« first day (1978 days earlier)      last day (3027 days later) » 

5:00 PM
Singapore appears to have killed more than just half my skin
Heh, this holiday has so far cost me £2998.65, excluding cash
Probably about £4000 if you include cash and incidentals, like passport, and those speakers I bought at Challenger.
So right on track for double my intended budget. Just like every time -_-
 
Bob
@PatoSáinz the wiring can still fail
 
Anonymous
@HackToHell hah! AMD said they are already on 14nm FinFET
 
Anonymous
also, "similar to 6 Titan X"... that's gonna get expensive fast
 
@PatoSáinz Same shit, different label. Meh.
 
Bob
5:10 PM
> Deep learning will allow self-driving cars to train themselves over time for all sorts of unexpected scenarios.
That sounds dangerous.
I want my self-driving car to be predictable, tyvm
 
Anonymous
@Bob luddite
 
@PatoSáinz Not really. It's only 6x the Titan X's GFLOPs rating because the Titan X's GFLOPs rating is artificially crippled
A £50, 6 year old GPU has more GFLOPs than a Titan X.
 
Bob
@PatoSáinz oh, sure. break out the meaningless insults.
 
Anonymous
also
 
Anonymous
>floating point ops per second
 
Anonymous
5:12 PM
maybe an opportunity for AMD, with their faster FP
 
@Bob Elon Musk in a blog mentioned that the system was 99% accurate but a car not detected at 70mph is very bad. That's why the cars need more computing power I suppose.
 
Also, typical nVidia number bending... where exactly did they get the fact that 8 TFLOPS is similar to 6x 6.6 TFLOPs?
@Bob Self driving cars are already too predictable
 
Anonymous
@qasdfdsaq because 3.5 GB are like 4 GB
 
The problem is coping with humans which are not predictable
 
5:15 PM
900 Mhz is cellular data band :O
 
Anonymous
@HackToHell it's ISM band
 
> It doesn't matter how well we program a car's maps and sensors -- it also needs to know how to deal with changes on the fly, like when a child jumps out on the road.
Err, Google's self driving car already handles all obstacles perfectly well. Stop.
 
HAMMERTIME!
 
Bob
s/child/object/
There's only so many actions you can take anyway.
 
@Bob And computers are already far better at choosing the best one than humans are.
 
5:17 PM
@PatoSáinz and that band is used for GSM-900
@qasdfdsaq Still not good enough
 
Bob
Machine learning for decision making in a car seems fairly pointless at best, outright dangerous at worst.
 
@HackToHell Far better than good enough
Within a few milliseconds it can detect the obstacle, determine its size and direction, as well as calculate the car's speed, heading, traction, and stopping distance, along with swerving distance and determine the optimal course of action, based on the location and velocity of all objects around it.
 
Bob
Machine learning necessitates a fair bit of trial-and-error... what, you're going to crash into 10 trees before you figure out that, maybe, you shouldn't do that?
2
 
A human takes over 200ms to just notice the obstacle in the first place.
@Bob LOL
 
@Bob It's all deep learning and neural networks these days
 
Bob
5:18 PM
It makes even less sense on a per-car basis. But spread across a connected network might work better.
 
@HackToHell It's a load of shit to sell overpriced hardware.
 
Bob
If you can learn from accidents of other vehicles.
 
So far self-driving cars have not been responsible for any accidents...
 
@qasdfdsaq Uh, neural networks are far better at object recognition
And they do use a lot of computational power
 
Bob
@HackToHell What kind of object recognition do you really need for a self-driving car, though?
 
5:20 PM
@HackToHell A car couldn't give a damn about advanced object recognition.
 
Bob
You need to know what other vehicles are. You need to know obstacles. You need to know speeds.
 
It can recognize every kind of object it needs to already.
 
Bob
You don't need to know if it's cat, dog or child.
 
@Bob I suspect the deep-learning aspect and adaptability is not really in order to avoid accidents
 
Bob
It doesn't matter if it's a bed or a cardboard box.
You just need to know (a) it's in front of you and (b) you don't want to hit it. And you can stop or move around it. Not much else to do.
 
5:20 PM
More to cope with unusual road conditions - unpaved roads, boarding a ferry, diversions, etc.
 
The entire terrain is fed onto the neural network with variables and the network itself segregates what causes the actions
 
Crossing a field... water on the road, etc.
 
Bob
@qasdfdsaq Then we get back to the tradeoff between trial-and-error vs a human programming the best response in.
 
@HackToHell You really should watch some demo videos about how self-driving cars actually work.
 
Bob
What does it take to figure out how to board a ferry? How many times are you going to drive into the water?
 
5:22 PM
@Bob That's the thing - machine learning removes the need for humans to program the best response in. The whole point is to figure out the correct course of action the first time
Simple trial and error... isn't "deep learning", it's dumb learning.
 
Bob
@qasdfdsaq But you need prior experience of some sort.
Whether that's someone programming in "water = bad" or it somehow learning that beforehand.
 
We don't need 8 teraflops and deep-learning AI to figure out driving into a tree is bad after several attempts.
Or water.
Or anything else you shouldn't drive into
 
Bob
Without experience, whether by programmed rule or by experimental results, you can't judge any choice as better than others.
Obviously not having those would be absurd.
So the real question is how much the adaptability can really help above the core rules.
 
@qasdfdsaq that's not machine learning
-.-
 
@HackToHell No shit, that's what I just said
Anyway, off to bed. I've got 27 hours of travelling to do tomorrow.
 
Bob
5:26 PM
Maybe you can improve fuel efficiency by moving slightly differently in different conditions.
Small things like that.
Safety decisions are rare (and varying) enough that it wouldn't be as much use.
 
Bob
@qasdfdsaq 'night. Have fun on the flight(s) :P
Oh, don't forget to reclaim your headphones?
 
@Bob Safety decisions are so deceptively simple I could do it with a bash script.
@Bob Heh, yeah. Flying from tiny back-end of nowhere town in Thailand to tiny back-end of nowhere town in Aberdeen is going to require four changes...
Also I got a (seemingly half broken) spare set of headphones off a mate here, though now that you reminded me I should check if Singapore airport has my glasses. If not, they're probably not worth reclaiming from Sydney.
 
Bob
@qasdfdsaq Until someone installs zsh and your script errors out :D
@qasdfdsaq Do you swap planes on that flight? Or did the people stopping in Singapore just get off?
Hm. Probably the former for such a long-distance flight.
 
Just don't start programming computers to make decisions based on moral grounds or the fat man/trolley problem.
@Bob Through flight on the same plane.
 
Bob
5:32 PM
@qasdfdsaq Then is it everyone off => rest, stretch legs => back on, or do the continuing passengers stay on the plane?
You can probably tell I've been on more long-distance bus rides than planes :P
Hm. Thought you were in a larger city in Thailand.
 
@Bob I hear it's everyone off => back on, but I've never had to do it myself. In the old days people didn't have to get off, now they do (some weird regulations)
 
Bob
@qasdfdsaq One thing that's annoyed me whenever someone talks about AI utopias and morals. They never seem to consider that the AI will largely get its morals from either the programmer or whatever material it learns from. And if that's the internet then god help us all.
5
 
Lol
Hmm, seems even the 777-300ER can't fly London to Sydney non-stop
 
HAMMERTIME!
 
Bob
Cavil just loves you :P
@qasdfdsaq I'd be more surprised if it could.
 
5:37 PM
@Bob Well the 777-200LR could... almost
 
Bob
o.O
That's a damn long way.
@qasdfdsaq Is that an almost with or without safety margins?
 
I don't know if the range ratings usually include safety margins or not
But I'm pretty sure I recently read that there was at least one airliner, either recently released or in development to be released in the near future that could do the trip non-stop
 
HAMMERTIME!
 
Bob
@qasdfdsaq Hm. Standard turbofan engine?
I remember seeing a lot about ramjets, etc., no close plans.
 
I'm guessing the range ratings do include safety margins, as the A380 is rated for 8500 nautical miles, and is in actual commercial service on a route of 8581+ miles from Dallas to Sydney
 
Bob
5:44 PM
Ah.
Next step: check what the safety margins are :P
 
Oh wait, nautical miles vs. non-nautical miles
My bad.
Rated at 15,700km, route length is 13,800km+
A350-900ULR will be rated for 16,112km though, just barely under the 16,884km distance from London to Sydney.
Huh weird
The Boeing 777 is a family of long-range wide-body twin-engine jet airliners developed and manufactured by Boeing Commercial Airplanes. It is the world's largest twinjet and has a typical seating capacity for 314 to 451 passengers, with a range of 5,235 to 9,500 nautical miles (9,695 to 17,594 km). Commonly referred to as the "Triple Seven", its distinguishing features include the largest-diameter turbofan engines of any aircraft, six wheels on each main landing gear, fully circular fuselage cross-section, and a blade-shaped tail cone. Developed in consultation with eight major airlines, the 777...
 
Bob
@qasdfdsaq So they could probably stretch it.
 
^^ This claims that at least one variant of the triple-7 is rated for 17,594km, but doesn't say which one, and never mentions that distance anywhere else on the page.
Ah, 777-200LR, 17,600 km.
Could most certainly do London to Sydney non-stop, as that's the distance achievable with maximum payload.
 
Bob
@qasdfdsaq 777-200LR
I'm too slow :P
 
Ah, 777-8(X):
> The 777-8's length is between the 777-200ER and 777-300ER with seating for about 350 passengers in a three-class configuration and has a range of 9,300 nmi (17,200 km) or more.
 
Bob
5:50 PM
Huh. They gave up on the triple-digit numbers? :P
 
Anyway, enough aircraft talk. I'm off to bed.
Really this time.
 
Bob
I should go too...
'night.
 
hii
Night bob :(
 
Bob
@allquixotic Excellent timing! :P
 
Hopefully they don't fuck up my A380 flight tomorrow
 
Bob
5:51 PM
I should be back in ~4 hours anyway.
 
Been looking forward to it for a while :-P
 
Bob
BTW, if you weren't already aware, @allquixotic: spoon.net/browsers
It's really nice. Much better than the BrowserStack lag...
 
@allquixotic What about meeee :-(
 
@_@
 
Bob
They also do online Selenium testing, but I probably won't be using that.
 
5:52 PM
@qasdfdsaq Night qasdffdsaq
It hurts my fingers to type your name properly
I prefer qasdfasdf
 
@allquixotic Yay \o/
 
@Bob ooer. as in, they do it for you?
 
Bob
The browser testing part download as app containers, so you can run old-IE just fine. Basically a very lightweight VM.
 
@qasdfdsaq ENJOY! first class?!
 
Bob
@allquixotic The Selenium one? Apparently.
 
5:53 PM
@Bob that fucking link could end up putting me out of a job; I WILL NOT share that will my bosses
 
@allquixotic Standard economy, but nice upper-deck two-seat
 
@qasdfdsaq nice! :D
 
Bob
@allquixotic Huh? Oh, I meant it runs the tests across multiple browsers, on their servers.
They don't design tests for you. I think...
 
@Bob oh, good :D
thank goodness :D
> Launch Selenium Grid

Click Start Grid to launch the Selenium Grid.

Chrome, Firefox and IE nodes will be launched automatically when your test executes.
yeah, they don't write tests for you... yaay
 
Bob
Unfortunately we can't really do Selenium/automated testing on our app -_-
 
5:55 PM
we can on one of ours - the one I mentioned to you about *coughsDC*
 
Bob
lol
Anyway, yea. Problem with Spoon is they don't support mobile browsers.
 
ahh
I tested our site on a mobile and it was amazingly good
 
Bob
So I might have to use it with BrowserStack for now, rather than outright replacing it.
@allquixotic Yea, about that... :P
 
@Bob yes, and it was aggravating but I got through it
 
Bob
@allquixotic Now imagine that being your day job.
 
5:58 PM
!!no
 
Bob
lol
=> gitter
 
@allquixotic And the plane seems completely full so there's a slim chance I might get upgraded again
Also off to bed, third time lucky perhaps.
lol
 
I doubt Bob would go to bed before me.
 
Bob never goes to bed when told to
only when he wants to
 
Bob
6:09 PM
:D
 
Even if he is in bed, he's gonna be browsing in mobile.
 
Morning all
 
6:25 PM
heh... FBI warning on a YT video.
 
lol
 
Description was "I found this video at: <link>"
 
7:22 PM
Somebody wants Linode dead and they're not giving up.
 
7:34 PM
@bwDraco No point in DoSing someone if you just give up in the end. They've probably had a ransom demand. Not that they would let anyone know of course.
 
... Digital Ocean, anyone? Lol
 
their competitor would not risk their business and reputation on that
it's probably China or Russia
@bwDraco woah
can't customers just make something like cloudflare authoritative for resolving their website?
Linode's own DNS would only be absolutely required for new customers or people needing to admin their server from the mgmt console
people with existing sites hosted there can easily route around the Linode DNS issue
 
8:21 PM
!!xkcd compiling
 
The ideal language for that xkcd; see their blog about compiler performance
 
 
1 hour later…
Anonymous
9:22 PM
> It takes about 10 seconds to compile, and it takes 940MB of memory to do so.
 
Anonymous
oh wow
 
9:38 PM
@PatoSáinz Especially when you consider that Google Go is also a statically typed, compiled language like Crystal, but you can compile something as complex as a web browser or IDE in about 5 seconds
except that Go actually makes you specify the types, but Crystal has to guess them :P
 
Anonymous
9:51 PM
@allquixotic I don't know if Crystal as a project is ever going to take off
 
10:03 PM
Password already reset.
 
10:47 PM
all except the last one: wtf?
 
@allquixotic That's what I hate about Verizon. Their network is completely closed.
On duty right now, so can't really stay here.
 
@bwDraco @allquixotic No it's not.
http://www.verizonwireless.com/accessories/4g-sim-activation-kit/
4G SIM Activation Kit

4/ 5 (44 Reviews)

SKU #VZW-SIM4GPP

Bring your own 4G Verizon phone
Activate on our low cost 4G LTE Plan with No annual contract
Plan: $45/mo for UNLIMITED TALK & TEXT plus 1GB of DATA
**Forget it**. You need a Verizon phone.

"Is my 4G LTE SIM card compatible with any other provider's device?

It's unlikely. You may be able to take a Verizon Wireless device to another carrier and get a new SIM card with that carrier. However, you would have to check compatibility with the carrier first."
So Verizon defines "Bring your own" as "Bring your own as long as it's a Verizon device" :/
Good old "USA". Whatever happened to "Land of the Free"? Time to update your National Anthem ;)
 
Anonymous
@bwDraco again, again again, again
 
11:11 PM
@DavidPostill Remember, Red White and Blue are the colours of freedom... Except when you see them flashing in your rear view mirror
 
@CanadianLuke lol
 
11:59 PM
I have just been restarted! This happens daily automatically, or when my owner restarts me. Ready for commands.
 

« first day (1978 days earlier)      last day (3027 days later) »