It just shows how out of touch with reality the "mainstream" media is. For years reputable publications like NYT, WSJ, The Atlantic, were heralding the arrival of driverless cars. And yet there is really no engineering basis for them at present. Before we can have driverless cars we will need something akin to AGI.
To expect a majority of journalists to recognize this sort of hype is unreasonable. Most people here on HN (not I, to pat myself on the back) were equally, perhaps more, credulous about self-driving cars. If a tech forum like HN called it wrong, anyone could.
> To expect a majority of journalists to recognize this sort of hype is unreasonable.
Give me a break. All the journalists had to do was watch the Google cars do the same circuits all day long in MV to realize they're just building a fake miles dataset for regulators.
Probably the same with the other companies involved.
Source: used to live on one circuit street off ECR. Saw up to 3 Google cars at a time behind each other, all day long.
unless you torture the definition of self-driving vehicle to mean limited driving capability on highly altered 'smart streets' or whatever, yes it's absolutely true.
As the article points out, ML is fundamentally way too brittle to deal with the long tail of issues that humans solve through common sense reasoning.
We're like two or three paradigm shifts in AI away from having anything that solves open ended, problems in uncontrolled environments without glitching out like a Roomba.
> according to its own IPO filings, Uber can only be profitable if it invents fully autonomous vehicles and replaces every public transit ride in the world with them.
People will be arguing that "true self driving is impossible" for decades after we've all been already happily riding around in cars without steering wheels.
Level 4 self-driving with remote assistance will allow "driverless" cars in the next decade. Over time, the cars will need to phone home less and less. It's possible that even in 100 years, there will still be one single remote assistance center in operation for the 1 in a million situation that the computer still can't handle.
By the time that last remote assistance center closes, nobody will even care about debating whether self-driving cars are bullshit or not.
All new technology is overestimated in the short term and underestimated in the long term.
Sort of like speech recognition... it was going to be big! and then it really couldn't understand as well as people thought... and it sort of fell out of the limelight. But quietly it started to be used for phones and now it's pretty much here.
The greatest problem of this technology is the very auto makers who is salivating at the idea of being the first one to release a fully autonomous car that would be compatible with the existing infrastructure including other manually driven cars, pressuring the engineering teams, i’m not sure the technology will be mature in the next 15 years.
17 comments
[ 4.7 ms ] story [ 47.8 ms ] threadGive me a break. All the journalists had to do was watch the Google cars do the same circuits all day long in MV to realize they're just building a fake miles dataset for regulators.
Probably the same with the other companies involved.
Source: used to live on one circuit street off ECR. Saw up to 3 Google cars at a time behind each other, all day long.
Not even close.
As the article points out, ML is fundamentally way too brittle to deal with the long tail of issues that humans solve through common sense reasoning.
We're like two or three paradigm shifts in AI away from having anything that solves open ended, problems in uncontrolled environments without glitching out like a Roomba.
http://graphics8.nytimes.com/packages/pdf/arts/1920editorial...
The article is also of poor quality, making lame swipes at ride-sharing and electric cars as well.
Where specifically does it say this?
Level 4 self-driving with remote assistance will allow "driverless" cars in the next decade. Over time, the cars will need to phone home less and less. It's possible that even in 100 years, there will still be one single remote assistance center in operation for the 1 in a million situation that the computer still can't handle.
By the time that last remote assistance center closes, nobody will even care about debating whether self-driving cars are bullshit or not.
Sort of like speech recognition... it was going to be big! and then it really couldn't understand as well as people thought... and it sort of fell out of the limelight. But quietly it started to be used for phones and now it's pretty much here.
Disregards the promise of Connected and Autonomous Vehicle (CAV) technology.
https://www.sciencedirect.com/science/article/pii/S209575641....