Submitted by brad on Fri, 2004-02-20 06:25.
I like to use our Rio Karma MP3 player in the car, but it's not nearly as good as it could be. So here are some jottings on what an ideal car dock would do for the player.
- Power and charge the player, of course
- Offer various options for sending audio to the car, including a built-in quality FM transmitter, a port for a special Cassette sized interface (more below) and various cables for car stereos that have an accessory jack (as mine has for a trunk CD-changer) or plain audio inputs.
- A wireless remote control to stick on the wheel (not needed if other remote control methods can work.)
- A microphone.
- To get really fancy, an 802.11 interface to allow it to sync up with computers inside the house while in the driveway. Though strictly, this would be even better inside the player, not in the dock.
The microphone would perform several roles. One, it would detect the ambient sound level in the car, and boost the music volume as the car gets noisier. No more super-loud when you start the car either.
Secondly, it would listen for the sound of the music the player is playing. It would try to tell if it was playing, so it could detect when the stereo is turned off or switched to something else, or when the car is turned off (if the loss of power from the accessory jack doesn't already reveal this.) When the sound stops (even if this takes 5 seconds to confirm) just pause the music back in time when the sound was first detected to stop. One could then from time to time send out pulses of the forthcoming audio, and if it hears them, treat that as a resumption of play. read more »
Submitted by brad on Wed, 2004-02-18 15:36.
Many people, trying to address concerns about the privacy implications of RFID tags have indicated that it can just become the norm, or even a requirement, to "burn" out the RFID tags in purchased products as they are sold.
I'll get to why that doesn't work in a moment, but first some background. RFID tags are cheap passive radio devices planned to go into most consumer products, replacing the bar-code. A reader can, within certain range of the tags, read the serial numbers of all tags in the area. Every tag has a unique number, so it makes a great bar-code for inventory control.
Soon your body will be covered with RFID tags everywhere you go. In your clothes, boots, watch, wallet, glasses, ID-badges, credit-cards etc. Scanners may show up everywhere. This provides the potential to put them on city streets, doorways, airports, train stations and so on, and, once you have scanned a person once, to track everybody's movements everywhere. Pretty 1984.
Here's the rub. We're going to want some of these tags. Not just to return products to the store. Today the readers are expensive, but soon they will be cheap, and we'll want to have something we can use to find our keys, wallet, glasses, watch or other losables. To let us know what's in our closets, on our bookshelves. We'll love it. So we won't burn out the IDs.
The only answer I have thought of (I don't think Rivest's jammer will work) is more expensive RFIDs that can be modified instead of burned out, so that they will no longer respond to any scanner, just to our personal one. So they work for us, not for others.
Even with this the IDs in credit cards, access cards and such will need to work in more scanners. How will we turn them off?
Submitted by brad on Tue, 2004-02-17 10:12.
I've just put up a new essay on my web site on whether challenge/response anti-spam systems are good or bad
As some may know, I've been running such a system longer than anybody, having written one in 1997. I wrote a white paper on best practices for such systems that some have found useful.
However, I also see a lot of complaints about C/R systems. Most of those complaints are about the new crop of C/R systems, many of which have annoying bugs. Because some of the concerns are real, however, I felt it was time for an article on those issues for a well-behaved C/R system.
Note that even my own system, in spite of being better behaved that most newer systems, does not meet all my own best practices, though it would if I were writing it again.
Submitted by brad on Tue, 2004-02-17 05:06.
Generally, I'm the last person to suggest we use technology to control people's lives and what they view. However, it's also the duty of parents to help teach their children how and when to use the media. Most commonly today you see things like the V-chip, which let parents block their unskilled children from seeing shows with certain "ratings."
A far more useful concept, I think, would be a device which limits the amount of time children can spend watching the TV. What they watch in that context can be mostly up to them, and if they understand the concept of a time budget, it will probably improve not just how much they watch but what they watch.
A PVR like the Tivo, or in particular, the DirecTivo, is the ideal platform for doing this. Children would get their own remotes, or a code to enter on the master remote to start using their weekly budget of TV hours. Once the budget was used up, they could not watch TV for a while. With a PVR, this would not block them from seeing a highly desired show, but it would delay it.
If two children wanted to watch the same show they could both enter their code to halve the amount of TV credit used, encouraging sharing and (minimal) socialization. Siblings would pretty quickly develop a market, trading TV hours like prison cigarettes with one another for real-world things, even money. This need not be discouraged. Random TV surfing would be discouraged, and commercial viewing strongly discouraged.
Adults would have to take the burden of having to enter their own code for unmetered viewing, a price they would pay to cut down their kid's TV hours.
Of course there are also some privacy considerations to consider. read more »
Submitted by brad on Thu, 2004-02-12 12:46.
As I noted earlier, there are all sorts of risks with remote voting over the internet, even if I suggest a way to make it doable. However, this is different from the question of voting machines. Like the folks at Verified Voting I believe that a voter-verifiable paper ballot is the simplest way to make computerized voting more secure. And I like voting machines because they can improve access and even make preferential ballot possible down the road.
But I look at the huge cost we are paying for voting machines. I propose breaking the voting machine process into two steps. The first is the ballot preparation machine. It helps the voter generate their ballot, and then prints it out on paper, in a human readable form that is also machine readable. You need lots of those.
With paper ballot in hand, you walk over to the scanning machine, which is stage 2. This machine reads the standard-format paper ballot, does OCR on the human readable text and confirms the ballot is readable as the voter desires. It also counts it. The ballot is then placed in a locked ballot box.
The scanning machine will be expensive, and secured, and built by an audited vendor. However, you need only a small number of those. The voting stations, which you need many of, can mostly be cheap. In fact, they can be free.
That's because you would generate a voting program that runs on standard PC hardware. On slow standard PC hardware. Probably an open source program, meant to run on Linux, and audited and verified by the open source community. They would love this job.
Then you ask the public to donate their old, slower PCs. Give them a small tax deduction if needed, but frankly I think you would get so many machines you wouldn't even need that. You could even be strict on the hardware requirements. Wipe the bios and put in a fresh one, possibly put in a cheap hard disk with the voting system installed. Get donated laser printers. You don't have a lot of security concerns with these machines because there is not a lot they can do to bollux the election. read more »
Submitted by brad on Tue, 2004-02-10 08:41.
In 2000, the Florida Presidential election ended up in a tie. Many people get offended at that remark, because they don't think of elections as being compatible with ties, they insist that their candidate really won.
However, to scientists, you have a tie when the results differ by less than the margin of error. And I refer not simply to the margin of error from the problems in the voting machines, but a much more bizarre margin caused by political pressure to interpret the results in different ways.
This is a curse because if you get a close result, there will be tremendous political and financial captial spent to try to push the results one way or another by redefining the rules. It's an unstable situation, solved that time only by damaging the surpreme court by forcing it to make such partisan rulings.
How to avoid this? One of the causes of the problem is that almost all states decide to give all their electoral votes to the winner of their state, rather than apportioning them. One reason they do this is that it makes the state a bigger prize in the election, and so the candidate works harder to please the state. (This is the same factor, as they will also work like crazy in a tie to be the prevailing party.)
To solve this I propose a slightly different formula for allocating the electoral votes. I'll flesh it out with 2 candiates in a state with 50 votes and 10M voters.
If one candidate gets, say, 51% of the vote (5.1M votes or more) then give them all the electoral votes. This, as before, keeps the candidate very interested in winning the state and pleasing its voters. If a candidate gets under 49%, they get zero as before.
If they get between 49% and 51%, apportion the votes on the pro-rata portion of this region. For example, if both candidates get 50%, they split the votes, 25 each. If one candidate gets 4.92M votes and the other 5.08M, we see 5 votes for the first candidate and 45 for the 2nd.
What this means is that if, by recounting or re-interpeting, you can add 4,000 votes to your total, you win one whopping electoral vote.
So it's worth fighting and recounting a bit, but not going crazy, because you aren't going to change the results a lot no matter what you do.
Yes, this means Gore and Bush would have split Florida's votes and Gore won the presidency, but of course it could easily benefit the other side under different circumstances. read more »
Submitted by brad on Fri, 2004-02-06 07:47.
Reading this NYT article about radar to cover car blind spots, which describes a system that will trigger lights in the rearview-mirror when cars are in the blind-spot, reminded me of an old idea I had some time ago I called “Eyes in the back of your head.”
The idea would be to wear a special collar while driving. This collar would contain small electrodes that could lightly stimulate the skin on the back of the neck. Perhaps just one row, but ideally a small 2-D image should be possible.
This would be connected to a camera, radar or sonar system pointing back from the vehicle. It would map where other vehicles are, and turn that into an image on the back of the neck.
Thus, as a car came up behind you and passed you, it would feel like something brushing the back of your neck on one side.
I was inspired to this by reading about a system for the blind that mapped a video camera image onto a 2000 pixel electrode map on the stomach. It was found that over time, the nerves would retrain and a sort of limited vision could develop. Might this have application in driving, or perhaps combat? read more »
Submitted by brad on Thu, 2004-02-05 05:59.
I hinted last week I would write about a peril from and to automatic cars, or actually any drive-by-wire cars.
That peril is they become highly useful terrorist weapons. Today terrorists get kamikazis to drive ordinary cars to attack targets and checkpoints. It will be easy to modify a drive-by-wire car (including the self-parking cars already on the market) to be controlled by the cheap remote controls found on toy cars and planes today, and easy to mount a wireless camera (X10, the terrorist's tool!) as well.
A remote control car can be a weapon on its own, just to smash into things, but more nastily it can be loaded with explosives or poison or other nasty things. If drive-by-wire cars become commonplace (and they will) this will be possible.
I present a problem without good solution, and I also fear some of the solutions even more than the problem. For example, one of the big advantages of the automatic self-parking car which I described earlier is the car that drops you off and picks you up right at the door of where you're going. However, just as false anti-terrorist security has made it almost impossible to park or pick people up at some airports, they will move to ban all vechicles from going just where we want them to go.
They may also start demanding government overrides for the automatic cars, so police can take control of our vehicles on demand, bypassing even manual control. They will try to tightly regulate the technology (stifling it) and only allow blessed companies to work on it. As I said, a problem without obvious solution.
Submitted by brad on Tue, 2004-02-03 16:54.
We've all seen public bathrooms where the women have a line snaking out into the hall, but we guys can just "whiz" in and out. We have sympathy (but not too much, see the joke)
Here's a good solution that will probably never fly because we're uptight. Two bathrooms: One small one with nothing but urinals for men, and another one with nothing but stalls to be shared among the sexes.
In effect, it would offer most of the building's entire pool of stalls to the women when they need it, instead of the 60% or so they get now.
Why not do this? Well, some women would prefer men not use their stalls, due to certain prejudicial (or justified) opinions they might hold about our collective restroom hygene. And women might also miss the private "just for women" chat and appearance adjustment space at the sink.
These concerns could be fixed one of two ways. Have men's and women's "ends" to the line of stalls, so only when it was full would women find themselves moving into the area we've used. And possibly have two private sink areas, since there is not as much contention there, and they take less space anyway.
I can credit the idea that the sexes want their private space.
Oh yeah, the toilets could have a little pin that pokes out when the flush lever is pulled that tips down the seat if it was left up. Does that make it better? :-)
Or are we just too uptight to do this? I know women who, facing the long line, waltz into the men's room. They obviously won't mind. I don't see men minding. It's sometimes means a longer wait for a stall, but it's fair.
So what about it, ladies? Would you prefer this arrangement?
Submitted by brad on Mon, 2004-02-02 07:39.
In my home I now have a "home computer", for a while in the kitchen, now in the living room. Of course I have had computers in my home since the 70s, but this one is different. It's an old cheap laptop I picked up, not powerful at all. What's different is how I use it.
I have connected a Visioneer sheetfed scanner to it. I can drop papers and business cards into it and it scans them. Then I drop them in a box. I have scans of all the receipets and other documents, but if for some reason I need the original I can see by date which ones were nearby and presumably find it quickly enough. A good idea might be to drop a coloured sheet in once a month.
This has inspired me to design a simple device which would be very cheap to build. It's a small sheetfed scanner like this one, which is the size of a 3-hole punch. It's battery powered so it can be stuck anywhere. It has no cables going into it, instead it has a memory card slot, such as compact flash.
When you scan, the data would just be written to the flash card. (Nicely this means the scans are fast.) A button or two on the scanner might set some basic options (like colour or gray, delete and rescan etc.) At most a small LCD panel would be all you get.
When the flash is full, you would take it to the computer, which would do all the real work. Scan and process the data. Convert grayscales to bitmaps at the right level as desired. OCR the text for searching and indexing. Detect orientation, tilt, business cards (by size) etc. all automatically.
And of course then let you view and organize your papers on the computer.
We've dreamed of a paperless office for some time but never gotten there (though we get a little closer all the time.) But can we get to a paperless home, or at least a lower-paper home? I find the paperwork and nitty-gritty of managing a home gets more frustrating with time and hope something can help it.
As noted, my design is extremely cheap. The scanner, a small controller and a flash interface is about all there is too it. Cheaper than the current scanners, which can be had well under $100. The flash card is the expensive part. read more »
Submitted by brad on Fri, 2004-01-30 17:18.
In general, I agree with the recommendations several security experts wrote condemning the new overseas military voting system SERVE, because voters used unsecure Windows PCs to vote.
However, in thinking over the matter, I suggest the following method and open it for criticism. It still has many of the flaws in such systems - no physical audit trail, and like all remote voting systems including mail-in absentee ballot, it allows non-secret ballot and vote buying, though it is not much worse than mail-in in that respect.
Here's the proposal. For each registered voter, generate a paper instruction book. In the book, list the choices they can vote for, and with each choice provide a multi-digit number to enter. Also provide a longer master number for the whole ballot. In addition, after each number, provide a second "ack" number.
Thus you might see a ballot with:
- George Bush: 8741 / 9832
- Al Gore: 9843 / 4382
- Ralph Nader: 0438 / 2833
The numbers are different on each ballot. The voter enters the master number and then the sub-numbers. The election server, combining the numbers can determine who the vote is for. Only the exact numbers will work (any other will generate an error, and only so many errors will be allowed.) It should not be possible for a program not knowing a secret known only to the master computer to map the numbers to a choice.
When the vote is cast, the master server responds with the ACK number, which again only it knows how to generate. The voter confirms the ACK number is correct. The voter -- if they trust the master voting web server -- can be assured that their vote was registered, as desired with the master voting web server.
There's nothing a man-in-the-middle, including a trojan program that has taken over the PC, can do to circumvent this. They can't change the vote, see who the vote was for, or stop the vote from being recorded without the secrets known only to the master vote computer.
And thus it should work from any unsecure web browser and in fact would work fine from a telephone. As long as the numbers are long enough to avoid any guessing attacks.
Though again, we are completely trusting the master web server and its security.
Vote buying is easy with all mail-in ballots. Just ask the bought voter to give you the ballot to mail (or to fill in) and you can check it first. It's also easy to do here. It is slightly easier because you can provide software to confirm it but it's really not a lot easier.
To the system, voting can still be anonymous, as there is no need to connect a registered voter with a particular ballot card. Let them, once confirmed, pull a random ballot card from the pile, or mail them one. Of course the ballot cards with the magic numbers must remain secure, as must all mail-in ballots.
Anybody find a window into this system?
Submitted by brad on Fri, 2004-01-30 09:01.
Ok, so the USA invaded Iraq with fear of WMD the announced reason for the extraordinary step of a pre-emptive war. The White House doesn't plan to apologize or say oops as evidence mounts the reasons were bogus.
Question: Would it be a wise move for the Democrats, for example, to issue their own apology. To say, "We voted for this war because we were told there was a serious threat. Learning there was not, we are deeply saddened, sorry and ashamed. How could we have done this?"
Saying this would ask the other side of the aisle, "Why aren't you ashamed, too?"
However, saying you're sorry or ashamed is, for better or worse, un-american. It's weak, too many feel, to admit remorse for errors. You don't do it unless you're forced into it.
You can still feel removing Saddam was the right thing to do. And you can still blame the bad intelligence you were given. But if I make a grave error, even if I make it in good faith because I trusted the wrong people, I still feel remorse over the error. Can the USA? Can some of its politicians?
Back in 1984 I participated in the World Debating Championships. (My team just missed the quarterfinals, alas.) During the finals, the proposition being debated was "The USA should apologize for the American Revolution." One of the most amusing "pro" arguments was that the USA should indeed apologize, and just not mean it. That this was the true American way.
Submitted by brad on Fri, 2004-01-30 06:12.
Perhaps I am too cynical, and after this you're going to think I hate MoveOn, but I'm wondering if the publicity over CBS's refusal to air their anti-Bush ad in the Superbowl isn't the result of very clever strategizing.
CBS claims they don't air controversial ads like election ads during the SuperBowl and never has. As far as I know, that's true. Not that I'm approving of the policy, I think they should take almost any ad that pays the fee.
However, the cyncial part of me wonders, did MoveOn know this all along? Did they try to air the ad during the Superbowl knowing they would get turned down, and that getting turned down would provide immense amounts of publicity for the ad campaign?
Perhaps not, but being turned down has definitely been the best thing that could have happened to the ad, which as I pointed out in an earlier post, was not the best pick from their contest. And even so, I hope CBS relents though I doubt it. (They only relent when airing a show about the Reagans.)
Submitted by brad on Thu, 2004-01-29 11:36.
I recently read it costs about $150,000 to put in traffic lights here in California. That's a heavy duty pole, the lights, the power, the connection to other lights and the vehicle sensors. Seems to me modern technology should be able to make that a lot cheaper. New lights are all LED (the energy savings pay for themselves quickly) but that should also mean smaller safer power lines with special digging under the road. (Wonder if they could do it with light pipes and have no electricity up there.)
However, as it turns out, we want electricity up there. It seems there would be a natural marraige between traffic lights and a wireless data network. Years ago I tried to suggest to Metricom that they buy a traffic light control software package and offer free traffic light computerization to any city.
That's because you could then mount your wireless nodes on the lights. You get everything you want -- you are at the major intersections (by definition, almost) and there is power there. A tiny amount of your bandwidth can control the lights from a computer. You're up reasonably high. Metricom used streetlights for their power and altitude.
And while I am loathe to suggest putting cameras at the intersections, ones watching the cars and not the people don't bother me nearly as much and are already there. These could allow human beings to direct traffic at important intersections during rush hour and unusual traffic flow.
A temporary light with wireless camera and control could also do this for special locations that need traffic control on weekends. Police forces can't afford to send an officer out, even though the time saved by the drivers would be huge, and besides it's risky work. However, remote tele-operators could easily do the job and save lots of time and bad traffic.
Submitted by brad on Wed, 2004-01-28 11:13.
Today, the U.S. Agriculture secretary issued a call that nations not ban beef imports from a country (like the USA) just because a single cow has been found with Mad Cow disease.
In other words, exactly what the USA did to Canadian beef when a single cow was found in Alberta with the disease.
And let's not get started on the Weapons of Mass Destruction question.
Submitted by brad on Tue, 2004-01-27 16:02.
For the second year in a row, I'm having a "Superbowl Commercials" party featuring my Tivo. Since the Tivo lets you watch a program while it's being recorded, you can watch the football at 20x speed and slow down for the commercials, and be finished when the game finishes.
To start and give the Tivo a chance to buffer up some game, we head for the hills in a renunciation of the couch potato for a hike. Then we return for the commercials and party.
Of course these commercials are hugely expensive and thus pretty good, and much more fun to watch with a crowd of cynical friends and perhaps some alcohol.
Have you got a Tivo or similar device? Hold a party of your own. I was asked, tongue-in-cheek, whether skipping the football part was unamerican. What could be more american than watching commercials?
Update: Party was a grand success, big crowd, all had a good time. After 2 years this is already a tradition. Tivo, bless their Orwellian hearts, tracks user clickstreams and reported in their press release that the Janet Jackson nipple was the most replayed event in their history. And yes, the folks at my party also wanted a slo-mo rewind.
Update: A new superbowl commercials party is planned for 2005. This time, it will be in HDTV, using the open source MythTV software. I'll blog a more detailed entry about it after the party.
Submitted by brad on Tue, 2004-01-27 06:49.
I often talk about Challenge Response spam filters because I wrote the first one. One complaint people make is that the filters will challenge even forged mail, causing a challenge to be sent to the forgery victim. While this is not a DOS attack window as some people believe (since you can as easily DOS the target directly as get others to do it for you) it does need more consideration.
However, there are some autoresponders who have no excuse in this, and it is them I am railing on today. With the latest worm program, I am getting "bounces" back from anti-viral mail filters which tell me, "The mail you sent contains a virus and was not delivered."
Of course I didn't send the mail, my address was forged. What bothers me is that the anti-virus program clearly knows there is a virus, and presumably then should know it is the sort of virus which puts in a fake address.
So why it feels the need to send an error to the address it knows is fake, I don't know. The bounces I can tolerate, the bouncing software has no way to know it was a virus, but the anti-virus software has no excuse.
Addon: I'm going to promote a note from the comments because naive me didn't think of it. The virus companies may be happy to send this "your virus was bounced" mail to the wrong address because it's an ad for their anti-virus service.
Submitted by brad on Tue, 2004-01-27 06:33.
I seem to be thinking a lot about the future of automatic cars these days. Already we're seeing cars in Japan that can park themselves in a tight parallel parking spot, and this leads me to think that the next market for the technology, after the basic automatic highway, won't be the city street but the parking lot.
Parking lots eat a lot of space, and where the land is expensive, automatic cars will offer automatic valet parking. Drive to the mall/office/whatever, enter the automatic lane and be whisked to the door. Get out and your car will run off and park itself efficiently, possibly some distance from the building. (In the future with more fully automatic cars trusted on city streets, it might rent itself out as an autotaxi.)
When ready to leave, use your cell phone to tell the car to come to the nearest door, and it will be waiting there. Obviously that's a great convenience, but the real reason this will happen is it saves a bundle for the building/parking lot, because they can park more cars in the same space, or even park cars offsite. Whatever cost is needed to bury guide-wires or other transponders is easily justified by the efficiency gain, especially in downtown multi-story lots, many of which already justify the cost of humans to do the work.
Later, however, I will reveal the big catch that may keep us from this.
Submitted by brad on Mon, 2004-01-26 10:23.
I went around and counted that we seem to have around 30 birick and wall-wart DC power supplies plugged in around the house, and many more that are not plugged in which charge or power various devices. More and more of what we buy is getting to be more efficient and lower power, which is good.
But it's time for standardization in DC power and battery charging. In fact, I would like to move to a world where DC devices don't come with a power supply by default, because you are expected to be able to power them at one of the standard voltage/current settings.
One early experiment is on airplanes. I have an adapter that takes the 12v from the airplane, and has many tips which put out different voltages for different laptops. These are expensive right now, but on the right track.
Our other early venture is USB, which provides up to 500ma at 5 volts. Many small devices now use USB for power if that's all they need. There are devices that plug in to USB only for power, they don't use the data lines. Some come with a small cigarette lighter plug that has a USB socket on it for car use. This includes cellular chargers, lights etc.
I think a good goal would be a standardized data+power bus with a small number of standard plugs. One would be very tiny for small devices and only provide minimal USB-level power, a couple of watts. Another would handle mid-level devices, up to a couple of amps. A third would be large and handle heavy duty devices up to say 20 amps, replacing our wall plugs eventually. There might be a 4th for industrial use.
In full form, the data bus would be used for the components to exchange just what power they want and have. Years ago that would have been ridiculous overkill, today such parts are cheap. However, to make it simple there would be a basic passive system -- perhaps as simple as a finely tuned resistor in place of the data components -- to make it easy and cheap to adapt today's components.
A fully smart component would plug into the smart power and get a small "carrier" voltage designed to run the power electronics only. A protocol would establish what power the supply can provide and what the component wants, and then that power would be provided. read more »
Submitted by brad on Wed, 2004-01-21 14:25.
Another transportion item, because last night the train I was on hit a car stalled on the tracks (the occupant is OK, though was hit by the car when the train bashed it.)
Since trains do hit things, why aren't solutions to this more common in our data network world? A laser detector over the grade crossings would be simple enough.
At dinner, my friend Kurth Reynolds made a suggestion that I have improved. How about a small robot, equipped with camera and other sensors, which travels far enough in front of the train that if it sees a problem on the track, can send a signal back to the train in time to stop it. Trains take a while to stop, which is one of the reasons they can't do anything when they see a car or person ahead on the tracks.
You can't be too far ahead or you enter the "space" of the earlier train on the track, though during any tight conflicts you can of course give up this "foresight" and bear through (or slow down.)
If you have a human driving the train, you can show them video of what's ahead of the robot and give them time for a decision. Some decisions (Robot hits something or derails) would be automatic. Of course the robot might hit the car stalled on the tracks (though it can stop much, much faster than a train) but do far less damage.
The robot would be tall enough to go over the suicides who are "sleeping" on the track, but light enough so a car hit by it would survive.
Simpler for shorter runs like commuter trains would just be cameras along the track beaming to the oncoming trains. The engineer could be seeing a mile ahead at all times. Hey, if x10 can sell 2 broadcasting video cameras for $80 (WARNING: Don't buy from their web site, you will be spammed to death) I bet this can be made affordable.
This is important because some people don't think we should have rail with grade crossings. Without grade crossings, rail becomes vastly more expensive.
Some updates five years later: Some have worried the robot could hit workers or cars. Today, we are more comfortable we can build robots which would use LIDAR and never hit anything that wasn't running onto the track. The robots would also be light and perhaps have airbags to soften the blow against something rushing onto the track. When coming to a grade crossing, the robot would actually stop at the crossing and wait for the guard to come down (for the train, if the path is clear) and continue to monitor the crossing and report if something stops in it. Then it would speed up again and start going down the track to assure it is far enough ahead of the train.