March 21st, 2015

The Flow of Used Books Is Now Smoothly Input to eBooks

Nate over at The Digital Reader/Ink, Bits & Pixels points out that Better World Books will sell you some of their used pbooks as individually scanned ebooks instead. Think of it as 1DollarScan tacked on top of BWB.

Japanese P-to-E book services pre-dated 1DollarScan, and they had less formally connected AmazonDirect to their service back in 2012.

Nakano of 1DollarScan said: "One interesting option we offer in both countries is Amazon Direct. A customer might buy a book on Amazon that isn’t available in an electronic format, and have it delivered directly to us. We scan it and then the customer receives the e-book. This is a good service since many books are not available as e-books, especially older books. The usage for Amazon Direct is really high."

For those with copyright questions, Nakano said earlier in that piece: "We also have a copyright management system on our website for authors and publishers. If publishers do not want us to scan their content, they can register with us. We are open and cooperative, and so far nobody has registered."

Debt Sausage Factory: Student Loan Edition

The Upshot at NYT has this article saying, how come we can't get detailed information about student debt?

It has a variety of reasons for wanting the detailed information, assertions that privacy can be preserved, it impugns the competency of the Education department, argues we need this data to "help" borrowers before they get into trouble (ha ha ha ha ha -- the way to help most student debt holders who get into trouble would have been _before_ they borrowed that money from that lender to go to that school. Problem is not fixable thereafter). At no point does it get anywhere near suggesting that perhaps the data do not actually exist. However, it has quotes like this:

"David A. Bergeron spent 34 years at the Education Department, ending up as an assistant secretary for postsecondary education. He left in 2013 and is now vice president for postsecondary education at the Center for American Progress, a Washington research organization. Mr. Bergeron says he is frustrated by the incomplete data. “The department releases only very high-level summary statistics, which do not allow us to understand the problems facing individual student borrowers,” he said. “Trying to solve a problem with incomplete information leads you to the wrong solutions.”"

Have we truly forgotten the recent structured debt debacle? Let me tell you my subbasement in the Rainier Bank Building story once again, because I like telling that story and I Am Old so listen up, pups, ya might learn something.

Once upon a time, in the world of paper (gasp! The Horror), every mortgage had a file and that file had to have certain documents meeting certain criteria in that file. When the mortgage moved from the originator (in this case, Rainier Bank) to Someone Else (I have No Fucking Clue), Every Last Freakin' File had to be checked, because That Is Due Diligence. My part of the story involved getting to wear shorts while working in a bank, because (a) there was no AC in these rooms and (b) it was a subbasement, so no customers. I had to painstakingly go through thousands of files that were about to be transferred to make sure they were all right and tight. Did I find errors? Yes! I did. My favorite errors involved neighboring houses in southern California owned by a half dozen or more names. This really did happen more than once, and I assume these were some kind of family business or family compound or extended family occupying two neighboring houses or whatever, but their files were predisposed to confusion and were incredibly difficult to fully disentangle because of similar addresses and similar names.

Or I should say, that _was_ due diligence.

Then, in the early world of electronic documentation, mortgages became entries in spreadsheets and the spreadsheets were transferred _without the files_. But, the law had not caught up with daily practice, so when this stuff landed in court, you had to track back through all the transfers to find the paper. A couple things happened. (1) A bunch of big operations (WaMu, I am looking at your corpse) went under and that made finding paperwork that may or may not have passed through them ... difficult to find and (2) I'm completely convinced that a whole lot of strip mall mortgage brokers just binned the original paperwork and is was never around to be found thereafter. There were efforts to recreate the paperwork, which were generally found to not actually be legal (that was the robosigning part of the scandal, IIRC).

I would like an explanation why this didn't also happen with student loans. Because for damn sure it did. The paperwork is not there to be found, so as long as your loan isn't actually being held by the Feds, you can probably slow-walk all collections attempts and demand the paperwork be produced. They'll never produce it so you never have to pay. (If the Feds have it, they might hold any taxes you were hoping to get back in a refund and do other things, so exercise due caution out there adult children!)

Voila. Problem. Solved. (Not really. But we can dream!)

I am not a lawyer or a financial advisor. This is not advice. This is political speech.

However, other people have been doing this for a while and it seems to work at least part of the time:

The usual debt collection rules apply: go get good advice from a lawyer who practices in this area before making _any_ payment or signing _anything_, especially if it goes to collection and ESPECIALLY if a debt that was removed gets reinstated. You lose a ton of legal rights to question the validity of the debt claim if you make a payment.

TRMS covers self-driving cars

Because Elon Musk did another announcement.

Rachel Maddow covered self-driving cars. Let's see how that went.

She talks about parking assist, the first models to have those over a decade ago. (Wikipedia mostly agrees.)

She starts by talking about the history of cruise control, giving pictures of an old AMC car one of the first "big American cars with" cruise control: 1965. She mentions adaptive cruise control, mentioning sensing cars ahead of you in the lane (that in the lane is Very Important, and when she says just pulled in front of you in the lane, she imputes better behavior to adaptive cruise control than I have experienced. If they just pulled in front of you, adaptive cruise control may not immediately notice it so you'd better be ready).

She mentions vibrating warnings that you are not staying in the lane. She talks about lane change/blind spot assist. She talks about forward and rear sensors to brake the car if getting too close to something else.

Then she talks about the Tesla AutoPilot on highways announcement.

The demo shown is clear: weather is good, there's no crap on the road, the lane markings are clear. This is all important. AutoPilot won't work if there isn't stuff to "see", either because weather snow, ice, crap) has compromised the sensors or because the road doesn't have the markings (faded or nonexistent lane markings, or lane markings covered up by leaves, snow, etc.). That's really exciting stuff (the video is over at The Verge).

But then she does this. Self-driving cars seemed so science fictional! "It is here. It's done. It's ready to go." Then she expresses surprise that Teslas are gonna get an over the air update that will give them AutoPilot (of course, people who have been buying Teslas for months now have been pissed that the feature was software disabled, so _they_ aren't surprised). She then goes down a legal tangent about cars having to have drivers (AutoPilot does not negate the need for a driver, so I don't see them immediately getting tangled up with the law. ETA: Slate on the legality issue for AutoPilot: Then there's a drone comparison. (This is a bad comparison.)

It gets really weird next. This is a HIGHWAY feature. Even Musk isn't crazy enough to deploy on roads with less maintained lane markings and similar support features. "But you know presumably the leap from your car being able to safely AutoPilot itself on the highway to your car being able to safely AutoPilot your kids to school and back, presumably that is just an incremental technological matter of upgrades."

Sure. It _is_ just an incremental technological matter of upgrades. Just like the iPhone in your pocket is just an incremental technological matter of upgrades from Babbage's Difference Engine, all munged up with something made by Marconi. "just an incremental technological matter of upgrades".


I paused, but this is the next sentence: "Upgrades that will unfold pretty fast once people are regularly knitting and doing their filing in the driver's seat while their car drives them to work. This is about to happen."

Discussion of, will it be normal or not, will it be required, will robot driving be safer, will people hate it and it will die.

And then a segue to media coverage of politics, retail meet-and-greet politics vs. delivery of info about candidates through broadcast, etc. Basically, Meerkat Will Change Politics. Which it very well may.

FWIW, she seemed to think that 1965 was quite a long time ago, so incremental upgrades in that time frame would not fit her idea of "it's right here".

ETA: The Slate article seems a little confused.

"if the software lets the Model S operate like a Level 3 car, letting the human “cede control of all safety-critical functions” to a machine that can, say, change lanes on its own, then it’s illegal. Cars with such capabilities, like the Audi A7 I piloted from Silicon Valley to Las Vegas, must be certified as test vehicles before they hit the pavement. You can’t sell them to the public."

Here's the definition of Level 3:

"Limited Self-Driving Automation (Level 3): Vehicles at this level of automation enable the driver to cede full control of all safety-critical functions under certain traffic or environmental conditions and in those conditions to rely heavily on the vehicle to monitor for changes in those conditions requiring transition back to driver control. The driver is expected to be available for occasional control, but with sufficiently comfortable transition time. The Google car is an example of limited self-driving automation."

There is no reason to believe AutoPilot will meet these criteria. Particularly not when every description and demo says that the driver is there with their hands ready to take control back at all times.

A _lot_ of online commenters have trouble grasping Level 3. They particularly fail to notice the "with sufficiently comfortable transition time". The Tesla Model is is no Google car.

[ETA: This is the doc referenced below: DOT HS 812 044 Human Factors Evaluation of Level 2 And Level 3 Automated Driving Concepts the pdf is named 812044_HF-Evaluation-Levels-2-3-Automated-Driving-Concepts-f-Operation.pdf hopefully that will enable you to find it I'm having trouble with the URL.]

[ ]

Elsewhere, NHTSA says an example of comfortable transition time is the car telling you about an upcoming construction zone.

Oooh, this would be cool!

"Automated Valet Parking
Audi20 has developed a prototype system that enables the driver to depart the vehicle at a parking garage entrance and to instruct the vehicle to “go park” via a smartphone interface. The vehicle then proceeds, empty, into the garage, travels until it detects an empty space, and parks. Upon being summoned by the owner via smartphone, the vehicle proceeds to the garage exit area where the driver re-enters the vehicle and resumes driving. Nissan21 has demonstrated a similar system."

NHTSA and I seem to agree that that project is the only Level 4 thing likely to wind up on the market in less than 5 years. I WANT IT!

Another illuminating paragraph about the difference between L2 and L3:

"In L2 vehicle systems, the operator subsystem, by definition, is expected to monitor the roadway and safe operation of the vehicle while remaining available to take full control at any time. Therefore, when the operator subsystem, potentially having full knowledge of this responsibility, chooses to engage in a secondary task that may disrupt or eliminate operator capability to effectively monitor the roadway and safe operation of the vehicle, this can qualify as or lead to misuse. Conversely, in an L3 system in which, by definition, the operator subsystem has sufficient time to transition from a separate task back to the driving task, the purposeful engagement of an alternative task is not necessarily misuse. However, in an L3 system, misuse could occur when the operator subsystem attempts to ignore warnings/alerts fully comprehended (including potential consequences for ignoring system communication) by the operator for reasons known to the operator."

As long as Tesla is clear that their systems require constant driver attention/readiness, they are not selling an L3 vehicle, and their car should be legal.

This whole document (which I haven't linked to yet) is reminding me of the RV joke where the man goes in the back, and then his wife joins him, no one is driving the RV any more because she misunderstood cruise control as an aviation like autopilot.

"Essentially, the operator may incorrectly assume the vehicle is still maintaining heading when, in fact, neither subsystem is actively steering the vehicle."

Chapter 8 has a great table showing where industry development is focused (so everyone in the Northeast who is hoping for cars that drive themselves? Fuhgeddaboudit. No one is interested in automating any element of car operation when the roadway is at all icy and/or it is snowing. Just no.).

Self-Driving Cars Need Smarter Roads: link fu

Obvs, lane keeping systems rely upon well marked lanes and nothing obstructing the markers.

But it could be a lot more complex than that, and we've done a lot of things to change infrastructure to reduce driving hazards, because while we _know_ the average driver isn't gonna change much, we might be able to change the road a lot (and we have! banked curves, divided highways, jersey barriers, etc.).

Let's start with deer!

In 2002, the Indiana Toll Road experimented with sensors to warn that deer were present.

By 2008, they were gone (and the start date on the system was apparently mis-remembered):

Dealing with deer is an ongoing issue, which you can research here:

None of it is working all that great. Volvo is/was working on an IR system for cars to spot deer at night:

A couple years ago, it was a couple years down the road:

In August 2014, they were still working on it, and their goals had been reduced a bit:

Now, it's about the very biggest animals, and about reducing speed just a bit, to reduce fatalities, rather than total avoidance.

If you've ever lived in a place like New Hampshire or Massachusetts which has a lot of deer running across roads in twilight, with woods on both side coming within yards of the roadway, you might be prepared to spend quite a lot to never, ever, ever hit a deer. But I think the technical problem is actually insanely difficult, because the real risk is when one of those suckers crossed the road _right in front of you_, giving you next to no time to stop. If they're standing in the middle of the road staring at you, not hitting them is no harder than not hitting anything else in the middle of the road.

This thing is a depressing list of roadway modifications tried with varying but generally expensive and not completely effective results:

RFID collars on deer in a herd, fences, crosswalks for deer, lighted signs for the drivers, laser detection of deer . . It's all here!

time to ipo for a cfo

I was reading this:

Basically because I've been thinking lately about how long companies are spending in VC These Years/Post Bust. And when I look at how much companies have raised, and the valuations implied by the more recent rounds at some companies, I can see why they aren't in any hurry to IPO. Altho that doesn't really explain why people bought into those later rounds. But whatever. You get excited. I get it.

Katie Benner says, "I’m told by lawyers and bankers who work on these deals that it typically takes a CFO a year or more to settle in, understand the books and figure out what needs to be done to get a company’s financial performance and accounting camera-ready for public markets."

Joy Covey started in December of 1996 (I still feel sad when I think about Joy). Amazon did its IPO in May of 1997. I know, I know, that's setting the bar unnaturally high, because Joy was very much the most amazing finance person I have ever encountered. Or probably ever _will_ encounter.

Clearly, my reference values for how long this stuff takes are not accurate.