London’s Legendary Taxi-Driver Test Puts Up a Fight in the Age of GPS
tmagazine.blogs.nytimes.com
tmagazine.blogs.nytimes.com
Uber drivers in hilly cities like San Francisco never, ever think to avoid climbing right up and down the steepest hills - the algorithms don't tell them not to, there's no traffic and it's not a low speed limit, it just sucks for other reasons.
I had a driver last week who had never heard of Valencia St, and then misunderstood the GPS voice and accidently got on the freeway, getting us stuck in cross-bay traffic for 20 minutes just trying to get to the next exit. Of course I rated that ride 5 stars, because I'm not actually a psychopath ( http://www.forbes.com/sites/jeffbercovici/2014/08/14/what-ar... )
It's going to be a long, long time before we put anything approaching the depth of information that years of experience brings into our algorithmic routing systems. And startups are structurally disincentivized from trying to, anyway.
I did a few years in taxis but owning a car that is CC exempt is cheaper including parking in NCP holes. Part of the reason its cheaper is after some experience driving around the city, you realise that at least half the taxi drivers are taking the piss.
"the knowledge" is imminently replaceable by technology and is merely a party trick IMHO.
Once I had gotten a text from my wife as to where to meet her for dinner with friends. She didn't know the restaurant name, and had gotten the street name wrong. I just read the text to my London black cab driver, he asked me: "Did you mean XXX restaurant on [correct street name]?" (Note: there were multiple restaurants on the street). And sure enough, he was 100% correct.
That's the day I learned to really respect drivers with The Knowledge.
Conversely, got a taxi from Angel to Leyton and the idiot went via Poplar and the A13 and the journey cost me £55 for a £30 journey.
As you rightly point out, it doesn't mean that drivers won't still take the piss, but it definitely limits it. (Particularly when I compare it to virtually any other large city I've traveled to.)
There's an expectation that drivers keep their knowledge current. Certainly the examiners keep their knowledge current and test based on that.
Anyway, back on topic, "the knowledge" also covers alternative routes for road closures. I've usually been pretty impressed by black cabs. They know obscure little streets on the opposite side of town. Rarely have to give instructions. Compared to NYC where cabbies ask me how to get somewhere.
Yes of course!! Some other party tricks include reciting hours of lyric poetry from memory, adding large numbers in your head quickly, etc.
You're talking like the world is fair or something. There's not necessarily an equivalence between effort expended and value created.
This is really not the case with my London cab experience. These guys are super professional. In many other cities I've found cabs taking the scenic route but never with black cabs. I feel your 'half' claim extremely inflated off probably what was very few bad experiences.
I work on a software for routing deliveries (https://www.fleetnavi.com), and although we don't have this functionality built into the platform yet, it's the first thing to do when we get volume.
I don't think it's as trivial because there are lots of things to consider. Some drivers are just SLOW and it will have nothing to do with traffic. Maybe all slow/terrible drivers use your app so all the route speed information is completely based on specific patterns set by your app.
I am in a major metropolis and I drive throughout the city, suburbs and adjacent cities regularly. I know the area and I have been driving for decades.
I have been using Waze for about a year now. The navigation routes it selects appear to be very good. They appear to be based on relevant factors too. I can only guess what variables they include in their calculation, such as weighted capacity, current traffic speed, etc. of the relevant route elements.
My experience with Waze is the opposite of yours.
It was (bar the web front end) a relatively modest sub-30kloc chunk of C++.
Now the scale is different but in 10 years, the problem isn't necessarily much more complicated, just larger.
In fact I think a derivative of it now runs some of Yodel's operations.
I also love this:
the problem isn't necessarily much more complicated, just larger
If a programmer ever said this to me, it'd be a massive red flag that the estimate they're about to give me is going to be very wrong.
Also, there is no need to get a perfect solution anything better than what's out there is useful.
Ummmm... yeah, scaling the kinds of algorithms that one might apply to this sort of problem up to larger data sets is often extremely difficult.
30kloc of COM/IDL as well that is. The actual routing engine was about 12kloc and the rest was the object model and deserializers. The routing engine was built on a two week binge of Red Bull, TAOCP and some papers printed out.
It really wasn't rocket science even though we charged them like it was.
From scratch by a non red bull enhanced team using c# instead it would probably about 3 months' work for two people to reproduce.
This is hardly a trivial problem. It might be trivial to deal with a simplified mathematical model of it, but throw people into the mix and it suddenly gets very complicated. http://en.wikipedia.org/wiki/Reflexivity_%28social_theory%29
Google Maps would also change my London commute route on a daily basis given the changes in traffic patterns.
Aren't they all? Until you actually try to do it...
[In seriousness, we are excessively bad at estimating the true difficulty of problems, and making them sound 'easy' when they're not only serves to devalue the work of the people who actually spend the time and effort to actually figure it out.]
0: http://en.wikipedia.org/wiki/Disruptive_innovation 1: http://upload.wikimedia.org/wikipedia/commons/8/8e/Disruptiv...
Google Maps directions is pretty disruptive tech, right? Add 10 years of experience, lots of roads, lots of time to gather data, real-time traffic info, and when I ask it for directions from my current location in inner London to a place in central London[1] it still gives driving directions first, even though I switch to transit, walking, or cycling everytime.
But hey, innovation.
0: http://en.wikipedia.org/wiki/Clueless 1: http://i.imgur.com/c7XAuUm.png
Of course not. Technology will eat the world and improve on and replace humans in all complex cases. Like deciding what mode of transport to use in central London. Because disruption.
Strongly disagreed. GPS has traffic information, a human driver does not.
Perhaps less anecdotally, local traffic is a major staple of local news radio and TV, which probably means it's important even for locals.
Its also just a matter of time until cars provide visual aids for navigation, and not just the voice on a GPS device.
You are doing everyone except the driver a disservice. By failing to give low ratings to bad drivers, you are precluding good drivers from entering the market (because the bad ones will never get phased out, and they will saturate the market). You are also preventing consumers from making a well-informed decision, which limits the effectiveness of market mechanisms.
B2C seems a bit different though you still see more 5s than deserved. Personally, I have no problem giving 4s for good to very good, but not truly outstanding.
You accidentally rating someone 3 for an average service can actually cost them their job. I'm hoping their algorithm uses some sort of a weighted average to offset the ratings of individuals such as yourself who consistently rate very low.
i.e. If I always rate 3, but sometimes rate 1; then that might be similar to someone always rating 5 but sometimes rating 1-3.
Ah, but then a user might throw an Internet Fit(tm) and cause a public outcry about ratings being manipulated by Uber etc. etc.
There are a lot of ways in which a ride can be unsatisfactory (smelly, creepy, crappy car, poor driving, loud music, etc.) but not that many in which it can be exceptional. Unless they show up in a Tesla Roadster or something.
Quoting from http://www.youtube.com/watch?feature=player_detailpage&v=C4r...:
"Engineers do not give out 5s. If you tell an engineer that the possible rating scale is 1 to 5 and they like you, they give you a 4. Because tomorrow they might love you and then they would give you a 5. They probably won't, but they might, so they give you a 4 so that they have room. We all are totally okay with this. But the world is also full of this other class of people who if they like you give you a 5. They're quite consistent about it. 'I like you. I see nothing wrong with that. I'm giving you a 5.'"
The graph (with, admittedly, artificial data meant to portray an example) at http://www.youtube.com/watch?feature=player_detailpage&v=C4r... shows it most clearly, with the satisfaction numbers from engineering clearly trending down. Apparently, engineers don't like to fill out forms, while admins do, so satisfaction survey results also have a sample bias in favor of the 5s of admin staff.
I believe this is yet another case of managers using data without actually taking the time to understand the domain, or use it to identify problems and develop solutions. The approach seems to be nothing more than a facade of customer service.
They let users crowd-source their own directions, along with much better turn-by-turn descriptions too. Eg, "turn sharp right just after the firehouse, go past the field and make a left just before the baseball field". Instead of "turn right, keep left and turn left".
"No, don't go to the lights, turn down this lane!". All of them had their brains explode that there was a better way that saved about 200 meters and 30 seconds of travel.
I'm not sure if there's an (unwritten?) rule that you get the rest of the fare for free if they're taking you a long way, but there was no resentment from them. All of them were incredulous there was a better way to do it.
(Just looked at Streetview. There's nothing special about that lane that should make it a big secret. I guess it must have been No Entry/One Way/something and changed recently before my journeys.)
As I said, the route seemed really obvious to me. It was a quicker way to get from Kingsland Rd to Hackney Rd, so it was two major roads in central London.
The short cut involved taking a narrow laneway, and i'm guessing something about that lane changed meaning it had only just become accessible.
I also had cabbies turn off the meter late at night. I'd light up a cigarette (back when that was allowed), they'd ask if I cared if they lit up too, and once a guy said "Ahhh, I was heading your way anyway" and he turned off the meter. I lived in East London surrounded by cab drivers, so maybe they were being neighbourly.
Folks with the "knowledge" could act as centralized route planners. The can provide many more routes in the time they could drive one (i.e. a single cabby driving might do 1 call every 20 minutes... but could easily do 10 route plans in the same time).
Essentially when someone asks the minicab driver "take me to X", the driver relays this information to a black cab driver at a desk who sends back a custom route. With GPS to identify the starting point and VOIP (via wireless) this could be as easy as a driver using a traditional radio (press the mike button, speak the destination, and viola the route appears on a dash mounted map along with spoken instructions).
Very much like the existing examination question/answer format.
Maybe you would also need to rotate the route planners out as drivers 1 week out of 4 so they keep up with the "knowledge".
Now most often the first question is "what's the address so I can put it into my GPS", even with metered cabs. I've been yelled at for asking to take routes that the GPS doesn't suggest, even though I know at 3pm on a Tuesday we shouldn't be on the BQE. Or there's construction on Flatbush and we should go around it.
I don't think we need a test quite like London's, but basic knowledge of neighborhoods and major intersections should be required. At least until we get those self driving cars that can optimize routes in real time.
And who cares if the cabbie doesn't like you. It's your right to demand an exact route as long as you're staying in the city. Tell him you're calling 311 and don't tip.
Black cab drivers know so much more about the city it's not even a contest. I have almost totally switched to using Hailo in London for that reason.
Drawing from experience trying to memorize a journey through the city of Bale, which whilst less large than London is quite winding and certainly suffers from the same compounding of historical and modern buildings. Going through my route, spots where i stopped to take in more details are very much more tangible in my mind. The wide park circled by a walkway where I stopped for a short break is quite vivid in my memory, and I can pull dozens and dozens of points from it. The path along the viaduktstraße leading up to it is far more faint in my memory, considering I only remember the Paulskirche and one of the bridges towards it. Considering only something akin to Google Streetview would be practical, there seems to be just to few points your mind can latch onto to create a detailed image of the location. In addition to that, your other senses do play a significant role in remembering locations.
Also, cab drivers need to know very recent information you're tempted to overlook designing a game like that. How fast does traffic go at this time of day at this location? Are they setting up construction over here? It's often not even realized consciously, but rather factors into decisions on a lower level, turning into the "hunch" you have about a certain route or even more generally a certain course of action.
So unless there is a way to have a perfect representation of the city that is also less than a day or two old, it's not really a replacement for actually going through those streets.
If you want to track the run (it's accurate!) the starting point is right next to this node in OSM: http://www.openstreetmap.org/node/279219489
> She has discovered that the posterior hippocampus, the area of the brain known to be important for memory, is bigger in London taxi drivers than in most people, and that a successful Knowledge candidate’s posterior hippocampus enlarges as he progresses through the test.
It seems that, like most things, there might be a tradeoff involved. According to this paper[0],
> However, they were significantly worse at forming and retaining new associations involving visual information. We consider possible reasons for this decreased performance including the reduced grey matter volume in the anterior hippocampus of taxi drivers...
The knowledge is a very useful tool, and in any city, a driver with good local knowledge is what you are looking for, regardless the middleman you use. Technology is only making the trade off of using amateur bearable, and in reality the real value of Uber and others is from the rest of the service provided.
A bit like in your example - if you had to go to the bookstore to load a new book on your reader and it could only hold one book at a time, the technological superiority of even the latest generation of kindle would not make them go pass novelty items. Real value of ebook comes from the rest of the services you get from it and that's the reason why even with the shit reading experience, the first gen kindle was a success.
The anachronism here is the segregation between black cab and minicab in London and the services the legal restrictions on the services they can offer. Although, IMO, that is not nearly as bad in London as people on HN experience in other places in the world. Black cab are meant to hailed on the street, now, without booking, review or other overhead by drunk people and tourists. Some licensing restriction to make sure you get a basic level of service in those circumstances is probably acceptable. If you have time to plan a bit in advance, there is Uber and plenty of competitors that have always been freed from the restrictions.
The Knowledge may make for better navigation, but as a customer, the amount of times it actually comes in useful (road works, accident, awkward route) is very low, 90% of journeys are straightforward.
It certainly doesn't justify the 50%-100% price increase for the privilege of hailing the cab on the street vs. by pre-booking, especially when most of the time the route you are taking is short (no Londoner in their right mind would use a black cab for a long journey, unless they have money to squander or it's so late at night that you're desperate)
The likelihood of getting any real value out of the knowledge when taking a cab in London is very low, and the fact I need to pay for idling in traffic with an un-capped fare potential on top of jacked rates just makes for a terrible proposition.
As a consumer, there's just no value in that, and Black Cabs need to evolve or compete themselves back into relevance, bring on Uber, Lyft, Wheel and any other newcomers.
Source: 14 years of being a grumpy Londoner.
Would you use our service again?
Was there anything we could have done better? (+comment)
Would you recommend us to your friends and family?
Also, let's not forget how people sometimes lose their ability to be objective when they're upset which may manifest itself in the form of a less than 5 star rating for a driver that made 1 small error or said something innocuous that we took offense to.
[1] http://www.bloomberg.com/news/2013-10-30/ups-uses-big-data-t...