Eliezer Yudkowsky's "Rationality: From AI to Zombies" is a must-read on this subject. He lays out very clearly and compellingly just why self-improving AIs are so dangerous, and what extremely narrow path the humanity needs to walk to avoid this danger.
Honestly, you can say it wasn't their fault, but nearly 10% of them in 6 months have been involved in accidents. Even if it wasn't the fault of the technology itself, why is the accident rate so high?
Because they are constantly on the road, for testing and data collection. I would imagine that they drove more in these 6 months than many cars do in 6 years.
Accident rate in general: 4-5%
Accident rate so far with only 48 vehicles: 8-9%
Ah, no, the meaningful rate is accidents per miles driven, not per car. These cars are driven constantly (for testing, data collection, etc), surely way more that twice the average.
It's not about not having free access - it's not about having access period. I would have gladly paid for an online-only access, but there is no such option. I've tried signing up for a TV service with Comcast (which owns NBC) - not only I have to pay something like $70/month for huge package of channels I would never watch, but the only way to buy it is to sign up for installation of equipment for TV I do not own, and I would only get access when such equipment arrives - there is no option to sign up for online access and not have their cable equipment shoven down my throat.
P.S. I looked into using a Canadian VPN service, but had trouble getting www.cbc.ca olympic coverage to show up correctly on Linux. Now I am planning to watch Olympics on eTVnet - a Russian-language site in Canada, which, unlike NBC, is happy to take my money to provide me with access to content I want to watch.
I would recommend people see a talk by Yaron Minsky at Jane Street Trading Company - http://ocaml.janestreet.com/?q=node/61 . In his talk he explains the fast trading approach that his company is taking (he claims they a part in about 10% of all stock transactions!) and talks about how and why they use the OCaml programming language for developing their trading algorithms.
The article is based on the IBM's press release and is misleading because of it. In fact, there are three competing teams - one lead by IBM, one lead by HP and one lead by HRL Laboratories.
See also the FBO website for more information about this program.
only RealVNC is affected, which is a crappy vnc anyway. TightVnc and better yet UltraVNC are far ahead of RealVNC, neither of which are affected btw.
Well, RealVNC comes with a XFree/XOrg driver (vnc.so) that gives a very natural way to share the "native" X session. This is extremely useful when you want to have the normal X session on your machine (running at a normal speed), but still want to preserve the ability to connect to it remotely via VNC. AFAIK TightVNC does not allow anything like that and UltraVNC is Windows-only.
Another really nice property of RealVNC is that it has a "reverse connection" mode, where the client listens and the server opens a connection to it (very handy in certain firewall configurations). Finally, RealVNC comes with a "vncconfig" that can be used to adjust configuration on-line, and with a viewer where you can adjust protocol details (bpp, etc) on-line.
As far as I can tell, both presuppose that you have already managed to convince your OS to present GPS to the software as a/dev/xyz that acts like a serial device. The problem is exactly that the Garmin GPS is not willing to talk serial-over-USB, they expect some proprietary way of talking over USB.
Well, I use a Gramin Geko, though that one has a serial cable that I use over USB
The GPS side is the one that matters - it seems that Garmin GPS that I have provides a choice of protocols to use over the serial cable, but insists on using their proprietary protocol over the USB cable.
I've recently gotten Garmin GPSMan 60CS as a gift and so far I was unable to get any Linux programs to talk to it (over the USB cable that came with it). Did anybody have any success with getting Linux talk to any Garmin GPS units over USB?
Do you know that SpamCop has a "quick reporting" option (you have to ask to get it enabled for you)? With quick reporting, you only need to submit the spam via email and the source IP gets automatically reported (but no reporting of spamvertized web sites this way). This way you do not have to go to clicking through their web site, and the bl.spamcop.net still gets all the data.
I filed an RFE in Red Hat's Bugzilla bug database asking them to look up into some of the (IMHO valid) points that ESR raised. Let's hope something actually comes out of it.
There are very many products out there that do version control. But, there are very few that provide robust Content Management, which includes version control, but also includes a system to quickly and directly retrieve content for a web site/application and other such ammenities described in the patent.
May be there are few of them, but it only takes one to create a prior art. CVS + CVSWeb "checkout" mode is a great example of a system that does exactly what you describe!
You would never do such a thing with CVS, unless you're insane.
Well, then may be I am insane;-) We run our project web site (http://metaprl.org/ using only the CVSWeb "checkout" functionality.
I am wondering how well protected the source code will be. If the history (such as Moscow police "white pages" database with all the unlisted numbers included quickly leaking out) is any indication, we might soon see CDs with full Windows sources being sold for a few bugs on every corner in Moscow...
In fact, we might see a variant of an "open-source windows" movement actaully happening there!
NDA would not necessary be ioncompatible with GPL. One can have an NDA that allows redustributing the GPLed packages and source, but would prohibit mentioning that the particular code is a part of some bigger distribution, that it was produced by the company X, etc.
In other words, NDA may allow redistributing the code, but not the meta-information (e.g any information on *where* the code came from).
Sometimes (often!) I wish Slashdot let you moderate the articles and not just the posts; this one would have been (-1, Troll) very quickly.
That what are you doing on Slashdot? Go to Kuro5hin!;-)
In 6 years I never had a "dependency hell"!
on
Is RPM Doomed?
·
· Score: 1
I've been a RedHat user since RedHat 3.x times ('96, I guess) and I've been using RedHat exclusively after upgrading my last Slackware box with RedHat 4.0 (still '96, probably) and currently I run 9 RedHat machines.
And guess what - I've never experienced an "RPM dependency hell" in my life! Just two simple guidelines save me every time:
If there exist a RedHat-build package (preferably for the release you have installed) of the software or library you are looking for, always use that and not some 3rd-party package.
If a binary package requires a version of the library that's different from what you currently have, then download the source package and do rpm --rebuild. Do not ever waste time upgrading libraries unless even the source package requires it (which should only happen if the package truly need the newer library and would never even compile with an older one).
In other words, whenever you have any kind of precompiled binaries, there always will be problems with the binaries requiring different libraries/etc. There are only two ways of dealing with these problems:
Have a small number of well-defined sets of library versions and only compile for those. This is (as I understand it) how Debian solves it.
Have some simple way of recompiling things from source. This is where RPM is so great. The RPM format should not ever be considered a binary-only format - one of the greatest features of RPM is its source format and how easy it is to rebuild any good quality source package on any RPM-based platform.
Slashdot Mirror
Eliezer Yudkowsky's "Rationality: From AI to Zombies" is a must-read on this subject. He lays out very clearly and compellingly just why self-improving AIs are so dangerous, and what extremely narrow path the humanity needs to walk to avoid this danger.
California is a "no fault" state
California is a "fault" state - see e.g. https://en.wikipedia.org/wiki/...
Honestly, you can say it wasn't their fault, but nearly 10% of them in 6 months have been involved in accidents. Even if it wasn't the fault of the technology itself, why is the accident rate so high?
Because they are constantly on the road, for testing and data collection. I would imagine that they drove more in these 6 months than many cars do in 6 years.
Accident rate in general: 4-5% Accident rate so far with only 48 vehicles: 8-9%
Ah, no, the meaningful rate is accidents per miles driven, not per car. These cars are driven constantly (for testing, data collection, etc), surely way more that twice the average.
And why is it that you are owed free content?
It's not about not having free access - it's not about having access period. I would have gladly paid for an online-only access, but there is no such option. I've tried signing up for a TV service with Comcast (which owns NBC) - not only I have to pay something like $70/month for huge package of channels I would never watch, but the only way to buy it is to sign up for installation of equipment for TV I do not own, and I would only get access when such equipment arrives - there is no option to sign up for online access and not have their cable equipment shoven down my throat.
P.S. I looked into using a Canadian VPN service, but had trouble getting www.cbc.ca olympic coverage to show up correctly on Linux. Now I am planning to watch Olympics on eTVnet - a Russian-language site in Canada, which, unlike NBC, is happy to take my money to provide me with access to content I want to watch.
1. Python. I thought all the quants liked C, assembler, and even VHDL for their high frequency stuff.
Not necessarily. For example, an HFT company Jane Street Capical uses OCaml, claiming it makes code reviews go a lot faster and Knight-style errors a lot less likely. https://queue.acm.org/detail.cfm?id=2038036
I would recommend people see a talk by Yaron Minsky at Jane Street Trading Company - http://ocaml.janestreet.com/?q=node/61 . In his talk he explains the fast trading approach that his company is taking (he claims they a part in about 10% of all stock transactions!) and talks about how and why they use the OCaml programming language for developing their trading algorithms.
It would help immensely if you would put in a link to the article instead of just saying you found a useful article ;-)
The article is based on the IBM's press release and is misleading because of it. In fact, there are three competing teams - one lead by IBM, one lead by HP and one lead by HRL Laboratories. See also the FBO website for more information about this program.
Well, RealVNC comes with a XFree/XOrg driver (vnc.so) that gives a very natural way to share the "native" X session. This is extremely useful when you want to have the normal X session on your machine (running at a normal speed), but still want to preserve the ability to connect to it remotely via VNC. AFAIK TightVNC does not allow anything like that and UltraVNC is Windows-only.
Another really nice property of RealVNC is that it has a "reverse connection" mode, where the client listens and the server opens a connection to it (very handy in certain firewall configurations). Finally, RealVNC comes with a "vncconfig" that can be used to adjust configuration on-line, and with a viewer where you can adjust protocol details (bpp, etc) on-line.
As far as I can tell, both presuppose that you have already managed to convince your OS to present GPS to the software as a /dev/xyz that acts like a serial device. The problem is exactly that the Garmin GPS is not willing to talk serial-over-USB, they expect some proprietary way of talking over USB.
Well, the compatibility list for GPS Connect states "GPS 60cs (serial only)", so it seems I will be out of luck with OS X as well...
I've recently gotten Garmin GPSMan 60CS as a gift and so far I was unable to get any Linux programs to talk to it (over the USB cable that came with it). Did anybody have any success with getting Linux talk to any Garmin GPS units over USB?
See my response to the parent post.
See my post above.
Do you know that SpamCop has a "quick reporting" option (you have to ask to get it enabled for you)? With quick reporting, you only need to submit the spam via email and the source IP gets automatically reported (but no reporting of spamvertized web sites this way). This way you do not have to go to clicking through their web site, and the bl.spamcop.net still gets all the data.
I filed an RFE in Red Hat's Bugzilla bug database asking them to look up into some of the (IMHO valid) points that ESR raised. Let's hope something actually comes out of it.
May be there are few of them, but it only takes one to create a prior art. CVS + CVSWeb "checkout" mode is a great example of a system that does exactly what you describe!
Well, then may be I am insane ;-) We run our project web site ( http://metaprl.org/ using only the CVSWeb "checkout" functionality.
I am wondering how well protected the source code will be. If the history (such as Moscow police "white pages" database with all the unlisted numbers included quickly leaking out) is any indication, we might soon see CDs with full Windows sources being sold for a few bugs on every corner in Moscow...
In fact, we might see a variant of an "open-source windows" movement actaully happening there!
Indeed, accrording to the IRS site the free E-File (for 60% of taxpayers) is coming on Jan 16th!
Mod the parent up!
NDA would not necessary be ioncompatible with GPL. One can have an NDA that allows redustributing the GPLed packages and source, but would prohibit mentioning that the particular code is a part of some bigger distribution, that it was produced by the company X, etc.
In other words, NDA may allow redistributing the code, but not the meta-information (e.g any information on *where* the code came from).
I've been a RedHat user since RedHat 3.x times ('96, I guess) and I've been using RedHat exclusively after upgrading my last Slackware box with RedHat 4.0 (still '96, probably) and currently I run 9 RedHat machines.
And guess what - I've never experienced an "RPM dependency hell" in my life! Just two simple guidelines save me every time:
In other words, whenever you have any kind of precompiled binaries, there always will be problems with the binaries requiring different libraries/etc. There are only two ways of dealing with these problems:
Take a look at ML - there everything (including modules) is completely defined by its type and there is no way to go around them.