Golly gee whiz, Mommy thought she could just leave little Declan and Aria parked at the computer for a few hours unsupervised so she could have a little Happy Juice and watch her soaps. But no, apparently there are 'disturbing' videos on YouTube. Did someone mention evolution?
This is an outrage! Something should be done! By someone who's not me!
Sure, sure, this is exactly what they were saying about Go before AlphaGo destroyed it. Now, like Creationists, you have to invoke incomplete knowledge to preserve mankind's special snowflakeness, like that's something only humans can deal with.
There was no 'writing it for Go' with AlphaGo Zero. The whole point of AlphaGo Zero (compared to AlphaGo Lee) was that there was no Go specific info other than the scoring. They just fed Go boards into a fresh deep net and used a new Monte Carlo evaluation algorithm that can be used for anything.
Oh you can't beat Checkers. Oh, you completely solved checkers? Well, you can't beat chess.
Uh, you beat chess, well you CERTAINLY can't beat Go.... You beat Go? Well, you couldn't have beat Go without human training.......... You beat Go without human training, and it can beat every human on the planet every time?
Well, you can't beat Starcraft ha ha ha!
AIs have done pretty well in poker, which is another partial-knowledge game, and there's no reason to think a deep learning AI will do any worse at partial knowledge than a stupid meatbag.
Everything you just said about Starcraft 2 is what people said about Go till Go was handily trounced and they had to retreat to the next barricade.
AGZ doesn't seem to make any fantastic, earthshattering, amazing moves... it trades points, seemingly nothing special. Yet it inexorably ends up ahead on points to the relentless outcome where no human player can beat it.
It'd be the same with SC2. It wouldn't crush you at the start, but inevitably it would win.
The same way meatbags 'solve' a starcraft game, except you have a million times more processing power?
AGZ can figure the best move at any time just from the current board setup. That should work in SC2 as well, though of course you need more TPUs and memory.
I don't get this in python. Recently, when a website changed its API slightly, I went back to some three year old python I wrote, and it was instantly obvious what was going on and I was able to modify it, test it, and check the code back into git in under a half hour.
Yes, I do get this with C++, which like perl is not designed to be read.
Yes, human players can still beat second tier AIs from Facebook and universities.
But if you turned the AlphaGo Zero team on it it would dominate it in a couple months max. AGZ learned, from scratch, how to beat every human on the planet at Go every single time in 3 days.
'This is the Year of the Linux Desktop' (aka every year since 1998) finally went down the tubes when Ubuntu went insane.
We went from being able to say 'Oh yeah, just install Ubuntu' if anyone expressed curiosity (even if we were personally using Debian or CentOS or whatever) to going uhhhh.... finally we could point to Mint, but by then it was too late.
Perl is great to write - you can just bang out some shit fast and it probably works. It's certainly better than using bash or sed for the same problem space.
The hell comes if you ever have to look at what you crapped out ever again. Six months later and it's unintelligible. Or worse, you have to use what someone else crapped out.
Yes, you can write really good, really clean, really maintainable perl if you want to put in some effort - I've seen it twice in my entire career.
Funny bit: slashdot launched 20 years ago this month. So if you were reading it back then most netizens would probably consider you an old fogey.
Anyhow, I'm still doing engineering which involves lots of design, coding, and tech lead. Adamantly refused to get into line management and that hasn't been a problem since there is a real shortage of skilled engineers. My depth and breadth of experience means lots of mentoring since I haven't turned into a cranky old bastard yet.
You have to learn new things, but I like that. Another key to keeping things fresh is switching jobs about every 5-7 years or so (which I've done my entire career) - about 5 years I start getting extremely bored and itchy feet.
Not sure what the hell I'd do if I got tired of it (go die on a farm?), but it's pretty much my choice when to retire.
The anti-Amazon crowd wouldn't mind if Amazon left and took half the tech jobs. Just makes it more of a 'hipster'* / counterculture paradise like Portland and Detroit - the inner bit, which is all that counts since that's where the non-brown people are.
* Provide your own definition since nobody can agree on one.
Oral sex in the/. Mile High Club is when you notice your seatmate is using a Windows laptop and you helpfully start berating them about using Macro$haft Winblowze, she says 'F@#$ you', you reply with 'F$#% you, you filthy vagina haver', and then one of you has your seat changed in the afterglow.
The code and techniques look like APT17 aka DeputyDog - hacking into tech firms, military and governments for the Chinese government for at least 10 years.
They realized CCleaner was a fantastic indirect vector into a whole lot of firms, and god knows what else they've got their fingers in that people haven't noticed since most firms are Equifax level incompetent with security.
Same reason we still have open offices even though they're hell for productivity (unless your job is just talking to people all day). Same reason people buy lottery tickets. Everybody knows better but want to believe they're the geniuses who can make it work this time for sure.
The UI people want to believe that if their design is just beautiful and elegant ENOUGH that it'll automatically have amazing usability. And if course it won't. They're no longer UX people, they're graphic designers.
'there is zero downside' - Absolutely, completely Wrong 'lower performance' - Wrong (depending on what you're doing)
I'm running 64-bit on my desktops because they've got tons of memory, but...
Besides 32-bit CPUs requiring 32-bit OS (which other people have covered extensively), if you've got a 64-bit CPU but only 512MB or 1GB of memory, like a Beaglebone Black or RasPi3, then running 32-bit OS and apps can save you quite a lot of memory and disk space because more things can be allocated in 4 byte chunks rather than 8 byte chunks.
Let's take a look at the venerable Notepad++, version 7.3.3. The 32-bit exe is 2.32mB and the 64-bit exe is 2.781mB. That's not 'omg', but it's 20% larger and that adds up over all of Windows's exes and anything you install. If I load langs.model.xml (about 281kB) then the 32-bit version uses 13.2MB of memory and the 64-bit version uses 14.7MB of memory, which is 11.4% more. Again, certainly not a doubling, but when you've only got 512MB of RAM this really adds up.
The 'lower performance' thing is hogwash too. If you're doing heavy 64-bit math then the 64-bit app is probably going to win. If you're doing lots of disk reads/writes and string processing the 32-bit app may well win (slightly) just because you're slinging less data around.
Each of these guys has his own little corner of the world and is totally ignorant of what's going on elsewhere.
Yes, in my industry we've replaced C++ (and good riddance) with C# on the PC side, but C is still going extremely strong on the lower end because when you've got a (by modern standards) little SoC with 48K of 'ROM' and 34K of RAM you can get a lot more out of C than you can Mono or Java. We even still occasionally bust out the asm, but for the most part modern C (with things like named structure initializers) is the sweet spot.
And people have been predicting the replacement of coders by AI for 30 years - it's like fusion power. I imagine at some point the mere code pigs (coders who just search stackexchange and copypasta code) might be replaced, but good luck replacing hands on problem solvers for a while.
But then, who am I to argue with a 'job leadership consultancy' *snicker*.
People working on COBOL aren't 'cowboys'. They're (metaphorically, like the cowboy thing) guys with suits and briefcases who might be briefly exuberant by getting a martini dirty style and hoping nobody notices this outrageous whimsy.
Yeah, I know, there's a company named that. Go look at the employee pics.
They're still looking at Apache Pass for 2018, but it's dead as a product/now/ for Skylake-EP/Purley and they're playing it down. At IDF 2015 it was 'Apache Pass Apache Pass Apache Pass' and at IDF 2016 it was 'Uh, yeah we'll have DIMMS', and now it's 'Wow, this is great for SSDs.'
I'm not surprised they brought some DIMMs to show (it's not dead dead, just not now), but I'm genuinely curious about this: Ask when you can buy some.
It's not great for reads though - Even the consumer m.2 SSD in my PC is faster for that. It's just really fast on small random writes, or small reads under LOW read/write load (small queue depth) because it really wins for latency there. So you have to be using it on something with lots of writes and low load small reads to get amazing benefit over current flash - and then it burns out fast.
I'm sure there will be some very targeted applications where this is perfect, like maybe mostly data acquisition and high speed db updates then at the end of the week you pop it out and put in a new one and you've got your snapshot and can go analyze the data on that.
Yeah, because it's not true. It doesn't work well as cache, and if you tried it you'd burn it out fast. It was SUPPOSED to, they promised a lot that you could use this as both storage as main memory, but they couldn't make it work and even Intel has stopped talking about Apache Pass.
Basically, this is a evolutionary step in flash, giving you much faster random write and maybe 2x the endurance for 2x the price, so not bad as a disk.
But it's not the revolutionary step Intel was promising.
Modern programmers have no idea how a linked list or a hash table are stored how, how sorts work, how garbage collection and memory allocation actually works...
This was really driven home to me when one of the Java guys was trying to look at some C code and thought the linked list stuff was a bug he'd found. 'It's just a reference back to itself!'
I have to admit that mostly it doesn't matter, but occasionally their assumptions of infinite memory and infinite processing power lead them to do some really dumb things that are inconceivable to anyone who actually knows how things work under the hood. I've given some of their Java code a 1000x speedup just by the simplest optimizations like not reading the entire file every time to process a single line.
As amusing as it is to contemplate the douchepocalypse an Apple vapor would entail, the patent is clearly for chip manufacturing and the guy who filed it is one of Apple's chip people.
Remember, Apple is very big into custom silicon, and like everything else they want control of every step of it and that includes how the chips are made.
Slashdot Mirror
Golly gee whiz, Mommy thought she could just leave little Declan and Aria parked at the computer for a few hours unsupervised so she could have a little Happy Juice and watch her soaps. But no, apparently there are 'disturbing' videos on YouTube. Did someone mention evolution?
This is an outrage! Something should be done! By someone who's not me!
Sure, sure, this is exactly what they were saying about Go before AlphaGo destroyed it. Now, like Creationists, you have to invoke incomplete knowledge to preserve mankind's special snowflakeness, like that's something only humans can deal with.
There was no 'writing it for Go' with AlphaGo Zero. The whole point of AlphaGo Zero (compared to AlphaGo Lee) was that there was no Go specific info other than the scoring. They just fed Go boards into a fresh deep net and used a new Monte Carlo evaluation algorithm that can be used for anything.
This is such the God of the Gaps argument.
Oh you can't beat Checkers. Oh, you completely solved checkers? Well, you can't beat chess.
Uh, you beat chess, well you CERTAINLY can't beat Go. ... You beat Go? Well, you couldn't have beat Go without human training. ... ... ... You beat Go without human training, and it can beat every human on the planet every time?
Well, you can't beat Starcraft ha ha ha!
AIs have done pretty well in poker, which is another partial-knowledge game, and there's no reason to think a deep learning AI will do any worse at partial knowledge than a stupid meatbag.
Everything you just said about Starcraft 2 is what people said about Go till Go was handily trounced and they had to retreat to the next barricade.
Yeah, I wish I could upvote you on this.
AGZ doesn't seem to make any fantastic, earthshattering, amazing moves... it trades points, seemingly nothing special. Yet it inexorably ends up ahead on points to the relentless outcome where no human player can beat it.
It'd be the same with SC2. It wouldn't crush you at the start, but inevitably it would win.
The same way meatbags 'solve' a starcraft game, except you have a million times more processing power?
AGZ can figure the best move at any time just from the current board setup. That should work in SC2 as well, though of course you need more TPUs and memory.
I don't get this in python. Recently, when a website changed its API slightly, I went back to some three year old python I wrote, and it was instantly obvious what was going on and I was able to modify it, test it, and check the code back into git in under a half hour.
Yes, I do get this with C++, which like perl is not designed to be read.
Yes, human players can still beat second tier AIs from Facebook and universities.
But if you turned the AlphaGo Zero team on it it would dominate it in a couple months max. AGZ learned, from scratch, how to beat every human on the planet at Go every single time in 3 days.
'This is the Year of the Linux Desktop' (aka every year since 1998) finally went down the tubes when Ubuntu went insane.
We went from being able to say 'Oh yeah, just install Ubuntu' if anyone expressed curiosity (even if we were personally using Debian or CentOS or whatever) to going uhhhh.... finally we could point to Mint, but by then it was too late.
Perl is great to write - you can just bang out some shit fast and it probably works. It's certainly better than using bash or sed for the same problem space.
The hell comes if you ever have to look at what you crapped out ever again. Six months later and it's unintelligible. Or worse, you have to use what someone else crapped out.
Yes, you can write really good, really clean, really maintainable perl if you want to put in some effort - I've seen it twice in my entire career.
Funny bit: slashdot launched 20 years ago this month. So if you were reading it back then most netizens would probably consider you an old fogey.
Anyhow, I'm still doing engineering which involves lots of design, coding, and tech lead. Adamantly refused to get into line management and that hasn't been a problem since there is a real shortage of skilled engineers. My depth and breadth of experience means lots of mentoring since I haven't turned into a cranky old bastard yet.
You have to learn new things, but I like that. Another key to keeping things fresh is switching jobs about every 5-7 years or so (which I've done my entire career) - about 5 years I start getting extremely bored and itchy feet.
Not sure what the hell I'd do if I got tired of it (go die on a farm?), but it's pretty much my choice when to retire.
The anti-Amazon crowd wouldn't mind if Amazon left and took half the tech jobs. Just makes it more of a 'hipster'* / counterculture paradise like Portland and Detroit - the inner bit, which is all that counts since that's where the non-brown people are.
* Provide your own definition since nobody can agree on one.
As he said, this is the Slashdot Mile High Club.
Oral sex in the /. Mile High Club is when you notice your seatmate is using a Windows laptop and you helpfully start berating them about using Macro$haft Winblowze, she says 'F@#$ you', you reply with 'F$#% you, you filthy vagina haver', and then one of you has your seat changed in the afterglow.
The code and techniques look like APT17 aka DeputyDog - hacking into tech firms, military and governments for the Chinese government for at least 10 years.
They realized CCleaner was a fantastic indirect vector into a whole lot of firms, and god knows what else they've got their fingers in that people haven't noticed since most firms are Equifax level incompetent with security.
Same reason we still have open offices even though they're hell for productivity (unless your job is just talking to people all day). Same reason people buy lottery tickets. Everybody knows better but want to believe they're the geniuses who can make it work this time for sure.
The UI people want to believe that if their design is just beautiful and elegant ENOUGH that it'll automatically have amazing usability. And if course it won't. They're no longer UX people, they're graphic designers.
'there is zero downside' - Absolutely, completely Wrong
'lower performance' - Wrong (depending on what you're doing)
I'm running 64-bit on my desktops because they've got tons of memory, but...
Besides 32-bit CPUs requiring 32-bit OS (which other people have covered extensively), if you've got a 64-bit CPU but only 512MB or 1GB of memory, like a Beaglebone Black or RasPi3, then running 32-bit OS and apps can save you quite a lot of memory and disk space because more things can be allocated in 4 byte chunks rather than 8 byte chunks.
Let's take a look at the venerable Notepad++, version 7.3.3. The 32-bit exe is 2.32mB and the 64-bit exe is 2.781mB. That's not 'omg', but it's 20% larger and that adds up over all of Windows's exes and anything you install. If I load langs.model.xml (about 281kB) then the 32-bit version uses 13.2MB of memory and the 64-bit version uses 14.7MB of memory, which is 11.4% more. Again, certainly not a doubling, but when you've only got 512MB of RAM this really adds up.
The 'lower performance' thing is hogwash too. If you're doing heavy 64-bit math then the 64-bit app is probably going to win. If you're doing lots of disk reads/writes and string processing the 32-bit app may well win (slightly) just because you're slinging less data around.
I guess I know who not to use for cloud backup.
Each of these guys has his own little corner of the world and is totally ignorant of what's going on elsewhere.
Yes, in my industry we've replaced C++ (and good riddance) with C# on the PC side, but C is still going extremely strong on the lower end because when you've got a (by modern standards) little SoC with 48K of 'ROM' and 34K of RAM you can get a lot more out of C than you can Mono or Java. We even still occasionally bust out the asm, but for the most part modern C (with things like named structure initializers) is the sweet spot.
And people have been predicting the replacement of coders by AI for 30 years - it's like fusion power. I imagine at some point the mere code pigs (coders who just search stackexchange and copypasta code) might be replaced, but good luck replacing hands on problem solvers for a while.
But then, who am I to argue with a 'job leadership consultancy' *snicker*.
People working on COBOL aren't 'cowboys'. They're (metaphorically, like the cowboy thing) guys with suits and briefcases who might be briefly exuberant by getting a martini dirty style and hoping nobody notices this outrageous whimsy.
Yeah, I know, there's a company named that. Go look at the employee pics.
They're still looking at Apache Pass for 2018, but it's dead as a product /now/ for Skylake-EP/Purley and they're playing it down. At IDF 2015 it was 'Apache Pass Apache Pass Apache Pass' and at IDF 2016 it was 'Uh, yeah we'll have DIMMS', and now it's 'Wow, this is great for SSDs.'
I'm not surprised they brought some DIMMs to show (it's not dead dead, just not now), but I'm genuinely curious about this: Ask when you can buy some.
It's not great for reads though - Even the consumer m.2 SSD in my PC is faster for that. It's just really fast on small random writes, or small reads under LOW read/write load (small queue depth) because it really wins for latency there. So you have to be using it on something with lots of writes and low load small reads to get amazing benefit over current flash - and then it burns out fast.
I'm sure there will be some very targeted applications where this is perfect, like maybe mostly data acquisition and high speed db updates then at the end of the week you pop it out and put in a new one and you've got your snapshot and can go analyze the data on that.
That's wrong, and it's not your fault, it's the article's fault. They bought Intel's weaseling about this.
They don't overprovision for /performance/ (which Intel will continually remind you), they overprovision for /endurance/.
The only way they hit their stated life numbers on a 375GB drive is to put 448GB of XPoint memory on it, which you can see by opening one up.
Yeah, because it's not true. It doesn't work well as cache, and if you tried it you'd burn it out fast. It was SUPPOSED to, they promised a lot that you could use this as both storage as main memory, but they couldn't make it work and even Intel has stopped talking about Apache Pass.
Basically, this is a evolutionary step in flash, giving you much faster random write and maybe 2x the endurance for 2x the price, so not bad as a disk.
But it's not the revolutionary step Intel was promising.
Modern programmers have no idea how a linked list or a hash table are stored how, how sorts work, how garbage collection and memory allocation actually works...
This was really driven home to me when one of the Java guys was trying to look at some C code and thought the linked list stuff was a bug he'd found. 'It's just a reference back to itself!'
I have to admit that mostly it doesn't matter, but occasionally their assumptions of infinite memory and infinite processing power lead them to do some really dumb things that are inconceivable to anyone who actually knows how things work under the hood. I've given some of their Java code a 1000x speedup just by the simplest optimizations like not reading the entire file every time to process a single line.
'unlimited' data plan.
This is especially bullshit because they HAD an unlimited data plan years ago and have spent the entire time since trying to kill it.
They're just freaked out that T-Mobile is now cheaper AND has a better network.
As amusing as it is to contemplate the douchepocalypse an Apple vapor would entail, the patent is clearly for chip manufacturing and the guy who filed it is one of Apple's chip people.
Remember, Apple is very big into custom silicon, and like everything else they want control of every step of it and that includes how the chips are made.
Yeah, Ruby is nice - the magpies were not its fault, or having its star hitched to Rails so tightly.