UAVs appear to be doing very well and expanding their commercial uses fast, in agriculture, search, line inspection, etc.
One problem with autonomous cars is that people may want to drive, even if they could be driven. EG HRH Prince Philip last week.
However our government would be much happier if we were unable to drive cars to places they don’t approve of. The “autonomous” vehicle will be Google controllable, not just observed. 
A guy I know just sold his company that had grown to include a substantial number European employees doing work in ‘mobility,’ very much including R&D on driverless vehicles for OEMs. With cash in hand he’s now looking at selling his current plane and buying a Mooney Ovation… obviously to fly himself for fun. So irony aside in that way I think driverless vehicles etc has worked out to be a very exciting field for him 
Maybe he sold at the right time to maximize his quality of life…
IMO all what these AIs do is to search through GBs of data in a structured manner, lumping together similar data for later processing of new data into the same similar clusters. Then these clusters typically represent certain modes, for instance broken or OK. The speed and the amounts of data is staggering, the intelligence is not.
An autonomous car requires intent. Without it, it’s just like a train on rails. Nothing new, only different kind of rails. The intent thing is far into the future.
Airborne_Again wrote:
There has been substantial scientific and practical development since the 70’s.
Yes, a lot of actually but the current success mainly sits on “internet-style” data and cheap compute being available rather than how smart the math in that scientific research…
My point on AI performance after internet, this time it is very different, there is a big difference between teaching a single car to drive using complex algorithms versus operating a simple math on real-time sensor data from 5 million cars in all sort of driver/road scenarios, the latter will outperform humans in all tasks, including flying
I bet Facebook computer vision and image recognition program with 5 billion pictures a day and enough compute heat to evaporate a small lake, that server have seen in the world yesterday more than all GA members combined during their entire life and we have some few who have seen a lot
Could still be 50 years away.
It has been just around the corner for that long.
Ibra wrote:
AI algorithms that people brag about are the same as those that failed in the 70’s (call them deep network instead of neural network),
This is not true. There has been substantial scientific and practical development since the 70’s.
Driverless cars and planes are the wet dream of bosses who would rather have companies without employees (and possibly customers) as they are too cumbersome and troublesome to manage. In other words, folks whose idea of business is an organisation where they can sit on their collective behinds while the money comes in without effort. Or, as it’s been described, the “Basil Fawlty Principle”.
From a social point of view I can not find anything positive about this development. Taxis without drivers, planes without pilots, parcel delivery without people, manufacturing without people, so who will buy all the goods if all those workers are out of work? Autopilots for driving on motorways, why not but as some Tesla and other drivers have found out at their or others expense, even with an AP a car needs constant vigilance, as much as autopilots in airplanes do, where no crew would go to sleep while the plane flies on AP.
Where this technology makes some sense is in dangerous environments.But even there, driverless is mostly not an option but rather remote control.
In the coldwar, there was more money pumped in the 60’s in AI than today in many research domains like aviation/defence that relies on voice/text/image processing…all has failed as you did not have internet (huge database, in the 70’s they call it skynet ;) ) and a cheap compute power
AI algorithms that people brag about are the same as those that failed in the 70’s (call them deep network instead of neural network), this has less to do with a 3D problem optimization but more to do with 5 billion tagged pictures that feed Facebook image servers or 2 million daily translation in Wikipedia….
Not really surprising since artificial intelligence has been progressing painfully slowly every since the earliest massive hopes in the 1960s, when the problem of representing a 3D world and working out which bit is in front of which bit, etc, was solved, resulting in all the optimism…
Food for thought. Thanks !
