Does it? A machine can indeed do a better job (in the sense of requiring negligible training) of knowing all the streets and democratized “the knowledge” which previously was the purview of a small guild.
I am sorry for their sunk cost, but really minicabs+gps and now uber/Lyft et al are as good, or would be if taxis didn’t have a pick up monopoly.
From previous articles on "the knowledge" I believe that no, a machine cannot really do "better" or not yet, at least.
IIRC at the exam you could get questions like "The passenger tells you that his cousin mentioned a building with a bas-relief of Holy Mary holding Jesus, it was right around the angle from his hotel, but he does not remember the hotel's name".
You are supposed to know the place and the best way to get there, based on just this (also, a good number of "knowledgeable" taxi drivers go for an official London Tour Guide exam soon after, with relatively little effort, considering they have already memorized thousands of places, monuments and architectural details).
I may concede that it can perform “adequately at a lower price point” (I use Google myself when I am not in my hometown), but the parent wasn’t talking about price.
The Knowledge has a very important secondary effect - it creates a huge sunk cost, vastly increasing the downside of losing your license. Getting a green badge takes three to four years of full-time study. You can lose it for relatively minor offences like short-changing a passenger or taking a tourist on the scenic route. Hackney cab drivers with a green badge have a far stronger incentive to do business honestly than a minicab or Uber driver.
>> Does it? A machine can indeed do a better job (in the sense of requiring negligible training) of knowing all the streets and democratized “the knowledge” which previously was the purview of a small guild.
A theoretical machine could do a better job. In practice, we can't build anything that can reason about routes anywhere near as well as London cabbies do. And I'd say that even without knowing about The Knowledge.
Pamar's comment above has a very good example of the kind of thing "a machine" is still very, very far from achieving.
People can learn, of course. I'm just saying that a an ever changing score of Uber drivers will not be comparable to drivers with "the Knowledge". Any other claim feels preposterous to me.
This can of course be replicated with a computer, but a simple GPS is not going to cut it. Advanced machine learning, voice recognition like nothing I have ever seen, very good A.I. to resolve misunderstandings about where tourists and other non frequent travellers might really want to go instead of where they actually said. And so on and so on. We can get there. An underpaid, recently started and soon to be replaced, stressed uber driver in his uncles car armed with a GPS app is just not going to cut it. Yet.
I haven’t seen any studies of what they might give up in the process, though it does involve a lot of sitting.
Example: https://www.scientificamerican.com/article/london-taxi-memor...