Tesla Rolls the Dice And Makes Autopilot A True Autopilot... Well Sort Of

With this move, in 2 years time Tesla could have over 1 billion road miles with "self-driving" hardware in the loop - not necessarily in control of the vehicle but always in data collection mode. They could also have AI software making real-time decision (again, not controlling the vehicle but more like a simulation with real world stimulus) and compare to actual driver inputs as a way to collectively enhance self-driving AI.

It would be a fun job to be an AI software developer at Tesla (I am in a different field and have my own kind of fun job). Other AI developers such as Google and Apple have far less real world data for their systems and therefore will be far behind Tesla as time goes on. Traditional automakers seem to be tepid in their commitment to full autonomous driving. They have a large base to collect huge amount of real world data, but also have many hurdles to overcome: drive-by-wire, inter-car connectivity, OTA software updates, sensor platforms, and (most importantly IMO) AI software development.

The last one (AI software) requires a huge amount of investment in terms of money and expertise. Tesla, with respect to autonomous driving, is no doubt leading the pack - and the gap is getting wider.

Do you think this might just be a ploy to get people to shift their focus from what Tesla's Autopilot is (a driver-assist system) and is not (an autonomous auto-pilot type of product) today, and to imagine how great it might be someday, maybe, in the future?

It's not "someday". They are doing it NOW. By the time their system has sufficient real world data (AI simulated decision versus actual driver's response) by the billions of miles, and get approval from regulators, they will already have had hundreds of thousands of cars on the road ready for a flip of the switch (so to speak - but it's actually that easy with OTA update).

HenryN wrote: It's not "someday". They are doing it NOW.

You're getting NOW mixed up with SOMEDAY.

NOW = Tesla's Autopilot is a driver-assist system.

SOMEDAY = Sometime in the future, Tesla's Autopilot might become a successful, fully autonomous navigation and driving system.

Thanks for providing us with a live demo of what I was saying in my original post :-/

@TheSteve: this interview will help your understanding of their plan

https://finance.yahoo.com/news/tesla-vehicles-being-produced-now-013938565.html

Watch the video: https://www.tesla.com/autopilot. Again, Tesla is doing NOW.

This does not mean that the system is perfected and ready for mass release. What they are doing NOW is building an platform of autonomous hardware and perfecting AI software. Once it's ready and approved, it will be an instant release rather than a slow roll out as seen with traditional car makers.

I wish this whole autonomous vehicle thing wasn't being shoved down our collective throats. It makes me thing of Jeff Goldblum's comment form Jurassic Park about bringing dinosaurs back to life - "(Tesla) is so preoccupied with whether or not they could that they didn't stop to think if they should."

Hmm ... there is no comparison - one is a sci-fi movie, the other is a serious effort to curb the traffic deaths (to the tune of 3000 a month in the US alone).

Your comment says a lot about the herd mentality against scientific progress. Let's think for a moment - without companies like Tesla or Google to lead the way, who is going to make the change ? regulators by themselves cannot force the hand if there is no technology to back it up.

It's commendable for Tesla to stick its neck out on this front - even Ford, MB, BMW are now joining the race. Where are the "industry leaders" such as GM, Toyota, VAG ?

Tesla is in the belief that this will save and improve lives, hence they definitely should. If they could, but don't, they become complicit in the short-comings of the status quo, basically liable for people dying while driving. The burden is on Tesla and the proponents to prove and convince the populace that these technologies are indeed safer.

If you're rushing new technology out to the public before it's ready then is that really progress and are you saving lives or gambling with them?

I am still of the opinion that the Tesla system has very similar capabilities to the systems of other manufactures - MB, Audi, Volvo, Bosch, Visteon, etc.

Only difference is that

(a) MAYBE/LIKELY Tesla has minimized some safety features/conservative business rules that are in place with other manufacturers systems, such that it SEEMS to be self-driving but is just less safe and more fast and loose than others

(b) Tesla is relentlessly promoting this technology (ala Donald Trump) in a desparate effort to stay relevant.

(c) Refer to (a) - Tesla has not backed up anything with real, reviewable, data - they have just "SAID" that the software SHOULD work. This should make us all nervous.

a.) The system they rolled out yesterday is so far beyond what any other manufacturer has in their cars today there is no comparison. No other company has a system combining 8 cameras, radar, and ultrasonics on the market today. I doubt we will see anything similar from other manufacturers for 1-2 years (although would love to be proven wrong).

b.) The Trump comparison is asinine. Tesla just became the first company to SHIP a car with Level 5 autonomous driving technology. There is no desperation here, they just nailed a major milestone that will further drive demand for their products. Any other car company would give their left nut to do what Tesla just did in 2016 (most are talking about 2020-2021 for their first product)

c.) The system works in shadow mode--meaning even if self-driving is not turned on the system is still working and noting the decisions it would have made versus a human driver. This will allow Tesla to gather billions of miles of data so that it can safely roll out functionality step by step. It is real and it is reviewable to law enforcement and regulators. The only people that should be nervous are execs at other car companies.

@SanJoseDriver: regarding your point (a), many people already tried to dismiss that the hardware for Tesla is commercially available, and any car maker can have the same. To counter that, I think what differentiate Tesla are their already advanced research in Artificial Intelligence (AI) in autonomous driving and the vast amount of data already collected on existing cars (200M miles by my crude estimate).

While any car company can equip their cars in similar fashion and can collect heaps of real world data, what they do with it is a very different story. That's where AI comes in, and that is what can make or break a company. AI is proprietary, it can't be simply copied or imitated and not easy to develop. Large companies can have research teams working on AI, but similar to tuning a chasis or a suspension, there will be subtle differences that will make the good ones stand apart.

As I mentioned in earlier post, traditional car makers have so many hurdles they need to overcome in order to catch up. These require major changes in how they design their cars from the ground up - some such as OTA software update even requires a complete change in their existing method. You as a Tesla owner knows first hand that OTA software update is a seamless process whereas most if not all other car makers still require car owners to bring their cars to dealers for installation.

While I believe car makers can do all these if they must, they don't show the determination and clear strategic goals as Tesla does. Even Google and to some extent Apple, have shown lukewarm commitment in their development.

That leaves Tesla at the forefront with literally hundreds of millions of (autonomous driving) miles ahead of competition.

If anything, other car companies are more prudent and are letting Tesla have the spotlight to see how said spotlight burns Tesla. Tesla simply does not have the engineering might of say Mercedes, Porsche, or Toyota (to name three) and by letting Tesla run with the ball, the flaws of the technology will be immediately apparent as Tesla most likely over promises and under delivers yet again. The simple fact is that market demand for autonomous cars is even smaller than the tiny EV market, so letting Tesla "win" is no win for Tesla. When the big companies jump in, Tesla will not have the market share to withstand the onslaught.

And when will that be? We've been waiting for a serious EV from the established car makers for what 5, 10, 25, over hundred years?

Once people can legally work, nap, and watch movies during their commutes, demand for self-driving cars will far exceed EVs. I spend ~2 hours a day driving, most of it stuck in miserable Silicon Valley traffic. That is 40 hour a month, a whole workweek of productivity lost, that I can have back if my car was self-driving.

If you drive half as much as I do and value your time at a pretty weak $20/hr, that is $400/mo wasted, $5k/year. Factor in that self-driving cars can easily be rented to carshare services when not in use to offset costs, and it will be a no-brainer a few years from now.