I think that Uber is making the wrong decisions in these disaster scenarios. At least if you have an algorithm that responds to demand, you can stand behind it and say "this works as it is supposed to, and is designed to get the most number of people rides as quickly as possible". Once they start turning it on and off (when people complain), it starts to feel a lot more arbitrary and less fair.
This is what you get when you decide to sell public space to the highest bidder.
If I were paying billions of dollars for the right to advertise, I would want to own it too.
Maybe this is an indication that we should stop selling our public spaces and events to such and extent.
Actually the original vision for Xbox One wasn't just gaming. Microsoft was pretty clear about this point.
In any case, good use of random quotation marks and fictitious statistics. Always makes a good argument better. Well executed.
I feel like MS has missed the fact that there are a lot of Xbox users out there that bought the machine specifically for the Kinect and associated content. If I wanted an ultimate gaming platform I would have bought a PC. I bought the Xbox because the Kinect is fun and gets you off the couch. Just because non-gamers don't rant and rave on the internet about Xbox vs PS4 specs doesn't mean they shouldn't have some say about how the platform develops.
A self-driving car should always be able to judge its stopping distance to a high degree of accuracy. None of these scenarios should ever happen. The car shouldn't be driving that fast to begin with.
I'm going to go out on limb here and suggest that it might have less to do with greed and more to do with the fact that infrastructure and people cost money.
As much as I would like it if everything in life were free, it isn't, which means that the people that build and maintain their infrastructure need a salary (not to mention what I imagine are non-trivial hosting costs), and that money has to come from somewhere.
Years ago I registered the site publicpatent.org. I had planned for people to just submitted random stuff (i.e. idea A + idea B), so that stupid, obvious stuff would be in the public domain and no one could try to patent it. Using NLP to automate the stupid idea generation is great!
As someone who has worked with young students, high school teachers, university students, and university faculty getting Pi's into the classroom (e.g. http://clean.energyscience.ca/..., http://rpi.science.uoit.ca/ I can say that the micro:bit may be a better starting point for really young kids and their tech phobic teachers than the Pi. From what I can tell, the micro:bit isn't really a computer (unlike the RaspberryPi), but rather a peripheral that enables some physical computing.
There are some ugly sides to the Pi for the uninitiated. I'm not saying one is better than the other (I really like the Pi), but I do think the micro:bit could be a welcome addition to the ecosystem.
I'm disappointed that BBC isn't making them available to the general community from the get go (or even before release to schools). We have a way better chance and troubleshooting (and populating stackoverflow) issues than they do. Despite the fact that this is intended to be plug-and-play, things never are (especially when they involve locked-down machines like those present at most schools).
In any case, I'm looking forward to getting one of these things!
I think it's worth mentioning that the Pi Zero likely has the ability to broadcast data, though not in the way you might expect. (I can't say for certain because I don't have one yet)
There are a few examples on the web about how to do this: http://www.instructables.com/i...
I tried this recently on the Model2, and it worked well. Obviously you would need something on the receiving end to listen + decode, but I don't think it would be too difficult. People have been doing this with HAM radio for ages.
If the antenna is a decent size, the range is pretty good. You could easily cover a house with low bitrate data. I'm sure this is in violation of lots of FCC rules, but meh.
I've been considering making a tutorial around python, where you generate a fake data set (with something like 7 billion values) and do the following:
Stage 1: loop over the whole data set and compute the true mean and the true standard deviation.
Stage 2: Take the same data set, but choose random samples (say N=1000) from the set, and compute estimates of the mean and the standard deviation. You can show how depending on which 1000 elements you choose, you'll get a spread of estimates for both which bracket the _true_ values which we know from stage 1.
You give the scenerio of an overpriced hard drive purchased from his brother. Obviously that would be bad, but so is a system that is stuck in the mud. Doing nothing also costs money (in lost productivity and output). These losses tend to be overlooked because they are harder to quantify. That doesn't make them any less real.
Don't get me wrong, I'm not in favour of this particular surveillance program, but it is frustrating to see minor IT issues becoming roadblocks. I have seen similar situations to this one.
A better mode of operation is to grant individuals with local decision making power and periodically check their decisions to ensure they make sense. The $5000 hard drive you imagined would show up on the books and would be difficult to justify. Sure, it might get through one time, but eventually that kind of behaviour would be caught (and the responsible individual would be held accountable). Most people aren't interested in risking their job so their brother can make a few grand.
Always assuming that everyone is trying to screw the system just tells workers that it is better to do NOTHING than to do what's right. You will have a much more productive workforce if you operate under the assumption that people want to make good decisions.
I've seen far too much money wasted on supposedly transparent processes.
If this data were at all convincing and held up to review, we wouldn't be reading a second hand report from Hacker News Magazine.
This would be the cover article of Nature or Science. The editors would be fighting to have it.
I wanted to buy the new retina Macbook Air, but the situation with ports, processor, etc was enough to keep me on this thing for the time being.
== Home desktop == i5-4690K, 16 GB RAM, 512 SSD Motherboard = I don't remember, one of the ones that works well with OS X (at some point I'll hackintosh it) GPU GTX 970 Went with the 970 b/c it's way cheaper than the 980 (and apparently has way less RAM...which was annoying to find out after the fact) I went with the K because at some point I'll goof around with overclocking
== Workstation 1 == i7-4820K, 32 GB RAM, 512 SSD Motherboard = I don't remember, something about 4 channel RAM... GPU Titan Black I went with the i7 to test hyper threading performance. This machine also has very good single threaded performance (it shuts down the other cores and ups the clock on one of them). This machine is due for a PCIe SSD upgrade (for testing performance where 32 GB RAM isn't enough) The Titan is for double precision CUDA work. I went with the K because at some point I'll goof around with overclocking
== Workstation 2 == Dual Xeon E5-2660 v2 (top says I have 40 threads), 28 GB RAM, 1 TB spinning disk (it was lying around...) Motherboard = I don't remember, it has two sockets... GPU GTX 980 The 980 works _very_ well for real time rendering.
I plan to switch to 4k monitors on the workstations relatively soon (I want a reasonably priced IPS panel to show up).
Compute efficiency between new and old machines (say 4 years) is a big deal in the power and cooling budget of a datacenter. I'm not talking about your personal desktop, I'm taking about big iron. Trust me, power costs matter at that scale.
Until we reach a point where compute per watt stabilizes, it is highly unlikely that anyone would be interested in using old components to build a cluster. The fact that the parts would all be slightly different would be a headache too.
Older gear typically uses more power / FLOP, and is slower, so your time-to-solution takes a hit too.
If we get to the point where the power usage / FLOP for an N+1 device is basically the same as N, then you might see people do this, so long as they are okay with waiting longer for a result. Until then, don't hold your breath
Slashdot Mirror
This robot has been in the worldwide news three times already. Whoever is planning the press campaign is doing a great job.
I agree. This feels fake
I think that Uber is making the wrong decisions in these disaster scenarios. At least if you have an algorithm that responds to demand, you can stand behind it and say "this works as it is supposed to, and is designed to get the most number of people rides as quickly as possible". Once they start turning it on and off (when people complain), it starts to feel a lot more arbitrary and less fair.
We have one at work and I've tried out some of the demos. It's a pretty impressive piece of kit.
If ever there was an example of ageism in tech...
This is what you get when you decide to sell public space to the highest bidder. If I were paying billions of dollars for the right to advertise, I would want to own it too. Maybe this is an indication that we should stop selling our public spaces and events to such and extent.
Actually the original vision for Xbox One wasn't just gaming. Microsoft was pretty clear about this point. In any case, good use of random quotation marks and fictitious statistics. Always makes a good argument better. Well executed.
I feel like MS has missed the fact that there are a lot of Xbox users out there that bought the machine specifically for the Kinect and associated content. If I wanted an ultimate gaming platform I would have bought a PC. I bought the Xbox because the Kinect is fun and gets you off the couch. Just because non-gamers don't rant and rave on the internet about Xbox vs PS4 specs doesn't mean they shouldn't have some say about how the platform develops.
A self-driving car should always be able to judge its stopping distance to a high degree of accuracy. None of these scenarios should ever happen. The car shouldn't be driving that fast to begin with.
If consumers could actually buy a priv at $290, things might have turned out differently. I've never seen one for sale anywhere close to that price.
I'm going to go out on limb here and suggest that it might have less to do with greed and more to do with the fact that infrastructure and people cost money. As much as I would like it if everything in life were free, it isn't, which means that the people that build and maintain their infrastructure need a salary (not to mention what I imagine are non-trivial hosting costs), and that money has to come from somewhere.
Years ago I registered the site publicpatent.org. I had planned for people to just submitted random stuff (i.e. idea A + idea B), so that stupid, obvious stuff would be in the public domain and no one could try to patent it. Using NLP to automate the stupid idea generation is great!
As someone who has worked with young students, high school teachers, university students, and university faculty getting Pi's into the classroom (e.g. http://clean.energyscience.ca/..., http://rpi.science.uoit.ca/ I can say that the micro:bit may be a better starting point for really young kids and their tech phobic teachers than the Pi. From what I can tell, the micro:bit isn't really a computer (unlike the RaspberryPi), but rather a peripheral that enables some physical computing.
There are some ugly sides to the Pi for the uninitiated. I'm not saying one is better than the other (I really like the Pi), but I do think the micro:bit could be a welcome addition to the ecosystem.
I'm disappointed that BBC isn't making them available to the general community from the get go (or even before release to schools). We have a way better chance and troubleshooting (and populating stackoverflow) issues than they do. Despite the fact that this is intended to be plug-and-play, things never are (especially when they involve locked-down machines like those present at most schools).
In any case, I'm looking forward to getting one of these things!
I think it's worth mentioning that the Pi Zero likely has the ability to broadcast data, though not in the way you might expect. (I can't say for certain because I don't have one yet) There are a few examples on the web about how to do this: http://www.instructables.com/i... I tried this recently on the Model2, and it worked well. Obviously you would need something on the receiving end to listen + decode, but I don't think it would be too difficult. People have been doing this with HAM radio for ages. If the antenna is a decent size, the range is pretty good. You could easily cover a house with low bitrate data. I'm sure this is in violation of lots of FCC rules, but meh.
I'm curious what they paid per user. I wouldn't mind moving to a co-op model. How much could it possibly cost to run this place per year?
We got rid of the penny here in Canada. It was no big deal. I've hardly noticed the difference.
I've been considering making a tutorial around python, where you generate a fake data set (with something like 7 billion values) and do the following:
Stage 1: loop over the whole data set and compute the true mean and the true standard deviation.
Stage 2: Take the same data set, but choose random samples (say N=1000) from the set, and compute estimates of the mean and the standard deviation. You can show how depending on which 1000 elements you choose, you'll get a spread of estimates for both which bracket the _true_ values which we know from stage 1.
Until Gabor Fekete weighs in on this, I'm unconvinced.
How is it that Google is being scrutinized for anti-trust (http://www.bloomberg.com/news/articles/2015-09-25/google-said-to-be-under-u-s-antitrust-scrutiny-over-android-iezf41sg) when Apple has been behaving like this for years? I don't want to start a flame war (I've used mac laptops forever), but I don't see how they continue to fly under the radar.
Don't get me wrong, I'm not in favour of this particular surveillance program, but it is frustrating to see minor IT issues becoming roadblocks. I have seen similar situations to this one.
A better mode of operation is to grant individuals with local decision making power and periodically check their decisions to ensure they make sense. The $5000 hard drive you imagined would show up on the books and would be difficult to justify. Sure, it might get through one time, but eventually that kind of behaviour would be caught (and the responsible individual would be held accountable). Most people aren't interested in risking their job so their brother can make a few grand.
Always assuming that everyone is trying to screw the system just tells workers that it is better to do NOTHING than to do what's right. You will have a much more productive workforce if you operate under the assumption that people want to make good decisions.
I've seen far too much money wasted on supposedly transparent processes.
If this data were at all convincing and held up to review, we wouldn't be reading a second hand report from Hacker News Magazine. This would be the cover article of Nature or Science. The editors would be fighting to have it.
I just went through an upgrade / refresh cycle, so normally I'm not this up-to-date with my hardware
== Laptop ==
2012 Macbook Air, 2 GHz i7, 8 GB RAM, 256 SSD
I wanted to buy the new retina Macbook Air, but the situation with ports, processor, etc was enough to keep me on this thing for the time being.
== Home desktop ==
i5-4690K, 16 GB RAM, 512 SSD
Motherboard = I don't remember, one of the ones that works well with OS X (at some point I'll hackintosh it)
GPU GTX 970
Went with the 970 b/c it's way cheaper than the 980 (and apparently has way less RAM...which was annoying to find out after the fact)
I went with the K because at some point I'll goof around with overclocking
== Workstation 1 ==
i7-4820K, 32 GB RAM, 512 SSD
Motherboard = I don't remember, something about 4 channel RAM...
GPU Titan Black
I went with the i7 to test hyper threading performance. This machine also has very good single threaded performance (it shuts down the other cores and ups the clock on one of them).
This machine is due for a PCIe SSD upgrade (for testing performance where 32 GB RAM isn't enough)
The Titan is for double precision CUDA work.
I went with the K because at some point I'll goof around with overclocking
== Workstation 2 ==
Dual Xeon E5-2660 v2 (top says I have 40 threads), 28 GB RAM, 1 TB spinning disk (it was lying around...)
Motherboard = I don't remember, it has two sockets...
GPU GTX 980
The 980 works _very_ well for real time rendering.
I plan to switch to 4k monitors on the workstations relatively soon (I want a reasonably priced IPS panel to show up).
If they don't know what they are doing, then why are they the leaders?
Compute efficiency between new and old machines (say 4 years) is a big deal in the power and cooling budget of a datacenter. I'm not talking about your personal desktop, I'm taking about big iron. Trust me, power costs matter at that scale.
Until we reach a point where compute per watt stabilizes, it is highly unlikely that anyone would be interested in using old components to build a cluster. The fact that the parts would all be slightly different would be a headache too.
Older gear typically uses more power / FLOP, and is slower, so your time-to-solution takes a hit too.
If we get to the point where the power usage / FLOP for an N+1 device is basically the same as N, then you might see people do this, so long as they are okay with waiting longer for a result. Until then, don't hold your breath
Given the purpose of these devices, cheaper would be better than faster.