Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
Check out the new SourceForge HTML5 internet speed test! No Flash necessary and runs on all devices. ×

Comment What's with the shitty headlines lately? (Score -1, Flamebait) 257

It seems like every other story hitting the front page lately is challenging the Betteridge law of headlines. It makes for shitty discussion because it's turning a news item into a binary proposition. There's not a lot of room for expansion or commentary.

Comment Re:FP16 isn't even meant for computation (Score 1) 55

So, one problem is that there is not always more data. In my field, we have a surplus of some sorts of data, but other data requires hundreds of thousands of hours of human input, and we only have so much of that to go around. Processing all of that is easy enough, getting more is not.

Also, by "effective", I should have made it clear that I meant "an effective overall solution to the problem", which includes all costs of training a wider, lower-precision network. This includes input data collection, storage and processing, all of the custom software to handle this odd floating point format, including FP16-specific test code and documentation, run time server costs and latency, any increased risks introduced by using code paths in training and , etc.

I'm not saying that I don't believe it's possible, I've just seen absolutely no evidence that this is a significant win in most or even a sizable fraction of cases, or that it represents a "best practice" in the field. Our own experiments have shown a severe degradation in performance when using these nets w/out a complete retraining, the software engineering costs will be nontrivial, and much of the hardware we are forced to run on does not even support this functionality.

As an analog, when we use integer based nets and switch between 16-bit and 8-bit integers, we see an unacceptable level of degradation, even though there is a modest speedup and we can use slightly larger neural nets. I'm very wary of anything with a mantissa much smaller than 16 bits for that reason--those few bits seem to make a significant difference, at least for what we're doing. We're solving a very difficult constrained optimization problem using markov chains in real time, and if the observational features are lower fidelity, the optimization search will run out of time to explore the search space effectively before the result is returned to the rest of the system. It's possible that the sensitivity of our optimization algorithm to input quality is the issue here, not the fundamental usefulness of FP16, but I'm still quite skeptical. If this were a "slam dunk", I'd expect to see it move through the literature in a wave like the Restricted Boltzmann Machine did.

Oh, and thank you for the like (great reading) and the thoughtful reply. Not always easy to find on niche topics online.

Comment Re:Exploitative by design? (Score 1) 153

It seems like these systems are exploitative by design, even if exploitation wasn't explicitly the goal. They're designed with every possible algorithm and available data to maximize labor output at the lowest possible cost. Individual workers are operating at extreme information asymmetry and against a system which does not negotiate and only offers a take it or leave it choice.

This is by far the best comment I've ever seen regarding this sort of algorithmic labor management.

Normally I'm all for this sort of thing--my company is a client and uses it to handle large bursts of data processing quickly--but the information symmetry argument is a powerful one. Also, there doesn't seem to be a lot of competition in this space, which might otherwise ameliorate a lot of the problems induced by the "take it or leave it" bargaining approach.

The analysis provided by the article is absurd, but yours seems to lead to the inescapable conclusion that some kind of regulation is necessary to prevent blatant exploitation. Maybe just reducing information asymmetry in some way, or requiring transparency in reports available to the public on the website regarding effective wages paid to workers as a fraction of the minimum and average wages of employees in their respective countries. Surely someone can find an answer to this.

Comment Re:FP16 isn't even meant for computation (Score 1) 55

Accidentally posted as anonymous coward, reposting under my actual name.

So they're all excited about the lowest-precision, smallest-size floating point math in IEEE 754?

FP16 is good enough for neural nets. Do you really think the output voltage of a biological neurons has 32 bits of precision and range? For any given speed, FP16 allows you to run NNs that are wider and deeper, and/or to use bigger datasets That is way more important than the precision of individual operations.

There's a lot of rounding error with FP16. The neural networks I use are 16-bit integers, which work much, much better, at least for the work I'm doing. Also, do you have a good citation that FP16 neural networks are, overall, more effective than FP32 networks, as you've described?

Comment Re:Sorry but (Score 2) 89

Why wouldn't Java be something a forward thinking CTO would be using? Java's been the intro CS language for a large number of schools for a few decades and the fact it's a "legacy" language means there's a huge developer pool and well tested tooling available. It's also not tied to the Windows platform so it's largely free from platform lock-in.

As for Apache's relevance...the Hadoop platform is huge and not going away any time soon. Lucene is the basis for two of the most popular enterprise search systems, one of which (Solr) is also an Apache project and written in Java.

Comment Fetch my fainting couch (Score 2) 310

Are declines in PC sales in any way surprising? Frost the past decade and a half a larger and larger portion of PC sales have been laptops. Schools from junior high through college practically (or actually) demand them. The proliferation of WiFi means just about anywhere with a roof is going to offer some internet connectivity. Besides ubiquitous internet access laptops have gotten way more consumer friendly by getting ever cheaper and lighter. For just about everyone a laptop is the form factor to buy.

For most of the past 15-16 years laptops were getting faster CPUs or way better GPUs every two years or so. Battery life didn't improve much but at least the machines got more powerful. The past 5-6 years though the landscape has changed. Fewer laptops ship with discrete GPUs as Intel's have increased in capability. Even low end laptops have SSDs and 8+ gigabytes of RAM. The usable lifespan of laptops has increased significantly. Even a change from an average of two to three years means fewer sales for manufacturers. There's a non-trivial portion of the laptop market that's seeing a replacement cycle of over three years.

In addition the sort of things people needed a laptop for ten years ago can be as effectively or more effectively done on a phone or tablet. Android and iOS tablets beat the shit out of Windows tablets and 2-in-1s because hey aren't saddled with a heavyweight OS that honestly is not designed to turn on and go and then back off just as easily.

Billions of smartphones and many millions of tablets have definitely sucked the oxygen out of the room for traditional PCs. With PCs not "needing" more regular upgrades is choking the PC industry. The PC market is saturated and is not likely to grow again. Emerging markets are not a savior because they don't have the same infrastructure as developed markets. They aren't going through a dial-up landline internet connected to a beige box phase. They're going right to smartphones, tablets, and other highly mobile devices that fit better in their infrastructure.

Comment Re: VR will help --- maybe (Score 1) 310

Unfortunately comes with a hardware dongle that's not really advertised. It's a room with enough free space to not break break a limb or get a concussion flailing about in a physical environment that doesn't match the one being presented to your eyes and ears. Kids and pets are also incompatible accessories.

If VR ever gets a non-trivial adoption the Wii-mote and Kinect injuries of yesteryear will seem quaint. We'll look back on the halcyon days before our TBIs from running into furniture or ducking below a virtual missile into the corner of the desk.

Maybe VR injuries will finally get more developers to support button remapping or other accessibility features.

Comment Re: Hello Wine (Score 2) 585

I've done native code on Windows in industrial safety and automation. You'd think that's an oxymoron, but it can be made sufficiently robust.

I've dealt with bugs in Microsoft's SDKs, and dealt with multiple generations of drawing APIs. Played WoW and other games on WINE on Gentoo. Watched the incessant scrolling of FIXMEs on the console.

I'd love it if I could get paid to hack on WINE...

Comment Re: So Slashdot is a blatant propaganda peddler no (Score 2) 496

Forcing labor intensive manufacturing to on-shore won't magically create a domestic blue collar workforce.

Say a company has a widget that costs a dollar to be manufactured in China and shipped to the US. That same widget if manufactured in the US today costs two dollars.

Adding tariffs such that the Chinese price would equal or exceed the domestic manufacturing cost would in theory incentivize domestic production. What it would do in reality is incentivize investments in automation to reduce the domestic production cost to any point below the two dollar mark. Removing the cheap option will just make companies move to the next cheapest option not just jump to the most expensive option.

Labor intensive low-skill production happens in places where the labor cost is low. There's no incentive in having human beings doing the work unless they are cheaper than machines.

Trying to force labor intensive manufacturing to return to developed first world countries will just hasten the adoption of automation. This will mean output and profit margins won't change for manufacturers and the number of manufacturing jobs will remain constant or decrease. Robots have less management overhead than humans and can be retrained for new positions much faster.

Comment Re:Q n A (Score 0, Redundant) 141

Unfortunately quoting the top played Steam games for the day does not really say what you think/want it to say. Of those top seven games, four share the same game engine (CS:GO, DoTA2, Gary's Mod, and TF2) and honestly can be looked at as mods for HL2. So your "Top 10" list is really a "Top 7" list. Of those seven games only a little better than half are available for Linux.

The second problem with those numbers is availability for Linux does not give any information about the number of those players running Linux. If you apply the Linux stats from the Hardware/Software survey you've got Linux at 0.84% of the Steam installed base. A full quarter of the Mac installed base.

You also can't simply point to Steam's stats and suggest Linux games are plentiful. The most obvious omission for those stats are competing platforms/publishers like EA and Blizzard. Since there's effectively zero Linux support for Origin or Battle.net using Steam's figures is a bad extrapolation of the games market. You're trying to pretend that Steam == the game market and the games the GP mentioned aren't by some metric "top" games.

I'd love Linux to get better support from game developers/publishers. I also really appreciate the fact Linux gaming today is far better than it has ever been in the past, including during Loki Software's heyday. It's a bit ridiculous though to point out obviously wrong or grossly misleading figures and make proclamations that Linux gaming is awesome and all the top games are available.

Comment Re:Declaring code is docs, but Android screwed Sun (Score 2) 405

That all being said, I hold an unpopular opinion. What Google did should be techinically legal, and it should obviously be possible to develop compatible implementations of operating systems and other software infrastructures. However, Googleâ(TM)s choice to usurp the Java empire totally fucked over Sun. Android started at a time when Sun was still Sun. They were making revenue from Java, and if that revenue stream had continued, the may have been able to avoid going under. Instead, Android totally ripped the rug out from under that part of Sun, and Sun had to liquidate and get sold to to the assholes at Oracle.

So while technically, within the law, Google doesnâ(TM)t owe a penny to Oracle (in my opinion), what Google did was morally wrong, and there were consequences (surely anticipated by Google to some degree or other) that lead to Sunâ(TM)s demise.

Put the brakes on that line of thought right there. Google tried to work with Sun in the beginning but could not come to terms. Whether Google wanted the Moon from Sun is immaterial, Sun had the opportunity to get the branded Java platform running on Android. My guess is Sun wanted Google to mate themselves to either J2ME or Swing at the UI layer and Google didn't like either of those options.

So at Android's inception we have Sun setting the stage for Sun getting fucked over. In addition Google used the Apache Harmony project because Sun at the time did not offer an Open Source implementation of the entirely of the Java Class Library.

If Google was "morally" wrong to use Apache Harmony then Apache themselves was "morally" wrong for writing it in the first place. Likewise Red Hat's IcedTea project is morally wrong as is GNU Classpath. I really don't want to have to defend the actions of Google but it is absurd to accuse them of being "morally" wrong in making a business decision for their platform.

Sun's worst enemy was often Sun. They had a specific vision for how they wanted Java to exist on mobile platforms and did not want to waver from it. Their vision was not compatible with what the market actually wanted. Google wanted to satisfy market demands, not Sun's vision for mobile devices.

Comment Re:"Habitable Zone" (Score 1) 267

Who are WE to determine that life has to be like US.

Your question comes up in one form or another every time this subject is discussed. It's not because it is insightful (despite the moderation) but because the questioner fails to think logically.

Firstly the question makes a logical leap by presupposing that it is life on Earth is a rare form in the universe. It suggests that forms of life completely alien in understanding exist throughout the universe and it is the Earth that is the odd ball by using carbon and water as the backbone of biological processes.

For the cosmic awesomeness that is our home planet and solar system, it's pretty average in a large number of ways. The Sun isn't super unique in many respects nor are the elements on which it and the Earth are composed. There's not a whole lot of Unobtanium or Raretonium to be found on Earth. We might be the only life in this part of the galaxy but that's likely not because the Earth and our solar system is especially unique on the galactic scale.

We'll miss extraterrestrial life because we were looking for ourselves the whole time.

The second major problem with the question is the scope is amazingly out of whack with the scope of the actual universe. Our galaxy alone has hundreds of billions of stars. If we could take a galactic census (say in just a 1kpc sphere) we would probably find millions of worlds with some form of life on them. Out of those millions your question presupposes that life forms unlike us would be the majority. Even if that was the case that still leaves a great many life forms that are enough like us for us to recognize as life forms.

Where your question fails in scope is not realizing that just a small chunk of the Milky Way has millions of stars and very likely millions of planets. Even if a majority host life forms wholly alien to us, there'll be enough Earth-like worlds for us to find life that is not wholly alien to us.

Slashdot Top Deals

"The algorithm to do that is extremely nasty. You might want to mug someone with it." -- M. Devine, Computer Science 340

Working...