Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google Will Become an AI Company

mattmaroon.com

Post reply on HN
Print view View original

71–80 of 168 posts

Re: Google Will Become an AI Company

#71

Post by pavel_lishin »

tgflynn wrote: post #30
DanielBMarkham wrote: post #2

This is a great article. Every now and then Matt can really knock one out of the park. Of all the tech that we talk about on here, there are only a few items that really catch my attention. Christmas tree machines are one of them. Auto-drive cars is the other. These two inventions, when complete, will massively change things. Good luck guessing when they'll be complete, though. Could be a decade. Could be a couple of…

The time-frame for auto driving cars is actually more difficult to predict than that for general adaptive robotics because of the regulatory and social issues involved. I'm sure it will be decades not centuries though. My prediction is that robotic servants will be common place in homes by the 2020's and that almost all manual occupations will be replaced with automation during that decade. I have no idea how society…

> almost all manual occupations will be replaced with automation during that decade

Unions will just go quietly into the night?

Re: Google Will Become an AI Company

#72

Post by pavel_lishin »

maweaver wrote: post #52
100k wrote: post #45

Traffic is caused by human error? I suppose flooding is caused by "water error", then? Traffic is caused by too many vehicles attempting to use a limited resource at the same time. Driverless cars may make this more tolerable (certainly riding the bus does) but the idea that this will make traffic obsolete is laughable.

In my experience most traffic jams are caused by bad merging, not lack of capacity. A few years ago I regularly drove a stretch of road that went from 2 lanes to 3 for a few miles and then back again, without gaining a significant amount of traffic in between (I have no idea why the road was designed that way, maybe one of the side roads feeding into it got a lot of traffic at a different time of day or something). E…

I've read several articles pointing out that traffic seems to be an emergent behavior, and can be predicted like the flow of a gas, regardless of the individuals involved.

Re: Google Will Become an AI Company

#73

Post by nowarninglabel »

Did no one stop to think about this? It is extremely far-fetched at best.

>safety regulations could be greatly relaxed.

No, at least not if the author's vision of 200mph average speeds is to be taken. When a mistake or malfunction happens at that speed, safety mechanisms will be imperative. Furthermore, having a mechanical car does not prevent: someone else running into you, a deer running in front of the car, etc.

> children could own cars

But they wouldn't, because the purchase would still be in the name of the parent. Furthermore, do you see parents sticking their 6 year olds on the subway just because they can? No. A very few do it and get ostracized by society.

> 3. The beverage industry will go.

False assumptions without supporting data, but I have no facts to counteract it.

>4. Speed limits will be unnecessary

Oh really? So we won't need limits for the existing drivers who aren't using driverless vehicles? How will the 200mph traveling car navigate around all the 60mph traveling ones? Furthermore, is every car going to be programmed to go slow in pedestrian zones? How do you enforce that without speed limits? The current Google Car wasn't jetting 200mph down the 101, it was driving under the speed limit in residential neighborhoods.

> The map will shrink greatly.

No. Fuel costs and traffic don't just magically disappear because of your fantasy land.

> Urbanization will reverse. Why pay $3,000/month for a flat in Manhattan when you can get from 100 miles upstate to work in 30 minutes?

I will. Just because you can live outside the city and travel to it at a faster rate does not make it a given that one would choose to. Urbanization has been the greatest driving factor of population trends in the last century. If anything, if what is proposed came to pass, you would see increased urbanization of small towns/suburbs.

>Airlines will be devastated. Why fly from New York to Chicago?

No. It will still be faster to fly. Are you serious? I mean gee why fly from New York to London when I can take a speed boat and have it take three days? I mean, seriously?

>9. Other forms of public transport won’t fare much better. A driverless cab won’t cost much more than a bus (which also will be driverless) but will be a hell of a lot nicer.

I'm sorry, I don't live in fantasy land where fuel costs suddenly become irrelevant. Fuel costs make up at least 16% of the overall cost. And there will still be a premium because people will be willing to pay it.

Yes, a driver less car will make someone a lot of money. Does this equate to the above points? No, the author's hypothesis has no basis in reality and no facts to back it up.

Re: Google Will Become an AI Company

#74

Post by khafra »

rimantas wrote: post #34
maeon3 wrote: post #14

Self driving cars are 15 years away. The self driving cars will have to deal with the chaotic human drivers, and this will require Strong AI. Once we have this, driving around will be one of the small issues of the day.

I hope they are further away. For me a bus is self driving car, but driving a car is also a pleasure not only the means to get from a point A to a point B.

Presumably autonomous cars would not replace the ones on closed courses, and you could still enjoy driving a car the way you enjoy all other pleasures which require specialized equipment and facilities.

Re: Google Will Become an AI Company

#75

Post by samtp »

randallsquared wrote: post #42
achompas wrote: post #22

Earlier quoted context omitted.

For many people, whatever nice benefits of density there are Here are a few that I came up with in 30sec: Varied options on things to do (like you said) Closer proximity to other people in general (more social opportunities) Commuting on foot Better access to more government services Greater selection of goods, food and produce

Except for "Greater selection of goods, food and produce", all the ones you add seem like they could be costs rather than benefits, for many people. Closer proximity to other people in general, for example, was one of the harms I was thinking of. But I'm slightly more misanthropic than most Americans, I guess. The one about greater selection isn't actually true, as far as I can tell. You can get pretty much anything…

I'm from the DC area, but now live in Columbia, SC. Every time I go back I'm reminded of how many different cultures, things to do, types of food, places to have a drink , listen to live music, etc... there are. I love Columbia, but you can't beat living in a major metro area if you like to enjoy new experiences.

Now keep in mind that I'm 22. Maybe as you get older, small towns with only a Waffle House and Jimbos Sports Bar (which closes @ 9pm) become more appealing. In most of rural America, the people are incredibly nice, but the poverty, lack of education, jobs, healthcare, and public services is appalling (theres a reason rural Tea Partiers are so angry). With that said, every place is what you make of it.

Re: Google Will Become an AI Company

#76

Post by Tichy »

mquander wrote: post #11
DanielBMarkham wrote: post #2

This is a great article. Every now and then Matt can really knock one out of the park. Of all the tech that we talk about on here, there are only a few items that really catch my attention. Christmas tree machines are one of them. Auto-drive cars is the other. These two inventions, when complete, will massively change things. Good luck guessing when they'll be complete, though. Could be a decade. Could be a couple of…

It's a little strange, because another trend directly competing with automatic cars is telecommunications. If it were possible and culturally acceptable for all white-collar workers to telecommute full-time, all of a sudden there would be half as many cars on the streets. If there were efficient, cost-effective, and prompt delivery systems for local goods like groceries, there would be half as many again. It's concei…

"If there were efficient, cost-effective, and prompt delivery systems for local goods like groceries"

Wouldn't trucks work? It sounds crazy at first, but I suspect one truck making the rounds every day, like the postman, would still be more energy efficient than hundreds of cars going back and forth to the supermarket?

Re: Google Will Become an AI Company

#77

Post by artmageddon »

100k wrote: post #45

Traffic is caused by human error? I suppose flooding is caused by "water error", then? Traffic is caused by too many vehicles attempting to use a limited resource at the same time. Driverless cars may make this more tolerable (certainly riding the bus does) but the idea that this will make traffic obsolete is laughable.

The truth is, we already have enough road. Compared to machines, our reaction times are very slow, and because of that we can't use them efficiently since we need to give each other space to slow down / change lanes. If the cars are able to communicate their intended destinations along with other things ("I'm speeding up, go" / "I'm slowing down, don't hit me" / "SYSTEM FAILURE") then we can pack more cars into the same amount of space.

Re: Google Will Become an AI Company

#78

Post by johnrob »

pavel_lishin wrote: post #71
tgflynn wrote: post #30

Earlier quoted context omitted.

The time-frame for auto driving cars is actually more difficult to predict than that for general adaptive robotics because of the regulatory and social issues involved. I'm sure it will be decades not centuries though. My prediction is that robotic servants will be common place in homes by the 2020's and that almost all manual occupations will be replaced with automation during that decade. I have no idea how society…

> almost all manual occupations will be replaced with automation during that decade Unions will just go quietly into the night?

If we can make food, medicine, and shelter almost free with the help of automation, then people would no longer need jobs to meet their survival needs.

Re: Google Will Become an AI Company

#79

Post by poet »

philwelch wrote: post #64
poet wrote: post #61

Earlier quoted context omitted.

Google's CEO doesn't think pilotless planes are possible by 2030: http://www.longbets.org/4 . Part of his reasoning: the FAA changes so slowly that if this were even all possible, the adoption and certification would all take at least 50 years. I don't see why the same wouldn't be true (perhaps on a smaller scale) for the auto industry.

Current airliners are exactly as driverless as Google's cars: they have an autopilot that does virtually all of the flying, with one or two human pilots set to take over in case they have to, just as the driverless cars Google is developing have a human driver set to take over in case they have to. (That's what makes them compatible with CA state law, as Google has been field testing their driverless cars for some ti…

This is a misunderstanding of the discussion at hand. The word driverless in the context of this article and the person I was replying to implies the lack of a human failsafe.

Schmidt and Matt address this in their arguments. For example, Matt mentions that the beverage industry will skyrocket because people can now "drive" home drunk. The human can't take over in an emergency in that situation. Pretty much every advantage of driverless cars that Matt listed are not possible if the human is required to be alert enough to take over. Further, the person I was responding to claimed he would never have to learn how to drive. If a human is going to take over in an emergency, s/he will still have to learn. Point being, the discussion at hand is about driverless vehicles without using a human as a backup.

Re: Google Will Become an AI Company

#80

Post by Micand »

Brad Templeton delivered a superb talk on robotic cars at the Singularity Summit 2009 (http://www.vimeo.com/7337628). It expounds on the technology's implications, supporting Maroon's assertion that even a small slice of the market will easily eclipse Google's stake in search. Of particular interest:

* Transportation is more dangerous than we think, and this is largely due to human factors. (Driver inattention is a factor in 80% of crashes; alcohol in 40%.)

* The purchase of private vehicles forces us into a "one size fits all" model. If someone goes skiing only once a year, he will purchase an SUV; if someone spends 90% of his mileage traveling alone to work, he'll still purchase a five-person sedan so he can haul around friends occasionally. By moving to a grid-like service that provides cars to us on demand, we will be able to choose the vehicle best suited to the type of trip we're making.

* Robotic cars could eliminate our dependence on foreign oil. Energy usage would be dramatically lower due to people using a vehicle only as large as they need for a given trip. Vehicles powered by alternative energy have a chicken-and-egg problem -- no one wants to build the infrastructure to deliver energy until people buy the vehicles, but no one wants to buy the vehicles until a ubiquitous energy infrastructure is in place. Robotic cars, however, would have no qualms with traveling halfway across a city to refuel, nor with waiting two hours in a lineup before refuelling.

* The transportation infrastructure will also become substantially more efficient, as cars will be able to travel much closer together without compromising safety. As a consequence, energy usage can be reduced another 30% by having cars draft one another.

* Before robotic cars would be accepted by the public, they'd have to meet much more stringent safety standards than we apply to human drivers. No one would accept a robotic car that killed a human, even if robotic cars on the whole were twice as safe as human drivers. Templeton figures we'll need cars on the order of 100 times safer than human drivers before they will be widely accepted. To convince people of the cars' safety, Templeton proposes the "school of fish" test -- imagine walking out onto a track swarming with cars travelling at 40 miles per hour, and having every car swerve around you no matter how hard you try to make them hit you.

* Robotic vehicles will record video everywhere they go, for this will prove invaluable in determining the cause of accidents. Any modifications to the driving software will then have the ability to be tested on the "trillion mile road test" -- they will have a corpus of testing data composed of the recorded footage of every trip ever made. New software will be tested against every vehicle accident that has ever occurred.

* The privacy implications of this universal recording are disconcerting. Templeton raises the spectre of a situation like that in Minority Report, where police can remotely override your control of a vehicle, locking you inside and transporting you to a destination of their choosing.

Post reply on HN