servodude wrote:We're probably talking at cross purposes though which I guess is my fault as I went in to the article as a geek with a long time spent in automation
Honestly I'm really glad that the legal side and paperwork are catching up with what we can do in the control space
- but I get more excited by the fancy toys
We probably are, so let's untangle. I'm not sure what it is about the headline which you think was oversold. It does contain one possible error in that it could be argued that Tesla's FSD was a more apt comparison, but what it says is entirely true, and doesn't seem sensationalist to me. Picks a low bar for comparison, but that's a different matter.
In 50 years time the answer to the quiz question 'Going back to before humans were banned from driving, which was the first commercially available autonomous driving system?' will be this. It's Yuri Gargarin, while others are trying to be Neil Armstrong, but we all still know who Yuri Gargarin is.
...and yes I said 'commercially available' because there are companies running L4 systems under special developmental and limited commercial licenses. But these aren't directly comparable systems. On top of the fact that they aren't licensed for sale to the public the L4/L5 race and the L2/L3 race have been completely different events (except for for Tesla who are competing in one, but identify as a competitor in the other (except when talking to the Californaia DMV), despite a significant point of cross over in companies like Mobileye and Aptiv who provide hardware to OEM's for one while competing in the other.
In one race serious people have sort to design top down a fully autonomous car, whilst in the other serious people and Tesla have sort to bolt functionality on to an adaptive cruise control to improve consumer experience and increase safety, mostly knowing that what this will get you is a snazzier cruise control not an autonomous car. The L4/L5 race is going rather well and frankly is far more interesting, and the L2/L3 race is successfully promoting the availability, uptake and even the legal requirement of more advanced safety features and what I guess you might describe as consumer 'toys'.
Moreover it appears that these two categories will remain separated in the way they roll out. L4/L5 systems will for the foreseeable future remain the purview of large TAAS fleets, largely run by the coalitions which built them, and added to on a city by city basis. L2/L3 systems will continue to rock up in cars at your local showroom. So again, no real point of comparison.
This is a lone point in the middle which makes it rather interesting. It has the child-like limited skillset of a L2 system, but is recognised as being a grown up like a L4 system. Not only do the company that builds it and regulators trust it to act unsupervised on public roads, but they trust it to be able to tell 10 seconds before it will need to hand back control to a human that in 10 seconds it will need to hand back to a human, and that it will behave appropriately if that human fails to take back control. This handover is unique. L4/5 systems don't need this functionality as the car is always driving, and L2 systems don't need it because the driver is always driving.
So I guess in summation:
- This isn't a small step up if your point of comparison is Tesla, this actually works.
- This isn't a small step up if your point of comparison is companies like Waymo and Motional, same sector different pathway.
- Actually allowing drivers to legally take their attention from the road in low to medium speed congestion is not only a notable first, but is likely to lead to significant reductions in driver stress and frustration, low speed accidents, and increased productivity. Seriously if this were available on the M25 it could add billions of usable hours to commuters' lives every year!
- If what you want is the 'toys' then how much better would this argument be if you could have it while stuck in a traffic jam which otherwise it turns out would lead to you wasting 35 minutes of your life paying attention to the bumper of the car in front of you moving forwards 3 car lengths every other minute (lorry overturned ahead).
Is it likely that if Mercedes hadn't got here nobody else would? No, like I said Audi had a similar system before it was legal to use it, but we have now done something which had not been done before. Is it likely that Mercedes have laid off some or all of the risk with an insurance policy? Probably, but this is actually a good thing, if Mercedes can't get insurance on the system then no-one will be able to which means legal use will remain technically achievable, but realistically unachievable.
So we have a cool new toy, with a potentially massive payoff, its manufacturer has removed a major stumbling block in its adoption, insurance, from the equation and autonomy on public roads has crossed form a 'definitely next year' to an 'actually that happened last year'.
It's not a time machine, but the headline didn't claim it was, it just said
Mercedes Drive Pilot Beats Tesla Autopilot By Taking Legal Responsibility
which they have.
ETA: The comparison made is with autopilot not FSD, autopilot is a really limited set of functionalities. Personally I think FSD is a better comparison because FSD aspires to autonomy, but the comparison to autopilot doesn't imply anything like the degree of functionality you might read in to a comparison to FSD, autopilot is similarly cruise control like, but not at all autonomous.