When you say it can operate without external data, that is exactly what you mean. You said, quote, "Autonomous cars don't need data channels of any sort. They're autonomous." That's a data vacuum, and it is a patently absurd claim to make. That's also a clear admission that you really don't know what "autonomous" means, since "don't need data" is not part of that definition. Every AV needs data to make decisions, and the better the data it can get the better decisions it will make. Like the example I already gave, where an AV that can get real-time data about a road closure could have chosen a different main route to conduct that four hour trip instead of getting to the crash site, deciding it can't get through, and then reverting to the back roads and taking an extra hour and a half.
It is not patently absurd. It is accurate. The vehicle must be able to drive safely without external data, or it is not autonomous. The fact that it can do a better job of getting you there in a timely manner doesn't mean that it needs a high-priority channel.
More significantly, in most major cities, traffic data is broadcast on sidebands of various radio stations, and can be received by any device for which the user has paid the fairly cheap lifetime subscription fee. Because traffic data is available from permanent RADAR stations along the highway, the receipt of that information can be strictly one-directional, and is thus well suited to a broadcast system. This is not to say that you can't do it over the Internet, just that it isn't strictly necessary to do so. Thus, making a high-priority bidirectional channel available for getting information that can be just as easily obtained by adding a $25 traffic data receiver would be patently absurd.
And even if you want to go full-on-Waze-style, with real-time data gathering to pinpoint slow spots and try to route people around on side streets (which may or may not actually reduce your drive time, depending on lots of other factors that are hard to predict, such as traffic light timing), there's no reason to believe that self-driving cars would benefit from that data arriving faster, with lower latency, because overall traffic patterns don't change enough from one second to the next for latency to matter anyway—even if it the latency is on the order of tens of seconds, much less when it is on the order of single-digit seconds.
Traffic mapping is simply not a real-time task and never will be. Nothing associated with self-driving cars benefits from real-time performance except what is happening entirely inside the car—processing data from sensors—and perhaps communication with nearby cars in some situations (which must necessarily be direct, not just because of latency, but also because of the high complexity of figuring out what cars are near you on the server side and routing data appropriately, and also because of the high probability that your cars will get their data from different companies whose data centers are on opposite sides of the country from each other).
No, the sorts of tasks that benefit from prioritization are things like live audio and video streaming. And these things are readily identifiable by port numbers, QoS flags, etc. without any need to use paid prioritization to give higher priority for traffic to/from a single company. Remember, this is not about prioritization, but rather paid prioritization, whose sole purpose, by its very definition, is to give an advantage to traffic from specific companies that pay over others that don't.
That last bit is what Comcast is trying to make people forget. Paid prioritization is, by definition, anti-competitive.