Yes, I read the article and others... other drivers slowed and avoided the truck.
This is actually one of the things that an (alert) driver has over an AI - you don't just watch traffic, you also watch for anomalies. I have a simple algorithm for driving at motorway speeds in that I have to be able to scan well ahead, and by that I don't mean two cars in front of me, it means several hundred yards, and all the lanes, not just the one I'm in. If I can't do that I don't feel I'm driving at a safe speed (and I do a fair bit of German Autobahn, I'm used to high speeds).
In this case, it would have helped the AI if it had been able to detect that there were unexplained avoidance manoeuvres ("unexplained" as the AI had not picked up on the truck), which should have been cause to alert the driver. This situation may improve in the future when auto-drive cars are able to signal each other (which also widens sensor range), but for now, a human driver would have picked up on the flow anomaly caused by people swerving around an object, even if the object itself wasn't that visible. I suspect that will be a hard one to embed in an AI, but if they could it would certainly improve its ability to spot problems - even if it's just to wake up the driver.
Personally I'm not comfortable with using something labelled "beta" at motorway speeds, but each to their own - if others didn't use it we would not get live data. However, I've seen a few sudden glitches that make me avoid the idea. Assist, fine, control, no, not yet, thanks.