They use IR (inferred transmitters aimed directly at the driver's eyes) and dedicated IR cameras (likely more IR sensitive only sensing IR with a smaller field of view aimed/focused directly at the driver's eyes). You could wear welder's glasses and they could track your pupils.Interesting. What’s the difference in how other OEMs track attentiveness?
Btw.. Not challenging your response,, I’m just not educated enough to understand it.
I put black electrical tape over those transmitters years ago.
As in why did it not continue to make a full stop when it was still showing red? There's been other cases where 12.x treats stops as yields, but it was stopping initially and continued with the green (with the incorrect belief that the pedestrian would continue to stay still):
No.
Right, that is mysterious to me. Even me, an inveterate California stopper, would have approximated a stop at the line.
No it accelerated from the position behind the lead car into the intersection. It just so happened that the light turned green before it got to the line. It monotonically increased speed until I broke the steering.
Yellow light was 0.5 seconds, 15 frames, prior video posted.
I mean, it doesn’t take a rocket scientist to make YouTube content when the manufacturer just gives you stuff to talk about. And then you can talk about the same topics on each update version.
I would check the Data bank size for the Ai model to see if anything changed version to version.
A single iteration of the network takes on the order of weeks to train, with additional weeks of testing at the very minimum and tons of new training footage. No way a newly-trained model comes out as a maintenance (third number) version.
And if there's nothing really to talk about, you can still put some clickbait thumbnail with huge letters "THIS CHANGES EVERYTHING"
You mean… two weeks?
Yep. This continued focus on the subtle differences in point releases is really odd when it appears to just be tweaks to inputs of the model.
Sorry poor wording on my part. The left screenshot shows a dark black line at the end of the blue path indicating FSD was planning to stop probably because the light was red. But indeed the acceleration was always positive as it was still pulling up to its stop line, and before it got there, it saw the green and maintained its acceleration without ever needing to decelerate for the stop.
There is something stopping the 3 of us. We aren't funny, interesting, or smart.
Like many, as a shareholder I’m hoping the next significant version comes during the trial period so those trying it out can gauge the rate of improvement.
