This is the leaders performance on actual production hardware outside of fancy (postprocessed) videos with their "cameras only" approach:

https://www.youtube.com/watch?v=UZ1XLqc5IUg

At this point I don't even care if they theorethically are the best, if they are willing to release this dangerous crap they are sooner going to go up in a big publicity fire than get actual autonomy to customers.

This is an old outdated and out of context video. I just drove 150 miles this weekend using AP2. It works pretty well.

It's February, hardly outdated. Did you drive on residential streets, as in the video, or highway?

See, highway "autopilot" debuted in upper class Mercedes Benz some 10 years ago..

Feb was before 8.1 software update which has significant AP updates. So it is outdated.

Industrial cities like Detroit and my beloved hometown of Pittsburgh have been beating Silicon Valley here pretty readily in the self-driving car space (one reason among many is that California's immaculate weather gives engineers a false sense of confidence when testing their vehicles on its perfectly-marked freeways, causing them to completely gloss over the majority of less-stellar American roads and weather-induced driving situations; indeed, the fatal Tesla autopilot crash took place on a Pennsylvania "turnpike" highway, and not a California freeway).

This happens the same day as Tesla's market cap surpasses Ford. Tesla's stock has been a real rollercoaster lately, and somehow, I just don't think the hype is justified. Remember, Tesla for a while used MobilEye (see anything geohot says about them!).

> indeed, the fatal Tesla autopilot crash took place on a Pennsylvania "turnpike" highway, and not a California freeway

The fatal autopilot crash occurred in Florida, not Pennsylvania. There was also another in China.

I see, so there are others that also support my argument (having been in Shanghai, I can definitely tell you that China's roads are sometimes questionable, the very erratic driver behavior that is the norm even more so). The Pennsylvania crash I was referring to in particular was this one:

http://pittsburgh.cbslocal.com/2016/07/07/police-investigati...

Thankfully, it does look like the PA one wasn't fatal after all though. For what it's worth, Tesla also directly and very publicly disputes the driver's own personal account by claiming that autopilot wasn't enabled during the crash.

¯\_(ツ)_/¯

Did you go to CMU? Just curious as I've seen your name pop up a few times with regards to self-driving cars.

No I didn't, but I did summer research internships there while I was in college. I was waitlisted from CMU SCS, so I "settled" for Cornell instead. ;-)

The self-driving car space is just something that I like to follow very closely, given that I've had picked up a bit of domain knowledge over time.

In rolled out they are, but what will next year bring. Ford, GM and the like are used to a competitive world - for years they have worked behind the scene to perfect something interesting before rolling it out. Ford has lost big in several lawsuits because their cars were not safe enough - as such I would expect them to be a little risk adverse. It would not surprise me if they had the same technology as Tesla but refused to release it because they were afraid of getting sued in one of the crash scenarios they know it is vulnerable to.

Tesla tech is level 3 - the driver has to be ready to take over. Can that eventually be turned into level 4? I don't know.

I am a pilot, and we've had autonomous airplanes for decades. Even entry level Cessnas have auto-pilots. The auto-pilot is a great tool for reducing pilot workload and increasing safety, especially en route, but you are still expected to be the pilot in command, especially during a tricky maneuver like landing in a crosswind with not so good weather.

I think Ford, GM, etc are right in the long run. There are so many edge cases in driving a car (blown tire, hydroplaning, uncertain intention of an oncoming driver), I would not trust a fully driverless car. That said, there is a lot of room for innovation with automatic braking and collision avoidance that will reduce accident rates and severity, maybe dramatically.

Once most cars are networked and semi-autonomous, lots of things become possible. Example, cars that automatically observe speed limits (which can be dialed down during bad weather). No single measure like this will eliminate all accidents, but each will add to safety.

The problem with too much automation is people become dependent on it and lose skills. This was a critical factor in the Air France flight that stalled and crashed off the coast of Brazil. I'd rather see them focus on helping people be better drivers than taking humans completely out of the loop except on protected roadways.

I'm not saying you're wrong but do you have anything other than anecdotal data?

Data is king when it comes to autonomous driving tech. Last time I checked Tesla had over a billion miles driven (may have passed 2 billion by now) in real world environment. Google has the second most data collected with 2-3 million miles driven which is orders of magnitude lower than Tesla. Not even gonna mention Ford.

Also Tesla is the only company that has the hardware setup for full autonomy to achieve level 5 on a production car. They are just working on software. Any other company couldn't even theoretically have level 5 autonomy even if they had the software ready to go

Tesla doesn't have billions of miles of data. You have been lied to and you bought it Hook line and sinker. The data Tesla collect are high precision map based on Gps logging and radar blacklist for false positives.

All that data is useless for anything other than level 2.

You're a portrait of the Tesla hype. But I don't blame you

And you're an astroturfing dummy account. This account was created just to make this comment?

> Also Tesla is the only company that has the hardware setup for full autonomy to achieve level 5 on a production car. They are just working on software. Any other company couldn't even theoretically have level 5 autonomy even if they had the software ready to go

And yet, aren't they -removing- hardware? And as a result, a whole new slew of their autopilot getting _worse_ is hitting the Internet. And some of those are really embarrassing - sunny day but not blinding brightness, clear line markings and the car will just veer into the oncoming lane, or entirely off road with little to no warning.

[citation needed]