The next 20 years will be a constant stream of half broken robots and ai systems. At least until we either get them to actually work or we figure out how bad they are and go back to using people again.
It will be like setting up video conferencing, except for everything.
Your car will go to the wrong place because it failed to account for a leap second, thus breaking gps.
Your fridge will put itself in eco mode because it thinks you're on vacation from misreading your calendar and spoils all your food.
Your front door will forget who you are after you get a haircut and lock you out.
After an automatic system upgrade your shower will no longer produce hot water, because a bug results in the system using Celsius instead of Fahrenheit. Your bug report is ignored for a week and after a dozen people complain you get in an argument with the engineer about whether the place you live in actually exists.
Wait. Who am I kidding... You won't be able to talk to an engineer. All support will be automated and it will be impossible to get the robot to understand your complaint. You'll just have to replace it. I suspect you're going to have to replace things a lot.
On the bright side it will probably be really cheap though.
And then there will be all the outages.
One of these days Uber is going to accidentally send all drivers in a big city to a single location. It'll be the great ai outage of 2018, the first of soon to be monthly occurrences of bizarre system failures we can't begin to imagine.
I feel like someone should write a pg wodehouseesque novel about the "smart" future.
> Your car will go to the wrong place because it failed to account for a leap second, thus breaking gps.
I do get the implication you're going for, but GPS actually ignores leap seconds for this reason. The time scale on the GPS system was fixed when it was deployed, and is 18 seconds ahead of UTC because it does not account for leap seconds.
Sometimes we design horrid systems that are harder to use than analog counterparts. Sometimes, our systems work because they required a lot of thought to go into them. AI hell is more likely to be a problem of decision trees than it is going to be anything to do with sensor disruption or time offsets.
Also GPS, now a part of all GNSS, is not an as essential system for fully automated driving, as pretty much all the other sensors (See also GNSS-denied navigation research).
If your navigation system didn't fail so far, it won't in the future. Navigation is the solved part.
I wonder why the fact is downvoted? It is easy to check that the statement is true, just click the link, run the code, see the results, read the authoritative sources linked there for additional information.
Because the GPS time scale is fixed. UTC drifts by leap seconds, GPS ignores them. I'm not sure what the point of the stackoverflow link is, because it certainly says as much itself. Yes, I understand that additional leap seconds are added as time goes on, which makes the difference between GPS and UTC larger as the years go by. But this doesn't mean that the GPS time scale isn't fixed.
My comment says explicitly that the difference the difference is not fixed. Even your comment aknowledges it (the difference increases with each intercalary leap second).
It is already here. Two days ago our HVAC controller apparently lost contact with one of the sensors (as technician told me) and we nearly froze at night (-10 C outside) only to came back later to a desert (+35 C in most rooms).
I just yanked the thing out and adjusted all valves by hand. Fixing was $200. Now i'm thinking about replacing it with Arduino and simple PID controllers...
I live in the Northeast US and I'd never get a Nest. The risk associated with an additional heating system failure mode in the winter just isn't worth it.
For traditional HVAC, because of the lag and few degree temp variance, you really only need the P part of a PID controller. But the same would happen of any controller. If the sensor goes bad it isn't going to work. Also, heat pump may require more sophisticated control.
It's reasonable to require a GPS to navigate, but not for moment-to-moment driving, e.g. knowing which lane you're in. Otherwise it won't work in tunnels.
Tall buildings can occlude GPS satellites. And current driving software decides I’ve gotten on the freeway, with no onramp, going the wrong way mind you, when I’m actually under it
Imagine yourself being transported to some unknown town blindfolded. Then blindfold is taken of and you are asked to drive around. You'd be still able to do that, just not drive to the particular destination. Or maybe you would be even able to do the latter due to the more or less naming of the streets.
Sibling comment makes the correct distinction between driving and navigating. I do not expect self-driving car to be able to navigate without GPS, but it should not be required for driving around.
However I'd say many humans would still be able to successfully navigate using road signs for directions.
It will be like setting up video conferencing, except for everything.
Your car will go to the wrong place because it failed to account for a leap second, thus breaking gps.
Your fridge will put itself in eco mode because it thinks you're on vacation from misreading your calendar and spoils all your food.
Your front door will forget who you are after you get a haircut and lock you out.
After an automatic system upgrade your shower will no longer produce hot water, because a bug results in the system using Celsius instead of Fahrenheit. Your bug report is ignored for a week and after a dozen people complain you get in an argument with the engineer about whether the place you live in actually exists.
Wait. Who am I kidding... You won't be able to talk to an engineer. All support will be automated and it will be impossible to get the robot to understand your complaint. You'll just have to replace it. I suspect you're going to have to replace things a lot.
On the bright side it will probably be really cheap though.
And then there will be all the outages.
One of these days Uber is going to accidentally send all drivers in a big city to a single location. It'll be the great ai outage of 2018, the first of soon to be monthly occurrences of bizarre system failures we can't begin to imagine.
I feel like someone should write a pg wodehouseesque novel about the "smart" future.