Turns out that the claims of a 40% reduction were complete bollocks, in fact it makes things far worse.
It’s important to read the whole article. The right takeaway is that the NHTSA’s conclusion was flawed, and they should do it again with a larger dataset.
3 Likes
Would also be nice to see what % of the crashes after autosteer was enabled happened with it on. Would be able to estimate how much of the difference is due to fluctuations caused by the relatively small samples vs autosteer’s functionality.
Judging by the lack of data, I’d hazard a guess that Tesla maybe don’t want it publicised. Not that it’s the fault of the software’s functionality, any more than it’s the fault of the drivers who think it means they can not pay attention.
We’ll never know for sure, but my feeling is the 2016 data was incomplete because Tesla simply didn’t have it. They sent along what they had.
What we do know is the NHTSA’s use of that data was criminally stupid, particularly where they counted 28 accidents with 0 miles driven pre-autosteer.
At any rate, who cares about the version of autopilot they stopped deploying in 2016? Lets see how it performs now.
The NHTSA (National Highway Transport Safety Authority?) had incomplete data and made judgements based on that data. They realised this and dropped the findings.
Then under freedom of information another group got the same incomplete data and is making judgements on the same incomplete data…
Neither has apparently gone back since and gotten complete and larger amounts of data and run the analysis correctly.
So one group is calling foul while the other tried to say “yes we know all ready”.
How is this news? Let alone the click bait bollocks being made out by op.
Edit: is this not the definition of low effort? It seems like op did not even read the linked article.