Yann Lecun has a nice comment on Deep Learning vs Supervised learning being the Heat Engine vs Steam Engine example, although they both seem to agree on hype.
I wonder if the 'Spam in a can' problem for underground transport can be resolved with VR or simply human absorption into BYOD.
I'm not so sure. A significant portion of the population will find the "Spam in a can" problem a claustrophobia trigger. There are people that would just never get into one. Like my wife -- who lasted about 5 minutes into a tour of a WW II submarine that was sitting on the surface with the hatches open. Which means we would never travel as a family in Spam can.
Of course there are ways of dealing with that. It will require thoughtful industrial design and human factors design, and certain minimum volume. A VR headset will itself become a claustrophobia trigger enhancer in a triggering environment. The human factors problem should not be trivialized.
Accurate speculation, which prediction is, is a skill that can be practiced and improved. This has benefits as it makes one better at estimating outcomes. Posting it online also helps keep people accountable. Yeah, there might be some promotion going on as well, but that can be true for anything one posts on line. If you find someone's post valuable, great. If not, that's fine, too.
I agree here and would underscore this point: in the hands of public experts, media talking heads, and many types of charlatan, predictions can be immensely powerful and have tremendous influence over public policy and social behavior. So I consider assessing them for accuracy and holding prediction-makers accountable a public good.
For a compelling book-length treatment of this question, see Philip Tetlock's Superforecasting:
I'm not against predicting things, there's great value in them but usually there is some science behind those predictions.
For example people who manufacture things usually offer a warranty period because they have a fairly accurate prediction model how long it will last before it breaks.
There is value in coming up with a mathematical model to predict the chances of a person paying back a loan used to buy a house.
Insurance companies run prediction models for all kinds of human behavior based on statistics.
But what is the value in random pointless predictions about when flying cars will be at thing? There's no math behind it other than intuition.
A friend of mine has a client who creates patents for future technologies 5-10 years down the line.
He understands that when this and this scientific breakthrough happens it will be applicable as a or to a technology and then he creates patents for that.
He has quite a lot of patents and make good money on it.
I don't understand why so many people are hung up on flying cars. It's a huge increase in the energy it takes to get from point A to point B, and for what reason? What do they hope to gain by wasting so much more fuel?
I'm sure fuel consumption isn't a concern, after all it didn't come up in Back to the Future, it just looked cool.
And I imagine people like the idea of going as the crow flies, shorter travel times, no traffic congestion and not having to fuss over road construction.
And since it's just a flight of fancy they don't have to consider the risks of falling out of the sky, mid air collisions, noise pollution, increased energy consumption, severe weather, running out of fuel, running out of fuel over Lack Michigan or the Atlantic Ocean or the human factor of letting people who can barely handle X and Y also deal with Z.
Maybe some people just have general optimism that the imagined benefits outweigh the obvious problems.
Flying cars are a terrible idea. People can't be trusted to not drive into things or run out of gas while driving _on land_. I wouldn't want the average person flying their car anywhere near me or my property.
The only conceivable way I could see flying cars being a 'thing' is if we carved out 'air roads' for them that didn't cross over buildings or other cars. At that point, what's the benefit?
What about self-driving flying cars? If people aren't driving the flying cars and they just navigate automatically then most of these sort of issues go away.
In a sense air roads could be safer. You have more flexibility for avoiding obstacles. I think it only makes sense if the vehicles on them are fully automated and operate in a highly regulated environment.
Maybe a big part of the solution is self-driving flying cars? Of course they would use Tesla batteries. And let's not forget to use Fuschia, while we're at it.
There's not much stuff to crash into in the air other than other air traffic which is far less densely distributed than things one shouldn't drive a car onto/into are on the earth's surface.
I don't get it either. We have no technology in view that would not produce a lot more noise and energy usage over current cars. It may work once energy costs almost nothing and can be produced in abundance (fusion?).
A Cessna 172 gets about 12-13 mpg at about 120 mph. [1] That may not be an unreasonable tradeoff of performance versus cost per mile if considerations other than fuel consumption are in play.
On the other hand, you'll never get a flying car because FAA regulations form a moat around widespread cheap aircraft. A 40 year old Cessna 172 can only have an engine provided by Lycoming, can only be overhauled or repaired with Lycoming and Cessna parts, and can only be serviced by FAA certified mechanics (even though the design is only slightly more sophisticated than a VW aircooled engine and is based on 1930's technology).
My Bonanza A36TN gets that mpg at over 200mph, with 6 total seats as well. Boston to Myrtle Beach area in 3.5 hours and ~60 gallons avgas vs 15 hours and 35 gallons of car gas; I find that tradeoff to be very much worth it.
FAA regulations for certified aircraft are indeed a huge hurdle on progress, but are not at all a hurdle for experimental aircraft (which is where flying cars are today). (Side note: private aircraft owners flying under part 91 [roughly: not charter, not airline] are allowed to produce or cause to be produced "owner produced parts", so not all parts need to come from the OEM. It plays out that most engine parts do, but many airframe, interior, etc parts do not.)
For me the flying car is a solution in search of a problem. When I fly somewhere, there are almost always rental cars (or rideshares) widely and easily available. I don't want to take my lightweight, expensive, and somewhat damage-sensitive airplane into traffic 600 miles from home where a fender bender leaves me unable to fly home. That's what a rental car is for.
Would it be nice to land in my own "car" and immediately drive to my ground-bound destination without having to unpack/repack? Sure.
Am I willing to accept the inherent compromises, expenses, and risks implied by that? Absolutely not.
Personally, I always assumed a "flying car" is something like a quadcopter, which does not have to drive at all. Just take off from your roof and land on the roof of where you want to be.
What happens when you want to have a barbeque and have 10 of your friends over? What happens in a high-rise building? Other circumstances where available roof space is less (to much less) than demand for quad-copter parking?
And quad copters are pretty dang inefficient (as are helicopters in hover mode), making the energy hurdle higher than fixed wing (or translational helicopter) flight.
Quads might end up being part of the solution (and I admit I hadn't considered them before, so thanks!), but I don't think they're well suited for covering significant distances, which is presumably where the appeal of a flying car is strongest.
My God. How did I miss that? Thank you for pointing out the blindingly obvious limitation in my thinking about cars vs autonomous cars. (This is a perfectly genuine response; zero sarcasm.)
It might be that the flying car appears the way it does because you have access to a private plane. I mean the functionality of the flying car is ubiquitous flying vehicles. There are ~250,000 general aviation aircraft in the US. [1] Or about the same number of general aviation aircraft in service as the number of model year 2015 and 2016 Kia Sorrentos sold in the US. [2]
Totally agree. The biggest gain aside from novelty would be increased speed/avoiding traffic, but if that's going to double or triple the cost of a trip, I'm sure most people would chose to sit in traffic.
The only one of those with even the remotest chance of happening is the Cryptocurrency one.
A Robocop by 2020? If you had a Robocop by 2090 I'd be shocked. (A remotely controlled Robocop is not a Robocop, you said helped by AI, and that's the part I'm focusing on.)
Just try to talking to one of those assistants like Google Home, and you'll realize just how far away we are from AI.
> [Driverless cars] Today, in 2017, I believe that at least 19 out of 20 people would agree.
Nope, not happening. Driverless cars (except in dedicated lanes) are not going to happen for 20 or 30 years if not longer.
What I do agree will happen is dedicated instrumented lanes for them on freeways (little point in doing it in cities). Kind of like train rails.
But your vision of them transforming cities is not going to happen for decades, if ever.
Personally, I think many of your time estimates are wildly pessimistic. Humanity is on an improvement graph that almost seems exponential - I don't think it would be crazy to see driverless cars in the next 5-10 years, considering how many people continue to get in to this space. There may be strict legislation around it, but that's a different discussion. Similarly, look how far we've come in "assistants" in the last 6 years since Siri was released... 2090 definitely seems like a stretch.
We've been able to do driverless cars since the 90s. It's doing it safely while mixed in with normal cars that's the problem. I'm with the other poster that we won't see driverless cars in non-dedicated lanes for at least 20 years.
Edit: actually the article mentions a self driving car from 1987, so even earlier.
look how far we've come in "assistants" in the last 6 years since Siri was released
Not far at all. I still struggle to do basic tasks with my Apple Watch 3 (like sending a text to one of my contacts, or automatically start/stop Run app when I start/stop running).
To be fair, look how far we've come in "assistants" in the last 6 years since Siri was released can be read as “look how much Siri has been surpassed in quality by other assistants since it was introduced” rather than “look how far Siri has advanced since it was introduced”.
What's the trigger event for the global recession you are predicting? You have listed some "worrying" factors - but there is no clear interconnection between them or trigger for any one to cause a global recession. Finding and predicting the trigger is the hard part IMO, as well as how governments will respond (more bailouts and QE?).
Can you point to a source for your prediction of a robotic cop killing someone by 2023?
Agree with Facebook "political party" instead of Zuckerberg running a campaign - not sure how this differs much from the traditional way corporations influence politics through donations, contributions, advertising, etc.
Your prediction of driverless cars doesn't mesh well with Rodney Brooks' prediction of driverless car saturation in 2045, he predicts we will have one dedicated lane for hands-free driving by 2021 and limited taxi/bus services by 2028. What is your basis for more aggressive forecast?
Is that question meant to be a joke? Ocean waves most definitely have external triggers.
The most common one is wind, but changes in atmospheric pressure can cause them, as can a sudden change in the sea floor due to an earthquake or volcano (this is what causes tsunamis). The rotation of the earth and the gravitational pull of the moon can cause waves when it moves water confined spaces - this kind of wave is usually called a tidal bore.
Is there some kind of ocean wave I'm unaware of that can spontaneously occur without some external force/trigger?
The facebook will probably happen. Facebook has already taken sides. There already is a Fox party and a CNN/MSNBC party. Before that there was a New York Times, Wall Street Journal Party.
Global Economic Recession is a safe one. The question is when?
I think Cryptcurrencies will crash with the recession. If a recession hits people are going to want to cash in their crypto for real money and guess what will happen to crypto?
Driverless / Flying / ... cars are probably the biggest change since the cell phone and internet. The only weird thing about them is their timming. People kind of travel less. Shopping is done over the internet, education is done over the internet, jobs are done over the internet. Socializing is done over the internet, Church over the internet.
Since he brings up a number of Elon Musk initiatives, it seems to relate in part to the public conversation that Elon has been having about AI.
For those who don't know Rodney Brooks, he was faculty at MIT for years, co-founded iRobot and Rethink Robotics, where he is working on collaborative robotics (robots that can work with humans without maiming them.)
It kind of gives you a sense of the sort of messaging that Dr. Brooks doesn't have a lot of respect for. I have big fan of Brooks' work from his MIT days on, it is very practical robotics which has done things that others haven't (Ghengis, the walking robot, really was a revolution in terms of mobility[1]).
But hype cycle or not, it would have been interesting to see where he had placed reusable rocketry in 2007. Would it have been 'by 2020 it will be common' or would it have been 'won't happen, the existing rockets work quite well and there isn't enough gain to justify the additional work.'
While Brooks picks on Elon a lot, the core of his essay is what makes it 'easy' to predict something will happen versus what makes it 'hard'. And I think he doesn't give enough credit to the fact that even if something is possible it needs a catalyst to get it to precipitate out of the realm of possibility into the realm of reality.
I think Dr. Brooks has a blind spot on space challenges, but I enjoyed his AI predictions. Of course it is entirely possible I just want easy access to space more than he does :-)
[1] I'm guessing @animats will disagree with that assessment :-) But he and Dr. Brooks approached the problem of inverse kinematics very differently and both got good results.
So would bundling large numbers of people into a single vehicle for the bulk of the journey and finding other ways to tackle the last-mile problem. Just imagine: transportation for the masses!
50 comments
[ 3.1 ms ] story [ 110 ms ] threadAbbreviations are:
* NIML meaning "Not In My Lifetime", i.e., not until after January 1st, 2050
* NET some date, meaning "No Earlier Than" that date.
* BY some date, meaning "By" that date.
I wonder if the 'Spam in a can' problem for underground transport can be resolved with VR or simply human absorption into BYOD.
Of course there are ways of dealing with that. It will require thoughtful industrial design and human factors design, and certain minimum volume. A VR headset will itself become a claustrophobia trigger enhancer in a triggering environment. The human factors problem should not be trivialized.
For a compelling book-length treatment of this question, see Philip Tetlock's Superforecasting:
https://en.wikipedia.org/wiki/Superforecasting
This project was spun off from the book and is my favorite online forecasting site:
https://en.wikipedia.org/wiki/The_Good_Judgment_Project
Your comment, on the other hand, is basically serving as a downvote-as-a-comment.
For example people who manufacture things usually offer a warranty period because they have a fairly accurate prediction model how long it will last before it breaks.
There is value in coming up with a mathematical model to predict the chances of a person paying back a loan used to buy a house.
Insurance companies run prediction models for all kinds of human behavior based on statistics.
But what is the value in random pointless predictions about when flying cars will be at thing? There's no math behind it other than intuition.
He understands that when this and this scientific breakthrough happens it will be applicable as a or to a technology and then he creates patents for that.
He has quite a lot of patents and make good money on it.
Not judging just saying.
Organize your own thoughts?
Teach readers new things in a way that directly ties the information to their current experience?
And so many more.
And I imagine people like the idea of going as the crow flies, shorter travel times, no traffic congestion and not having to fuss over road construction.
And since it's just a flight of fancy they don't have to consider the risks of falling out of the sky, mid air collisions, noise pollution, increased energy consumption, severe weather, running out of fuel, running out of fuel over Lack Michigan or the Atlantic Ocean or the human factor of letting people who can barely handle X and Y also deal with Z.
Maybe some people just have general optimism that the imagined benefits outweigh the obvious problems.
The only conceivable way I could see flying cars being a 'thing' is if we carved out 'air roads' for them that didn't cross over buildings or other cars. At that point, what's the benefit?
On the other hand, you'll never get a flying car because FAA regulations form a moat around widespread cheap aircraft. A 40 year old Cessna 172 can only have an engine provided by Lycoming, can only be overhauled or repaired with Lycoming and Cessna parts, and can only be serviced by FAA certified mechanics (even though the design is only slightly more sophisticated than a VW aircooled engine and is based on 1930's technology).
[1]: https://www.quora.com/How-many-miles-per-gallon-mpg-does-a-C...
FAA regulations for certified aircraft are indeed a huge hurdle on progress, but are not at all a hurdle for experimental aircraft (which is where flying cars are today). (Side note: private aircraft owners flying under part 91 [roughly: not charter, not airline] are allowed to produce or cause to be produced "owner produced parts", so not all parts need to come from the OEM. It plays out that most engine parts do, but many airframe, interior, etc parts do not.)
For me the flying car is a solution in search of a problem. When I fly somewhere, there are almost always rental cars (or rideshares) widely and easily available. I don't want to take my lightweight, expensive, and somewhat damage-sensitive airplane into traffic 600 miles from home where a fender bender leaves me unable to fly home. That's what a rental car is for.
Would it be nice to land in my own "car" and immediately drive to my ground-bound destination without having to unpack/repack? Sure.
Am I willing to accept the inherent compromises, expenses, and risks implied by that? Absolutely not.
Do I think the market is? Nope.
And quad copters are pretty dang inefficient (as are helicopters in hover mode), making the energy hurdle higher than fixed wing (or translational helicopter) flight.
Quads might end up being part of the solution (and I admit I hadn't considered them before, so thanks!), but I don't think they're well suited for covering significant distances, which is presumably where the appeal of a flying car is strongest.
As for efficiency, yes, no simple solutions there. Note that they can still have fixed wings, the hover mode is only necessary for take off/landing.
[1]: https://www.aopa.org/about/general-aviation-statistics/activ...
[2]: https://en.wikipedia.org/wiki/Kia_Sorento#Sales
Happy to hear comments about it.
A Robocop by 2020? If you had a Robocop by 2090 I'd be shocked. (A remotely controlled Robocop is not a Robocop, you said helped by AI, and that's the part I'm focusing on.)
Just try to talking to one of those assistants like Google Home, and you'll realize just how far away we are from AI.
> [Driverless cars] Today, in 2017, I believe that at least 19 out of 20 people would agree.
Nope, not happening. Driverless cars (except in dedicated lanes) are not going to happen for 20 or 30 years if not longer.
What I do agree will happen is dedicated instrumented lanes for them on freeways (little point in doing it in cities). Kind of like train rails.
But your vision of them transforming cities is not going to happen for decades, if ever.
Edit: actually the article mentions a self driving car from 1987, so even earlier.
Not far at all. I still struggle to do basic tasks with my Apple Watch 3 (like sending a text to one of my contacts, or automatically start/stop Run app when I start/stop running).
Can you point to a source for your prediction of a robotic cop killing someone by 2023?
Agree with Facebook "political party" instead of Zuckerberg running a campaign - not sure how this differs much from the traditional way corporations influence politics through donations, contributions, advertising, etc.
Your prediction of driverless cars doesn't mesh well with Rodney Brooks' prediction of driverless car saturation in 2045, he predicts we will have one dedicated lane for hands-free driving by 2021 and limited taxi/bus services by 2028. What is your basis for more aggressive forecast?
The most common one is wind, but changes in atmospheric pressure can cause them, as can a sudden change in the sea floor due to an earthquake or volcano (this is what causes tsunamis). The rotation of the earth and the gravitational pull of the moon can cause waves when it moves water confined spaces - this kind of wave is usually called a tidal bore.
Is there some kind of ocean wave I'm unaware of that can spontaneously occur without some external force/trigger?
Global Economic Recession is a safe one. The question is when?
I think Cryptcurrencies will crash with the recession. If a recession hits people are going to want to cash in their crypto for real money and guess what will happen to crypto?
Driverless / Flying / ... cars are probably the biggest change since the cell phone and internet. The only weird thing about them is their timming. People kind of travel less. Shopping is done over the internet, education is done over the internet, jobs are done over the internet. Socializing is done over the internet, Church over the internet.
The pairing of the latter two is something few would have predicted not that long ago.
For those who don't know Rodney Brooks, he was faculty at MIT for years, co-founded iRobot and Rethink Robotics, where he is working on collaborative robotics (robots that can work with humans without maiming them.)
But hype cycle or not, it would have been interesting to see where he had placed reusable rocketry in 2007. Would it have been 'by 2020 it will be common' or would it have been 'won't happen, the existing rockets work quite well and there isn't enough gain to justify the additional work.'
While Brooks picks on Elon a lot, the core of his essay is what makes it 'easy' to predict something will happen versus what makes it 'hard'. And I think he doesn't give enough credit to the fact that even if something is possible it needs a catalyst to get it to precipitate out of the realm of possibility into the realm of reality.
I think Dr. Brooks has a blind spot on space challenges, but I enjoyed his AI predictions. Of course it is entirely possible I just want easy access to space more than he does :-)
[1] I'm guessing @animats will disagree with that assessment :-) But he and Dr. Brooks approached the problem of inverse kinematics very differently and both got good results.
Could you expand on this? Thanks.
I like this essay, it seems to condense a lot of his thinking over some time.