Blurgle is my Trigger

From DCTVpedia
Jump to navigation Jump to search
Blurgle is my Trigger
Number 2794
Broadcast Date JUNE 24, 2016
Episode Length 43:19
Hosts Tom Merritt
Guests Darren Kitchen, Len Peralta

The UK has voted to leave the European Union. Does it matter to the tech consumer? Darren Kitchen and Tom Merritt talk about why it doesn’t, but when it will. Plus Len Peralta illustrates!

Guest

Top Stories

Apple announced it is discontinuing its Thunderbolt Display. Apple will sell out current stock through their stores. The company pointed users to 3rd party monitors going forward, as no replacement model has been announced. Ars Technica previously reported on rumors that Apple would release 5k monitors driven by an external GPU later this year. BuzzFeed managing editor John Paczkowski says sources telling him a next-gen display with an integrated GPU is still coming.
Oculus updated its hardware-specific runtime Friday and removed the DRM code that prevented games from working with other headsets like the HTC Vive. The change was not mentioned in its update notes. The developer of Revive, which enables other headsets, noticed the change and posted a note on the project’s Github page. Revive now no longer disables game DRM to provide compatibility. Oculus confirmed to Ars Technica that it "will not use hardware checks as part of DRM on PC in the future.”
Submitted by Kylde
In a blog post yesterday, YouTube announced live streaming is coming to their mobile app. The live stream will feature comments overlayed on the video stream and users will be able to search live video. The feature has rolled out to select creators attending VidCon, with wider availability coming "soon." Youtube also announced a Creator Hub with a benefits program and a promise to have actual people respond by email to Creator support questions within one business day.
Uber is testing a new version of its app that presents a dynamically priced fare when users put in a destination. If surge pricing is in effect text under the fare will note ‘increased demand’. Previously Uber displayed a lightning bolt on the amount fares were multiplied by. Uber began upfront pricing in UberPool and has been testing it for UberX in five Indian cities and six in the US since April. Uber expects to roll out the changes to the rest of the world over the next few months.
In a cool proof of concept researchers demo, Ben-Gurion University of the Negev have been testing ways to exfiltrate data from air-gapped computers that don’t have speakers, using the cooling fan. Helpnetsecurity says they call attack the fansmitter. Software controls the fan’s speed in order to control the acoustic waveform coming from the computer. They demonstrated the transmission of keys and passwords at distances up to 8 meters at a bitrate of 900 bits/hour. That’s 1/4 of a bit per second. Of course the trick is getting the software on the target machine in the first place.
Wired’s Davey Alba has a story on how Amazon Echo is winning the race to a screenless future. The Echo passed more than 1,000 skills this month. Skills are third party app-like things the Echo can do like turn on lights, read RSS feeds etc. The key advantage for Amazon is AWS which means Echo only has to run a small amount of code, and the rest of it is cloud powered. It also has been collecting data since 2014 making its machine learning algorithms better.
Revisions to Privacy Shield, an agreement on data transfer between the EU and the US, was sent for review to European member states Thursday night. A vote is expected in early July. The previous safe harbor system was struck down by a European court over concerns with US surveillance practices. The new revision provides more detail on the conditions under which US intelligence services collect and safeguard data.

Discussion

Pick of the Day

I'm a fan of creative magic systems in fiction, and the novel Nameless by Matthew Rossi delivered that in an entertaining urban fantasy setting peopled with characters who were pleasantly not one-dimensional (including villains whose motivations weren't comic-book-thin).


Cheers,
Submitted by KiwiRed from T-Shirt-Ready-Mid-Winter, Christchurch NZ

Messages

IT’S AN ODD QUESTION

Josh compares it to a grenade. You might fall in it to save people but you wouldn’t buy one to jump on.

MAYBE EACH CAR COULD HAVE THE OPTION

Joe: What if that was an option and we got to choose when we were calibrating your car?

RJ agrees “The decisions of self driving cars is done in software so there is no reason that the choice made to protect the occupants of the vehicle at all costs can't be a user selected mode.”

DEPENDS ON WHO IS IN THE OTHER CAR

Ron: Who is responsible if somebody thinks the "system" made (or has a pattern of making) the wrong choice, because someone will likely claim the "wrong" party was saved.

SHOULD'T BE THE PRIORITY ANYWAY

Jake: At least for now, their priorities are reducing the risk factors in ever getting into a situation where moral decisions are needed


BUT THAT’S NOT HOW IT WORKS

Alan: I think maybe at best you can say they may have to program a tendency. can you tell me what the car will do? And the answer will be: if you tell me the readings on these hundreds of sensors and the car's driving history, and thousands of factors from other cars in the network…


I'm a software engineer with quite a lot of experience in machine learning (ML). I just listened to the discussion of self driving cars in ep 2793 and the post-show.

I don't think an engineer will need to explicitly program the car to save passengers or pedestrians. Instead it's behaviour in new situations will be an emergent property of its ML model.

It's not possible to hard code responses to every situation. So the ML model will be trained on a large data set of situations, and told how to behave in each case. Then in the real world, when it's being bombarded with situations that don't fit nicely into any of the training examples, it will do its best to interpolate the correct behavior from what it knows.

I suppose it's possible that one of the training examples involves choosing between saving 10 pedestrians or one passenger, and that it has been told the 'correct' behavior in that specific case. But there are always going to be cases that the ML model hasn't seen, and in these cases the model itself will have to make the decision.

Realistically I think it's more likely that when the car is loosing control, the model will just keep trying to regain control, and never really make that sort of decision.
Sent by Liam in barometrically stochastic Sydney


First, most drivers believe that UberPOOL is a ripoff, in part because of the guaranteed fares; even at published UberPOOL rates for Los Angeles County, the quoted fare always comes out to less than what it would cost to just run the "meter" normally, sometimes significantly so. I don't think they'd take too kindly to guaranteed fares for all Uber services that are less than what they should be.

Secondly, expect there to be some friction between drivers and passengers once a request to make a quick stop at a Taco Bell or 7-Eleven gets turned down because the driver is no longer being paid to deviate the few blocks it would take to get there, or wait the few minutes it would take to make a purchase. Trips with multiple stops could get even messier.

On the positive side, this could cut down on the number of passengers who don't put in a destination when they call for an Uber or Lyft. If you're one of those people, believe me when I say that your driver hates it when you do that.
Sent by Sekani

YouTube

Links



Preceded by:
"The Car-bayashi Maru"
Blurgle is my Trigger
Followed by:
"Feel the Learn"