of memory in their phone as a way to invalidate this line of reasoning.
There are two flaws with this. First, in a couple years, everyone's mobile phone will double as a high resolution digital camera and camcorder with hours and hours of capacity. So just because it you couldn't record the whole movie today doesn't mean you can't tomorrow.
Second, there is no provision that you need to intend to record the *whole* movie. Do you think they waited to make sure he caught the whole 2 hour movie before they decided he was breaking the law? Of course not.
So the original argument was quite on the mark, in fact, overly paranoid or not. If your cellphone in 2007 has a built in camcorder, and you pull it out to talk, and by the position of your hand the lens is pointed at the screen, does this mean you should be arrested and sent to jail for a year?
> One major issue, every time I booted it would alternate between thinking my Orinoco gold pcmcia wireless card would be not present, or thing a new one had been put in and try to reinstall it
Weird. I have that issue with the previous Mandrake. (9.2, right?) I "solved" it (loose usage, I know) by just ejecting the Orinoco card during bootup and re-inserting after it passed the eth0/1 initialization. Kinda stupid, but it stopped complaining.
It does something similar with the floppy drive, though. I haven't put the floppy drive in the bay since before I ever installed Mandrake on the thing, and yet every other reboot it tries to install it or remove it.
> SDL is a reasonable answer to portability while still accomplishing the integration that MS has achieved, but SDL isn't really as mainstream as OpenGL is.
SDL and OpenGL are not mutually exclusive. I have very successfully used SDL to handle joystick input, window creation, and sound output, with OpenGL for 3D. SDL in fact is designed to work this way, since SDL will create OpenGL rendering contexts for you when you create windows, and it handles the fullscreen video modes far easier than any other method of creating OpenGL contexts that I have used.
And just for comparison: I ported a DirectX/3D game from windows to SDL/OpenGL on Linux, and cut out about 1000 lines of code in the process because the DirectX API is so ugly. (I realize that might imply DirectX is more powerful, and I also realize my abhorrence of the MS DX API may be simply personal preference, and to boot, this was a DX5 program, so the API may be simpler now. I'm just relating the facts as I found them.)
Oh, and wasn't SDL used by Loki quite often for porting games to Linux?
For me it's slashdot as well.
But to make things better, my top four slashboxes are:
1. Mine, with links to commonly visited sites like dictionary.com (would the editors please add one for this!) and gamespot.
2. Google!
3. AllMusic
4. Pricewatch
Therefore, I have google and slashdot, as well as some other sites, as my homepage.
I would not call that disinformation at all. I assume you are attempting to make a subtle distinction between management of the labs, and what one typically means by "belong to", not that you are attempting to troll or anything.
Prime Contract W-7405-ENG-48 (Contract 48 or Prime Contract) is executed between the University of California (UC or University) and the U.S. Department of Energy (DOE), and it is under this contract that LANL and LLNL function.
In other words, UC manages the labs under contract from DOE.
You fell victim to one of the classic blunders! The first is never get involved in a land war in Asia. The second, only slightly less well known, is this: never go up against a nuclear weapons laboratory when death is on the line!
a "dubious 4 month hole in your resume" isn't going to ruin your IT career
Just want to second that opinion.
If I were looking to hire someone for an IT job, I would have no qualms about hiring someone who just took a half-year off to follow a dream.
You might want to see if you can just "take a leave of absence" from your current job.
things like decent refraction, area lights, global illumination, caustics, raytracing. We can expect to see at least some of those implemented in hardware somewhere in the future.
Good points, but I'm not sure I agree with you on everything. First, you can implement refraction and reflection with current generation (as least current for the PC world) texture mapping techniques. See sphere maps for an example. What we might see is more native support for them.
Raytracing is a technique, not an effect, and while there are implementations of hardware raytracers using vertex/pixel shader programs, I don't see any need to use raytracing directly as an algorithm in real time for quite a while. Most effects can be handled without having to resort to it.
There are several techniques for caustics, e.g. photon maps, but I have yet to see those used dynamically. Same for global illumination, where radiosity has been used for quite a while for static lighting. But these are phenomenally expensive, and while I expect that we will see them some day, there probably won't be any real-time implementations of them in the next couple years in time for the next generation of consoles.
Coincidentally, what I think of as junebugs are apparently also called "maybeetles". (e.g. see the filename for the JPEG I linked to.) This is the first I ever heard of it, though.
But I agree; those little dragonflies in the original post I'm pretty sure are most often known by the name mayflies.