Send News. Want a reply? Read this. More in the FAQ.   News Forum - All Forums - Mobile - PDA - RSS Headlines  RSS Headlines   Twitter  Twitter
User Settings
LAN Parties
Upcoming one-time events:
Chicago, IL, USA, IL 01/27
Chicago, IL USA, IL 10/19

Regularly scheduled events

User information for SMA

Real Name SMA   
Search for:
Sort results:   Ascending Descending
Limit results:
Nickname Scottish Martial Arts
Email Concealed by request
ICQ None given.
Homepage http://
Signed On Jun 16, 2002, 23:16
Total Comments 3189 (Veteran)
User ID 13410
User comment history
< Newer [ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 ] Older >

News Comments > IL-2 Sturmovik: Cliffs of Dover Blitz Edition
4. Re: Morning Mobilization Dec 15, 2017, 07:35 Scottish Martial Arts
It's been forever since I played Cliffs of Dover so I might be mistaken, but I believe there is a "cheat" key to autostart the engine/etc. See if there's a section in the key bindings labeled Cheats.

In terms of actually starting up the engine, the manual has a checklist for what to do. It's pretty brief and straightforward but you have to do it in the right sequence and not skip anything.

P.S. The Fuel Cocks are what open the fuel valves for flow into the engine. I believe the name is related to the usage of cock that we see in phrases such as "cocked the rifle".
Reply Quote Edit Delete Report
News Comments > Out of the Blue
7. Re: Morning Mobilization Dec 9, 2017, 16:55 Scottish Martial Arts
Acleacius wrote on Dec 9, 2017, 15:17:
So I can't begin to imagine what it's going to do in NY. Better pull out those foot warmers and good luck!

This is the view from my apartment as of an hour ago. The New Yorkers around me today seemed non-plussed, but as a native Californian seeing my first New York winter, it felt like a lot of snow.
Reply Quote Edit Delete Report
News Comments > Square Enix on the Future of Deus Ex
5. Re: Square Enix on the Future of Deus Ex Nov 28, 2017, 11:23 Scottish Martial Arts
Iím very reassured to hear this. I absolutely loved Mankind Divided, although the criticism of a weak plot and characters is entirely fair. My biggest disappointment with the game however was the fear that there wouldnít be a proper follow up. If one is indeed coming, albeit on likely a 5+ year timescale, then Iím very glad to hear it.  
Reply Quote Edit Delete Report
News Comments > Evening Safety Dance
12. Re: Morning Mobilization Nov 22, 2017, 09:14 Scottish Martial Arts
yuastnav wrote on Nov 22, 2017, 07:02:
jdreyer wrote on Nov 22, 2017, 02:12:
Scottish Martial Arts wrote on Nov 21, 2017, 23:08:
We all know Cutter almost exclusively talks out of his ass but itís particularly amusing when you are on the inside of something he can only speculate about. Sadly, I canít elaborate.

Don't be a tease! Expound, please.

Yeah, give us all of this juicy insider intel. :>

Hehe, as an engineer I'm just supposed to let PR handle those sorts of questions. A few general observations, though:

Location is actually a pretty hard problem. Cellphones collect a lot of signals but at the same time they were never specifically designed to be position trackers. In theory, they can provide precise user location; in practice, it's a lot harder than that, particularly in urban environments or indoors.

Data collection is nearly entirely about improving the accuracy, precision, and efficiency of existing services, or to support the development of new ones. To the extent that data collection is malicious, it's to anonymously target you for advertising.

Some of the data that's collected are, in the abstract, creepy. In terms of its concrete use, "your" data are just structured collections of numbers in a near infinite sea, being processed by massive distributed server clusters without any observation by humans.

Along the same lines, no one can look you up and find out what's known about you. To the extent that it could be revealed in a data breach, hackers would need to steal ALL of Google's data and reverse engineer it. Given the scale at which Google operates, it's highly questionable whether any other actor has enough storage capacity to actually be able to store all of that data for later offline analysis.

As another poster alluded, the big thing you should be worried about is the fact that the technology exists in the first place. If Google, Facebook, et al., can collect "personal" data, so can criminals and government security and intelligence agencies.

Reply Quote Edit Delete Report
News Comments > Evening Safety Dance
6. Re: Morning Mobilization Nov 21, 2017, 23:08 Scottish Martial Arts
We all know Cutter almost exclusively talks out of his ass but itís particularly amusing when you are on the inside of something he can only speculate about. Sadly, I canít elaborate.  
Reply Quote Edit Delete Report
News Comments > Golden Joystick Winners
14. Re: Morning Mobilization Nov 19, 2017, 20:59 Scottish Martial Arts
Illumin wrote on Nov 19, 2017, 18:56:
Pretty damn boring list of Winners.

The new Zelda fully deserves the praise it gets and by itself justifies the purchase of a Switch. The rest of the list, however, definitely falls into the category of "I've heard good things, but nothing so good as to make me actually want to play it".
Reply Quote Edit Delete Report
News Comments > Morning Mobilization
5. Re: Morning Mobilization Nov 9, 2017, 12:43 Scottish Martial Arts
Verno wrote on Nov 9, 2017, 11:36:
I've rarely seen worse traffic than in Vegas so probably a good testing ground for self-driving vehicles.

I'm in the SF Bay Area this week for work and the traffic on 101 N, on a Sunday afternoon when I got in, erased all of my nostalgia for this place, even though I grew up around here.
Reply Quote Edit Delete Report
News Comments > Out of the Blue
14. Re: Out of the Blue Nov 5, 2017, 20:14 Scottish Martial Arts
The Half Elf wrote on Nov 5, 2017, 17:29:

But as pointed out, look at her bumper, and realize it's in NYC.

Yeah, drivers honking about stuff that no one can do anything about, like a red light, seems to be the NYC standard. I'm just glad I don't have to drive here.
Reply Quote Edit Delete Report
News Comments > Morning Consolidation
16. Re: Morning Consolidation Oct 30, 2017, 21:59 Scottish Martial Arts
Prez wrote on Oct 30, 2017, 18:19:
As for the Switch itself, I'm kinda indifferent. Do I want to play the new Zelda and Mario? Yeah. Enough to buy a whole new system for them? Not really.

Personally, I think it's worth the price of admission for those two games alone. Like Verno, Breath of the Wild is one of the best games I've ever played, and Super Mario Odyssey seems just as good after a weekend of playing it (although I suppose there is plenty of playtime left for my opinion on that to change). The only other first party game I have is Mario Kart 8, which is a ton of fun, but as a Mario Kart game it just isn't going to surprise you or fill you with wonder the way Zelda or Mario will. Add in some great indie support and you won't lack for quality content, as long as you understand that this console is going to offer different experiences than a PS4/XBOX or a PC.
Reply Quote Edit Delete Report
News Comments > Saturday Tech Bits
7. Re: Saturday Tech Bits Oct 15, 2017, 17:27 Scottish Martial Arts
jdreyer wrote on Oct 15, 2017, 14:56:
My experience with MS has revealed it to be the case. Former H1-Bs applied to positions at my company and I was part of the hiring team. We've also worked with MS tech workers on projects, many of whom were H1-B holders. And we've been told MS has hundreds and hundreds of H1-Bs working for them.

However, I don't know about Google and Apple specifically. So maybe they're different. Just a couple of questions:

1. Can everyone at Apple and Google be superstars? Don't they need a certain number of mid-level grunts to do the tedious work?

2. Given that both Apple and Google colluded to reduce salaries, why wouldn't they use H1-B status of even excellent engineers to push salaries down? Or is your contention that the very best can name their price regardless of immigration status?

It wouldn't surprise me if Microsoft has hundreds of H1-Bs working for them, but it would surprise me if they specifically hire H1-Bs over citizens/permanent residents simply to pay them less than. My experience working in one of these companies suggests H1-Bs are treated just as a means to get work authorization for foreign engineers who do not have a green card, for example a recent graduate who went to an American university and was hired out of school, but whose temporary student work authorization is set to expire. The sort of scenario that is typically conjured by mentioning H1-Bs, i.e. a department getting laid off and replaced with barely competent Indian contractors from an outsourcing firm, simply doesn't exist, or if it does, it is extremely well hidden.

1. Not everyone at Google or Apple is a super star, but everyone is a top performer for where they are in their career. A mid-level or senior engineer will surely out perform a new grad, but compared to other new computer science/engineering graduates in their cohort, the Apple or Google hire will be a far stronger performer than the, for example, Equifax hire.

2. They colluded to prevent poaching of talent. I don't mean to defend what they did, but coming to an (illegal) agreement to prevent key engineers from being headhunted to another firm is distinct from agreeing to suppress salaries. The effect on individual engineers largely manifests in lost opportunities to jump around for pay raises, but the emails that were revealed in court made it explicit that Jobs, Schmidt, Zuckerberg, Whitman, et al. were agreeing not to poach, headhunt, or otherwise hire away their engineering talent. Don't get me wrong, I in no way think these companies are altruistic, but I do think their incentives, as I understand them, lead them to prize engineering talent and to either compete strongly to secure it -- as in the case of offering generous salaries -- or in this case to engage in anticompetitive behavior to secure it.

At any rate, my contention is that these companies view engineering as the source of profitability, and not as an operating cost to be minimized by any means necessary. Likewise, they have the name brand value, generous salary, perks, and prestige to ensure that they have an essentially infinite candidate pool. Accordingly, they have very tough hiring standards that filter out the vast majority of candidates but retain those that will be most productive, which in turn means that in order to meet hiring targets they recruit globally and continuously. Consequently, they view immigration status not as something to exploit to reduce costs, but as an obstacle to be worked around for engineers who have made it through the hiring gauntlet. For some that will mean getting an H1-B or winning the green card lottery; for others it will mean transferring to an office in a country where they have work authorization. But as far as I can tell, it does not mean hiring foreigners to undercut domestic salaries.
Reply Quote Edit Delete Report
News Comments > Saturday Tech Bits
5. Re: Saturday Tech Bits Oct 15, 2017, 09:22 Scottish Martial Arts
jdreyer wrote on Oct 14, 2017, 16:51:
Or be allowed to hire H1-Bs out the ass at half salary.

I strongly doubt either Apple or Google do that. They surely both hire H1-Bs but not at half salary. Unlike Disney, or Bell Edison, or any of the other myriad companies that have famously axed their technical staff in favor of H1-B-based outsourcing firms, both Apple and Google view engineering as major profit centers, not cost centers. Likewise, both companies have an essentially infinite talent pool to draw on -- every software engineer on the planet would like to work there -- and the more they invest in engineering the more money they earn. In other words, their incentive structure is to find the best engineers on the planet, and then pay them lavishly to get them to join and to stay and to make the company money. Anecdotally, two of the engineers on my team have Chinese passports, and while I don't know their immigration status (H1-B vs. greencard) I can say for certain that they both make substantially more money than I do.
Reply Quote Edit Delete Report
News Comments > Saturday Metaverse
10. Re: Morning Mobilization Oct 1, 2017, 11:35 Scottish Martial Arts
Mr. Tact wrote on Oct 1, 2017, 07:38:
jdreyer wrote on Oct 1, 2017, 00:54:
I've come around on this. When Dems take over congress in 2018,
What in the world makes you think that's going to happen? Last time I looked at the polling, the chances of that happening were slim to none.

The generic ballot ("Would you rather vote for a Democrat or a Republican for the next Congress?") have pretty consistently showed a substantial lead for the Democrats for months now. FiveThirtyEight's polling aggregation of the generic ballot is currently showing an 8 point lead, which is consistent with generic ballot leads in the past two wave elections, e.g. the Republican takeover in 2010 and the Democratic takeover in 2006. All that said, the generic ballot is an imperfect measure and a lot can happen in 13 months, so Democrats and anti-Trump Independents and Republicans shouldn't get their hopes up or speak in certainties, particularly with regards to the Senate where Democrats are defending about three times as many seats as the Republicans are this cycle. Still, there's evidence to support an expectation of Democratic gains in 2018, potentially very large ones.
Reply Quote Edit Delete Report
News Comments > Morning Tech Bits
15. Re: Morning Mobilization Sep 28, 2017, 19:36 Scottish Martial Arts
DrSquick wrote on Sep 28, 2017, 16:39:
Can GPUs replace CPUs? My limited understanding was that GPUs were highly specific and could only do a few things but ultra fast and the equivalent of thousands of threads at once, whereas CPUs were more all-purpose.

Yeah that's basically correct. The headline doesn't really capture what the article says to be honest: Huang seems to be suggesting that GPUs will become far more prevalent in the machine learning and cloud computing spaces, which is definitely an in progress trend. Deep Learning in particular is particularly amenable to GPU acceleration, as Deep Learning models are essentially large computational graphs with many layers of computational nodes which can be run in parallel as the input traverses the graph. In fact, a new class of hardware, TPUs or Tensor Processing Units, are increasingly being added to data centers, and a TPU is essentially a GPU that's been optimized for processing tensors, i.e. vectors or matrices of arbitrary dimensionality, which is typically the means by which a Deep Learning computational graph is implemented for processing.
Reply Quote Edit Delete Report
News Comments > Saturday Mobilization
7. Re: Morning Mobilization Sep 23, 2017, 18:19 Scottish Martial Arts
WaltC wrote on Sep 23, 2017, 11:55:
For most Mac users, booting up is a "complex" endeavor...;)

Maybe, but there's a not insignificant subset of Mac users for whom a complex endeavor consists of writing a feature extractor that computes the cosine similarity of the vector space of WiFi MAC address/normalized RSSI tuples for use in a reinforcement learning ML model to infer optimal reporting rates in another module. That subset of Mac users are called software engineers, and they use a Mac because they need a Unix/Linux system in order to be able to do the aforementioned complex endeavor, and underneath the glossy exterior, MacOS is a standards compliant Unix system.
Reply Quote Edit Delete Report
News Comments > Out of the Blue
10. Re: Morning Mobilization Sep 19, 2017, 18:16 Scottish Martial Arts
Yarr, it's Talk Like a Pirate Day!

At work today, one of our command-line utilities for working with our internal distributed file system had its usually progress bar replaced with one that filled in with "Yaaarrr!". It was glorious. I'm glad engineer hours get spent on the little details that make life worth living.
Reply Quote Edit Delete Report
News Comments > Morning Mobilization
13. Re: Morning Mobilization Sep 13, 2017, 16:39 Scottish Martial Arts
Verno wrote on Sep 13, 2017, 16:04:
TouchID is fast and efficient. What is the big improvement over that system?

Same thing with the removal of the home button, whether software or hardware.

I believe the rationale in both cases is to free up space on the front of the phone for an edge to edge screen. If I remember correctly, they were working on some sort of TouchID sensor embedded in the screen itself, but couldn't get it working well enough in time for launch, and as a result they fell back on face sensing as a compromise.
Reply Quote Edit Delete Report
News Comments > Morning Tech Bits
10. Re: Morning Mobilization Sep 8, 2017, 17:35 Scottish Martial Arts
The Half Elf wrote on Sep 8, 2017, 11:10:
why would this tech even exist?

Another reason is that it's a problem/question that is almost ideally suited to machine learning. Machine learning is generally pretty good at classification problems based upon visual input. If you're a researcher experimenting with a new deep learning model, you'd want to choose a problem that you think you have a reasonable chance of actually being able to train the model to solve. And if that problem has some semi-interesting social/human implications -- i.e. that there are some facial signifiers of sexuality that are mostly invisible to humans but perceptible to a statistical model -- then all the better. It did after all get the researchers some mainstream press coverage.
Reply Quote Edit Delete Report
News Comments > Saturday Mobilization
11. Re: Morning Mobilization Sep 3, 2017, 10:54 Scottish Martial Arts
eRe4s3r wrote on Sep 2, 2017, 23:18:
Although I guess this limits what you can or should do with your phone, especially when it starts to call the cops on you for a suspected "criminal habit" o.O Or refuses to call the ambulance when you injure yourself in an illegal act...

That sort of emergent behavior isn't possible with present AI techniques. When articles written for laymen discuss "researchers being surprised at the resulting behavior" or "engineers not knowing which action the AI will take", they end up leaving the impression that your typical AI agent can take any action. In fact, an agent can only take an action within the output range of the problem being solved, an output range that has been defined by the engineers that designed the agent. Within that output range, you, as the engineer, don't know what action the agent will take: sometimes it will come as a complete surprise, prompting you to double check that, yes, everything is in fact working, and other times it will be basically what you expected. But the agent cannot define its own output range, or define the problem its meant to solve, or define the input features it makes decisions about, etc.; that's all decided by the engineers who settled on using an AI to solve whatever problem they're working on.
Reply Quote Edit Delete Report
News Comments > Saturday Mobilization
9. Re: Morning Mobilization Sep 3, 2017, 07:29 Scottish Martial Arts
Quboid wrote on Sep 2, 2017, 22:12:
Do companies typically actually want this, rather than grabbing as much data as they can?

As with all things, it depends. With some applications, they genuinely do need the data on their servers in order for the application to work and to be able to assess and improve it; with others however, it's merely been the case that up until now the most practical way to process the data is on a distributed server infrastructure. In that latter case, there is a lot of incentive to figure out a way to shift data processing to the user device and obviate the need -- and the associated costs -- to collect, store it, and process it on their own computing infrastructure.
Reply Quote Edit Delete Report
News Comments > Saturday Mobilization
3. Re: Morning Mobilization Sep 2, 2017, 19:15 Scottish Martial Arts
RedEye9 wrote on Sep 2, 2017, 15:14:
It's no longer spying on you, your habits and what you do if it's in your hand.

To a certain extent it's not: if all the data used to train the AI agent never leaves the phone, and the AI agent (whether it be a recommender or a face recognizer or whatever) is trained on the phone with that data, then the company in question knows no more about you than before you owned the device.

Machine learning on device is a big push right now for exactly this reason: if the ML model can be trained on device, then no sensitive data needs to be sent to, stored, and processed on third party servers. This saves companies money and limits the amount of sensitive data they have to handle, and potentially be liable for. The challenge heretofore has been that the limited computational resources on phones have severely limited the kind and quality of on-device training you can do, but that will probably change substantially in the next 4 or 5 years as both hardware and ML algorithms and ML engineering techniques improve.
Reply Quote Edit Delete Report
3189 Comments. 160 pages. Viewing page 1.
< Newer [ 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 ] Older >


Blue's News logo