True. Removing the fog is very expensive. It can be done, though. Watch Fog, Intensive, Dispersal Of (FIDO)

Everyone who deals with financial statements professionally does that.

AI as a field suffers from the delusion that we're one breakthrough away from strong AI. There were people saying that at Stanford in the mid-1980s when I was there, just as the "expert system" hype was failing. There is progress; the current generation of machine learning can do some things quite well. But it's not leading to strong AI Real Soon Now. We can't even do dog-level AI. Or mouse-level AI. Insect-level AI, yes. (It was bacteria-level AI in the 1980s. There is progress.)

More likely, we'll get robots that can sort of deal with the real world, and they'll be improved over time. (It's embarrassing how lame robotics really is, after 50 years of R&D. DARPA is trying to kick some ass with the DARPA Humanoid Challenge to get machines that can do something useful other than work an assembly line.) We'll get programs which can deal with most business problems ("Microsoft Middle Manager 3.0"), and they'll be improved over time.

Hardware is not the problem. If it were, we'd have things that were very smart, but very slow. Then someone would rent enough Amazon AWS instances to make them fast.

PC boards aren't that expensive. What's the point?

I'd rather have fewer connectors. Fewer points of failure.

The SEC started requring companies to file their earnings reports in the Extensible Business Reporting Language a few years ago. At first, it was only for big companies; now it's everybody. The SEC displays this info in a standard format on line. Here are the latest earnings for DICE Holdings, Slashdot's parent. Here's the raw XML behind that data. Turning that into verbiage isn't that hard.

I've been doing this for years at Downside.com, extracting the raw data from the human-readable text. This is now obsolete, but it's still running. Here's the same DICE financial statement as processed by Downside. That's Perl code that's been running for 15 years now. When it started, nobody was doing that. Now that everybody in finance has that data, it's probably time to retire Downside's old extraction engine.

It's another materials article. They do not have a "walking robot". They have one piece of synthetic muscle fiber hooked up to some supports. If they hook up an oscillator to the power, it jerks along. There are other artificial muscle technologies. This new one is supposedly powered by the chemical solution in which it operates, not by the electric field that triggers it. That's new. But if it's chemically powered, there must be waste products from the reaction that have to be flushed out. You need a whole circulatory system for the thing.

1990 called. It wants its IR LAN back.

No, it does not work that way. Natural gas power plants are big gas turbines, like aircraft engines. (Some are derived from aircraft engines.) When they are turned off, fuel consumption goes to zero.

Here's a wordless animation from China showing the details of how a big gas turbine plant starts up and runs. It's very clear, and not dumbed down.

So they had a bad meeting. It happens. It's even worse across language barriers. Most successful business teams get over that.

Google gets that automatic driving can kill people. The guys from social pushing "Cruise" put shiny plastic on lane keeping and adaptive cruise control and call it automatic driiving. They're right in the middle of the "deadly valley" - it's good enough you can take your hands off the wheel, but not good enough you can trust it. Those guys are going to be a problem.

GM is in serious legal and PR trouble right now because they have an ignition switch problem which causes cars to stop if people have a keychain with too much stuff on it. 13 GM people have already been fired. Google has never faced having to take responsibility like that.

The software industry is used to being able to dump its product liability on the customer. This will not work in the "Internet of Things".

In the real world, they're not. Here's the current CAISO output graph for all of California (which is 800 miles long and has a wide range of climate zones, with wind farms hundreds of miles apart) in the last 24 hours. Max wind generation today: 3600MW. Min: 300MW. That's over a 10:1 ratio. Checking PJM (the power grid for the northeastern US), today's max was 3200MW. Min: 900MW. About 3.5:1. Most days, those ratios are around 4:1.

So you still need a lot of natural gas plants that can be started up when the wind fails. Understand that load varies about 3:1 over the course of a day, in a predictable way, with peak load in midafternoon. Solar power output matches air conditioning load very nicely. Wind, not so much.

The price of bulk power goes way down late at night. Once in a while it goes negative for an hour or two. This happens on PJM when load is low, Ontario Hydro has excess water they're running through generators, the nuclear plants are running smoothly and don't want to shut down, and the wind turbines are getting good wind. The hydro and nuclear guys have a slow response time, so they'll pay to generate power rather than shut down for a few hours. So the wind guys, who can stop in a minute or two, drop out rather than pay. The turbine blades go to zero pitch and feather, the brakes come on, and the turbines slow and stop.

Read the entire paper, not the summary. There are some interesting points there. One is that NSA does not have a shortage of cybersecurity experts. That's because they train them. It takes three years of full-time training. The agencies that complain that they can't find anybody aren't investing in their people in the way that NSA does. Other agencies don't invest in their people like that.

This is typical of employer whining about not being able to get the people they want. Sure, the companies who want people with some very specific skill set, right now, often at low pay, can't find them. Organizations that are willing to train people don't have those problems.

One unexpected item from the paper: "One operating system, having been installed in almost a billion devices, has yet to attract malware in any significant way -- although it is falls short of being provably secure." What are they talking about? QNX? VxWorks?

Doing this is called Simultaneous Localization and Mapping, or SLAM. There's been enormous progress in that in the last decade. The basic idea is to take a large number of images of the same scene, possibly with inacccurate data about where they were taken, and build up a 3D model. It sort of works most of the time. Some algorithms do well indoors, especially where there are lots of strong edges and corners. Those are easy features to lock onto. Outdoors is tougher, although outdoors you can usually use GPS. It's a basic capabiilty robots need.

The video is frustrating. There's no comparison with previous work. Is this an advance, or did they just use known algorithms.

Why should the Linux kernel have a compression algorithm in it?

There's a way to do video compression so that frame rate doesn't matter. It's called Framefree. (PowerPoint, unfortunately). With that, you can crank up the playback frame rate as high as the output device can go.

Framefree was developed at Kerner Optical, which was spun off from Lucasfilm. Kerner went out of business a few years ago, and although there was a web site "framefree.us" and even a browser plug-in, it never caught on.

The idea is that the intermediate frames between key frames are mesh-based morphs, rather than MPEG-type block updates. Compression is compute-intensive, and playback requires a GPU. You can generate as many intermediate frames between keyframes as you want. Intermediate frame generation means interpolating the mesh points and then warping the image pieces to fit. So not only can you have very high display frame rates, you can also have ultra-slow slow motion. No MPEG-type blockiness, either.

While Framefree compression never caught on (probably because a high performance GPU in every set top box and DVD player was too expensive back then) the technology is used in sports programming to generate ultra-slow slow motion without using ultra-high frame rate cameras. Maybe it will make a comeback in the era of "4K" video with 60FPS frame rates.

That's the way it should be. If you want to subscribe to something, use RSS. That's totally under the control of the recipient. If you unsubscrbe from an RSS feed, there's no way the sender can keep sending to you.

It's easy to follow an RSS feed if you're using Thunderbird; a bit harder if you're a Google slave.