Brad Templeton is Chairman Emeritus of the EFF
, Singularity U
computing chair, software architect and internet entrepreneur, robotic car strategist, futurist lecturer, photographer and Burning Man artist.
This is an "ideas" blog rather than a "cool thing I saw today" blog. Many of the items are not topical. If you like what you read, I recommend you also browse back in the archives, starting with the best of blog section. It also has various "topic" and "tag" sections (see menu on right) and some are sub blogs like Robocars, photography and Going Green. Try my home page for more info and contact data.
Submitted by brad on Mon, 2005-04-04 14:38.
Perhaps this is one of those ideas that some car has implemented and I haven't yet seen it. As many people know, in several years ago a number of cars arranged so that their interior lights would not go off immediately when you closed up the car. This gives you the ability to still see shortly after closing up the car and walking away.
Of course this also drives people nuts, because in many cases you can't tell if the lights stayed on because you didn't close a door properly, and you would end up waiting around to see if they would go off.
Some cars fixed this by having the light fade out, but that's still pretty slow and of course elminates the light you were hoping for.
I would suggest that cars develop some more overt signal, to be triggered immediately when the car has decided that all doors are closed and the car is off, and the lights will be going off in 20 seconds. Such as a quick blink pattern when you close the door, or a flash of the headlights, or a quiet sound or bright internal LED.
Seeing this blink pattern, you would be 100% confident the car is closed and you haven't left the lights on, and could walk away, lit for a few seconds like you want.
Submitted by brad on Tue, 2005-03-29 13:10.
Death Valley normally gets 1.5" of rain a year, but this year it got over six, so we headed down the greatest spring wildflower show in 50 years and were not disappointed.
My preliminary gallery of Death Valley Wildflower Photos is now up. Of course I also shot many panoramas but have not yet assembled them. (I've been barely using Windows of late so I need to get a box rebuilt.) I will announce when the panoramas are available.
Submitted by brad on Fri, 2005-03-25 07:10.
Here's a business idea for both mobile phone companies and people who operate those giant digital signs in public places (such as malls and the Times Square jumbotron.)
Let people text a message to the sign for a lucrative but affordable fee. It would then display ASAP, though possibly a human would have to check for "offensive" messages, whatever that means. You could see people putting up love notes to their valentines as they both go by the sign, rivals having battles and debates in their messages etc. Could be both entertaining and lucrative. The texted billboards (or from a web form with graphics) would contain a bar with the texting number or URL to enter your own. If it were cheap enough you might see crowds stopping to enjoy the battles on the jumbotron.
Submitted by brad on Tue, 2005-03-22 06:12.
For the past couple of years, I've been mulling over an idea for a different kind of DVD "rental" company, similar in ways to the popular NetFlix. Now I have encountered a new company called Peerflix which is doing something similar. Is it annoying or vindicating to see somebody else run with something? :-)
So instead I will comment on Peerflix, which I am going to try out, and what I planned to do differently.
The rough idea is a movie network that doesn't own the movies. The members do. The members declare what disks they have that are available to go out (key in or scan UPC codes or just put disks in drives) and, just like netflix, they also browse the list of DVDs and pick what they would like to rent. For each disk you have out, you are entitled to one in (approximately), and somebody close to you, who has the disk you want, is told to mail it to you.
Once scaled up, it's faster than netflix (the disk is mailed to you directly from the last person to have it, rather than going through the warehouse) but mainly it's vastly cheaper. In theory it could even run for free, with postage and mailers being the only cost -- plus of course the initial disks you introduce into the system. Netflix 3-at-a-time is $216/year, the one at a time is $120 per year.
There are, however, a number of interesting problems to solve in doing this, and some special factors you may not know about Netflix. read more »
Submitted by brad on Mon, 2005-03-21 07:26.
Here John Dunn suggests sending an AI to negotiate with any aliens we discover via SETI.
This raises an interesting question. If SETI worked, and we got a signal from an alien intelligence, and the signal was understood to be a description of a computer architecture and then a big long, and undecipherably complex computer program -- possibly an AI -- could we dare run it?
Oh, it would be so tempting to run it. Contact with an alien species, possible untold wealths of knowledge, solutions to all our problems and more. But if it can contain those things it's probably smarter than us. And as an alien, it has its own goals which are alien to ours.
AI pundit Eliezer Yudkowsky spends much of his time warning about the dangers of even a human-designed AI, and has developed a convincing argument that it's next to impossible to keep something much smarter than you locked up in a box no matter how much you resolve to do so. It's probable we couldn't keep the alien AI in a box either as it does a superhumanly good job of convincing us just what wonderful things it could do for humanity (or just the people with keys to the box) if released.
Indeed, a good strategy for a growth-oriented AI creature would be to broadcast itself out at lightspeed, in the hope that other creatures would run it, and it could then use their resources to build more computers on which to run itself and transmitters with which to transmit itself. It might even do that at the same time as providing wonderful benefits for the host culture, or of course it could toss them by the wayside as it saw fit.
Remind you of Pandora? In Contact by Carl Sagan, the aliens send plans for an FTL transporter, which presumably is a physical device with no AI, so they are able to build it. They debate building even that, worrying if it's a weapon, but the debate would be much more on an AI, and probably end up in the negative.
Submitted by brad on Sun, 2005-03-20 10:35.
I have looked at a lot of image management programs, though not all of them, and been surprised that none match what I think should be a very common workflow. Sure, they all let you browse your photos and thumbnails of them, move them around, and rename them. And some let you do the functions I describe but usually doing them to a lot of photos is cumbersome because they only have a slow mouse interface or a poor keyboard interface.
Here's what I want to do, and right now use a combination of programs to make happen.
- First, pick the "potential winners" from a set of photos. That means letting me with a single keystroke copy the selected photo or mark it for later copying to a directory of the best shots I will actually put on the web. Two keystrokes here is two many. This must be done from full-screen view, not from thumbnails or reduced views. You can only truly judge a winner in full screen view. Thus, in this view, we should have basic movement on keys (space for next photo, backspace for previous is common) and a keystroke to tag/copy and go to the next, or at least to tag/copy and then I will hit space for the next. A way to go back and undo it would be nice. xzgv almost does this.
- Then scan the winners again and remove the duplicates. Often you will have 2 or 3 good shots of a subject that all were potential winners. So now it's time to quickly delete (no confirmations here, these are just copies) the other candidates and leave the winner. Quick switch between full screen view and a multi-photo view is a plus here.
Because serious photographers take several shots of everything interesting, scanning for the winner often involves comparison with the other shots in the photo sequence. A perfect UI for this is hard, though a clever program could spot images bunched together in time or even (with advanced algorithms) similar in composition. A strip of thumbnails to get a sense of all the shots of an item while picking the one winner would be good. A quick switch to a tiled view of all the potential winners at maximum size, with a way to pick the winner (here mouse click makes sense) also could be good. This ability is of use not just in duplicate scanning but also initial winner picking. I tend to find that I will see an image, tag it as a winner, then move on to next image to notice the next one is even better. It would be nice to know in advance that might be so (thus the thumbnail strip.)
- Once I have the winners, put them into categories. Create a series of named directories, and quickly move the photos into them. Here's where a traditional thumbnail browswer which lets you select multiple photos and move them works well. Most programs do this step OK.
- Once I have the winners in categories, caption them. Again, it should be really fast. View photo (at least 1/4 screen size, not a thumbnail) and type in the caption. Then a single keystroke to go to next photo to caption it. Caption should go into jpeg caption, or a simple file that can be worked with later. ACDsee comes close to doing this but they use ugly keystrokes.
- Next, order them for presentation on a web page. Not necessarily by date or sequence number or caption.
- Finally, generate a web gallery or slide show based on the order and captions and sorting. Or, in my case leave available the data for my own scripts to do this.
Some programs as I note, come close. However often they use cumbersome keys (alt keys and ctrl-keys when regular letters would do) or they require confirmations on frequently performed acts (useless as you quickly learn to automatically confirm, just wasting your time and providing now protection.)
But does any system do all this, for linux or windows? Let me know.
Submitted by brad on Fri, 2005-03-18 11:10.
On a recent roadtrip, I did some “wardriving” where you scan for 802.11 (wifi) access points. Today they are everywhere. The scanning program lists the network name (SSID) as well as other information like the model of access point and whether it has encryption on. Often the SSIDs are informative, with the names of families and companies. Mine is an web address that would let a neighbour contact me.
All this happens because most access points transmit a regular “beacon” packet which lists their SSID and other information needed to connect to them. Seeing that the SSIDs were sometimes interesting, I wondered if we might do much more with a special beacon.
This beacon would deliberately tell you a bit about the access or location. It would contain a mixed XML/HTML packet with a variety of useful fields and general text. These could range from simple descriptions (“This access point belongs to Joe Smith, I’m a programmer”) to information (“On this site, Paul Revere stopped on his ride to consult with local minutemen”) to street directions (“Turn right to get to highway 101, left for downtown”) to, of course, advertising (“We sell fresh fruit and have a special on plums today.”)
In other words, a replacement for signs and billboards and markers. And perhaps much more. Access points would also talk about themselves, declaring, for example, if the owner is offering open internet access for free or for fee, or has a local database of information, and what classes of information are in the main text. The local lattitude and longitude for those without a GPS could be useful, along with local map data in a compact form.
Users could quickly get a program for their laptop (such as Netstumbler) to read and display such virtual annotations to the world as they drive. Primarily for passengers to use, of course. Eventually dedicated boxes would become available, and onboard car computers and GPS units could understand the protocol. Mass market access points would include a set-up screen in their web interface to let the owner enter the information beacon text and enable it. (Today some APs have open source firmware and an energetic programmer could do this right away.)
All of this might be both useful and entertaining. Children might enjoy reading all the random bits of information that flow by and stop asking “are we there yet?” The journey can become the reward. (Of course remember to look out the window sometimes.)
I can imagine vendors making a cheap solar powered access point that, during the day at least, sends out information beacons as soon as enough power is stored in the capacitors to send one. These could operate on a small, cheap solar cell (the more power, the more frequent the beacon) and be placed anywhere. “I’m an oak tree!”
Below, I will get into some technical issues and discuss the unanswered question, which is how to avoid abuse by excessive advertisement, spam and falsehoods. read more »
Submitted by brad on Thu, 2005-03-10 18:36.
Update: Well, clearly this was already being done when I asked for it, just not at the airports I flew from. It's now close to universal.
Airport pickup is becoming another nightmare in some cities, with police barring cars from waiting for passengers, causing people to circle.
Airports should take a piece of parking lot and turn it into a marshalling area for people with cell phones. If you are picking somebody up, and you have a cell phone, you go to the marshalling area, where cars wait in parked lines like a parking lot. When you get a call from your passenger saying they have bags and are ready to go to the curb, then you go out and get into a special passenger pickup lane. You go right to the numbered spot your passenger told you. For passengers without cell phones, phone booths will sit at the exits of course.
For those not with cell phones, there is of course short term parking or the endless circling, though generally you want less of that.
Submitted by brad on Sat, 2005-02-26 03:16.
Ok, I don't publish too many of my telecom ideas here since I am working on revolutionizing the phone call for my next business, but here's a simple one.
If you have a large carrier voice mail, such as the voice mail for a wireless company, you should notice if I call somebody and they are not simply busy, but in the act of leaving a voice mail for me. If so, you should break into their voice mail dialog and connect us.
Right now a common occurance is this. Somebody calls, you can't quite get to your phone, so it stops ringing and goes to your voice mail. But there on the phone is the caller-ID. You call them back, and you get their voice mail right away. Why? Because they are leaving you the voice mail from the failed call.
Similar thing when a call disconnects. Who has the duty of calling back. Often we both call back and we both get voice mails or busy signals. While one could develop an ethos (original caller calls back) we've never managed, so the tech could help us.
Easiest when we are both on the same carrier, but in fact this could be done in any tightly integrated carrier voice mail, though you do have to be careful about PBX lines in this case because you might be calling from or to a shared trunk.
Submitted by brad on Wed, 2005-02-23 06:23.
Ok, it's strange because I think one of the whole points of the hard disk video recorder / PVR is that you are not supposed to watch live TV any more, not supposed to channel surf -- but I keep coming up with ideas relating to it. Maybe I have a secret desire to surf again.
As many people know, with digital recording, the no-surf rule is enforced because it's harder to do. The digital delay introduces a long channel change delay, intolerable when combined with another delay (satellite/cable box).
Here's a surfing algorithm that could give instant channel change. Surf slightly delayed TV. It works particularly well if the box has multiple tuners.
Using the spare tuner, grab short snippets of every other show that's on right now, or at least everything in the surf list/favourite channels. Just a few seconds of each. As available, update these snippets, with focus on the adjacent channels. If the program changes, you need to grab a new snippet as you must always have the current program.
When the viewer wants to surf, surf not the current live TV but the saved snippets, which will be anwhere from a few seconds to a few minutes old. You will be able to move through them instantly, like the old instant-channel change from an analog TV. You will see the program guide info as well, but the visual clues that we draw upon in surfing will still work fine.
If the user dwells on a channel for longer than the usual surf interval, you will switch the tuner to that channel. You will need to do a nice graphic transition from the surf buffer to the live TV.
Now admittedly, that will sometimes frustrate. It may be the particular scene that attracts your interest -- the bad guy is holding a gun on the good guy, about to shoot, and suddenly it disolves to something a minute later. However, the alternative, which is what we currently get, is that you get black screen for 3 seconds, and then it shows you the later (live) scene. Instead of black screen you get some sample video from an earlier time in the show. The key thing is that the viewer should be aware they are surfing old snippets.
One could also keep snippets of varying lengths from different times, depending on the surfing speed desired. Though usually you would play the longest. You could also develop "smart snippets" which tried to grab the action after coming back from the first commercial break etc. (Problem is those happen on a lot of stations at once.) read more »
Submitted by brad on Mon, 2005-02-21 12:34.
In writing an essay I'm working on about why hard disk video recorders are as novel as they are, I explored a concept I think is worthy of its own blog entry. This is the concept of Telepathic User Interface or TUI.
A TUI is a user interface that you use so much that it becomes unconscious. Perhaps the classic TUI is touch typewriter keyboard. I just think letters and they simply come out. I am no longer consious of the mechanism. In many cases I think sets of letters and even words and they just come out. From the mind to the computer -- telepathic.
Other examples include the car. After you drive a car for a while it becomes an extension of yourself. Learning the clutch is hard but soon you are not thinking about it at all. And the remote control on a Tivo, I write in the essay, has aspects of a TUI -- you learn how to move around a program without thinking.
A TUI is not always a natural interface or even a good interface. It's just one you use often enough to make it subconsious. It doesn't have to be intuitive -- an intuitive interface is simply one that's easy to guess the operation of.
When it comes to computer software, this helps us understand the dichotomy between the GUI/WIMP style and the command line and keyboard style which still has many devotees.
GUI interfaces are easy to learn, and easy to guess. And of course for positional inputs they are markedly superior and often the only choice. But by and large, the story of Mice and Menus took a path away from the TUI. You have to focus your eyes on the pointer in order to use a GUI, and you have to read to use a menu. It's much more difficult to use such a system unconsiously. (Mouse gesture interfaces change that a bit.)
Fans of text editors like VI and Emacs, with complex, non-intuitive keyboard interfaces love them because they have reached TUI state, at least in part. Many of the operations have become unconsious, and thus much faster and easier as far as the user is concerned.
Command line interfaces are never completely TUI, but they take advantage of the TUI nature of touch-typing. Because touch typing maps words from brain to screen, complex commands can have a fair bit of TUI to them.
It is a rare technology that can earn a TUI. You need to be using it a great deal, and regularly. Video games also develop TUIs because of their devotion. And while it doesn't seem to matter how intuitive the interface is, since many users will never attain the TUI state with a program, that's no excuse for trying to be more intuitive and easy to handle.
On the other hand, programs that don't provide keyboard shortcuts and other muscle-memory schemes for doing things will never develop a TUI, no matter how heavily used they are. Who changes a font in the Excel spreadsheet without being conscious of all the steps they are taking?
Submitted by brad on Thu, 2005-02-17 07:24.
You'll recall an earlier post about the Silicon Valley 100 and getting stuff for free. I promised I had something to say about toilets anyway, so I will describe my experience with the Swash I was given as well as the Daelim Cleanlet which I bought a few years ago.
If you've gone to Japan, you have probably seen these fancy high-tech toilet seats, which try for a bit of bidet function in a seat. Their prime function is to have a heated water reservoir and a little wand that comes out to squirt water at what the Daelim manual calls the "personal area" and the "feminine area." They also tend to heat the seat, and make it descend slowly so it doesn't make a noise when you put it down. Both of these also have the optional feature of fan to blow heated air to dry your personal or feminine area.
I've got these two units, and I have tried various others in Japan. None of them can really compete with the water flow and cleaning ability of a real bidet, but most people don't have the space in their bathrooms for one of those. I was going to suggest the slogan "Every asshole needs one" but I don't think they are likely to use it.
These bidet-seats are about the only high-tech toilet invention to get a decent market, which is surprising because if you ask the patent office, toilet inventions are among the most common patent applications. I guess people spend a lot of time on toilets with nothing else to think about. read more »
Submitted by brad on Wed, 2005-02-16 11:35.
Earlier, I wrote some proposals for improving ebay style feedback, including not having feedback revealed until both have left it. That has some flaws, but the main reason eBay is unlikely to do this is that eBay likes feedback to be positive, they want to convince buyers it is safe to shop there.
So here’s an alternate idea to prevent revenge feedback. Revenge feedback is only vaguely in eBay’s interests, in that the fear of it keeps feedback positive, but the existence of it adds to the negatives.
To solve this, attempt to detect revenge feedback and print statistics on it. What would be detected is negative feedback left by a seller on a buyer after the buyer has left negative feedback, but not if the buyer left this feedback immediately.
In theory the buyer has just one duty — to pay promptly. Indeed, since eBay owns PayPal they could also just report about buyers whether they paid promptly with PP and that should be all you need to know. Sellers might want to tag a “troublesome buyer” who has a lot of complaints after getting the item but I think that’s in an entirely different class of feedback anyway.
So really, a seller should leave feedback once the buyer has paid, and negative feedback only if the buyer pays slowly, pays falsely or doesn’t contact the seller.
Under my system above, if the seller waits to give feedback, in particular waits until after the buyer gives feedback, she’s taking a risk that her own negative feedback will get counted in the revenge count. And a high revenge count will scare away deals, deservedly.
More simply, the system could also just count how often the feedback came in the expected order (Seller’s first, then Buyer’s) and how often the other way around. This would strongly encourage sellers to feedback first. You would see when bidding that a seller always or rarely feedbacks on payment, and again, stay away from those who don’t.
Now admittedly, with the fear of revenge feedback gone, buyers would be more honest, and reputations would drop a bit. eBay might still want to avoid this, but with luck it would not be a big change.
Updated thoughts: It may be time for a 3rd party company to begin offering more detailed reputation information. Since eBay has stopped robots it doesn’t like, this would have to be on-client software which extracts results of transactions from eBay to another database that a browser add-on (like ShortShip) can display. All the useful information could be stored — feedback order, possible revenges, feedback based on dollar volume etc. Counting no-feedback transactions is harder and probably requires a blockable spidering operation or some complex shared network. To this one could add more feedback done outside of ebay, including revenge claims and full text stories that eBay doesn’t allow in feedback comments.
Useful hint: eBay doesn’t allow URLs in feedback, but if you invent a random string you can put ‘Search for randomstring’ in the feedback comment, and make a web page with that string in it that Google and the rest will find. Then people wanting to know more than 80 characters can learn it. Of course, the other party can also make a web page with that string so searchers see both sides, which is fine. A good non-random string might be something like eBay followed by the item number, as in ‘eBay130064299000’ — in fact, if such a method became common you could search for it without even needing it in the feedback.
Submitted by brad on Tue, 2005-02-15 07:31.
I notice from this chart in advertising age (which requires an annoying complex if free registration form) that there's been a giant jump in the CPM (cost per thousand viewers) of a 30 second Superbowl ad. From 1968 to 1998 it hovered close to $10 in constant dollars -- or about a penny per impression.
Then there's a big jump (thanks in part to Fox and the dot-com boom) and now it's up to $25. But they're still paying, even though the $10 figure is more common for regular TV.
As I have noted elsewhere, part of the revolt against TV advertising is that it's a terrible deal for the viewer. They get a penny (2.5 cents in the superbowl) to show you an ad. So if you watch a full hour of ads (ie. almost 4 hours of TV) you get in return $1.20 worth of programming (wholesale price.) $1.20/hour is an illegally low wage. Even with superbowl ads it's still below minimum wage but we watch them because they are more entertaining.
This increase I think comes from several factors. First of all these ads have a cachet -- the fact that you advertised in the superbowl expands your reputation more than an ordinary ad does.
But mostly, because these ads have a reputation for being above average and less repetitive, many go out of their way to see them rather than go out of their way to avoid them. And that certainly seems to be worth the premium. It is another pointer to the prediction I made earlier this week that we might eventually see TV shows that demand better quality from their advertisers.
That pays off because if all the ads on a show are filtered for how useful or entertaining they are, it improves the odds the viewer will watch all of them. Google surprised everybody by demanding that its adwords customers do good ads that get clicked on. The idea of a publisher refusing an advertisers money was shocking (though it makes perfect sense here, because they are charging per click, and if you aren't getting clicks, they aren't getting good revenue from their space -- effectively enforcing a CPM while calling it pay per click.)
I keep waiting for adcritic to return using bittorrent.
Submitted by brad on Mon, 2005-02-14 19:16.
I've disabled the trackback feature that allows other blogs to tell this blog they have pointed to it. It's being abused by spammers. Unlike comments, which can have a little turing test on them, trackback needs to be automatic to work, so it's a lot harder to protect against spam except through complex blacklists etc.
I'll see people who link to me in my weekly referer logs. I wrote a nice script that keeps track of all referers, and figures out which ones are new, and which ones suddenly brought in a lot of traffic, and I have it generating a summary once a week. I should release it some day...
Submitted by brad on Sun, 2005-02-13 08:33.
Many people have old, low resolution digital cameras lying around from the previous generation. Here's a good use to put them to, particularly if you have a housekeeper.
When somebody needs to put something away, and they don't know where you like it to go and have to figure out where, have them pull out the old digital camera and take a picture of the item and a picture of where it was put.
Then every so often you can pull out the images into an online folder, ideally with a thumbnail browser.
Even though the camera is probably low res, like a megapixel, you probably want to set it to an even lower res to get a ton of photos on it.
As your memory fades in later life, you might even find this handy to do for yourself. You could just organize your stuff so everything is in an obvious place, or just take photos of all your things where they are. Good for insurance purposes, too. It won't work for the car keys or glasses, but it might work for some stuff.
Submitted by brad on Wed, 2005-02-09 08:13.
For the third year in a row, I held a "superbowl commercials" party. In this case, we used MythTV to record the game in HDTV for a while, then sat down to fast forward through the football and slow down to watch the commercials.
I've written a page about the Superbowl Commercials Party here, including fun stories from the party and a bit about the MythTV software that did the recording and playback on two different floors in HDTV.
Also included is a discussion of how the equipment I used will become illegal in just a few months, and what we're doing to try and stop that. It's an interesting lesson in how regulation can stop innovation and what's really happening in the future of TV.
Submitted by brad on Fri, 2005-02-04 11:30.
Here's a simple though not too exciting idea. Make bells for cat collars in different pitches. Thus you can always know which cat is coming just by sound.
Submitted by brad on Sat, 2005-01-29 14:46.
One thing I've noticed when you get a TV broadcast that has 5-channel sound, it that you get the voices on the center channel. Particularly with things like voice-overs, sportscasters etc. If you can mute the center channel, you can watch a game, for example, with no commentators.
But sometimes you do want to hear what they said, if there is something on the field you don't understand. If you have a PVR, you can rewind and turn on the center channel audio (though rarely is there a good UI to do this) but here's another idea.
If the center channel is muted, offer a button that shrinks the display and puts up the last minute or two of history of the closed captions. Of course if there are already bars on the screen use those to display this. You can read quickly (let one of the scrolling controls let you move around) and figure out what was going on, faster than you could through replaying.
Submitted by brad on Fri, 2005-01-28 10:31.
You may have run into the story of a fireman charged with burning down his own home. They charged him because his Safeway Club card records showed he had purchased the type of firestarter that was used in the arson on his house.
Sounds like a good case? Problem is somebody else confessed to the arson. He's now a free man.
People often wonder why privacy advocates get up in arms about things like the Safeway database. I mean, how can it harm you, especially if you're not doing anything suspicious?
The problem is that police are attracted to the evidence that is easy to find. But when databases become more and more comprehensive, the chance that they will contain something interesting grows.
In an old-time investigation, finding receipts for the firestarters would be a major clue, and mind convict somebody. That's because searches of what you bought weren't so easy. If you bought the very tool used in the crime, and it was prominent enough that they found it, it looked bad for you.
But the cops aren't aware they are falling into one of the traps of bad science. When you have a lot of data, you can always find something that matches what you are looking for. When you find it, your intuition tells you "this is too strange to be coincidence." But in fact math tells us that it is. That's why you must never start with the conclusion and dig around in a big pool of data looking for evidence of your conclusion. Good scientists have known not to do this for years. Cops haven't.