Elon isn't promising this for consumer cars in 2020. He's talking about internal development vehicles. The author seems to have missed this distinction, and then based large parts of the article on this misunderstanding.<p>For me that's really the strongest plausible interpretation. There are other, weaker interpretations such as the author is twisting the facts to suit an agenda. I'm not even going there. I think the author has just bought into a common meme about Elon "promising" things when he's really talking about in-house development cars.<p>Also:<p>>But for the time being, deep learning algorithms don’t have such capabilities<p>I don't know how he is so confident about the absence of capabilities, when he's not an insider. And the phrase "for the time being" is a tip-off that even the author acknowledges that deep learning is still a developing field.<p>Tesla does have some interesting approaches. The article also doesn't mention the hydranet, something Tesla has talked about publicly, which sounds to be a fusion of multiple deep learning networks. In any case, I would look elsewhere for insights about Tesla, unless the author can evolve his view to be more open minded.<p>The article does have some good points about the challenges. I'm a skeptic as well. But also very optimistic and curious to see where the tech and the massive data will take us.<p>The mentioned problem of having a dearth of China data, well, that is being chipped away at now.<p>It's certainly a more massive problem than the US driving environment… China is insane. We visited a "this used to be secret, now it's not" huge super-wide emergency backup military runway that was also used as a road and 1) for pedestrian traffic 2) for ox carts 3) for freeway-speed traffic 4) for normal speed local traffic 5) for bicycles and 6) for threshing wheat… they lay the wheat out and let passing vehicles do the work of threshing it, with people tending the process and walking out into the middle of the road/runway to fetch the partially processed wheat. Just think all the edge cases in that one environment alone. And that was just coastal China. It's a big country. The data is going to be interesting.