Same goes for vision processing. The reason most robots today use LIDARs as their primary "vision" sensor is that image processing is very slow and not very good. With a 3d LIDAR you can do mapping, object detection and recognition, people detection, grasp planning, obstacle avoidance, etc. all very easily in polynomial time. Most of these tasks are computationally intractable using images, if they're possible at all (for example, night time operations).
10 years ago, that was true. Not any longer. Take a look at
DARPA's current robot manipulation project. Watch a vision-guided robot with two three-fingered hands put a key in a lock, turn the key, and open a door. That's being done with stereo vision and force feedback. The gold-colored device with two lenses is a Point Grey stereo vision device.
Much of early vision processing is done in custom hardware or GPUs now. It's not "computationally intractable". It's fairly expensive computationally, because the amount of work per pixel is high. But it's not exponential.
When I ran a DARPA Grand Challenge team a decade ago, we relied on LIDAR too much, and vision not enough. The winning teams relied almost entirely on vision when they were going fast - the LIDARs didn't have the range or angular accuracy to read terrain out to the stopping distance. What they got from vision was basically "distant part of road looks like near part of road - OK to go fast if near part of road is good", obtained from a classifier system. So they could out-drive their LIDARs. We couldn't exceed 17mph.
That's fine for the software... what about the hardware? Do we have actuators (hydraulic, pneumatic, whatever) that are good enough? And most important, do we have a power supply that's good enough?
Yes. They're not common items, but the right stuff does exist. There are military robots powered by small Diesel engines. The LS3 (the militarized successor to BigDog) is powered by a small Diesel engine, which drives both a generator and a hydraulic pump. (The U.S. military is all-Diesel; one standard fuel runs everything.)
Miniature hydraulic cylinders and proportional valves provide actuators with very good power to weight ratios. Modern brushless servomotors aren't bad, either. Pneumatics are also an option. (Precise position control with pneumatics is quite possible if the servovalve is close to the cylinder.)
This is very ambitious. Yet it's almost within reach, if enough money is thrown at the problem. A lot of money.
First, this requires solving low-speed legged locomotion over difficult terrain. Walking over rubble and climbing a ladder will be very tough. But it's not totally out of reach. A machine which can walk on either two or four limbs would be an advantage over rubble. Ladder climbing is a four-limb problem, and it's been done at least once. Grabbing the rungs with all limbs will make it easier.
Replacing a pump is an interesting problem. I don't think anyone has yet demonstrated complex part replacement in an unstructured environment. That's probably the least well developed part of the problem.
Building a suitable machine will be expensive. Each Boston Dynamics robot has cost upwards of $20 million. Even the Willow Robotics machine is over $100K per unit, and it's just two arms on a wheeled base. This thing has to have roughly human dimensions and be self contained. Petman isn't self-contained; it needs external power.
Yet, looking at the problem, I can see how to approach it. Modern control theory is good enough. Machine learning and vision processing are good enough. Simulators are good enough to allow debugging in simulation. Enough people know this stuff that the job can be staffed.
Now, if Slashdot wanted to have cool videos, they could have NMA do it. NMA is a video production house and the fastest animation shop in the industry. They turn out animations in hours.
I suspect their results are more along the lines of "what proportion are willing to tell us they're gay" rather than "what proportion is gay."
The designers of the study were concerned with that, and their methodology involved both hour-long interviews and sealed questionnaires not seen by the interviewers. The actual question they asked, though, is "had sex with... in the last year", which requires finding someone with the desired inclination. Opportunity declines with population density.
No, only about 1 in 40 in the US. See "Sex in America, a Definitive Survey" 2.8% of men, 1.4% of women. 1 in 10 in the largest 5 cities, much lower in more sparsely populated areas. Data based on detailed interviews with 3400 people randomly selected, not self-selected.
By the end of the 1960s, rock amps had achieved enough power to reach the threshold of pain. From then on, much of the "wall of amps" thing was fake. You just didn't need that much speaker area to hit the threshold of pain.
A friend of mine was a roadie for metal groups years ago, and she discovered this when setting up for Metallica. Most of the "amps" were empty boxes. At least they were enclosed boxes. In the picture above, the low-budget metal band just used fake fronts.
It's helpful to register trademarks on your important domains, if they're unique enough. This means a quick win in a UDRP proceeding, and gives you the option of suing anyone who ended up with your domain. It's about $400 per domain.
More importantly, own your domains. If WHOIS doesn't have your name and address in "Registrant", you do not own the domain. You're just renting it from somebody. Your hosting provider should never have their name in there. This really matters when there's a dispute. Deal directly with your domain registrar. Do not deal with them through a hosting service.
"Private registration" works the same way. The "private registration" service owns the domain, and you have a contractual relationship with them, at best. See what happened when RegisterFly went bust.
Organizing that stuff is hard work. Work continues getting 1960s protest info cataloged. Stanford had a group trying to organize Martin Luther King's stuff. That took years. Then they got the archives of the Black Panther Party, and are now grinding through that. The archives of Students for a Democratic Society (SDS) are at Kent State.
Much of the plder stuff is too variable for fast scanning. Somebody has to put posters, handouts, and brochures through a flatbed, slowly. The fast book scanners need more structure.
I had an exhibit at Maker Faire once. I realized I was being used as free entertainment for a flea market. Haven't been there since. Maker Faire is a for-profit operation run by O'Reilly Media, whose main business is running overpriced conferences.
First, "being persistent" (common dating advice for men in earlier decades) became "stalking". Telling a woman she looked hot became "sexual harassment", even when the man had no power over the woman. Now, asking for sex makes you a "sexual predator". And if a woman agrees to sex, men have to worry that she may later claim she was raped.
The location, 150 miles south of the Haida Gwaii islands, is about 200 miles west of Vancouver. The currents along the Pacific coast of North America run southward, so the ship is going to drift into US waters soon. The U.S. will probably get stuck with the bill for towing the thing before it hits something. It's floating high in the water, so the hull is in good shape.
Automatic hot nightclub recognition would be fun. Crunch on Facebook data to see who's popular. Use Face Beaury Rank (see "Automatic Classification of Chinese Female Facial
Beauty using Support Vector Machine" for the theory) to see who's good looking. Use Foursquare data to see what places fill up with hot women. Compute the male/female ratio for locations. Discard places which are almost all female (those are probably beauty salons, etc.) Display on map.
OK, so they have a web site.
It's hosted in Malaysia, runs Joomla content management, and uses Gmail for replies. They have Facebook and Twitter links. Their videos are on Youtube, and they have a movie site to provide a front for them. The video isn't too useful without translation.
The Navy wants this so that, when they're dealing with a small boat that's causing a problem, they have an option between "ignore" and "blow them out of the water". Somalia pirates, smugglers, boats getting too close (see USS Cole) - things like that.
How many more years will slashdot have an off-by-one error on your Score in your profile?
Right. The people behind Slashdot failed "Web 2.0."
Could be worse. Take a look at tribe.net . It was cool once. Then they went Web 2.0. They tried to emulate Myspace's user-redesignable pages. They botched it so bad that everybody left. Looking at my old account today, of the 20 tribes to which I subscribed, one has updates: "Tribe.net Bug Reports - 2264 new".
From an add-on developer perspective, Firefox's frantic updates are a pain. I have the same add-on for Firefox and Google Chrome. Most of the code is common. On the Firefox side, I have work-arounds for two bugs in Firefox, and they've been open bug reports in Bugzilla for many months. There's a new bug this week because the last update to the Mozilla add-on SDK broke something in message passing. That's supposedly fixed in the next version of the SDK being released today. Now I have to rebuild, update and test my add-on, then run it through the Mozilla approval bureaucracy again. (Yes, the AMO web site says this happens automatically. That's only true if you let them host the source code.)
Over on Google Chrome, it just works. No workarounds needed. A stable API. No updates needed from my side.
I get far more downloads of the Firefox version, though.
Gawker already uses tracking from Google, Facebook, Quantcast, Dedicated Networks, Comscore Beacons, Google Analytics, ChartBeat, DoubleClick, Parse.ly, New Relic. (Abine.com has a tool to detect and block such things.)
The Raspberry Pi is already obsolete. Rhombus-Tech is coming out with a board based on the Allwinner ARM implementation, 3x as fast as the obsolete CPU the Pi crowd is using. "Mass-volume pricing (just for the CPU card, and therefore excluding tax, shipping, profit, a case and a power supply) looks to be on target for around $15:" They're also looking at reusing the BeagleBoard form factor (which is much like an Arduno) and coming out with a fast Linux board in that format.
By the time the Raspberry Pi crowd delivers, they'll be obsolete. Much like the OLPC.
Look at the 50 years 1912-1962. Science produced the theory for radio, tubes, transistors, large scale electric systems, plastics, worldwide communications systems, radar, jet aircraft, antibiotics, computers, and nuclear power. Then look at the next 50 years, 1962-2012. It's all improvements on that stuff.
There hasn't been a new energy source in the last 50 years. Solar cells and atomic power are more than 50 years old now. Fusion was a bust. Yes, more volumes of Physical Review come out each year, but they don't contain breakthroughs like fission. There are more scientists than ever, but output has declined, because most of the easy results have been found. Innovation peaked around 1870, during Edison's lifetime. That was when steel, steam, and electricity came together and much of the modern world was developed. Edison's lab had a goal of a minor invention every 3 days, and a major breakthrough every two weeks. That was with a staff of about 30. Nobody has productivity like that today. There's progress in the bio area, but the level of effort required for even small progress is quite high.
Science is a non-renewable resource. The effort required for new discovery increases over time. The easy discoveries have already been made. The cost of new discovery increases steadily, and eventually becomes uneconomic. This is why big company research labs disappeared in the 1980s.
Startups do not help much. Today's "innovations" are things like Twitter. I go to venture capital meetings, and most of the ideas are at that level. (Or much lower; I've heard a pitch for a social network for cats.) It's all about applying too much technology to banal tasks.
There is such a stack: Open JAUS. JAUS is the Joint Architecture for Unmanned Systems used by many military robotic and unmanned systems. It's somewhat dated, and has a more open-loop approach more suited to teleoperators than fully autonomous systems.
Dealing with the time constraints in robotics rules out some of the approaches used in other software. Microsoft's Robotics Studio was built on a web-like approach, and it was a flop. Game programs tend to be tied to the display refresh rate, which isn't helpful in robotics. In robotic systems, there may be several subsystems with their own cyclic rate and processing delay, and they need to talk to each other. The inputs which have processing delays, like vision systems, produce outputs which represent the situation at some time in the past. Updates to the world model based on multiple sensors must all be synchronized to the time of the observation, not the time the data became available. This matters when you're moving fast. For slow robots, not so much. Many research robots are slow and pause a lot because they don't do this. That was the norm a few years ago, but it's not any more.
Robotic systems tend to need hard real time control. That control can be quite complex, not just a simple servo loop. Inside the more advanced and agile robots, like BigDog, you tend to find QNX, not Linux. (Typical test for a hard real-time OS: hook up a square wave oscillator to an input, and a scope to an output. Put a high-priority program in the system which turns on the output when the input comes on. Watch the input to output delay on the scope. Load up the system with lower-priority tasks. If the input to output delay is ever substantially longer, (more than a few microseconds) the system is not hard real time. The "real time" variants of Linux have trouble getting down to 1ms, and 10ms of jitter is observed. In hard real time systems, 10us is more like it. Servo control in BigDog executes every 1ms, balance every 10ms.) However, as CPUs get faster, the limitations of Linux have become less of an issue.
They mean it. If you're logged in, you show up on their map, visible to other users. They can access your Facebook account. The Capitol of Panem knows all, sees all. May the odds be ever in your favor.
Right. Better Place is "all hat and no cattle", as they say in Texas. Their original scheme, to deploy swap stations in areas where you can't drive too far, like Hawaii and Israel, made sense. But they didn't actually do it. Instead, they just hyped up new deals.
Even the Tokyo taxicab demo was only 3 cabs running for 6 months before it closed down. That's the ideal situation - a uniform fleet of vehicles which stay in a small area and return daily to a central location. If it didn't work economically there, it's probably a failure.
The NYC Taxi and Limousine Commission has a bigger electric taxi trial than that. They're now trying round 2, with Nissan Leafs. Round 1 tried an electric PT Cruiser. Didn't work. The NYC Taxi and Limo Commissioner growled "It got to spend a lot of time on the back of a flatbed tow truck and not a lot of time as a taxicab". Better Place didn't compete for that market. They probably didn't want to face the NYC "it had better work" attitude.
Microsoft is aggressive towards their competitors, as was IBM in its day. Both had antitrust problems. Google and Facebook are aggressive towards their users. They have privacy-invasion problems.
This is the price of ad-supported "free". Microsoft wants you to buy their stuff. You're the customer. With Google and Facebook, you're the product.
Slashdot Mirror
Same goes for vision processing. The reason most robots today use LIDARs as their primary "vision" sensor is that image processing is very slow and not very good. With a 3d LIDAR you can do mapping, object detection and recognition, people detection, grasp planning, obstacle avoidance, etc. all very easily in polynomial time. Most of these tasks are computationally intractable using images, if they're possible at all (for example, night time operations).
10 years ago, that was true. Not any longer. Take a look at DARPA's current robot manipulation project. Watch a vision-guided robot with two three-fingered hands put a key in a lock, turn the key, and open a door. That's being done with stereo vision and force feedback. The gold-colored device with two lenses is a Point Grey stereo vision device.
Much of early vision processing is done in custom hardware or GPUs now. It's not "computationally intractable". It's fairly expensive computationally, because the amount of work per pixel is high. But it's not exponential.
When I ran a DARPA Grand Challenge team a decade ago, we relied on LIDAR too much, and vision not enough. The winning teams relied almost entirely on vision when they were going fast - the LIDARs didn't have the range or angular accuracy to read terrain out to the stopping distance. What they got from vision was basically "distant part of road looks like near part of road - OK to go fast if near part of road is good", obtained from a classifier system. So they could out-drive their LIDARs. We couldn't exceed 17mph.
That's fine for the software... what about the hardware? Do we have actuators (hydraulic, pneumatic, whatever) that are good enough? And most important, do we have a power supply that's good enough?
Yes. They're not common items, but the right stuff does exist. There are military robots powered by small Diesel engines. The LS3 (the militarized successor to BigDog) is powered by a small Diesel engine, which drives both a generator and a hydraulic pump. (The U.S. military is all-Diesel; one standard fuel runs everything.)
Miniature hydraulic cylinders and proportional valves provide actuators with very good power to weight ratios. Modern brushless servomotors aren't bad, either. Pneumatics are also an option. (Precise position control with pneumatics is quite possible if the servovalve is close to the cylinder.)
This is very ambitious. Yet it's almost within reach, if enough money is thrown at the problem. A lot of money.
First, this requires solving low-speed legged locomotion over difficult terrain. Walking over rubble and climbing a ladder will be very tough. But it's not totally out of reach. A machine which can walk on either two or four limbs would be an advantage over rubble. Ladder climbing is a four-limb problem, and it's been done at least once. Grabbing the rungs with all limbs will make it easier.
Replacing a pump is an interesting problem. I don't think anyone has yet demonstrated complex part replacement in an unstructured environment. That's probably the least well developed part of the problem.
Building a suitable machine will be expensive. Each Boston Dynamics robot has cost upwards of $20 million. Even the Willow Robotics machine is over $100K per unit, and it's just two arms on a wheeled base. This thing has to have roughly human dimensions and be self contained. Petman isn't self-contained; it needs external power.
Yet, looking at the problem, I can see how to approach it. Modern control theory is good enough. Machine learning and vision processing are good enough. Simulators are good enough to allow debugging in simulation. Enough people know this stuff that the job can be staffed.
Bonus points if you can find a hot East Asian girl willing to bash China.
NMA's Li Anne, out of Tapei, does that.
Now, if Slashdot wanted to have cool videos, they could have NMA do it. NMA is a video production house and the fastest animation shop in the industry. They turn out animations in hours.
I suspect their results are more along the lines of "what proportion are willing to tell us they're gay" rather than "what proportion is gay."
The designers of the study were concerned with that, and their methodology involved both hour-long interviews and sealed questionnaires not seen by the interviewers. The actual question they asked, though, is "had sex with ... in the last year", which requires finding someone with the desired inclination. Opportunity declines with population density.
Isn't the figure that 1 in 10 people is LGBorT?
No, only about 1 in 40 in the US. See "Sex in America, a Definitive Survey" 2.8% of men, 1.4% of women. 1 in 10 in the largest 5 cities, much lower in more sparsely populated areas. Data based on detailed interviews with 3400 people randomly selected, not self-selected.
By the end of the 1960s, rock amps had achieved enough power to reach the threshold of pain. From then on, much of the "wall of amps" thing was fake. You just didn't need that much speaker area to hit the threshold of pain.
A friend of mine was a roadie for metal groups years ago, and she discovered this when setting up for Metallica. Most of the "amps" were empty boxes. At least they were enclosed boxes. In the picture above, the low-budget metal band just used fake fronts.
It's helpful to register trademarks on your important domains, if they're unique enough. This means a quick win in a UDRP proceeding, and gives you the option of suing anyone who ended up with your domain. It's about $400 per domain.
More importantly, own your domains. If WHOIS doesn't have your name and address in "Registrant", you do not own the domain. You're just renting it from somebody. Your hosting provider should never have their name in there. This really matters when there's a dispute. Deal directly with your domain registrar. Do not deal with them through a hosting service.
"Private registration" works the same way. The "private registration" service owns the domain, and you have a contractual relationship with them, at best. See what happened when RegisterFly went bust.
Organizing that stuff is hard work. Work continues getting 1960s protest info cataloged. Stanford had a group trying to organize Martin Luther King's stuff. That took years. Then they got the archives of the Black Panther Party, and are now grinding through that. The archives of Students for a Democratic Society (SDS) are at Kent State.
Much of the plder stuff is too variable for fast scanning. Somebody has to put posters, handouts, and brochures through a flatbed, slowly. The fast book scanners need more structure.
I had an exhibit at Maker Faire once. I realized I was being used as free entertainment for a flea market. Haven't been there since. Maker Faire is a for-profit operation run by O'Reilly Media, whose main business is running overpriced conferences.
First, "being persistent" (common dating advice for men in earlier decades) became "stalking". Telling a woman she looked hot became "sexual harassment", even when the man had no power over the woman. Now, asking for sex makes you a "sexual predator". And if a woman agrees to sex, men have to worry that she may later claim she was raped.
The "sex is bad" side has won.
The location, 150 miles south of the Haida Gwaii islands, is about 200 miles west of Vancouver. The currents along the Pacific coast of North America run southward, so the ship is going to drift into US waters soon. The U.S. will probably get stuck with the bill for towing the thing before it hits something. It's floating high in the water, so the hull is in good shape.
Automatic hot nightclub recognition would be fun. Crunch on Facebook data to see who's popular. Use Face Beaury Rank (see "Automatic Classification of Chinese Female Facial Beauty using Support Vector Machine" for the theory) to see who's good looking. Use Foursquare data to see what places fill up with hot women. Compute the male/female ratio for locations. Discard places which are almost all female (those are probably beauty salons, etc.) Display on map.
OK, so they have a web site. It's hosted in Malaysia, runs Joomla content management, and uses Gmail for replies. They have Facebook and Twitter links. Their videos are on Youtube, and they have a movie site to provide a front for them. The video isn't too useful without translation.
"The Afghanistan Of Islam Rejects Pollution of Western Democracy" is interesting reading. It's a good summary of the theocratic position, and gives some insight into why this is such a tough war to end.
The Navy wants this so that, when they're dealing with a small boat that's causing a problem, they have an option between "ignore" and "blow them out of the water". Somalia pirates, smugglers, boats getting too close (see USS Cole) - things like that.
How many more years will slashdot have an off-by-one error on your Score in your profile?
Right. The people behind Slashdot failed "Web 2.0."
Could be worse. Take a look at tribe.net . It was cool once. Then they went Web 2.0. They tried to emulate Myspace's user-redesignable pages. They botched it so bad that everybody left. Looking at my old account today, of the 20 tribes to which I subscribed, one has updates: "Tribe.net Bug Reports - 2264 new".
From an add-on developer perspective, Firefox's frantic updates are a pain. I have the same add-on for Firefox and Google Chrome. Most of the code is common. On the Firefox side, I have work-arounds for two bugs in Firefox, and they've been open bug reports in Bugzilla for many months. There's a new bug this week because the last update to the Mozilla add-on SDK broke something in message passing. That's supposedly fixed in the next version of the SDK being released today. Now I have to rebuild, update and test my add-on, then run it through the Mozilla approval bureaucracy again. (Yes, the AMO web site says this happens automatically. That's only true if you let them host the source code.)
Over on Google Chrome, it just works. No workarounds needed. A stable API. No updates needed from my side.
I get far more downloads of the Firefox version, though.
Now this is a site that insists on registration.
Gawker already uses tracking from Google, Facebook, Quantcast, Dedicated Networks, Comscore Beacons, Google Analytics, ChartBeat, DoubleClick, Parse.ly, New Relic. (Abine.com has a tool to detect and block such things.)
Now Gawker wants an anal probe, too?
The Raspberry Pi is already obsolete. Rhombus-Tech is coming out with a board based on the Allwinner ARM implementation, 3x as fast as the obsolete CPU the Pi crowd is using. "Mass-volume pricing (just for the CPU card, and therefore excluding tax, shipping, profit, a case and a power supply) looks to be on target for around $15:" They're also looking at reusing the BeagleBoard form factor (which is much like an Arduno) and coming out with a fast Linux board in that format.
By the time the Raspberry Pi crowd delivers, they'll be obsolete. Much like the OLPC.
Look at the 50 years 1912-1962. Science produced the theory for radio, tubes, transistors, large scale electric systems, plastics, worldwide communications systems, radar, jet aircraft, antibiotics, computers, and nuclear power. Then look at the next 50 years, 1962-2012. It's all improvements on that stuff.
There hasn't been a new energy source in the last 50 years. Solar cells and atomic power are more than 50 years old now. Fusion was a bust. Yes, more volumes of Physical Review come out each year, but they don't contain breakthroughs like fission. There are more scientists than ever, but output has declined, because most of the easy results have been found. Innovation peaked around 1870, during Edison's lifetime. That was when steel, steam, and electricity came together and much of the modern world was developed. Edison's lab had a goal of a minor invention every 3 days, and a major breakthrough every two weeks. That was with a staff of about 30. Nobody has productivity like that today. There's progress in the bio area, but the level of effort required for even small progress is quite high.
Science is a non-renewable resource. The effort required for new discovery increases over time. The easy discoveries have already been made. The cost of new discovery increases steadily, and eventually becomes uneconomic. This is why big company research labs disappeared in the 1980s.
Startups do not help much. Today's "innovations" are things like Twitter. I go to venture capital meetings, and most of the ideas are at that level. (Or much lower; I've heard a pitch for a social network for cats.) It's all about applying too much technology to banal tasks.
There is such a stack: Open JAUS. JAUS is the Joint Architecture for Unmanned Systems used by many military robotic and unmanned systems. It's somewhat dated, and has a more open-loop approach more suited to teleoperators than fully autonomous systems.
Dealing with the time constraints in robotics rules out some of the approaches used in other software. Microsoft's Robotics Studio was built on a web-like approach, and it was a flop. Game programs tend to be tied to the display refresh rate, which isn't helpful in robotics. In robotic systems, there may be several subsystems with their own cyclic rate and processing delay, and they need to talk to each other. The inputs which have processing delays, like vision systems, produce outputs which represent the situation at some time in the past. Updates to the world model based on multiple sensors must all be synchronized to the time of the observation, not the time the data became available. This matters when you're moving fast. For slow robots, not so much. Many research robots are slow and pause a lot because they don't do this. That was the norm a few years ago, but it's not any more.
Robotic systems tend to need hard real time control. That control can be quite complex, not just a simple servo loop. Inside the more advanced and agile robots, like BigDog, you tend to find QNX, not Linux. (Typical test for a hard real-time OS: hook up a square wave oscillator to an input, and a scope to an output. Put a high-priority program in the system which turns on the output when the input comes on. Watch the input to output delay on the scope. Load up the system with lower-priority tasks. If the input to output delay is ever substantially longer, (more than a few microseconds) the system is not hard real time. The "real time" variants of Linux have trouble getting down to 1ms, and 10ms of jitter is observed. In hard real time systems, 10us is more like it. Servo control in BigDog executes every 1ms, balance every 10ms.) However, as CPUs get faster, the limitations of Linux have become less of an issue.
Want to see control by the content creator? Check out The Capitol.pn This is a web site that's up front about being about control. It requires both Internet Explorer and Facebook. "Even you are being monitored as we speak."
They mean it. If you're logged in, you show up on their map, visible to other users. They can access your Facebook account. The Capitol of Panem knows all, sees all. May the odds be ever in your favor.
Right. Better Place is "all hat and no cattle", as they say in Texas. Their original scheme, to deploy swap stations in areas where you can't drive too far, like Hawaii and Israel, made sense. But they didn't actually do it. Instead, they just hyped up new deals.
Even the Tokyo taxicab demo was only 3 cabs running for 6 months before it closed down. That's the ideal situation - a uniform fleet of vehicles which stay in a small area and return daily to a central location. If it didn't work economically there, it's probably a failure.
The NYC Taxi and Limousine Commission has a bigger electric taxi trial than that. They're now trying round 2, with Nissan Leafs. Round 1 tried an electric PT Cruiser. Didn't work. The NYC Taxi and Limo Commissioner growled "It got to spend a lot of time on the back of a flatbed tow truck and not a lot of time as a taxicab". Better Place didn't compete for that market. They probably didn't want to face the NYC "it had better work" attitude.
Microsoft is aggressive towards their competitors, as was IBM in its day. Both had antitrust problems. Google and Facebook are aggressive towards their users. They have privacy-invasion problems.
This is the price of ad-supported "free". Microsoft wants you to buy their stuff. You're the customer. With Google and Facebook, you're the product.