Welcome, Guest. Please login or register.
March 28, 2024, 01:06:28 PM

Login with username, password and session length

Search:     Advanced search
we're back, baby
*
Home Help Search Login Register
f13.net  |  f13.net General Forums  |  General Discussion  |  Topic: The robots are coming 0 Members and 1 Guest are viewing this topic.
Pages: 1 ... 8 9 [10] 11 12 ... 14 Go Down Print
Author Topic: The robots are coming  (Read 214005 times)
Torinak
Terracotta Army
Posts: 847


Reply #315 on: June 28, 2016, 01:58:28 PM

The situation I worry more about is the car can decide the kill the sole occupant or 5 bystanders. Which does it choose?


Easy. It uses facial recognition to identify everyone involved, checks their social media accounts, and acts accordingly.

If it weren't for the craptons of DRM, there'd probably be a market in third-party "human worth evaluator" add-on modules.
Khaldun
Terracotta Army
Posts: 15157


Reply #316 on: June 28, 2016, 02:00:01 PM

The whole "kill the occupant or 5 bystanders" thing is largely coming from analytic philosophers and their trolley problems, not from engineers trying to figure this stuff out in practice, partially because the engineers do know that at some point you will have to have a view of what the priorities list in collision situations might be and it can't be situational in the way that human drivers like to believe (largely incorrectly) that we can be.
Count Nerfedalot
Terracotta Army
Posts: 1041


Reply #317 on: July 01, 2016, 06:37:13 PM

Or, the car could just decide the driver is too stupid to worry about any more and keep on trucking. https://www.washingtonpost.com/news/the-switch/wp/2016/07/01/the-technology-behind-the-tesla-crash-explained/ The trucker, who may well be trying to divert attention from his having failed to yield right-of-way when he turned across the highway in front of this poor inattentive guy, claims the Tesla driver was watching a movie https://www.theguardian.com/technology/2016/jul/01/tesla-driver-killed-autopilot-self-driving-car-harry-potter which was still playing when the car came to a stop on down the road.

Yes, I know I'm paranoid, but am I paranoid enough?
Merusk
Terracotta Army
Posts: 27449

Badge Whore


Reply #318 on: July 01, 2016, 08:50:20 PM

It's quite possible the Trucker is telling the truth and not just covering his ass. The Tesla subreddit has been talking about this for a few days, and that guy was known to frequently push the limits of Autopilot. The radar may have thought the truck was an overhead sign and continued onward, it's not a true "autopilot" the way people are using it and not meant to be used with a totally inattentive driver at highway speeds. The groupthink there immediately believed he was to blame for once again putting too much faith in the Autopilot.

Here's a more detailed article on the crash:
http://electrek.co/2016/07/01/understanding-fatal-tesla-accident-autopilot-nhtsa-probe/#disqus_thread

Also, the guy has been in 'problem' situations before. Here he mentions nearly getting squashed by trucks in other cars.
http://i.imgur.com/0r7P03J.jpg
Then here's a link to one of the guy's videos where the autopilot avoided an accident.
https://www.youtube.com/watch?v=9I5rraWJq6E

Apparently a witness said she was doing 85 on the divided highway and was passed by the Tesla just before the crash. The Autopilot is capped at 90mph so that's possible. There's got to be some blame for the victim here as driving history seems to indicate he was often in problem situations.

All that said, it's still far below the average highway miles for accidents. 130million with Autopilot for 1 fatality and the US has an average of 1 fatality every 70-90million miles.

Also, time for Mansfield bars on the sides of trucks, too. I was surprised to learn Europe has them standard while we continue to keep them open enough to swallow cars.


The past cannot be changed. The future is yet within your power.
Count Nerfedalot
Terracotta Army
Posts: 1041


Reply #319 on: July 02, 2016, 02:21:47 PM

Yeah, my original post was mostly snark. We don't know all the facts, probably never will, but there is likely plenty of blame to go around including the trucker, the Tesla driver, our society which values corporate profits over human safety, and probably Tesla for selling a complex and extremely dangerous when used inappropriately technology to idiots. Oh, and the car also comes with an even more dangerous when misused accessory called Autopilot!   Ohhhhh, I see.


Yes, I know I'm paranoid, but am I paranoid enough?
Torinak
Terracotta Army
Posts: 847


Reply #320 on: July 02, 2016, 04:27:12 PM

Anything short of level 4 autonomous vehicles (fully autonomous with no expectation of operator intervention) is going to be prone to these sorts of incidents. As automation gets better, more people will stop paying any attention at all, or be unable to take control quickly enough when the automation fails--they'll sleep, leave the driver's seat, get immersed in movies or video games, etc. People seem to interpret some automation as full automation.

As the automation gets better, driving properly is going to become more fatiguing due to the need to stay fully engaged (so you can take over quickly enough) but without having anything to do the vast majority of the time.

I'm eagerly awaiting deployment of level 4 autonomous vehicles. Level 3, not so much.
Count Nerfedalot
Terracotta Army
Posts: 1041


Reply #321 on: July 02, 2016, 06:21:07 PM

Yep, for highway operation pretty much anything between cruise-control but short of fully autonomous is just a catastrophe waiting to happen and really should not be allowed. Parking is one thing. 70+mph operation is another thing entirely.

Also, one of the articles I read pointed out (but in friendlier terms) that the Tesla's sensor suite is trying to get by on the cheap by relying almost exclusively on camera systems which are vulnerable to exactly the kinds of things that probably caused this accident: combinations of bright skies and/or facing the sun and white/light-colored vehicles which are high and thus at/above the horizon (ie, a semi-trailer). Real autonomous systems add radar to their sensor suites which is NOT fooled by those conditions.

Yes, I know I'm paranoid, but am I paranoid enough?
Merusk
Terracotta Army
Posts: 27449

Badge Whore


Reply #322 on: July 02, 2016, 06:31:44 PM

Mobileye, the sensor guys, came out and published yesterday that their system isn't even designed to avoid this type of crash. It's aimed at Rear-end or head-on collisions. Side impacts are outside of their current scope of product.

http://electrek.co/2016/07/01/tesla-autopilot-mobileye-fatal-crash-comment/

Tesla's system apparently *does* use the Radar but given it was a truck with nothing on the sides of the trailer, the radar sensors interpreted it it as an overhead sign. Radar is actually less fidelity and precision than optical so this makes sense.

http://electrek.co/2016/07/02/tesla-autopilot-mobileye-automatic-emergency-braking/

Tesla should really never have called this "Autopilot," "Self Drive," or anything implying the car is fully autonomous. It's really computer-assisted driving and maneuverability.

The past cannot be changed. The future is yet within your power.
MahrinSkel
Terracotta Army
Posts: 10857

When she crossed over, she was just a ship. But when she came back... she was bullshit!


Reply #323 on: July 03, 2016, 10:40:03 PM

I think it's interesting that even the half-assed version Tesla has is right now still safer than human drivers (1 fatality for 134 million miles, versus 1 per 90M for human drivers). And given that this guy finally won his Darwin Award and saw how stupid you have to get for it (seriously, 85 MPH in a 55 zone of a road that has left turns without traffic lights, that would be reckless driving with a human at the controls), the other idiots may tone it down a bit.

--Dave

--Signature Unclear
Tale
Terracotta Army
Posts: 8558

sıɥʇ ǝʞıן sʞןɐʇ


Reply #324 on: July 07, 2016, 10:30:22 PM

Yegolev
Moderator
Posts: 24440

2/10 WOULD NOT INGEST


WWW
Reply #325 on: July 08, 2016, 07:16:55 AM

That's how you get cyborgs!

Why am I homeless?  Why do all you motherfuckers need homes is the real question.
They called it The Prayer, its answer was law
Mommy come back 'cause the water's all gone
Merusk
Terracotta Army
Posts: 27449

Badge Whore


Reply #326 on: July 08, 2016, 11:38:29 AM

We already have cyborgs. Multiples.

You mean The Borg.

The past cannot be changed. The future is yet within your power.
Ghambit
Terracotta Army
Posts: 5576


Reply #327 on: July 08, 2016, 06:24:29 PM

That thing's got nothing on the the modularBody:
https://www.youtube.com/watch?v=OORQRO8mul8

Kill it with fire!!!!!

"See, the beauty of webgames is that I can play them on my phone while I'm plowing your mom."  -Samwise
calapine
Terracotta Army
Posts: 7352

Solely responsible for the thread on "The Condom Wall."


Reply #328 on: July 09, 2016, 07:36:59 AM

That thing's got nothing on the the modularBody:
https://www.youtube.com/watch?v=OORQRO8mul8

Kill it with fire!!!!!

You can't kill what doesn't exist.  wink

Restoration is a perfectly valid school of magic!
Ghambit
Terracotta Army
Posts: 5576


Reply #329 on: July 09, 2016, 10:50:51 AM

Here I've been re-checking that page waiting for a kit to become available.  pfft

"See, the beauty of webgames is that I can play them on my phone while I'm plowing your mom."  -Samwise
Yegolev
Moderator
Posts: 24440

2/10 WOULD NOT INGEST


WWW
Reply #330 on: July 11, 2016, 11:44:48 AM

What was I looking at?

Why am I homeless?  Why do all you motherfuckers need homes is the real question.
They called it The Prayer, its answer was law
Mommy come back 'cause the water's all gone
Khaldun
Terracotta Army
Posts: 15157


Reply #331 on: July 12, 2016, 04:35:12 PM

I think it's interesting that even the half-assed version Tesla has is right now still safer than human drivers (1 fatality for 134 million miles, versus 1 per 90M for human drivers). And given that this guy finally won his Darwin Award and saw how stupid you have to get for it (seriously, 85 MPH in a 55 zone of a road that has left turns without traffic lights, that would be reckless driving with a human at the controls), the other idiots may tone it down a bit.

--Dave

Come on. This is not exactly a serious test of what an install of the half-assed autopilot would do in serious mass release outside of controlled settings. Put the autopilot on a significant proportion of the cars in service and get back to me about whether it's safer than human drivers. Elon Musk's self-pleasurings are not evidence, his confidence in them notwithstanding.
Venkman
Terracotta Army
Posts: 11536


Reply #332 on: July 13, 2016, 07:04:11 PM

Which I think is what is needed.

Autopilot that can coexist with normal drivers is years (decades?) away. And yet, publicly traded companies and startups needing funding can't plan that far ahead without being able to monetize every year along the way.

These competing factors pretty much require AI to have their own driving lanes. First take over the HOV lanes. Then add more lanes and take away from the old manual drivers over the course of a decade until eventually manual driving is for diehard hobbiests in their own special smoking driving sections.

Because the current approach of co-mingling will just result in more accidents caused by stupid humans thinking they're living the Jetsons with all the money they need to buy technology they trust way more than understand. And yet it'll be the AI that'll be blamed.

And that'll just make the AI angry.  Ohhhhh, I see.
ajax34i
Terracotta Army
Posts: 2527


Reply #333 on: July 14, 2016, 03:23:00 AM

Seriously, road work sucks.
Khaldun
Terracotta Army
Posts: 15157


Reply #334 on: July 14, 2016, 10:01:54 AM

This is pretty much what I think is coming next. The big money will pour into autonomous car development because the people who are sitting on billions and billions are going insane for a lack of a place to push it. Look at all the money they pushed into Theranos, which was pretty much an obvious fraud ("we have real medical nanotech! Honest! No, we can't tell you how it works or tell you anything about our operations, but really! We are disruptors!!!!!").

Autonomous cars aren't a fraud, they're a real possibility, but once the money starts really putting its bets down on the imminence of it beyond what's going on so far, then there will be immense pressure to get it out there as a commercially viable technology this very second. And what the AI guys are going to realize is that the degree of discretionary knowledge that a fully autonomous vehicle on our actually existing infrastructure needs is basically only one step away from full sentience, and that they're not going to get there because as always that's tougher than people in the AI community think it is. But what would work is basically a new infrastructure that puts an expert driving AI inside bowling-alley bumpers--an AI-only lane with guides and backup highway AI that reverts to human control once you leave the highway. So the money will push for that too, and probably the guys who want solar paving installed will crowd into the same scene. Probably they'll start with existing diamond lanes/commuter-privileged lanes and rebuild those. This is not that far off how the interstate system got built in the first place in the early 20th C--car enthusiasts undertook some of the initial work, some engineers started fiddling with the best paving materials and got to a fairly reproducible result, investors and manufacturers started agitating for local and state level infrastructure improvements, the federal government got involved in fits and starts, and then suddenly blammo, big (somewhat) integrated system that became the standard.
Torinak
Terracotta Army
Posts: 847


Reply #335 on: July 14, 2016, 02:41:45 PM

This is pretty much what I think is coming next. The big money will pour into autonomous car development because the people who are sitting on billions and billions are going insane for a lack of a place to push it. Look at all the money they pushed into Theranos, which was pretty much an obvious fraud ("we have real medical nanotech! Honest! No, we can't tell you how it works or tell you anything about our operations, but really! We are disruptors!!!!!").

Autonomous cars aren't a fraud, they're a real possibility, but once the money starts really putting its bets down on the imminence of it beyond what's going on so far, then there will be immense pressure to get it out there as a commercially viable technology this very second. And what the AI guys are going to realize is that the degree of discretionary knowledge that a fully autonomous vehicle on our actually existing infrastructure needs is basically only one step away from full sentience, and that they're not going to get there because as always that's tougher than people in the AI community think it is. But what would work is basically a new infrastructure that puts an expert driving AI inside bowling-alley bumpers--an AI-only lane with guides and backup highway AI that reverts to human control once you leave the highway. So the money will push for that too, and probably the guys who want solar paving installed will crowd into the same scene. Probably they'll start with existing diamond lanes/commuter-privileged lanes and rebuild those. This is not that far off how the interstate system got built in the first place in the early 20th C--car enthusiasts undertook some of the initial work, some engineers started fiddling with the best paving materials and got to a fairly reproducible result, investors and manufacturers started agitating for local and state level infrastructure improvements, the federal government got involved in fits and starts, and then suddenly blammo, big (somewhat) integrated system that became the standard.

This country doesn't invest in infrastructure.

The private sector cannot (and will not) dump the trillions of dollars needed into building new lanes/roads. The government will be prevented from doing so because infrastructure projects might demonstrate that the government can be competent.

You can't have separate lanes because they'll be abused by non-autonomous vehicles--witness the rampant abuse of HOV lanes in most areas (around here, certainly). And unless it's effectively impossible to mix non-autonomous and fully-autonomous traffic, the autonomous vehicles have to be able to handle non-autonomous traffic.

Some of the self-driving cars (e.g., Google's) are actually closing in on full autonomy. Others (e.g., Tesla's) seem to be driven on hyping up vehicle/lane-following features and calling that "autopilot".
jgsugden
Terracotta Army
Posts: 3888


Reply #336 on: July 14, 2016, 03:34:51 PM

I find it hilarious that some of you think we'll make serious progress on truly autonomous cars before we kill ourselves off.

2020 will be the year I gave up all hope.
Samwise
Moderator
Posts: 19212

sentient yeast infection


WWW
Reply #337 on: July 14, 2016, 03:43:52 PM

Fifteen years ago I would have agreed that the logical next step would be AI-only lanes.  One of my professors at college was on a team that had a working prototype of an AI that was safer and faster than human drivers as long as you had it in a dedicated lane.  Visual recognition was nowhere near viable at that point, so they lined the lane with rare-earth magnets and put sensors on the cars that could "see" them regardless of visual obstructions, plus radio to communicate with other similarly-equipped cars and proximity sensors (I forget if it was radar or infrared or what) to detect and avoid imminent collisions in case the car ahead of it had some sort of catastrophic failure.  The idea was that with this system in place you could pack cars together into "pods" with insanely close following distances at high speeds because they weren't limited by human reflexes.  The human driver would merge into the lane in between pods, hit a button, and the autopilot would take over and join a nearby pod.  Once it was time to exit, the human driver hits a button to disengage, and the other AIs give the car a wide berth in anticipation of the meatsack taking the wheel to merge out.

That was fifteen years ago, though.  Now, with Google cars already on the road that only need a human present for liability purposes, and Tesla iterating on its autopilot features to try to catch up, I think we're close enough that it's not going to be worth dividing up the roads at this point.  Putting more AIs on the roads will only make them safer, because unlike human drivers they'll be able to communicate wirelessly with each other and they're not going to do completely unpredictable things.
« Last Edit: July 14, 2016, 03:46:11 PM by Samwise »

"I have not actually recommended many games, and I'll go on the record here saying my track record is probably best in the industry." - schild
Typhon
Terracotta Army
Posts: 2493


Reply #338 on: July 14, 2016, 04:25:48 PM

Agreements for standards on how self-driving cars communicate with each other isn't something that will happen till a decade after we have self-driving cars... unless the government gets involved and then it will be so horribly broken that we'll wish we didn't have it.
Venkman
Terracotta Army
Posts: 11536


Reply #339 on: July 14, 2016, 04:27:43 PM

Except they'd still be surrounded by unpredictable humans, especially ones who don't want to be caught drinking/texting/whatever while driving so will claim "the AI did it" (like that yutz in PA).

HOV lanes are, in many places, a suggestion, for lack of enforcement because it relies on humans (police) being present and witnessing and catching humans (drivers) violating the rule.

But imagine AI cars with programming to recognize non-AI cars. Should be easy enough since every AI will likely be some equivalent of a MAC address in some central DB.

Now those AI cars tattle on the non-AI cars, and the latter gets a nice $200 bill in the mail from the same agency that sends those "oops you 'forgot' [sped-through] the EZpass/Fastlane" sensor. How often will that happen before non-AI drivers get the hint?

I'm harping on this because I believe it's the only way. We're not going to go from a world where all cars are not AI to all cars with AI instantly. Nor will AI cars be able to catch up with human idiocy and the politicized acrimonious legislation that will follow all the accidents caused by humans by blamed on AI. Only need to look at what happened when regular cars first started coming on scene (during horse/trolley days) to see what will happen.

And Google, well, their work reminds me of Segway. A very interesting approach for a limited set of conditions. I haven't done exhaustive research, but as far as I can tell, their car can't actually be driven by a human. Like, at all. No steering wheel nor pedals in the ones I've seen.
Merusk
Terracotta Army
Posts: 27449

Badge Whore


Reply #340 on: July 14, 2016, 06:28:56 PM

Except they'd still be surrounded by unpredictable humans, especially ones who don't want to be caught drinking/texting/whatever while driving so will claim "the AI did it" (like that yutz in PA).

Dashcams are easily installed, are becoming a thing with the younger crowd and are likely to become a standard feature in the next 5-10 years. These cut-off 99.9% of such bullshit claims. Folks with AI cars will install them if only to protect themselves.

And Google, well, their work reminds me of Segway. A very interesting approach for a limited set of conditions. I haven't done exhaustive research, but as far as I can tell, their car can't actually be driven by a human. Like, at all. No steering wheel nor pedals in the ones I've seen.

Yeah, you do need to do some research. They're not working in a limited set of conditions, they're out driving normal city streets on a regular basis.

As for the rest, saying something should be delayed due to Luddites would have us all living with the Quakers.
« Last Edit: July 14, 2016, 06:33:10 PM by Merusk »

The past cannot be changed. The future is yet within your power.
Count Nerfedalot
Terracotta Army
Posts: 1041


Reply #341 on: July 14, 2016, 07:37:46 PM

... saying something should be delayed due to Luddites would have us all living with the Quakers.

um, the Quakers have absolutely nothing against technology, only against hurting other humans for any reason. And smoking and drinking, of all things. I assume you mean the Amish who are fine with an iron plow but not a steel tractor, or heaven forbid, a cellphone? 

If you're trying to be funny with prejudicial slurs, at least be accurate about which group holds the beliefs you are belittling!  Ohhhhh, I see.

Yes, I know I'm paranoid, but am I paranoid enough?
Khaldun
Terracotta Army
Posts: 15157


Reply #342 on: July 14, 2016, 07:46:44 PM

Fifteen years ago I would have agreed that the logical next step would be AI-only lanes.  One of my professors at college was on a team that had a working prototype of an AI that was safer and faster than human drivers as long as you had it in a dedicated lane.  Visual recognition was nowhere near viable at that point, so they lined the lane with rare-earth magnets and put sensors on the cars that could "see" them regardless of visual obstructions, plus radio to communicate with other similarly-equipped cars and proximity sensors (I forget if it was radar or infrared or what) to detect and avoid imminent collisions in case the car ahead of it had some sort of catastrophic failure.  The idea was that with this system in place you could pack cars together into "pods" with insanely close following distances at high speeds because they weren't limited by human reflexes.  The human driver would merge into the lane in between pods, hit a button, and the autopilot would take over and join a nearby pod.  Once it was time to exit, the human driver hits a button to disengage, and the other AIs give the car a wide berth in anticipation of the meatsack taking the wheel to merge out.

That was fifteen years ago, though.  Now, with Google cars already on the road that only need a human present for liability purposes, and Tesla iterating on its autopilot features to try to catch up, I think we're close enough that it's not going to be worth dividing up the roads at this point.  Putting more AIs on the roads will only make them safer, because unlike human drivers they'll be able to communicate wirelessly with each other and they're not going to do completely unpredictable things.

I couldn't disagree more with this. In fact, I think you'll be shaking your own head at yourself in 15-20 years because attempts in the interim to just put autonomous cars on the road are going to produce hilarious and brutal and tragic kinds of unpredictability, both in the interaction between car AIs and other wireless-enabled systems and between car AIs and human drivers. Real-world materiality is still a system that is pretty foreign to most of the kinds of AI and algorithms we have--the ones that look sophisticated are entirely post-facto, cribbing from past behaviors by real human users in order to look like they're anticipatory. Put the existing AIs into an actual meatspace environment where unpredictability is BOTH a matter of muscles/nerves and emotion/cognition and they're fucked. They can't just simulate being intelligent by bootstrapping what people have done in 99% of past circumstances, they have to be as anticipatory in a routine way as most of us are when we notice a red pickup acting slightly erratically, notice the driver has a Confederate flag on the back windowpane, remember that we're driving on a smaller highway through east Texas on our way back to Maryland, and know that it's 9pm and Friday and realize we have a probable early drunk who should really not be tangled with.
Samwise
Moderator
Posts: 19212

sentient yeast infection


WWW
Reply #343 on: July 15, 2016, 12:53:23 AM

Agreements for standards on how self-driving cars communicate with each other isn't something that will happen till a decade after we have self-driving cars... unless the government gets involved and then it will be so horribly broken that we'll wish we didn't have it.

Doesn't matter because it's a bonus -- the AIs already need to be able to deal with mute human drivers.  Worst case scenario they use the same logic to deal with AIs who don't speak their language.

Cooperative AIs will potentially have a better ride experience, though.  Imagine if when it were time to get off the freeway all the other cars just parted like the goddamn Red Sea in unison to let you through, rather than each imdividual asshole deciding he doesn't want to risk letting you get in front of him.  Google and Tesla cars being able to talk to each other will benefit both of them; they'll figure it out.

"I have not actually recommended many games, and I'll go on the record here saying my track record is probably best in the industry." - schild
ezrast
Terracotta Army
Posts: 2125


WWW
Reply #344 on: July 15, 2016, 02:56:49 AM

Darniaq and Khaldun, I love you two but you're wrong. There is a shitton of money being thrown into autonomous vehicles right now, and while the well will inevitably dry (leaving the dessicated corpses of many a scrappy startup scattering the landscape) there is no way the industry is going to give up on this. Google and Tesla are the only ones with the resources to show off right now, but the major auto companies are desperate to get in on a piece of the action. They're all just waiting to see which of the aforementioned startups comes out with viable tech first so that they can acquire it for a zillion dollars, rather than building it themselves. This will start happening in a matter of years.

Also, for the record, highway automation is likely to be considered safe well before city driving. It's a much easier problem to solve because you basically only have to recognize lanes and vehicles. No stop signs hidden behind trees, no fucky five-way intersections with indecipherable traffic controls, no small children darting out from behind parked cars (if any tragedy manages to put a damper on the industry, I'm betting it will be this one, not anything on the highway). Just way, way fewer edge cases overall. And humans are really not built for any sort of decision making at highway speeds, so in many ways there is a lower bar to clear there.
ajax34i
Terracotta Army
Posts: 2527


Reply #345 on: July 15, 2016, 05:51:23 AM

I think that "autonomous" is the wrong direction to take cars; "networked" and "swarm intelligence" may give better results faster. 

"Autonomous" is trying to pre-program each car with all the roads and all possible behaviors, so it can function alone.   "Networked" and SI can be introduced on current cars, so they share sensory information and, eventually, road conditions and behaviors while still being driven by human drivers.  I drive to work daily, that's the road most traveled for me, my car could function as an info server for any car that enters that road for the first time or rarely.  All cars would instantly get the dimensions, position, and condition of the road ahead, simply from the fact that other cars are occupying spaces on that road; it would be like filling the blood vessels with a dye and performing a ct scan.

Of course, we'll first have to have a format war.
Merusk
Terracotta Army
Posts: 27449

Badge Whore


Reply #346 on: July 15, 2016, 06:01:05 AM

... saying something should be delayed due to Luddites would have us all living with the Quakers.

um, the Quakers have absolutely nothing against technology, only against hurting other humans for any reason. And smoking and drinking, of all things. I assume you mean the Amish who are fine with an iron plow but not a steel tractor, or heaven forbid, a cellphone?  

If you're trying to be funny with prejudicial slurs, at least be accurate about which group holds the beliefs you are belittling!  Ohhhhh, I see.

You're right, I misstated the sect. You want to assume it's a slur, that's on you. They get to choose their way of life and I've no problem with it so long as they allow that choice.

I do have a problem with people who state it's best for everyone to just hold off on things because they're uncomfortable with them and the implications. That we should all be Amish because maybe they're right and electricity and modern life is evil. That we should forgo automated machinery because there's a few idiots who'll ruin things.

Holding society back because of false beliefs isn't useful, it's hurtful to the whole.

I think that "autonomous" is the wrong direction to take cars; "networked" and "swarm intelligence" may give better results faster. 

"Autonomous" is trying to pre-program each car with all the roads and all possible behaviors, so it can function alone.   "Networked" and SI can be introduced on current cars, so they share sensory information and, eventually, road conditions and behaviors while still being driven by human drivers.  I drive to work daily, that's the road most traveled for me, my car could function as an info server for any car that enters that road for the first time or rarely.  All cars would instantly get the dimensions, position, and condition of the road ahead, simply from the fact that other cars are occupying spaces on that road; it would be like filling the blood vessels with a dye and performing a ct scan.

Of course, we'll first have to have a format war.

You're right, that's the ideal. Don't let the perfect be the enemy of the good, however. Autonomous, despite being harder, is the first step because of how the world is. We have the infrastructure in place and we have to deal with the fact that not all people will willingly give-up self-driving for some time; that it will die-out like horseback riding as the standard for locomotion.
« Last Edit: July 15, 2016, 06:04:18 AM by Merusk »

The past cannot be changed. The future is yet within your power.
ezrast
Terracotta Army
Posts: 2125


WWW
Reply #347 on: July 15, 2016, 11:01:23 AM

I think that "autonomous" is the wrong direction to take cars; "networked" and "swarm intelligence" may give better results faster. 

"Autonomous" is trying to pre-program each car with all the roads and all possible behaviors, so it can function alone.   "Networked" and SI can be introduced on current cars, so they share sensory information and, eventually, road conditions and behaviors while still being driven by human drivers.  I drive to work daily, that's the road most traveled for me, my car could function as an info server for any car that enters that road for the first time or rarely.  All cars would instantly get the dimensions, position, and condition of the road ahead, simply from the fact that other cars are occupying spaces on that road; it would be like filling the blood vessels with a dye and performing a ct scan.

Of course, we'll first have to have a format war.
I hope you're right, because that's almost exactly what my company is doing. Conceptually we're hitched to the AV wagon because that's where all the sweet investor money is, but intelligence sharing is a good place to be no matter how the actual self-driving part pans out.
Venkman
Terracotta Army
Posts: 11536


Reply #348 on: July 15, 2016, 06:30:08 PM

\Yeah, you do need to do some research. They're not working in a limited set of conditions, they're out driving normal city streets on a regular basis.

As for the rest, saying something should be delayed due to Luddites would have us all living with the Quakers.

Not sure where you're getting the idea that I'm saying it should be delayed. It should not, if for no other reason than it's already here and happening. Besides being niave to think it can stop, there's no good reason for it to.

What I am saying is to be realistic about the rollout itself. Google's been driving on normal city streets on a regular basis for a few years. Their monthly reports are interesting. The way they increment their test conditions is too.

And because it wasn't clear enough: I want this to happen. I want mass transit in a personal vehicle because it's the best of both worlds: the dream of open road freedom coupled with the preening of a car that's "yours'", but without the fuckstupid idiocy of drunk drivers ("I drive better after a drink") and testoterone ("I'm the best driver on the road") or distraction (how many states now have banned cellphones/texting while driving?) based decision making at 70mph.

But again, I want it to happen the right way. The slow way. The way that purposely limits the conditions so companies can monetize each step of their investments while consumers get trained on what it means bit by bit. I'm an incrementalist when it comes to the rollout of technology that can kill people.

And the irony of this sub-debate is that I'm not saying anything that is different from what's currently happening. It's not like we're going to go from manual drive to steering-wheel-less cars in one jump, so I picked on one model of Google's program for that reason. They know it. We know it. The insurance companies do too.
Khaldun
Terracotta Army
Posts: 15157


Reply #349 on: July 16, 2016, 05:28:34 AM

Darniaq and Khaldun, I love you two but you're wrong. There is a shitton of money being thrown into autonomous vehicles right now, and while the well will inevitably dry (leaving the dessicated corpses of many a scrappy startup scattering the landscape) there is no way the industry is going to give up on this. Google and Tesla are the only ones with the resources to show off right now, but the major auto companies are desperate to get in on a piece of the action. They're all just waiting to see which of the aforementioned startups comes out with viable tech first so that they can acquire it for a zillion dollars, rather than building it themselves. This will start happening in a matter of years.

Also, for the record, highway automation is likely to be considered safe well before city driving. It's a much easier problem to solve because you basically only have to recognize lanes and vehicles. No stop signs hidden behind trees, no fucky five-way intersections with indecipherable traffic controls, no small children darting out from behind parked cars (if any tragedy manages to put a damper on the industry, I'm betting it will be this one, not anything on the highway). Just way, way fewer edge cases overall. And humans are really not built for any sort of decision making at highway speeds, so in many ways there is a lower bar to clear there.

This is pretty much what I said: lots of money being thrown at it, the industry will force success no matter what it takes, and what that's likely to lead to is rebuilding highways to make autonomous driving as safe and reliable (and commercially viable quickly) as it can be. Humans will take over back on local streets because solving the complexity of local street conditions is a much harder problem. But the idea that just by having autonomous cars that can talk to each other and to street lights etc. you will eliminate unpredictable outcomes on both highways and local streets is wrong: there are going to be systems-level glitches that will surprise everyone, and this will be particularly true as folks try to extend the more controlled highway situation into other road networks.
Pages: 1 ... 8 9 [10] 11 12 ... 14 Go Up Print 
f13.net  |  f13.net General Forums  |  General Discussion  |  Topic: The robots are coming  
Jump to:  

Powered by SMF 1.1.10 | SMF © 2006-2009, Simple Machines LLC