Watch this news being used to spin that the politicians solved the problem and it's all good now. Please ignore the half billion dollars they handed out with no tender, see it was totally a good idea, it fixed the reef before they even spent it!
They cover this in the paper, under Monte Carlo rendering. Based on the timings they report, a trained CNN was producing results close to equivalent to a much higher ray count, for a low count, real-time Monte Carlo rendered scene. 2000 times faster. I don't know what current de-noising for Monte Carlo rendering looks like, but this is quite interesting. I've also seen some work combining CNN with RNN/LTSM that might also apply to this.
It's been a few years, but my last experience in contracting like this was that the recruiter MADE themselves non-optional. In order to access the recruitment service (primarily to reduce the time invested in screening applicants), companies have to agree to make the payments to the recruiter. Advertisements for positions are made by the recruiter and don't include details of who the actual company is, until after you're under contract.
Additionally, companies and contractors are in breach of contract if they tell each other what they are paying/being paid by the recruiter. My last employer "accidentally" let me find out this information, and suffice to say the recruiter was making a very nice profit. AND you're required to not work for the employer directly for a period of time after the contract ends.
Our PM is so useless, that him publically supporting it was probably what killed it. The opposition would oppose it on principle, and the factions inside his party would oppose anything he does that they haven't told him explicitly to do.
I understand there is also some sillyness involving UPNP in some devices, so you can connect to the device "from your phone", as in, from the wider Internet. This probably includes the initial connection brokered through a central service, but much of the bulk data via direct connection.
OAuth2 isn't uselessly complicated. OAuth (version 1) was, because they wanted to not require HTTPS, but wanted all the security mechanisms HTTPS would have provided. OAuth2 requires HTTPS, and removed the complex handshaking required in version 1.
Reading through the published paper, it's a flaw with the implementations, not the protocol itself, which is reassuring. It can be fixed by adding the missing checks, rather than having to replace OAuth2.
Interestingly, this requires DRM (dialling home to validate keys periodically). So if the company does the "good" thing and releases their game without DRM, they are set up for exactly this kind of rip off.
I agree the screen tearing issue with the Pebble OG really sucks (although to be fair, it's plausible it didn't show up in QA, since it's an interaction with the case slightly changing shape after multiple warm/cool cycles). Given that it's been fixed with the Time models, I assume it's also fixed for the 2's.
Fortunately for me, I'm in Australia, where it's law to have a minimum 2 year warranty.
At this point I think Amazon considers actual books a sideline. Textbooks, on the other hand, are more of a required artefact to prove you have made your sacrifice at the alter of consumerism, than something intended to share knowledge.
I found out that the text chat in Doom used a standard IPX broadcast mechanism - when my father (a network engineer) came in and told me off for my choice of language.
It might depend what country you're in, but here in Australia they already have adverts up to 30 second long (usually 15) that are unskippable. I don't know what proportion these are in comparison to skippable adverts, but I -wish- they'd go down to 6 seconds long.
(Yes, I have an adblocker, I leave it off for Youtube because ad impressions = income for the creators of the series I'm watching.)
I'm waiting to see if/how this affects Pebble Time. We've been wanting access to the Google Voice API for ages now. Personally I want it mostly for Google Now integration, which may or may not be separate.
Thank you, I didn't know about Safe. At one point I had to recover a backed up EncFS and ended up using a Linux VM to mount the encfs then copied everything out unencrypted.
I honestly thought the Liberal Democratic Party was part of the "Liberal Party", since what is referred to as the Liberal Party is actually a coalition of 4 (I think) different parties with names of a similar structure.
And they sure as heck aren't liberal in policy, either. In fact they are the more conservative of the two main parties.
Actually they already use GPUs - and there are companies making ASICs now. Dedicated Bitcoin mining boxes. The people who purchased GPUs specifically with mining in mind are apparently already annoyed, because the new computational power coming online means they are seeing less return, due to the increasing requirements as the number of mined bitcoins increases.
I'm not a fan of white noise, but at various times and moods I've found other noise types (pink and brown in particular) to be effective. I've used http://simplynoise.com/ in the past as the generator - combined with a set of good headphones, it will block out most sound.
Another alternative I use is orchestral music - specifically, no vocals. This makes it less attention grabbing. But I can't use it when I am trying to be creative/problem solving, for some reason.
Actually, Valve is mentioned in the article as one that people wanted investigated - but not as one that required a subpoena to provide information. This suggests that Valve voluntarily told them how their pricing works.
Which, as far as I know, is "We set what price the producer wants us to, or they refuse to sell on Steam at all."
You're right, it's not just CSG - although the system does contain CSG-esque boolean operations.
Our currently live product (Symvol) shows how our software does keep a tree - and we use point evaluation for producing output. MeshUp is going to be more targeted/simplified to use, so the tree possibly wont be shown/will be an advanced view.
Disclaimer: I work for Uformia, the company running the Kickstarter.
It's definitely marketing hype, but that's the kind of thing marketing is about.
There's kind of two processes being described in the product. Mesh repair is working with polygons and adding/modifying the polygons to produce a better mesh. This is intended to make better meshes for the second part, which is mesh mixing. Mesh mixing is functional, so can be sampled at any resolution.
Disclaimer: I work for Uformia, the company running the Kickstarter.
No, it's boundary is just a side effect of the definition. We use zero value to be the boundary, but that's just a convenient convention.
NURBS and other parametric surfaces still have limitations. It's very difficult to define complex shapes with them, and the boolean operations often break, leaving you with gaps between patches or surfaces that have no matching other side. The staff and beta testers include people experienced with polygon and NURBS, and there are things people are used to not being able to do (or have to do certain ways) with NURBS that "just work" with this system.
Functional objects aren't new, by any means (they predate polygons in fact). We're using some new functions (rather then the boolean ones used in existing CSG modelers).
3D printing resolution IS limited by the polygons. If you try to put a mesh into any existing printing software that has individual polygons sized at the precision of the motors, the software does break (and the mesh file will be measured in gigabytes).
And yes, "resolution of nature" is a marketing phrase. Don't ask me, I just work here.
Disclaimer: I work for Uformia, the company running the Kickstarter. We have an existing product (which isn't mention in the summary, but is in the Kickstarter) but we want to make a more specific user friendly version.
G-Code is closer to the role PostScript fills in 2d printing, but it's only loosely standardized, and every printer seems to need their own sub dialect of it. That's why there's still a common interchange format being used, with printer specific software/settings being used to produce the G-Code that actually goes to the printer.
We do support outputting to slice oriented formats (Bitmap and another that I've just gone completely blank on), but we don't use these internally. Instead, our software is using 3D functions. There is existing software that does this (CSG modellers) however we use both a different set of functions to those traditionally used, that have better properties for smooth objects (in our opinion).
Being mathematical, they are accurate to whatever numerical precision is used, so we can produce slices (our preferred format) or meshes at whatever resolution is wanted, as appropriate for the printer.
(The Fine Summary doesn't include that we already have an existing product based on the technology, but we want to make a more targeted version that doesn't require the customer to also have the Rhino 3D modelling software.)
The trouble is, everyone working with the new media is trying to bring the old workflow directly over. For example, the generic file format for 3D printing models is STL - a truly terrible file format that is a simple list of free floating triangles (no shared vertices or anything nice).
We (the people are Uformia) aren't really aiming for visual modelling. It's the printing we're aiming for. And we think just adding more and more complex methods of describing surfaces isn't enough, especially with the new printers coming out that combine materials.
Disclaimer: I work for Uformia, the company running the Kickstarter.
Slashdot Mirror
Watch this news being used to spin that the politicians solved the problem and it's all good now. Please ignore the half billion dollars they handed out with no tender, see it was totally a good idea, it fixed the reef before they even spent it!
They cover this in the paper, under Monte Carlo rendering. Based on the timings they report, a trained CNN was producing results close to equivalent to a much higher ray count, for a low count, real-time Monte Carlo rendered scene. 2000 times faster.
I don't know what current de-noising for Monte Carlo rendering looks like, but this is quite interesting. I've also seen some work combining CNN with RNN/LTSM that might also apply to this.
It's been a few years, but my last experience in contracting like this was that the recruiter MADE themselves non-optional. In order to access the recruitment service (primarily to reduce the time invested in screening applicants), companies have to agree to make the payments to the recruiter. Advertisements for positions are made by the recruiter and don't include details of who the actual company is, until after you're under contract.
Additionally, companies and contractors are in breach of contract if they tell each other what they are paying/being paid by the recruiter. My last employer "accidentally" let me find out this information, and suffice to say the recruiter was making a very nice profit. AND you're required to not work for the employer directly for a period of time after the contract ends.
Our PM is so useless, that him publically supporting it was probably what killed it. The opposition would oppose it on principle, and the factions inside his party would oppose anything he does that they haven't told him explicitly to do.
I understand there is also some sillyness involving UPNP in some devices, so you can connect to the device "from your phone", as in, from the wider Internet. This probably includes the initial connection brokered through a central service, but much of the bulk data via direct connection.
OAuth2 isn't uselessly complicated. OAuth (version 1) was, because they wanted to not require HTTPS, but wanted all the security mechanisms HTTPS would have provided. OAuth2 requires HTTPS, and removed the complex handshaking required in version 1.
Reading through the published paper, it's a flaw with the implementations, not the protocol itself, which is reassuring. It can be fixed by adding the missing checks, rather than having to replace OAuth2.
Interestingly, this requires DRM (dialling home to validate keys periodically). So if the company does the "good" thing and releases their game without DRM, they are set up for exactly this kind of rip off.
I agree the screen tearing issue with the Pebble OG really sucks (although to be fair, it's plausible it didn't show up in QA, since it's an interaction with the case slightly changing shape after multiple warm/cool cycles). Given that it's been fixed with the Time models, I assume it's also fixed for the 2's.
Fortunately for me, I'm in Australia, where it's law to have a minimum 2 year warranty.
At this point I think Amazon considers actual books a sideline. Textbooks, on the other hand, are more of a required artefact to prove you have made your sacrifice at the alter of consumerism, than something intended to share knowledge.
I found out that the text chat in Doom used a standard IPX broadcast mechanism - when my father (a network engineer) came in and told me off for my choice of language.
It might depend what country you're in, but here in Australia they already have adverts up to 30 second long (usually 15) that are unskippable. I don't know what proportion these are in comparison to skippable adverts, but I -wish- they'd go down to 6 seconds long.
(Yes, I have an adblocker, I leave it off for Youtube because ad impressions = income for the creators of the series I'm watching.)
Electric semi's are already in development and testing. E.G. http://www.popularmechanics.co...
Presumably this concept is aimed more for suburban areas, where hopefully semi's aren't a significant portion of the traffic.
I'm waiting to see if/how this affects Pebble Time. We've been wanting access to the Google Voice API for ages now. Personally I want it mostly for Google Now integration, which may or may not be separate.
Thank you, I didn't know about Safe. At one point I had to recover a backed up EncFS and ended up using a Linux VM to mount the encfs then copied everything out unencrypted.
It's still the goal of PNaCl variant of Native Client ( https://developer.chrome.com/n... ). But I haven't heard of anyone actually using this.
I honestly thought the Liberal Democratic Party was part of the "Liberal Party", since what is referred to as the Liberal Party is actually a coalition of 4 (I think) different parties with names of a similar structure.
And they sure as heck aren't liberal in policy, either. In fact they are the more conservative of the two main parties.
Actually they already use GPUs - and there are companies making ASICs now. Dedicated Bitcoin mining boxes. The people who purchased GPUs specifically with mining in mind are apparently already annoyed, because the new computational power coming online means they are seeing less return, due to the increasing requirements as the number of mined bitcoins increases.
I'm not a fan of white noise, but at various times and moods I've found other noise types (pink and brown in particular) to be effective. I've used http://simplynoise.com/ in the past as the generator - combined with a set of good headphones, it will block out most sound.
Another alternative I use is orchestral music - specifically, no vocals. This makes it less attention grabbing. But I can't use it when I am trying to be creative/problem solving, for some reason.
Actually, Valve is mentioned in the article as one that people wanted investigated - but not as one that required a subpoena to provide information. This suggests that Valve voluntarily told them how their pricing works.
Which, as far as I know, is "We set what price the producer wants us to, or they refuse to sell on Steam at all."
You're right, it's not just CSG - although the system does contain CSG-esque boolean operations.
Our currently live product (Symvol) shows how our software does keep a tree - and we use point evaluation for producing output. MeshUp is going to be more targeted/simplified to use, so the tree possibly wont be shown/will be an advanced view.
Disclaimer: I work for Uformia, the company running the Kickstarter.
It's definitely marketing hype, but that's the kind of thing marketing is about.
There's kind of two processes being described in the product. Mesh repair is working with polygons and adding/modifying the polygons to produce a better mesh. This is intended to make better meshes for the second part, which is mesh mixing. Mesh mixing is functional, so can be sampled at any resolution.
Disclaimer: I work for Uformia, the company running the Kickstarter.
Without getting into too much detail:
No, it's boundary is just a side effect of the definition. We use zero value to be the boundary, but that's just a convenient convention.
NURBS and other parametric surfaces still have limitations. It's very difficult to define complex shapes with them, and the boolean operations often break, leaving you with gaps between patches or surfaces that have no matching other side. The staff and beta testers include people experienced with polygon and NURBS, and there are things people are used to not being able to do (or have to do certain ways) with NURBS that "just work" with this system.
Functional objects aren't new, by any means (they predate polygons in fact). We're using some new functions (rather then the boolean ones used in existing CSG modelers).
3D printing resolution IS limited by the polygons. If you try to put a mesh into any existing printing software that has individual polygons sized at the precision of the motors, the software does break (and the mesh file will be measured in gigabytes).
And yes, "resolution of nature" is a marketing phrase. Don't ask me, I just work here.
Disclaimer: I work for Uformia, the company running the Kickstarter. We have an existing product (which isn't mention in the summary, but is in the Kickstarter) but we want to make a more specific user friendly version.
G-Code is closer to the role PostScript fills in 2d printing, but it's only loosely standardized, and every printer seems to need their own sub dialect of it. That's why there's still a common interchange format being used, with printer specific software/settings being used to produce the G-Code that actually goes to the printer.
We do support outputting to slice oriented formats (Bitmap and another that I've just gone completely blank on), but we don't use these internally. Instead, our software is using 3D functions. There is existing software that does this (CSG modellers) however we use both a different set of functions to those traditionally used, that have better properties for smooth objects (in our opinion).
Being mathematical, they are accurate to whatever numerical precision is used, so we can produce slices (our preferred format) or meshes at whatever resolution is wanted, as appropriate for the printer.
(The Fine Summary doesn't include that we already have an existing product based on the technology, but we want to make a more targeted version that doesn't require the customer to also have the Rhino 3D modelling software.)
The trouble is, everyone working with the new media is trying to bring the old workflow directly over. For example, the generic file format for 3D printing models is STL - a truly terrible file format that is a simple list of free floating triangles (no shared vertices or anything nice).
We (the people are Uformia) aren't really aiming for visual modelling. It's the printing we're aiming for. And we think just adding more and more complex methods of describing surfaces isn't enough, especially with the new printers coming out that combine materials.
Disclaimer: I work for Uformia, the company running the Kickstarter.