Redefining what a map can be with new information and AI • Hacker News Board

Re: Redefining what a map can be with new information and AI

#271

Post by Loughla » Thu, Apr 01, 2021, 1:54 PM UTC

cam_l wrote: post #163

oceliker wrote: post #46

Earlier quoted context omitted.

> If I could, I'd usually choose the lowest stress route! Apple Maps sometimes features alternate routes that are "simpler" or with "fewer turns", which I think is a very nice feature. I'd rather not cut through small side streets just to save 1-2 minutes.

Google maps often sends me on unsealed roads to save a few minutes, when driving in the country. Don't get me wrong, I love this feature but I can understand others would not.

Unsealed. What a fascinating and more specific word I just learned today for what we call gravel roads.

That being said, you really have to watch that. There is a popular google maps cut-through near me that takes you down a road we call "bridge out" for, well, obvious reasons. At least once or twice a year the local tow truck operator has to figure out how to haul a car out of the creek down there.

There is another road that has been closed long enough that there is now a 30-40' oak tree in the middle of it, and yet google maps thinks it is an appropriate cut through between highways.

You really have to watch that in the country. The data is not great sometimes.

Re: Redefining what a map can be with new information and AI

#272

Post by eru » Thu, Apr 01, 2021, 2:09 PM UTC

lovemenot wrote: post #197

eru wrote: post #185

Earlier quoted context omitted.

> On its face, this argument is clearly wrong. "Absolutely nothing", is not equivalent to extremely little. Especially when, as in the case of Google Maps, there is a large multiplier. It makes sense to look at these things normalised per-capita. So for Google, making a small change has a big impact per-Googler, but it's still a small change per-capita. Making a small per-capita change is still a small change overall…

Though we might agree on the first premise below, there is an error in subsequent logical consequences: "small changes alone are not enough" -> "only big changes can work" -> "small change X ought not to be undertaken" To me parts of this looks like the fallacy of the excluded middle. To claim that X has some merit is not to claim that X on its own has sufficient merit. I apologise if you are not making such an argum…

> "small changes alone are not enough" -> "only big changes can work" -> "small change X ought not to be undertaken"

I agree that this is wrong.

To make small changes do a lot, you have to have a lot of them.

You can still judge small changes by how much bang they provide for the buck. (Or more formally, employ the standard marginalist framework of economics.) A small impact is fine, if the cost was small, too.

Wind-power (and other renewables) are a bit complicated too judge on these historic efficiency measures, because there was so much government interference.

(I think government should perhaps tax CO2 and other emissions, but not offer any subsidies. I want eg turning the lights on less be on the same footing as switching to 'green' energy suppliers.)

(I also don't think the government subsidies for wind and photovoltaic actually helped that much in the long run. Eg as you can see, once the subsidies ran out, the photovoltaic industry mostly left Germany for China.)

Re: Redefining what a map can be with new information and AI

#273

Post by eru » Thu, Apr 01, 2021, 2:14 PM UTC

chrisco255 wrote: post #237

eru wrote: post #177

Earlier quoted context omitted.

Keep in mind that your car emits CO2 mainly from burning fuel. Even if you don't care about CO2, you might want to care about burning less fuel. That stuff ain't free. (And the beauty of CO2 taxes would be that they make worrying about fuel costs equivalent (or equivalent enough) to worrying about emissions.)

We have taxes already in the form of taxes on every gallon of gas that's sold. 50 cents or more per gallon depending on the area. It's great if you're rich and successful and work from home in tech to argue for higher taxes. Not so much when you have no choice but to travel for work and you have to spend several hundred dollars a month to do so.

I grew up in Europe. American petros prices (and petrol taxes) look extremely cheap to me in any case.

> Not so much when you have no choice but to travel for work and you have to spend several hundred dollars a month to do so.

People always have choices on the margin.

In any case, I am not arguing for a higher total tax burden. Eg I'd be very happy for the CO2 tax to be distributed equally to all voters. Or for petrol taxes to substitute for other government revenue.

Eg in my adopted home of Singapore there's a cap of about one million cars on the road total in the whole country. To operate a car, you need a Certificate of Entitlement (CoE). Each CoE is valid for ten years. Each month approximately 1% of CoE run out and are thus re-issued via an auction.

We also have pretty high pretrol taxes.

But overall government revenue is only about 15% of GDP.

Re: Redefining what a map can be with new information and AI

#274

Post by red0point » Thu, Apr 01, 2021, 2:34 PM UTC

I find it hilarious that the GIF showing it in action at the airport is giving totally wrong directions.

It sais "Go down one level to Floor 2", to apparently reach gate E27. However, if you go down this escalator, all you will find is 2 train tracks with trains taking you out of the airport. You indeed need to take a train to gates E, but those depart from a totally different location.

Re: Redefining what a map can be with new information and AI

#275

Post by logshipper » Thu, Apr 01, 2021, 2:45 PM UTC

SXX wrote: post #69

logshipper wrote: post #47

Earlier quoted context omitted.

I'm sorry, this is is dangerously close to conspiracy theory territory. I am by no means, an apologist for Google, but I seriously doubt they have the ability/resources, or even motivation to manage route planning at such a microscopic level. I recognize that they have been caught with their pants down with things like altering search results [1], but making broad speculative claims like this about the real world wit…

My bad. I guess I really should've made it clear enough that I not suggest there is some conspiracy going on. I just replied with few ideas on how this can be used in more of black mirror scenario.

Thanks for clarifying :)

I suppose it is not _entirely_ out of the realm of possibility, but I'd be willing to bet that they won't end up taking that "route".

Re: Redefining what a map can be with new information and AI

#276

Post by jozvolskyef » Thu, Apr 01, 2021, 3:28 PM UTC

jozvolskyef wrote: post #233

bayindirh wrote: post #230

Earlier quoted context omitted.

I've heard only two things about start/stop systems: 1. I don't like it (Folks didn't have a clear answer. Maybe ego?). 2. Batteries are too expensive (Which I think is acceptable for such tech).

I happen to be one of the people who didn't like it until I got used to it. The reason is simple - if you'd been dependent on an unreliable car in the past, starting the engine is associated with all sorts of uncertainties. Starters in particular seem to be the part that fails the most often in pre-stop-start cars. With each start there's a chance something will go wrong. This then translates to a bit of anxiety ever…

I should have mentioned that my experience with stop-start is only with manual transmission cars that start the engine upon touching the clutch, sometimes before it is in full contact and usually before I'm ready to release it. If I'm going for a fast start I start the engine preemptively by lightly touching and releasing the clutch.

The other commenter's frustration with automatic transmission vehicles that start the engine upon releasing the brake sounds valid to me, especially if there isn't a convenient way to start the engine preemptively.

Re: Redefining what a map can be with new information and AI

#277

Post by bayindirh » Thu, Apr 01, 2021, 3:40 PM UTC

jozvolskyef wrote: post #276

jozvolskyef wrote: post #233

Earlier quoted context omitted.

I happen to be one of the people who didn't like it until I got used to it. The reason is simple - if you'd been dependent on an unreliable car in the past, starting the engine is associated with all sorts of uncertainties. Starters in particular seem to be the part that fails the most often in pre-stop-start cars. With each start there's a chance something will go wrong. This then translates to a bit of anxiety ever…

I should have mentioned that my experience with stop-start is only with manual transmission cars that start the engine upon touching the clutch, sometimes before it is in full contact and usually before I'm ready to release it. If I'm going for a fast start I start the engine preemptively by lightly touching and releasing the clutch. The other commenter's frustration with automatic transmission vehicles that start th…

I have used two cars with start/stop: Renault Megane and VW Polo. Both were diesels and had automatic transmission.

Both of them started the engine just before my foot leave the brake pedal. I think they sensed the rate I eased them and decided that I'm going to roll.

I also drive an automatic Focus as my daily driver which doesn't have start/stop and didn't have a noticeable difference in their driving experience during stops, TBH.

Maybe some cars have different implementations and algorithms. A brand had on demand stop IIRC. You pressed brake a little harder to command the car to stop the engine.

Re: Redefining what a map can be with new information and AI

#278

Post by ehsankia » Thu, Apr 01, 2021, 4:03 PM UTC

dkdk8283 wrote: post #137

ehsankia wrote: post #88

Earlier quoted context omitted.

Google also explicitly says that it only does it when the ETAs are very close, so realistically, it will still be doing what you want. It feels pretty extreme to write such an angry comment about maybe getting a route than 1-2% slower.

Define “close”. Is it a fixed threshold or a percentage of trip length?

I agree it's a bit vague, but realistically I would hope/think it's never more than min(2%, 5m). It obviously won't be 2m slower on a 4m ride, or 30m slower on a 5 hour ride. I don't think any reasonable person would say those are "close".

Re: Redefining what a map can be with new information and AI

#279

Post by uoaei » Thu, Apr 01, 2021, 4:38 PM UTC

aaron695 wrote: post #246

uoaei wrote: post #198

Earlier quoted context omitted.

What about this is AI? This is all perfectly doable with "classical" algorithms.

> What about this is AI? The PR Title - "Redefining what a map can be with new information and AI". And the PR last line - "All of these updates are possible thanks to AI advancements" And that might be it. > "classical" algorithms They probably just need to change a constant on the time to tell the user to change lanes and they'd make a difference on screw ups. The fact Apple now says "sorry" when it makes a mistake…

You have found yourself on a comment thread only discussing the eco-friendly routing option, which requires nothing but a fancier version of Dijkstra's algorithm.

Re: Redefining what a map can be with new information and AI

#280

Post by jachee » Thu, Apr 01, 2021, 6:25 PM UTC

nightski wrote: post #162

jachee wrote: post #144

Earlier quoted context omitted.

Hypothetically, given the choice between a 10 minute, higher-environmental-impact route and an 11-minute, lower-impact one: Are you really going to take the literally selfish , worse-for-the-planet route to save one minute? Wouldn't you like to make a small "sacrifice" (the choice is only given when the ETAs are similar) that might help slow the destruction of the only environment we have?

I don't think it's fair to call it "selfish". If I am someone who only drives a few thousand miles a year or even less (this is actually true for me), then wanting to save a minute or two isn't exactly selfish compared to a person who drives 10-20k miles a year. You can't really judge on the scale of one trip. I'm fine with the feature though as long as it gives an option. In fact, I would enjoy having the new capabi…

IF there are 10-20k "people like you" making the "not really selfish" choice, they're doing as much harm as the person who drives 10-20k miles a year.

It's about network effects. The more people we get to make the more eco-friendly choice, the better the place gets for all of us.