![NG-[C:G.Prom,B:TWTW15,Wk:5015,Dim:320X50]:12:12](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_uUh9ACqbjmjB2mb-dSlHEt4aRxcmoMDQEjy0_xZrwmT9b0LomsVBxipu7jmjyHmEueyIfk_4mal0DTJ0vc4fy2JA47R-MlRClRTemeH-ZyEH5oNbecTudqeRULV_ZrJWsdasZ3c0JEIaft0RWvvtD9XTMVwv6HHS6tVMeR3GsGdm2_Izox_A=s0-d)
3/15/2016
How Google’s DeepMind beat the best in Go, and why that matters
South Korea just finished
hosting an historic tournament of the ancient game of Go. On one side,
Google DeepMind’s AlphaGo program. On the other, South Korean Go
champion Lee Sedol, widely considered one of the best players of all
time. The contest was to be held over five days, with one match per day,
but only three days were needed. AlphaGo won in three resounding
victories, each more dominant than the last, and then went on to win 4-1
total. Google DeepMind’s achievement has been called bigger than
IBM’s 1997 win against chess champion Garry Kasparov — just like
Kasparov, Sedol went in predicting a clean sweep for the human side, but
unlike Kasparov, ended up getting almost swept himself.
To be fair to the now-defunct human masters, Go is a game that seems to truly stump humans as well. It originated in China at least several thousand years ago, and while the exact story behind it isn’t known, its mechanics seem strikingly similar to medieval battlefield tactics. The original Chinese name literally means “encircling game,” and it involves laying your pieces so you trap your opponent’s pieces entirely — while, of course, trying to prevent that from happening to your own.
Rather
than shaving down the possibility space for moves, for instance by
having a limited number of very specific moves associated with each
piece, as in chess, a Go board represents a wide open field of
possibility. It’s much larger than a chess board, 19 by 19 spaces
according to tournament rules; the combination of a large number of
spaces and non-restrictive rules governing moves in those spaces means
that there are a uselessly large number of total possible moves. There
are more possible positions for all stones on a Go board than there are
atoms in the universe — though I don’t think this oft-cited figure
includes WIMPs!
Computers have classically worked by quick-searching large numbers of possibilities, so if the number of possibilities becomes unsearchably large, as it does in Go, then a classical computer ought to be fundamentally incapable of playing well. But that’s just the first problem, because unlike chess, Go is all about attrition.
On a Go board, you have to constantly work toward both territory-right-now and a more abstract concept referred to as position.
It’s all about boxing out your opponent, opening up territory for
yourself while cutting them off from the best opportunities for future
play. How do you tell whether this position has better long term potential than that one, when your ability to predict specific outcomes hits a mathematical wall a dozen or so moves into the future?
To
answer these questions, the team at DeepMind created two basic
components to the system. One neural network tries to get a feel for the
overall — how is the board laid out, who’s in a better position, and
what areas needs to be threatened and defended? This is the source of
much of the difficulty, both for computers and humans, and with human
players it’s where we most commonly run into annoying words like “hunch”
and “intuition.”
The other major component uses knowledge about past games of Go to look a certain number of moves into the future, given different possible decisions for its own move — each possible path is considered for likely counter-moves, and good responses to each of these, and probable counter-moves to those… and on and on. On a 19-by-19 board, this sort of process needs to be directed by the high-level game- and board-awareness of the first algorithm, or else flail about with poor results.
The overall process
is not unlike imagination: The system essentially plays out little mini
games as far as it can, due to computational limitations, looks around
in this imaginary space, and judges the desirability of this possible
path versus all the other imaginary paths it has tested. That’s roughly
similar to how some human beings describe their approach to the game.
One DeepMind engineer described it in a live interview aired during Game
2 that the heatmap of interesting moves AlphaGo creates of any given
snapshot is much like human intuition. “As a Go player, if I look at the
board, there are immediately some points that jump out at me as moves
that I might want to play. AlphaGo does the same thing.”
Perhaps that’s why even human masters often refer to a “feeling” when asked precisely why they choose one strategy over another — they no more “understand” the truly deterministic outcomes of their decisions than AlphaGo itself.
So, do human Go masters truly understand it? And if so, does that mean AlphaGo does, too? Sedol was surprised to have to deal with a number of moves in the first game (in particular, move 102 to the right, if you’re interested) and seemed to come to think of AlphaGo as a truly unpredictable and dangerous opponent. The program seemed to display genuine creativity — an illusion of course, but an important one.
In the second game, played on March 9, AlphaGo took an unusual opening approach, causing Lee pause. In the first game AlphaGo used far more of its allotted time than Lee, but in the second the human champion used up far more of his time than the computer. He ran out of time first, and continued hesitation brought him down to his final one-minute allotment of overtime before forfeiting due to timing out. It’s incredible — in a very real way, the computer psyched Lee out. What’s more incredible is that it’s actually possible that those early moves were made for specifically that reason. During that game Michael Redmond, the only Westerner to ever reach the top “9-dan” rank in Go, said that this latest version of AlphaGo plays an innovative game. He said its style was “already something I could learn from.”
At the end of the day, the fall-back explanation for human skill at Go is experience — we think it will work because it has worked in the past, even in a very abstract, “This move sort of looks like that move from six months ago!” kind of way. That’s precisely how a deep neural network learns. The difference is that while a human might be able to play at most a few thousand games a year, AlphaGo can play millions every day. That’s how it has acquired such astounding ability, fully a decade before even optimistic projections. By combining a computer’s brute force ability with a truly novel network-of-networks approach to breaking down the possibility space much the way a human mind does, DeepMind has been able to (at least) rival the best humanity has to offer.
This is an incredible achievement, regardless of whether AlphaGo had won the tournament overall. It shows that neural networks really can take us past previously impassable barriers. All Skynet nightmares aside, in the end Alphabet CEO Eric Schmidt said it best: “The winner here, no matter what happens, is humanity.”
To be fair to the now-defunct human masters, Go is a game that seems to truly stump humans as well. It originated in China at least several thousand years ago, and while the exact story behind it isn’t known, its mechanics seem strikingly similar to medieval battlefield tactics. The original Chinese name literally means “encircling game,” and it involves laying your pieces so you trap your opponent’s pieces entirely — while, of course, trying to prevent that from happening to your own.
A fancy Go board, with a few pieces laid out.
Rather
than shaving down the possibility space for moves, for instance by
having a limited number of very specific moves associated with each
piece, as in chess, a Go board represents a wide open field of
possibility. It’s much larger than a chess board, 19 by 19 spaces
according to tournament rules; the combination of a large number of
spaces and non-restrictive rules governing moves in those spaces means
that there are a uselessly large number of total possible moves. There
are more possible positions for all stones on a Go board than there are
atoms in the universe — though I don’t think this oft-cited figure
includes WIMPs!Computers have classically worked by quick-searching large numbers of possibilities, so if the number of possibilities becomes unsearchably large, as it does in Go, then a classical computer ought to be fundamentally incapable of playing well. But that’s just the first problem, because unlike chess, Go is all about attrition.
Chess
Grandmaster Gerry Kasparov despairs in the face of IBM’s early attempt
at machine learning — complete with Japanese surrogate to make the
computer’s moves, just like today.
The moves from Game 1 of 5, March 8.
The other major component uses knowledge about past games of Go to look a certain number of moves into the future, given different possible decisions for its own move — each possible path is considered for likely counter-moves, and good responses to each of these, and probable counter-moves to those… and on and on. On a 19-by-19 board, this sort of process needs to be directed by the high-level game- and board-awareness of the first algorithm, or else flail about with poor results.
The overall process
is not unlike imagination: The system essentially plays out little mini
games as far as it can, due to computational limitations, looks around
in this imaginary space, and judges the desirability of this possible
path versus all the other imaginary paths it has tested. That’s roughly
similar to how some human beings describe their approach to the game.
One DeepMind engineer described it in a live interview aired during Game
2 that the heatmap of interesting moves AlphaGo creates of any given
snapshot is much like human intuition. “As a Go player, if I look at the
board, there are immediately some points that jump out at me as moves
that I might want to play. AlphaGo does the same thing.”Perhaps that’s why even human masters often refer to a “feeling” when asked precisely why they choose one strategy over another — they no more “understand” the truly deterministic outcomes of their decisions than AlphaGo itself.
So, do human Go masters truly understand it? And if so, does that mean AlphaGo does, too? Sedol was surprised to have to deal with a number of moves in the first game (in particular, move 102 to the right, if you’re interested) and seemed to come to think of AlphaGo as a truly unpredictable and dangerous opponent. The program seemed to display genuine creativity — an illusion of course, but an important one.
In the second game, played on March 9, AlphaGo took an unusual opening approach, causing Lee pause. In the first game AlphaGo used far more of its allotted time than Lee, but in the second the human champion used up far more of his time than the computer. He ran out of time first, and continued hesitation brought him down to his final one-minute allotment of overtime before forfeiting due to timing out. It’s incredible — in a very real way, the computer psyched Lee out. What’s more incredible is that it’s actually possible that those early moves were made for specifically that reason. During that game Michael Redmond, the only Westerner to ever reach the top “9-dan” rank in Go, said that this latest version of AlphaGo plays an innovative game. He said its style was “already something I could learn from.”
At the end of the day, the fall-back explanation for human skill at Go is experience — we think it will work because it has worked in the past, even in a very abstract, “This move sort of looks like that move from six months ago!” kind of way. That’s precisely how a deep neural network learns. The difference is that while a human might be able to play at most a few thousand games a year, AlphaGo can play millions every day. That’s how it has acquired such astounding ability, fully a decade before even optimistic projections. By combining a computer’s brute force ability with a truly novel network-of-networks approach to breaking down the possibility space much the way a human mind does, DeepMind has been able to (at least) rival the best humanity has to offer.
This is an incredible achievement, regardless of whether AlphaGo had won the tournament overall. It shows that neural networks really can take us past previously impassable barriers. All Skynet nightmares aside, in the end Alphabet CEO Eric Schmidt said it best: “The winner here, no matter what happens, is humanity.”
AMD unveils 2016 GPU roadmap; Polaris offers 2.5x performance per watt, may utilize GDDR5
At its Capsaicin event yesterday, AMD didn’t just unveil its $1,500 Fiji GPU
— it also showed off its next-generation Polaris GPU and gave an update
on its longer-term GPU roadmap. The new roadmap isn’t just significant
for what it says, but what it doesn’t mention as well.
First, a bit of background. 14nm is expected to be a major new node for both AMD and Nvidia, but it’s particularly important to AMD’s situation. Not only is Team Red fighting to regain market share lost to Nvidia over the last few years, its GPUs have been less power-efficient and noisier than Team Green’s counterparts. (The Radeon Nano, it must be noted, was an exception to these trends, with markedly better power efficiency than any previous AMD GCN GPU.) This is particularly problematic in mobile, where every watt and decibel count.
Previously, AMD had claimed that it would deliver a
2x performance-per-watt improvement with Polaris as compared with its
28nm-class hardware. Now, the company has bumped that prediction to 2.5x
improved performance-per-watt. Readers should keep in mind that like
all metrics, performance-per-watt is not an absolute. Because silicon
power consumption is not a simple linear curve, these figures are likely
best-case estimates based on midrange hardware, not the worst-case
scenario when comparing top-end parts clocked at maximum frequency. Even
so, a 2.5x efficiency bump is a noteworthy generational leap.
There are several interesting bits of information we’d like to discuss. First, AMD’s next-generation GPU family, codenamed Vega, is now forecast to arrive hard on the heels of Polaris. After measuring the red boundary boxes and Polaris’ relative position compared to the summer 2016 launch time frame, the position of the Vega box suggests a launch at the tail end of 2016 or very early 2017 timeframe.
The fact that Vega is specifically labeled as using HBM2 also suggests that this is the major high-end architecture revamp that Team Red fans have been waiting for, with much larger frame buffers and, if rumors are true, a true next-generation GPU architecture from AMD (Polaris is still based on GCN, albeit a heavily improved version).
The fact that Vega is labeled as HBM2 whereas Polaris isn’t labeled with any type of memory at all suggests that Polaris will either use GDDR5 or GDDR5X. I’m inclined to think GDDR5 is more likely — GDDR5X still requires significantly more power than GDDR5, and isn’t expected to be ready for a volume launch in a matter of months. PC Perspective believes Polaris will still use HBM1, which is also possible.
If AMD is using
conventional GDDR5 for Polaris, it’s probably for practical reasons.
When the company debuted Fiji last year, it took pains to note that
GDDR5 made progressively less sense as frame buffer sizes and clock
speeds ramped up. AMD chose to adopt HBM in 2015, partly because it
wanted to deliver an enormous amount of memory bandwidth and was willing
to accept a smaller maximum VRAM frame buffer to do it. The company has
always maintained that HBM made more sense than GDDR5 for the Fury X
and Fury Nano, but never claimed that it would adopt HBM across the
entirety of its product stack.
One of the rumors we’ve heard about Polaris is that AMD will target the mainstream and upper-midrange of the GPU market. This is where the money is in enthusiast gaming; Steam’s Hardware Survey is currently broken and not displaying any GPUs from NV at all, but archival data from September showed the GTX 970 as the single-most popular GPU from AMD and Nvidia, with 3.97% of the total install base. The Steam Hardware survey is inexact at best — it often takes months to be updated with new cards and at present all of the Nvidia data seems to have been rolled into Intel’s data — but it still offers objective proof that this is the sweet spot for AMD to target if it wants maximum revenue for its efforts. If AMD intends to deliver a GPU with vastly improved performance per dollar, performance-per-watt, and hold open the option for 8GB cards, GDDR5 may still be the better choice. HBM2 may not be ready for launch and HBM was never more than a niche launch, meant to be cost-effective above a high price point but not to take the technology mainstream.
If we assume that Greenland existed at all, this data suggests AMD canceled the part in favor of pulling Vega in and launching an all-new architecture more quickly. Alternately, the code name could still have some bearing on future Zen APUs. We know AMD is working on these parts, but the company has said nothing about a launch date or what features and capabilities they’ll offer.
I’ve speculated in several stories that AMD might use HBM2 memory in APUs to finally break the bandwidth limitations that have held integrated graphics back for nearly twenty years. I still think this is likely in the long run, but whether or not AMD rolls the technology out will depend on both HBM/HBM2 costs and Zen’s ability to command favorable margins and higher absolute prices compared to Carrizo and Kaveri/Godavari.
As of this writing, the cheapest DDR4-2133 on Newegg is $59, while DDR4-3200 is $85. Given that AMD’s integrated GPUs have often been entirely memory bandwidth-bound, it’s entirely possible that the first wave of Zen APUs will lean on the 50% improved memory bandwidth offered by high-speed DDR4-3200 — and 51.2GB/s of memory bandwidth is nothing to sneeze at in a dual-channel system.
First, a bit of background. 14nm is expected to be a major new node for both AMD and Nvidia, but it’s particularly important to AMD’s situation. Not only is Team Red fighting to regain market share lost to Nvidia over the last few years, its GPUs have been less power-efficient and noisier than Team Green’s counterparts. (The Radeon Nano, it must be noted, was an exception to these trends, with markedly better power efficiency than any previous AMD GCN GPU.) This is particularly problematic in mobile, where every watt and decibel count.
Radeon
Nano power efficiency at default and 150% board power level. The Nano
was the first high-end GCN card to land within spitting distance of
Maxwell.
AMD’s GPU roadmap
AMD’s GPU roadmap for the next 18 months is shown below:There are several interesting bits of information we’d like to discuss. First, AMD’s next-generation GPU family, codenamed Vega, is now forecast to arrive hard on the heels of Polaris. After measuring the red boundary boxes and Polaris’ relative position compared to the summer 2016 launch time frame, the position of the Vega box suggests a launch at the tail end of 2016 or very early 2017 timeframe.
The fact that Vega is specifically labeled as using HBM2 also suggests that this is the major high-end architecture revamp that Team Red fans have been waiting for, with much larger frame buffers and, if rumors are true, a true next-generation GPU architecture from AMD (Polaris is still based on GCN, albeit a heavily improved version).
The fact that Vega is labeled as HBM2 whereas Polaris isn’t labeled with any type of memory at all suggests that Polaris will either use GDDR5 or GDDR5X. I’m inclined to think GDDR5 is more likely — GDDR5X still requires significantly more power than GDDR5, and isn’t expected to be ready for a volume launch in a matter of months. PC Perspective believes Polaris will still use HBM1, which is also possible.
This
slide shows the problem of stacking 2D RAM configurations indefinitely.
Polaris could still offer 8GB of RAM in 8x integrated circuits,
however, as opposed to the 16 seen here.
One of the rumors we’ve heard about Polaris is that AMD will target the mainstream and upper-midrange of the GPU market. This is where the money is in enthusiast gaming; Steam’s Hardware Survey is currently broken and not displaying any GPUs from NV at all, but archival data from September showed the GTX 970 as the single-most popular GPU from AMD and Nvidia, with 3.97% of the total install base. The Steam Hardware survey is inexact at best — it often takes months to be updated with new cards and at present all of the Nvidia data seems to have been rolled into Intel’s data — but it still offers objective proof that this is the sweet spot for AMD to target if it wants maximum revenue for its efforts. If AMD intends to deliver a GPU with vastly improved performance per dollar, performance-per-watt, and hold open the option for 8GB cards, GDDR5 may still be the better choice. HBM2 may not be ready for launch and HBM was never more than a niche launch, meant to be cost-effective above a high price point but not to take the technology mainstream.
Whither Greenland?
If you’ve paid attention to the various AMD “roadmaps” passed around in enthusiast circles, you’ve probably seen references to Greenland sprinkled across them in the past year. It’s never been clear exactly what the specs of Greenland were, and the various claims made about it transformed over time. Either way, it’s not listed on this roadmap.If we assume that Greenland existed at all, this data suggests AMD canceled the part in favor of pulling Vega in and launching an all-new architecture more quickly. Alternately, the code name could still have some bearing on future Zen APUs. We know AMD is working on these parts, but the company has said nothing about a launch date or what features and capabilities they’ll offer.
I’ve speculated in several stories that AMD might use HBM2 memory in APUs to finally break the bandwidth limitations that have held integrated graphics back for nearly twenty years. I still think this is likely in the long run, but whether or not AMD rolls the technology out will depend on both HBM/HBM2 costs and Zen’s ability to command favorable margins and higher absolute prices compared to Carrizo and Kaveri/Godavari.
As of this writing, the cheapest DDR4-2133 on Newegg is $59, while DDR4-3200 is $85. Given that AMD’s integrated GPUs have often been entirely memory bandwidth-bound, it’s entirely possible that the first wave of Zen APUs will lean on the 50% improved memory bandwidth offered by high-speed DDR4-3200 — and 51.2GB/s of memory bandwidth is nothing to sneeze at in a dual-channel system.
2016 is turning out to be an amazing year for augmented reality
One way to tell that a new
market is coming of age here in Silicon Valley is when special-purpose
venture funds are formed to focus on it. Augmented reality has just
achieved that milestone, with the launch of Super Ventures. The fund
itself is small, but the event served as a great touch point for members
of the still-close-knit AR community to come together and provide some
insight on the future of AR, both for consumers and developers. It also
serves as a sneak peak at the state of the industry ahead of the
much-larger annual flagship Augmented World Expo this coming June.
Running
untethered is even more important for AR than for VR. While some AR
applications, like CastAR’s tabletop gaming, are confined to a small
area, most involve allowing the user to move around in their environment
and get an annotated (or augmented, if you will) view of reality.
Microsoft’s HoloLens (now available to developers) and Magic Leap’s
device (now being demoed privately) have attracted the most press among
standalone solutions, but startups like ODG have been shipping
untethered AR “smart glasses” for a while now.
Another
AR app that can run on a standard mobile device, WayGo, is one of the
first companies to be backed by the new AR-focused Super Ventures fund.
It allows the automatic identification (and translation in place) of
text in the surrounding environment, as it is captured in real time by a
device’s camera. Similar in concept to Google Translate, WayGo says it
has better support for automatic text recognition in non-Western
languages, and if focusing much of its marketing efforts on selling into
commercial applications.
Narrow
field of view (FOV) has been one of the factors that most limits what
can be accomplished with existing AR devices. Part of what made
Microsoft’s first HoloLens demo so compelling was its relatively-wide
field of view (rumored to have been provided by Waveguide vendor Lumus).
Subsequent versions, though, have fallen back to providing only a more
narrow image (Microsoft has compared it to looking at a 15-inch monitor
from 2 feet away).
Short-term, vendors like Meta are addressing the FOV problem by using a wider reflective surface to form the image. Meta, for example, is demoing its Meta 2 — claiming a 90-degree FOV for the glasses which are now available for pre-order. Unfortunately, the Meta 2 glasses are tethered, making them a prime candidate for a next-generation computer UI — which Meta promotes heavily, and the Meta 1 essentially worked as a second monitor for your PC — or for lab or studio applications, but not as a great option for outdoors or mobile. For developers on a limited budget the Meta 2 developer kit is $950, compared to $2,750 for ODG’s R-7, and $3,000 for Microsoft’s Hololens. Game and VR developers will also like that the Meta 2 supports Unity for development.
The long-term future for AR displays looks even-more promising. In addition to better wave guide modules, a new technology that uses laser projection directly into the retina (it isn’t as scary as it sounds) is on the way. Glyph by Avegant is one of the better-known laser-projecting AR devices, but not the only one. Massively-funded (and hyped) startup Magic Leap is using a type of laser projection technology (also referred to as Virtual Retinal Display or VRD) for its upcoming AR system. Judging by demo videos it has released, the VRD in Magic Leap’s system also does eye tracking to allow selective refocusing of the image based on where the user is looking. I was able to demo a research project at Stanford that allowed refocusing based on eye tracking, and it is a very powerful concept. VRDs can also completely saturate the rods and cones in your eye, allowing for completely opaque objects — and a nearly VR experience if their field of view is wide enough — in addition to supplying an augmented overlay of the real world. That allows what have sometimes been called “mixed reality” applications — ones that can combine the best of VR and AR.
VRDs also have two other potential advantages over more traditional display-based interfaces. Because the image is formed on your retina, and not on a fixed display in front of you, the issue of accommodation and vergence mis-match (simply put, the problem where your eyes think they are looking at something far away while they are focusing on a tiny display an inch in front of your head), which is a big factor in VR-induced motion sickness, is solved. Unlike waveguides, VRDs also don’t need to grow in size as the field of view increases. It is likely that they will become the display technology of choice for at least high-end AR, and even some VR, applications. VR goggle makers are developing their own answers to this problem, including the Lightfield Sereoscope developed by Stanford’s Gordon Wetzstein in cooperation with Nvidia.
Analyst Digi-Capital predicts an AR market of $90 billion annually by 2020, compared to $30 billion for VR. That is the sort of promise that made Super Ventures founding partner, Ori Inbar, explain to me that after years of organizing AR-specific conferences, he was “ready to put his money where his mouth is.” More surprisingly, given the relative amounts of buzz, AR is already a much larger market than VR, with hundreds of commercial, industrial, and military applications provided by literally dozens of hardware and software vendors. Before it goes mainstream, though, at least another generation or two of advances in hardware will be needed.
What makes AR special?
Augmented reality is characterized by combining views of the “real world” with computer-generated content. I put real world in quotes because some AR solutions actually allow the user to see their surroundings and project generated content onto that view, while others use a live camera feed of the surroundings with an overlay of generated objects. The former have the advantage that they are much more natural to use, and allow a better sense of context for the user. The later are often simpler to implement, lower on processor power, and can run on standard-format mobile devices — without requiring special glasses.Running
untethered is even more important for AR than for VR. While some AR
applications, like CastAR’s tabletop gaming, are confined to a small
area, most involve allowing the user to move around in their environment
and get an annotated (or augmented, if you will) view of reality.
Microsoft’s HoloLens (now available to developers) and Magic Leap’s
device (now being demoed privately) have attracted the most press among
standalone solutions, but startups like ODG have been shipping
untethered AR “smart glasses” for a while now.AR doesn’t always require a geeky, new wearable
Unlike with VR, many AR applications run on either standard, or slightly-enhanced, mobile devices. For example, ScopeAR was demoing “over the shoulder” industrial coaching applications, where experts can be called in to help a worker in the field by illustrating the steps to take in performing a maintenance or repair operation — all as an overlay on the actual scene being captured by the mobile device’s own video camera. Pre-recorded tutorials can also be used “offline” to walk a user through a process — as long as the system has an accurate baseline image of the mechanism being worked on, initialized by the placement of a small marker device.
ScopeAR
showed me how its industrial maintenance applications — like this one
for replacing a circuit breaker — could be run on a smartphone or
tablet, as long as it has rear-facing video capability
Processing power remains an issue: The cloud provides one answer
One of the biggest problems with mobile-device-based AR (and VR) is processing power. Demoing the creation of a 3D model from simply sweeping across a room with a device is one thing, but having it work in real life is another. We found this when we tried to replicate those amazing Google Project Tango demos in a typical apartment. The results were nowhere near as polished. Startup GeoCV is addressing this issue by adding a cloud dimension to the mix. Your depth-measurement-equipped mobile device (like a Tango or RealSense) is used to gather the initial data, but the processing is done in the cloud — allowing for more complete and more accurate models. Startup Gridraster takes a different approach, and hopes to allow mobile devices to harness the GPU of nearby computers to augment their native capabilities.Field of view needs solving for AR to go mainstream
Narrow
field of view (FOV) has been one of the factors that most limits what
can be accomplished with existing AR devices. Part of what made
Microsoft’s first HoloLens demo so compelling was its relatively-wide
field of view (rumored to have been provided by Waveguide vendor Lumus).
Subsequent versions, though, have fallen back to providing only a more
narrow image (Microsoft has compared it to looking at a 15-inch monitor
from 2 feet away).Short-term, vendors like Meta are addressing the FOV problem by using a wider reflective surface to form the image. Meta, for example, is demoing its Meta 2 — claiming a 90-degree FOV for the glasses which are now available for pre-order. Unfortunately, the Meta 2 glasses are tethered, making them a prime candidate for a next-generation computer UI — which Meta promotes heavily, and the Meta 1 essentially worked as a second monitor for your PC — or for lab or studio applications, but not as a great option for outdoors or mobile. For developers on a limited budget the Meta 2 developer kit is $950, compared to $2,750 for ODG’s R-7, and $3,000 for Microsoft’s Hololens. Game and VR developers will also like that the Meta 2 supports Unity for development.
The long-term future for AR displays looks even-more promising. In addition to better wave guide modules, a new technology that uses laser projection directly into the retina (it isn’t as scary as it sounds) is on the way. Glyph by Avegant is one of the better-known laser-projecting AR devices, but not the only one. Massively-funded (and hyped) startup Magic Leap is using a type of laser projection technology (also referred to as Virtual Retinal Display or VRD) for its upcoming AR system. Judging by demo videos it has released, the VRD in Magic Leap’s system also does eye tracking to allow selective refocusing of the image based on where the user is looking. I was able to demo a research project at Stanford that allowed refocusing based on eye tracking, and it is a very powerful concept. VRDs can also completely saturate the rods and cones in your eye, allowing for completely opaque objects — and a nearly VR experience if their field of view is wide enough — in addition to supplying an augmented overlay of the real world. That allows what have sometimes been called “mixed reality” applications — ones that can combine the best of VR and AR.
VRDs also have two other potential advantages over more traditional display-based interfaces. Because the image is formed on your retina, and not on a fixed display in front of you, the issue of accommodation and vergence mis-match (simply put, the problem where your eyes think they are looking at something far away while they are focusing on a tiny display an inch in front of your head), which is a big factor in VR-induced motion sickness, is solved. Unlike waveguides, VRDs also don’t need to grow in size as the field of view increases. It is likely that they will become the display technology of choice for at least high-end AR, and even some VR, applications. VR goggle makers are developing their own answers to this problem, including the Lightfield Sereoscope developed by Stanford’s Gordon Wetzstein in cooperation with Nvidia.
Zspace
uses a custom 3D monitor, augmented with an active laser pointer and
head tracking, to provide an AR-like experience floating over the
display. Here, classroom software from Corinth is shown interactively
illustrating how volcanoes work.
Given time AR is likely to trump VR
Not surprisingly, most AR advocates (including me) make the case that eventually AR will greatly outpace VR as a technology and a market. After all, there is a limit to how much of the day we are likely to want to spend “head down” in a pair of googles using VR, while subtle AR devices will eventually be no more troublesome than a pair of designer sunglasses — and will be able to provide useful information throughout the day.Analyst Digi-Capital predicts an AR market of $90 billion annually by 2020, compared to $30 billion for VR. That is the sort of promise that made Super Ventures founding partner, Ori Inbar, explain to me that after years of organizing AR-specific conferences, he was “ready to put his money where his mouth is.” More surprisingly, given the relative amounts of buzz, AR is already a much larger market than VR, with hundreds of commercial, industrial, and military applications provided by literally dozens of hardware and software vendors. Before it goes mainstream, though, at least another generation or two of advances in hardware will be needed.
Plastic-eating bacteria set to revolutionize waste disposal
Few of us are likely to spend
much time meditating on the problem of plastic, at least until a
tranquil afternoon is violently disrupted by the vision of a stray
Walmart bag sailing across the azure sky. Having lived a few streets
down from a Walmart, I can personally attest to the menace of runaway
shopping bags. But the problem goes far deeper than aesthetics — many
plastics leach chemicals
that act like the sex hormone estrogen when introduced into the body,
increasing the likelihood of birth defects and other health
complications.
For these and many other reasons, finding innovative ways to remove plastic waste from the environment has become an issue of increasing importance. Thankfully a group of researchers at Kyoto University in Japan may have seized upon a solution, one involving the humble bacterium Ideonella sakaiensis.
One
of the reasons we find plastics so troublesome is that they are not
rapidly biodegradable; they stay resident in the environment long after
they have served their purpose. Enter Ideonella sakaiensis:
Japanese scientists have shown this bacteria is capable of digesting a
chemical called polyethylene terephthalate, the substrate of many
plastics we find in household products like bottled drinks, cosmetics
and household cleaners. This could be revolutionary – imagine carrying a
super soaker charged with an aqueous solution of the bacterium and
zapping down those Walmart shopping bags as they careened overhead.
The above equivalent of environmental skeet shooting might even become a popular leisure activity, except for one catch: The bacteria in question digest plastic at an epically slow rate, approximately six weeks to eat through a thin layer of PET. If this method is ever to succeed at disposing of the megalithic Styrofoam structure that my laptop computer shipped in, clearly something faster acting is required.
Thanks to progress being made in genetic engineering, though, a way to rev up the speed at which the bacteria digest plastic could soon be at hand. The scientists performing the study have already sequenced the bacteria’s genome and using gene editing techniques, are well on their way to figuring out a way to increase its plastic eating potency. Technologies like CRISPR, the subject of much hand wringing recently when Chinese scientists used it to edit the genomes of human embryos, are in bad need of a success story.
Thanks to sensationalist media outlets and unquestioning audiences, genetic engineering catches more than its share of flack. In reality, if humanity is to persist very far into the twenty-first century without reverting to primitive hunter-gatherer societies, it will require the aid of advanced forms of genetic engineering – if only to remove the immense amount of plastic waste we are currently producing.
We have previously reported on the success of genetic engineering in combatting the Zika virus, producing bananas enriched with beta-carotene to combat and malnutrition, and making salmon available to a wider range of people. If the removal of plastic waste could be added to that list of accomplishments, it might just succeed in turning the tides of public opinion in favor of genetic engineering. Until then, activities like plastic bag skeet shooting are likely to remain the stuff of imagination (well, at least mine).
For these and many other reasons, finding innovative ways to remove plastic waste from the environment has become an issue of increasing importance. Thankfully a group of researchers at Kyoto University in Japan may have seized upon a solution, one involving the humble bacterium Ideonella sakaiensis.
Bottle breakdown. Illustration: P. Huey. Reprinted with permission from U.T. Bornscheuer, Science 351:1154 (2016)
The above equivalent of environmental skeet shooting might even become a popular leisure activity, except for one catch: The bacteria in question digest plastic at an epically slow rate, approximately six weeks to eat through a thin layer of PET. If this method is ever to succeed at disposing of the megalithic Styrofoam structure that my laptop computer shipped in, clearly something faster acting is required.
Thanks to progress being made in genetic engineering, though, a way to rev up the speed at which the bacteria digest plastic could soon be at hand. The scientists performing the study have already sequenced the bacteria’s genome and using gene editing techniques, are well on their way to figuring out a way to increase its plastic eating potency. Technologies like CRISPR, the subject of much hand wringing recently when Chinese scientists used it to edit the genomes of human embryos, are in bad need of a success story.
Thanks to sensationalist media outlets and unquestioning audiences, genetic engineering catches more than its share of flack. In reality, if humanity is to persist very far into the twenty-first century without reverting to primitive hunter-gatherer societies, it will require the aid of advanced forms of genetic engineering – if only to remove the immense amount of plastic waste we are currently producing.
We have previously reported on the success of genetic engineering in combatting the Zika virus, producing bananas enriched with beta-carotene to combat and malnutrition, and making salmon available to a wider range of people. If the removal of plastic waste could be added to that list of accomplishments, it might just succeed in turning the tides of public opinion in favor of genetic engineering. Until then, activities like plastic bag skeet shooting are likely to remain the stuff of imagination (well, at least mine).
North Korean submarine missing, presumed sunk
The US and South Korea are
reporting that North Korea is believed to have lost one of its
submarines. The vessel was operating off the coast of North Korea when
it disappeared, and the North Korean navy has been observed engaged in
search and rescue operations in the area where the sub is believed to
have been stationed.
North Korea’s navy (formally called the Korean People’s Navy) fields a wide range of vessels, including amphibious landing craft, over 400 patrol boats, and 70 submarines. Most of the country’s subs, however, could charitably be described as “ancient.”
North Korea’s arsenal includes:
50 Romeo-class submarines: These are based on a post-WW2 1950s Soviet design, which was in turn based on the WW2-era Type-XXI German U-Boat. History buffs may recognize the Type XXI as the absolute cutting-edge of WW2 submarine warfare. It was the first submarine designed to remain submerged rather than as a surface vessel that submerged to attack or evade the enemy. It could travel for days on battery, recharge in less than five hours via snorkel, and featured advanced hull designs and silent running capabilities that exceeded the other submarines of its day.
The
Romeo enhanced some of these capabilities further and the KPN is using
Chinese designs built between 1973 and 1995, but the oldest of these
vessels is still over 40 years old.
40 Sang-O submarines: These are much smaller than the Romeo class, at just 300 tons. These are the largest submarines that North Korea is known to have put into domestic mass production. These are small ships, suitable for local patrols and domestic defense, but not for significant force projection. A boat of this type was captured by the South Koreans in the 1996 Gangneung submarine infiltration incident.
10 Yono-class submarines: These are midget subs, at 130 tons. They have a crew of two with 6-7 special forces onboard. A ship of this class is believed to have been involved in an October 2010 incident in which a South Korean corvette was destroyed by an unknown North Korean submarine.
In addition, North Korea is believed to have fielded a new submarine as recently as 2014. Little is known about this new, Sinpo-class design — it may be based on older vessels from Yugoslavia or incorporate technology from newer Russian vessels. South Korean sources suggested at one point that it could be built from old Golf II-class hulls that North Korea imported in the early 1990s, but this has not been confirmed. If rumors are true, it would be North Korea’s first ballistic missile submarine, though it lacks the ability to project force much outside of NK’s territorial waters.
As of this writing, the KPN has not requested outside assistance or aid with any rescue efforts. It could be some time before we know which submarine was lost and whether or not any survivors were rescued by the recovery crews.
North Korea’s navy (formally called the Korean People’s Navy) fields a wide range of vessels, including amphibious landing craft, over 400 patrol boats, and 70 submarines. Most of the country’s subs, however, could charitably be described as “ancient.”
North Korea’s arsenal includes:
50 Romeo-class submarines: These are based on a post-WW2 1950s Soviet design, which was in turn based on the WW2-era Type-XXI German U-Boat. History buffs may recognize the Type XXI as the absolute cutting-edge of WW2 submarine warfare. It was the first submarine designed to remain submerged rather than as a surface vessel that submerged to attack or evade the enemy. It could travel for days on battery, recharge in less than five hours via snorkel, and featured advanced hull designs and silent running capabilities that exceeded the other submarines of its day.
North Korea’s Romeo-class subs are believed to be based on the Chinese Type 033.
40 Sang-O submarines: These are much smaller than the Romeo class, at just 300 tons. These are the largest submarines that North Korea is known to have put into domestic mass production. These are small ships, suitable for local patrols and domestic defense, but not for significant force projection. A boat of this type was captured by the South Koreans in the 1996 Gangneung submarine infiltration incident.
10 Yono-class submarines: These are midget subs, at 130 tons. They have a crew of two with 6-7 special forces onboard. A ship of this class is believed to have been involved in an October 2010 incident in which a South Korean corvette was destroyed by an unknown North Korean submarine.
In addition, North Korea is believed to have fielded a new submarine as recently as 2014. Little is known about this new, Sinpo-class design — it may be based on older vessels from Yugoslavia or incorporate technology from newer Russian vessels. South Korean sources suggested at one point that it could be built from old Golf II-class hulls that North Korea imported in the early 1990s, but this has not been confirmed. If rumors are true, it would be North Korea’s first ballistic missile submarine, though it lacks the ability to project force much outside of NK’s territorial waters.
As of this writing, the KPN has not requested outside assistance or aid with any rescue efforts. It could be some time before we know which submarine was lost and whether or not any survivors were rescued by the recovery crews.
Subscribe to:
Posts (Atom)
![NG-[C:G.Vide,B:Allbrand,Wk:5015,Dim:160X600]:Televisions](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_tng2keZfIa9nTf66vUehU9iY_IIZloqa46moIGQuX8uSWbQBu2Qyi_thCJSezcCEHE5Wyyk5lyy7gtZFff6GEW8eCj4HjvCoFSsSYaJmGqRPBH1r_CRSj0sQrRnwhIxwOk2a_QHcanulReXFh0OHChUsj57IpjeeVvL0fgHXp51pHZy2UhjA=s0-d)
![NG-[C:G.Comp,B:Allbrand,Wk:5015,Dim:336X280]:Tablets](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_s54-lcJKjPFIwd7CBuiPEvNVGnrFTkIL5ikxboMlt1XxYZ4_sKVtUewzTuDOX-dgTOAOuKejrEoCrnQ95QWL976QgfG39mAvvGsILnIon_6Rl_EjpaNnnWc1bPqHta7ZfxvfolQANEOmuJJ3G-d6F2SUYkKr5LX60ujPGRfS9ADxu5MsI1Wg=s0-d)
![NG-[C:G.Prom,B:Allbrand,Wk:4915,Dim:300X250]:Home_Page](https://lh3.googleusercontent.com/blogger_img_proxy/AEn0k_tZeK5cWUtbxV-05t--ktnAZCY5cxbMAECR941MDij8LaFs_D-NJAj7aSrOXFmLSyJQfF6P9W0VkzAkyjfUH4IAR4v7U_jk21jdFtBEhi2sWnKxXz8pHOna5dbtmnAcukGF5l0sJUcbNbk6Vm-UdEu9oqVTBNrmbn1iRnOekhn7HEtRlngwiQ=s0-d)