Re: AI?
It sounds like the weights of the various biomarkers is determined via machine learning, i.e. feed lots of examples through a neural network. This is typically called AI these days.
130 publicly visible posts • joined 28 Feb 2011
There's something weird:
"Lee said that although the false positive and negative rates of the algorithms were low at 0.024 and 0.037 respectively, the team needed to verify their results with many more patients."
With 23 people tested, why wouldn't all of your error rates be multiples of .043 (1/23)?
I think you're right. The problems with this plane stem from the fact that the engines are too far forward, in front of the center of gravity. They did this because the fuel efficient engines Boeing wanted to use were too big to fit further back. Unfortunately, if the nose rises, wind hitting the front of the engine pushes the nose even higher, an example of positive feedback. Other planes don't have this problem, and so don't require an MCAS to be stable when the nose drifts upwards a bit.
Fundamentally, this is a flawed design, and the MCAS is a complex bandaid designed to hide this fact. The only solution is to put smaller engines on the plane, and live with the worse fuel economy.
As the pilots quoted here noted, once the MCAS fires and the plane dives, the pilots probably aren't strong enough to adjust the trim against the wind. It's pretty scary watching the FAA and Boeing make the same potentially fatal mistakes again, this time in public. Let's hope the BAA, CAA or European equivalent put a stop to this crap.
Great summary, not only of Windows, but of MS in its entirety. Their approach was, until Azure, to just throw a random collection of poorly implemented functionality into Windows, all bundled together for whatever they charged (around $100 for the weakest tea, IIRC), cutting into the market for anyone who wanted to do a better version of whatever Windows threw in for free. I believe Ballmer called it 'cutting off the air supply' of their competitors, in an accurate turn of phrase that I believe he probably regretted.
Windows NT was the first system whose internals you could study without becoming ill, and that only if you didn't look at its VM or file system interfaces. I haven't dared look at it for years, but I have no reason to believe it's improved. SunOS did the file system / VM system much better than anyone else at the time (thanks, Steve).
MS definitely violated all sorts of anti-trust laws, but what really did them in was the Internet. They just didn't get it, and their browser's attempts at doing things proprietarily continued to hurt them, as their attempts to innovate were labeled, not incorrectly, as 'extend and extinguish'. It took Amazon's success with AWS to define the market well enough for MSFT to actually start competing again, with Azure, in an environment where, for the first time, they weren't leveraging their position in Windows to get an unfair advantage for mediocre technology.
I remember reading that the 737 MAX is just generally less stable than its competitors because the engines are so far forward that they're ahead of the center of gravity of the plane. So, if the nose starts going up, the air catching the engine cowling pushes the nose even further up. That's why there's an MCAS in the first place -- to prevent that instability.
Myself, I think I'm going to do my best to avoid flying in these things. It sounds like their fundamental stability depends upon the correct operation of the MCAS software, and the pilot's correct interaction with that software, unlike nearly any other commercial plane flying today. No thanks.
It seems to assume that everyone from Asia shares a similar culture, whether they're from Japan, Korea, Viet Nam, Cambodia, India, Pakistan, Afghanistan, Turkey, Siberia, China (and China itself obviously has many cultures internally).
How this became the SJW term of art for people from all the above countries is a mystery to me, but it certainly seems disrespectful. Or at least stupid.
An honest question. Clearly, the flawed software is required to make the 737MAX fly like earlier 737s, which it apparently doesn't do accurately enough to be safe for untrained (on the new plane) pilots to fly.
What I'm wondering about is whether the plane, with its forward placed engines, will always be significantly more unstable than other planes, even after training pilots on accurate simulators for the MAX. I'm wondering this, because it sounds like the plane has a serious tendency to pitch upwards in scenarios that wouldn't be a problem for other planes, and correcting for this behavior with software doesn't seem to be working particularly well. After all the misinformation Boeing has spouted so far, I certainly wouldn't fly on a MAX, no matter what software it's running.
If the plane is never going to be as stable as an older 737, I'm guessing that the only solution will be to put older engines back on the plane, with Boeing paying some compensation to companies that already purchased a MAX. If thats' the case, Boeing better be designing a new, better plane starting about 5-10 years ago.
I'm reading Bad Blood (about Theranos) and it strikes me how far people went to convince themselves not to notice that Theranos was a scam.
And that's what this sounds like, too. They refuse to answer reasonable questions. That's usually because they know the real answer is worse than you can imagine.
The Nova Scotia government is just showing how technologically incompetent they are by prosecuting someone for enumerating public documents.
Hopefully the prosecutor will drop the charges if they can find someone in New Scotland who knows anything about web servers, and can explain it to them.
There's so little we know of our own civilization that extrapolating to extra solar civilizations is a joke. Two questions make that pretty clear:
1 -- If humans disappeared today, how long would it take for another technologically advanced species to arise? I don't mean intelligent -- it is quite possible that dolphins or whales are more intelligent than us. But neither one has managed to build a telescope or a radio. AFAICT, it took 4.5 billion years for a civilization capable of building a radio to arise on Earth. It has been about 120 years since the first radio transmissions.
2 -- How long will humanity, or our direct descendants, exist at least at a level capable of broadcasting radio.
We don't know the answers to these questions, and they're about us and Earth. It would be even harder to extrapolate these answers to other worlds.
The Sparc may or may not be vulnerable, but I don't think this explanation covers it.
The Intel issue comes from a speculative code path loading data into the L1 cache from mapped but protected memory. This has nothing to do with TLBs -- the real question is whether the kernel address space is accessible to speculatively executed instructions.
Intel's error came from allowing a speculative load to proceed even when the ring # was wrong to allow the access to complete.
Yup, even remembering exactly what the interface is to a string package you haven't used for a year takes a search for me. Javascript, Objective C, C++, C, and Python all have slightly different versions, and I'm just not going to remember the parameters to substr in something I haven't used in > 12 months.
Perhaps I dreamed the whole thing -- Google's been no help here, but I recall a TOPS-10 program running at Carnegie Mellon, called "tingle" that, when run, would generate a nerdy pornographic one liner, of the sort "Shove a DECtape up my ass, I need to obstruct justice, cried the President!"
That would have been late 1970s through the early '80s. Anyone else remember that? Bueller?
Several years ago, I took a track from a Decemberists CD, encoded it at 64Kbits, 128Kbits, 256 KBits, 320 Kbits, both MP3 and AAC, along with the original song's bits, and then made a new CD with the resulting songs. Played it on my decent component stereo (probably cost about $2000 in today's USD). 64Kbits definitely sounded flat, and I thought that 128 was still perceptibly worse than lossless. Both 256 Kbits and 320 Kbits were indistinguishable to me from the original CD. So, I've been happy with 320Kbits rips since then; 256 would probably be fine, but I figured a little extra wouldn't hurt.
Actually, I did a few other things as well -- some classical piece I've forgotten, and something from Phillip Glass. All basically behaved the same as the Decemberists.
With regard to MP3 or AAC, I really couldn't tell any difference once I got to 256 Kbits or above.
You pay about 5-10 cents/GB for transfers between regions. You even pay 1 or 2 cents per GB for transfers out of one availability zone into another in the same region. So redundancy isn't free.
That's probably another reason it doesn't get done as often. But I suspect the biggest issue is the complexity, and dealing with additional latency when going between regions.
One thing you should know, if you'd never used it, was that the ITS "shell" was a binary debugger.
You'd login by typing "mlk$u" where "mlk" was your user name, and "$" was the Escape character. No password was required (or even allowed; thanks RMS). You could then type something like "4/" and see what's in location 4 (which was a register on the PDP-10). There were ways of running commands like macsyma, too (":macsyma", i.e. put a ":" in front of the command line).
Even stranger, you could type $$^R, and it would unprotect the OS, so that you could patch the running kernel from your shell.
I am not making any of this up.
The thing is, I like OS/X. But I have to admit that I'm also disappointed with the choices Apple made. I've using a late 2011 MacBook Pro, which I've upgraded to 8GB of memory. I've been holding off putting a new SSD disk in it, since I was hoping I'd be able to get a really light MacBook Pro 15". And I guess I can -- the new 15" model is 1.6 lbs lighter than what I'm using. But it's expensive, and I still have to see how badly they've messed up the keyboard. And I need dongles for everything.
But I'm half tempted to go down to a 13" display, and if I do, I'll go to the one that doesn't have a touch bar, and save $300.
But I think I'm also going to have to at least look at what the Linux ecosystem looks like these days.
Sitting here in a building in Pittsburgh, and measured 49Mbps download (T Mobile). Yesterday got 33Mbps at this location. I almost always get LTE speeds > 15 Mbps.
T mobile has a smaller coverage area than some, but I certainly don't see speeds like 2 Mbps; measured speeds are way higher.
Everyone knows that the current Macbook Air / Pro machines are due for a refresh. I'm typing this on a late 2011 machine Macbook Pro, which works fine; there's certainly no reason why I'd buy a replacement for it now when I expect a new, lighter, version within 6 months.
I *should* buy a large SSD for it, but have been holding off, thinking that perhaps the next rev of the Macbook Pro will be cool enough to make it worth upgrading.
If Apple wants its sales to improve, it'll have to ship new hardware. And not just a home use machine like the (relatively) new 12" Macbook.
I don't use Windows very often, but my wife calls me in when there are problems with her work machine. I was surprised by how badly Windows 10 handles the basics: it is far more buggy than I would have expected. To wit:
1 -- We haven't been able to configure it to run two displays at once. She has a largish Acer VGA display that worked fine with Windows 8, but when plugged into Windows 10, didn't appear to work, at any resolution. As a matter of fact, you can't even see the display in any of the configuration programs (why are there two?)
I assumed I needed a specific driver (but why not get a default configuration from a *VGA* display). However, by accident we booted it with the cover closed, and then it used the big display! But stopped using it when you open the cover.
Really, supporting a laptop screen + a VGA display shouldn't require any special configuration in 2016.
2 -- My wife has had to install printer drivers nearly a half dozen times, just to print on an HP 6180 ink jet printer. Again, is this rocket science?
Her personal macbook required zero manual configuration to use either of these devices.
You portray Apple's position as black and white, but it really isn't. Apple is being asked to spend their own money breaking into their own OS. No matter what they do, there'll always be *some* attack that can work against even future phones, even if it requires taking the phone apart atom-by-atom.
Apple is saying "No, we won't do this," and wants to stop now, even though the costs are probably not prohibitive today for a single iPhone 5C.
In other words, Apple doesn't want to be ordered to spend their own money to subvert their own security. It will *never* be *impossible* for them to get keys out of a phone. But it will be increasingly expensive, time consuming and likely have an increasing likelihood that an error will accidentally destroy the data on the phone.
I think Apple's most concerned about setting a bad precedent that it can be ordered to work to remove the security protections it inserted in iPhones. Right now, the request is that they generate firmware that sends keys to a piece of hardware without extra rate limiting. In 5S and later phones, there's a secure enclave that makes it much harder to do the passcode testing, since the enclave itself performs rate limiting.
But in the future, Apple MIGHT find itself ordered to make use of any bugs it later discovers in the enclave, or even to physically modify an enclave, to allow the same types of attacks on a more secure phone. And they might be required to turn over the resulting firmware to the FBI to use on other phones (along with keys to allow the FBI to sign it for other phones). Apple no doubt wants to stop from even starting going down this slippery slope.
Also, note that although the keys used to encrypt flash data are highly random AES keys, those keys are protected only by the passcode, and for most people, those passcode are way too small (4-6 digits). So, while someone who's really careful could make it very hard for the FBI to break into a phone, even with Apple's help, a novice (99.99999% of Apple users) would use a passcode from a much smaller space, conceivably small enough that with Apple's help, and a few bugs in the enclave here or there, the FBI could break into the phone.
Most significantly, this level of detail is way too technical for nearly everyone following this story. To them, the question is going to be a very simple "Are iPhones secure from the government, or are they not?" Apple wants the answer to that to be "Yes, they're secure."
Optical lasers wavelengths are in the few thousands of angstroms (angstrom = 10^-10 meters). But a proton is 10^-15 meters in diameter, and 1/10,000 of that is 10^-19 meters, or one billionth of a photon of visible light.
So, how can you use such large photons to see such a small change in length?
Also, how do you know the size of the black holes involved, at such a distance?
The 480p that video gets reduced to is still better than decent quality, and it's very nice to be able to watch Netflix or whatever even when your hotel Wifi is crap.
No one is being held up for cash to get binge-on's benefits, and users can disable it if they want. As far as I can tell, it is just an automated tool to do something that I used to wish to be able to do anyway: reduce the bandwidth used by a video stream when using 3G/4G/LTE.
The risk of violating net neutrality is that smaller companies might not be able to afford to pay off someone like Verizon for good access to its customers. That's not a problem here -- anyone can participate in Binge-on, and it doesn't cost the server's company any extra money.
Wow, so VW didn't come clean about the Audi and Porsches, they waited for the EPA to discover the same software was running on these other cars. Did VW really not realize that all VW owned brands will be tested?
Honestly, VW should have its imports banned for sale in the US for 5 years, just for stupidity.
I hope those bozos at the FCC figure out that routers are just low power computers. If someone really wants to violate the FCC regulations, they can just buy at small PC, Raspberry Pi, a dumb storage appliance, or pretty much anything, and drop their own kernel on it. All they're doing with these regulations is making life harder for people just trying to get some decent router software.
The FCC should concentrate on people who actually violate their regulations, not worry about "pre-crime."
I use it with Plex for ripped DVDs, Netflix, Hulu (Plus), HBO Go, Amazon Prime videos (via Chrome tab casting, which works quite well, even with a 2011 Macbook Pro doing the casting), CBS.COM (for Colbert, also via tab casting), FXX and Vudu purchased videos.
It's cheap, it fits into a TV that has no other reasonable connections, and it works well for everything but Apple media. So, I don't buy Apple media.
Of course, those losers at Amazon don't provide a native iOS casting app, but tab casting works well enough that I really don't mind.