Woz has been irrelevant for many years and all of his involvement into various startups are all meaningless. Once a guy on Shark Tank wanted to impress the Sharks and told them that he knew Woz and perhaps even called Woz on the phone at the time. Uhhh...the Sharks were unimpressed as was I. Now if he had called Jobs and Jobs vouched for him, that would make it far more meaningful.
I don’t care about the level of importance people put on Woz’s claims and ideas.
I DO think that fully autonomous cars are a pathological technology, which is a large step beyond his current statement.
Being pathological hasn’t stopped billions of dollars and decades of time being invested and wasted pursuing other irrational and impractical things, and it won’t stop this pathological pursuit either. Ultimately, time will exhaust the notion, as the fundamental flaws in the concept continue to persist. Or it won’t. There are still people pursuing personal jet packs and flying cars.
How is pursuing a better machine-based solution to a fundamentally flawed human-based task irrational? What if it is practical, but maybe just not right now, is it still irrational to get closer? We don’t achieve anything without trying, and the notion that humans absolutely must physically control automobiles without pursuing ways to make them actually “auto” is absurd and sounds like the people who thought you couldn’t travel faster than a horse or that we’d never fly. You think maybe the Wright Brothers saw the future of fly by wire airliners? Give me a break.
He's a smart guy, but apparently initially bought into the hype (and probably doesn't have much philosophy or philosophy of mind experience). I'm glad to see he's now thinking more clearly and isn't just parroting the industry baloney. We need more people like him to speak up.
... I could add, with 100% certainty, we are not ready to colonize the Moon just yet either. Do I make the front page too?
But, the technology and foundations for that are possible. AI cars rest on technology that may never be ready, or on sci-fi concepts that aren't possible.
Waymo are going to launch their service very soon even though I have very much the same reservations/concerns as he does.
I won't be using one any time soon.
Hopefully people like Steve can bring people to their senses before going ahead with this foolishness. People are going to die needlessly. But it's $$$ (not safety) driving this, so they won't let clear-thought stand in their way.
georgie01 said: ... With our culture’s obsession with feeling like we’re so smart and feeling so proud of ourselves (even though it’s just a tiny fraction of us who actually make AI things), we forget that the human brain is extremely sophisticated. Just because we can’t do math as quickly as a computer doesn’t mean AI can do a better job than us, or that we even understand the capabilities of the brain to such a degree that we can replace the extent of processing necessary to have confidence in AI’s judgement.
There is no judgement, nor is it acting like our mind. That's the sci-fi version of AI. What it is doing is adding to an elaborate knowledge base that will continue to index more and more situations... except that what is being indexed here is insanely massive. The scope of the job is being grossly underestimated. Yes, sensors will get better, and computers will get faster. But that does little to advance AI in a way many imagine. It won't just get magically better because it has more speed/capacity/detail.
SpamSandwich said: Please. There is a very large population of drivers in the US (Baby Boomers) who are not getting any younger and they, possibly more than any other segment of the population, want and need this kind of service.
Wanting doesn't make it so. But, yes, that and commercial interests are what is pushing this blindly forward at an alarming pace.
I'll bet that when cars get more autonomous ....it will soon develop into everyone will be required to either own an autonomous vehicle or ride in autonomous busses.
That's really the only way I can see it kind of working. All AI (which makes the AI a bit irrelevant) vehicles networked together and working together with an extremely well-mapped world. But, even then, unless we re-do our entire infrastructure, it still has to deal with weather, pedestrians, cyclists, animals, etc. But AI-only vehicles at least solves the vehicle-vehicle interaction.
Here’s a mind experiment... imagine trying to create autonomous vehicles that move no more than 3mph. All of the requirements for sensing their immediate environment would remain, but the issue of them having accidents resulting in deaths would be greatly eliminated. I’m thinking this would be doable today, from what we’ve seen from the results of the DARPA challenge, from Waymo’s advances, Tesla autopilot, and others. Can we agree on that?
Okay, how about 4mph? See where this is going? A couple more turns of Moore’s law and cancer starts to get cured, weather predictions might be accurate enough that we don’t need more accuracy (do you need to know if it’s going to rain in two weeks?), crime heat maps become good enough to pre-position law enforcement in advance (already there for some types of crime), and our mind experiment becomes safe at any speed (to paraphrase Ralph Nader in the 1960s).
christopher126 said: Currently, there are approximately 36,000 highway deaths per year or about 102 per day. 50% are due to alcohol or some other drug induced impairment.
It's about 1 fatality per 100M miles driven. In that regard, I don't think all AI vehicles combined have booked enough miles for their even to be 1 death, and there have already been a few.
If we were really serious about improving safety, we'd do something about that impairment thing... or even give some reasonable driver training. This isn't about safety, though. That's just the sales-pitch.
DAalseth said: After all who objects to anti-lock breaks? They were the first automotive AI technology. Who objects to cruise controls that maintain a distance from the car ahead of you? That's another one.
There's nothing AI about either of those.
SpamSandwich said: Also, autonomous vehicles as we know them today are not examples of "artificial intelligence", they are examples of "machine learning".
Good point. That's a better way to talk about it... how hard it is to solve the machine learning challenge faced here. AI starts people thinking about sci-fi and terminator and 'self-aware' or systems that actually think, etc.
radarthekat said: As for when, did Woz the visionary speak to how many more turns of Moore’s Law it will take before processing capability of machine learning systems will far surpass the capabilities of humans for discrete tasks like driving?
Moore's Law has nothing to do with this. It's a quantitative problem, not a quantitative one.
...
GM already has a 2019 model with ZERO CONTROLS....no steering wheel or pedals........
No. They don't.
They've made press releases saying they PLAN to... but they don't yet... and likely won't.
No doubt. I think, was it Nissan made a bunch of futuristic videos too. But, I can make futuristic videos, as well. That doesn't mean I have a clue about how to actually do what is in the videos. It's marketing BS.
From a legal perspective I think states should limit autonomous vehicles to interstate highways for 3-5 years so the manufacturers can gain more real world experience.
I don't think they should be allowed on public streets/roads at all. If they can't drive 20 mph, why let them go 60mph?
I think the Woz has an overabundance of skepticism coupled with a high level of risk aversion.
In other words, he's far more wise than most of the commenters in this thread, and like 3/4 of the tech industry.
Did you mean “qualitative not quantitative?” Because if you did, you would be only half correct. It’s both. On one hand the system needs to be able to make appropriate decisions based upon data it’s presented (continue forward, speed up, slow down, turn, etc), but it won’t have time to carry out those decisions in time to evade an obstacle or make a necessary adjustment if the speed is too fast for the data provided, such as if the data provided only covers the ten feet in front of the vehicle. More data is needed. The system needs to be processing what’s out 100 feet, 200 feet, and farther depending upon its speed and other factors. Lots of data is needed (quantitative) and appropriate decisions need to be made against that data (qualitative). This is a simplistic representation of the problem, which is multi-variant and intractable, but not unsolvable. Start at 3mph, as I suggested in my mind experiment from my previous comment, processing data from just 50 or 100 feet around the vehicle. Learn from that, advance the tech, layer on new data, from farther out away from the vehicle, let the system learn that, increase safe drivable speed, just like a human driving student.
My Tesla drives itself on highways and even local roads with bends and curves without full turns. It changes lanes and enters and exits on ramps. This problem is solved. What’s left is recognizing stop signs and lights and turning appropriately on them. I’m sure that will be released to general public within two years. Woz may have used autopilot incorrectly or maybe some ancient version but it’s hard to think it’s a decade away or never happening.
... I could add, with 100% certainty, we are not ready to colonize the Moon just yet either. Do I make the front page too?
Actually setting up a base on the moon would be far easier than getting an automobile to operate with human levels of safety. The moon might involve more expensive hardware but most of the technical issues are solved.
As has has already been seen, robot cars fail miserably when it comes to accident avoidance. AI tech simply is t advance enough to perform emergency maneuvers beyond breaking or giving up. Actually a lot of humans can’t handle this well either so we shouldn’t blame robot cars. The classic problem if what to do when someone runs out in front of you is something that requires a quick reaction and often selecting the least of many evils.
2019 will be a big year for autonomy on the roads. It won't be "better" than a very good driver right away, but it'll get there.
Not a chance in hell. The unfortunate thing here is the starry eye politicians are not evaluating the safety properly. It is what the cars do when an exception occurs that should be the deciding factor on their safety. So far self driving cars have proven to be far worse at handling emergency responses to road hazzards. Getting from point A to B isn’t a huge problem with today’s tech when nothing out of the ordinary happens. It is having AI hardware and software complex enough and fast enough to do as well as the average human when faced with a sudden hazard.
I think letting a care drive for you will turn your brain into mush, Driving is problem solving activity, you need to be at you best and be able to solve problems in real time. If you let a car drive you everywhere, it just another sedentary activity which will cause you to decline in your ability to solve problems.
Also, if you think mechanization/automation of the US and shipping manufacturing jobs over seas killed jobs. Making all transportation autonomous will kill more jobs than you could imagine.
And yet in the recent election most of the nation voted for job killing politicians. One recently elected politician is upon arm because Amazon chooses her city.! I can’t even begin to grasp how good paying jobs can be a problem.
Unfortunately you you are right because industries like trucking hate paying a proper wage and will quickly adopt such tech. We could quickly see robots moving millions of tons of goods across our interstates. So yeah a serious problem.
The only recourse we really really have to joblessness is immigration controls and population controls. Unfortunately half the population won’t accept one and the other half will not accept the other. Sad times really.
In 1895 there were two automobiles in the whole state of Ohio...
Due to the lack of cars, as one would imagine, there were no car accidents.
Except once, when these two, aforementioned, automobiles (again, the only two in Ohio) ran into each other!
Currently, there are approximately 36,000 highway deaths per year or about 102 per day. 50% are due to alcohol or some other drug induced impairment.
The first part (Ohio) above may be untrue, but it does make the point, albeit, in a rather circuitous way...if we could snap our fingers and have instantly driverless cars we would take the drunks out of the equation and save over 18,000 lives per year...just doing that would be worth it.
Imagine if it could reduce highway deaths completely....Hmmm.
Best.
P.S. In 2016, 2,433 teens in the United States ages 16–19 were killed and 292,742 were treated in emergency departments for injuries suffered in motor vehicle crashes. That means that six teens ages 16–19 died every day due to motor vehicle crashes and hundreds more were injured.
In 2016, young people ages 15-19 represented 6.5% of the U.S. population. However, they accounted for an estimated $13.6 billion (8.4%) of the total costs of motor vehicle injuries.
Or you could just shoot the drunks when prosecuted for their defect. One of the greatest problems in this country is accepting people with mental defects. Frankly the only place I can see where the death penalty makes sense is in dealing with the mentally ill. You could eliminate millions of innocent deaths simply by getting these people off the streets for good.
There is a big difference between what is needed for a self driving car in a warm climate urban environment and one that can take a back country road in rural Montana in the snow or one buried in leaves in Vermont. Or black ice in the panhandle of Texas in January.
Anyone who has driven and paid attention to the many cars with driver assistance technologies knows that there are many gaps as there are with almost any human developed software logic.
I do not want one, do not want to pay the extra amount of money these systems will cost, do not want to be the victim of this unnecessary technology, etc.
I like driving and am fairly good at at- even in pretty crappy weather. I do not see a self driving car as a bonus, but as a danger. The fact that GM has shown a prototype of a car without controls just shows another reason why I would not buy their products anyhow. This same company sold knowingly defective ignition switches for well over a decade that caused many "accidents" and killed many people.
Then there is the whole legal problem. If an autonomous car that belongs to you kills someone is the liability on the manufacturer or the owner? Are you going to be able to sue GM or the person stupid enough to trust GM in the first place?
One final post this morning. Machine Learning is a subset of AI research so you really can’t disconnect the two. Even today you really need abilities to make decisions that ML isn’t really prepared to handle. This is where really hot cars are at these r weakest as the AI tech simply isn’t there yet. It is how a machine handles something it has never dealt with before that is really key to safe robot cars. Personally I think we will need breakthroughs for in hardware design to better support AI. We aren’t talking every faster machine learning hardware here.
Astor allot this noise about self driving cars s and Apple what if Apple isn’t even working on a self driving car. Instead we are seeing a ruse taking our focus off more humanoid robots that Applekay be working on. Autonomous robots are even more interesting than self driving cars bing cars if you ask me. Frankly such machines would fit in with Apples health care interests
I think the Woz has an overabundance of skepticism coupled with a high level of risk aversion.
In other words, he's far more wise than most of the commenters in this thread, and like 3/4 of the tech industry.
Yes indeed. In my opinion, it's beneficial to have older & wiser leaders with those traits to help balance out the young & reckless. A combination of personalities is best, with thoughtful thinkers in-between to add to the collective mix. It would be like having Tim Cook and Elon Musk on the same team, but with plenty of folks in-between to smooth out the peaks of those two. The worst possible leader you could possibly have would be one who is older & reckless, willing to sacrifice anyone below them because the leader already has it all and nothing to lose no matter how ruinous their decisions/actions are on others. It happens.
As long as men are driven by the fantasy of having sex in a car without needing to control the vehicle in any way, there definitely will be a self-driving car. And probably sooner.
I'll bet that when cars get more autonomous ....it will soon develop into everyone will be required to either own an autonomous vehicle or ride in autonomous busses.
By then Uncle Sam will make Human Driving totally illegal unless it's in an autonomous vehicles.....I'll be dead and stinkin' by then so it won't matter to me...
GM already has a 2019 model with ZERO CONTROLS....no steering wheel or pedals........
Spot on. In the same way the automobile lobby drove horses off the roads and the truck/bus lobby killed rail, the self-driving cars lobby will eventually get human driven cars banned. It will start with dedicated lanes for self-driven, then dedicated lanes for human-driven, then full no-go zones for human-driven, and finally, an outright ban. All in the name of safety, but really the almighty $.
Posters here seem to be as gung-ho about self driving cars as people were about social media a few years ago. Facebook sure turned out to be the positive social force that its starry-eyed advocates said it would be. Or if social media isn't enough to convince you, consider block chain currency. That turned out really well. For felons, Russian intelligence, and the environment.
Tech enthusiasts refuse to think or hear about the possible negative effects of the new technologies they are championing. Here's one that nobody likes to hear: Self-driving cars will not solve the urban road congestion problem. In fact it might make it worse. Let's not even talk about the liability problem. You and I know that the manufacturers will move heaven and earth to get a liability dodge passed into law.
There is a big difference between what is needed for a self driving car in a warm climate urban environment and one that can take a back country road in rural Montana in the snow or one buried in leaves in Vermont. Or black ice in the panhandle of Texas in January.
"Today at @Google IO we shared for the first time this before-and-after view of Waymo self-driving in snow. Using ML we can filter out the noise in our sensors that snow creates, allowing us to see better in bad weather. #io18"
I see everyone here disagrees with Woz. But Woz is where he is because he is a visionary. And sometimes being a visionary includes seeing things that can't be done. Eg, 50 years ago many people thought that a helicopter would be on everyone's front lawn. A visionary would have said "not likely" and been right.
Woz is nothing close to a visionary. Steve Jobs was a visionary. Woz is a computer nerd who did the nitty gritty work early in Apple's life. He understood electronics. Steve understood vision, presentation.
I am wholly unimpressed with Woz's career. I think he's a nerd who pretends he can see the future. Really, he's just a guy that used to build computers in the garage.
Yeah WOZ we love you, but your behind the times...
Perhaps we should wait a bit (and I don't mean a month) until after the service is in operation before it is judged an unqualified success...
There have been multiple articles related to Waymo's fleet in Phoenix (interviewing locals, observing the fleet), and found there are still many issues out there. Starting a self-driving fleet (which will have a driver behind the wheel it sounds like), with a few vehicles, in one of the best mapped and easiest climates in the world, is a great first start, but it isn't just another couple of years and new chips that will make this widespread.
Lots of possible detours ahead if you can take the blinders off. All it takes is one very public death from this new service and it sets the whole timeline back a few years.
A minimum of 7 years before such autonomous driving (with no backup driver) is widely available...
Sorry, but why does the media hang on Wozniak's every word? Why do we care what he thinks? He hasn't been relevant in decades. What has he produced since he left Apple, while living on his AAPL stock? He's never been shown to have even a shred of insight into where things are going, or even a fundamental understand of what makes products and companies successful. He's certainly never under what has made Apple successful.
I agree with you that SW is over-rated, especially on the topic of Apple.
But on this one, I agree with him wholeheartedly. Fully self-driving cars are at least 7-10 years away. They have to first fix insurance/liability issues, and then hundreds (if not thousands) of state and local regulations related to vehicles and vehicle traffic have to be worked on one by one. On top of which, the US Congress will have to pass legislation.
I think it will all ultimately happen, but not before 2025.
Comments
Okay, how about 4mph? See where this is going? A couple more turns of Moore’s law and cancer starts to get cured, weather predictions might be accurate enough that we don’t need more accuracy (do you need to know if it’s going to rain in two weeks?), crime heat maps become good enough to pre-position law enforcement in advance (already there for some types of crime), and our mind experiment becomes safe at any speed (to paraphrase Ralph Nader in the 1960s).
Did you mean “qualitative not quantitative?” Because if you did, you would be only half correct. It’s both. On one hand the system needs to be able to make appropriate decisions based upon data it’s presented (continue forward, speed up, slow down, turn, etc), but it won’t have time to carry out those decisions in time to evade an obstacle or make a necessary adjustment if the speed is too fast for the data provided, such as if the data provided only covers the ten feet in front of the vehicle. More data is needed. The system needs to be processing what’s out 100 feet, 200 feet, and farther depending upon its speed and other factors. Lots of data is needed (quantitative) and appropriate decisions need to be made against that data (qualitative). This is a simplistic representation of the problem, which is multi-variant and intractable, but not unsolvable. Start at 3mph, as I suggested in my mind experiment from
my previous comment, processing data from just 50 or 100 feet around the vehicle. Learn from that, advance the tech, layer on new data, from farther out away from the vehicle, let the system learn that, increase safe drivable speed, just like a human driving student.
As has has already been seen, robot cars fail miserably when it comes to accident avoidance. AI tech simply is t advance enough to perform emergency maneuvers beyond breaking or giving up. Actually a lot of humans can’t handle this well either so we shouldn’t blame robot cars. The classic problem if what to do when someone runs out in front of you is something that requires a quick reaction and often selecting the least of many evils.
Unfortunately you you are right because industries like trucking hate paying a proper wage and will quickly adopt such tech. We could quickly see robots moving millions of tons of goods across our interstates. So yeah a serious problem.
The only recourse we really really have to joblessness is immigration controls and population controls. Unfortunately half the population won’t accept one and the other half will not accept the other. Sad times really.
Anyone who has driven and paid attention to the many cars with driver assistance technologies knows that there are many gaps as there are with almost any human developed software logic.
I do not want one, do not want to pay the extra amount of money these systems will cost, do not want to be the victim of this unnecessary technology, etc.
I like driving and am fairly good at at- even in pretty crappy weather. I do not see a self driving car as a bonus, but as a danger. The fact that GM has shown a prototype of a car without controls just shows another reason why I would not buy their products anyhow. This same company sold knowingly defective ignition switches for well over a decade that caused many "accidents" and killed many people.
Then there is the whole legal problem. If an autonomous car that belongs to you kills someone is the liability on the manufacturer or the owner? Are you going to be able to sue GM or the person stupid enough to trust GM in the first place?
Astor allot this noise about self driving cars s and Apple what if Apple isn’t even working on a self driving car. Instead we are seeing a ruse taking our focus off more humanoid robots that Applekay be working on. Autonomous robots are even more interesting than self driving cars bing cars if you ask me. Frankly such machines would fit in with Apples health care interests
Tech enthusiasts refuse to think or hear about the possible negative effects of the new technologies they are championing. Here's one that nobody likes to hear: Self-driving cars will not solve the urban road congestion problem. In fact it might make it worse. Let's not even talk about the liability problem. You and I know that the manufacturers will move heaven and earth to get a liability dodge passed into law.
Waymo to Start First Driverless Car Service Next Month
https://www.bloomberg.com/news/articles/2018-11-13/waymo-to-start-first-driverless-car-service-next-monthYeah WOZ we love you, but your behind the times...
and this:
Woz is nothing close to a visionary. Steve Jobs was a visionary. Woz is a computer nerd who did the nitty gritty work early in Apple's life. He understood electronics. Steve understood vision, presentation.
I am wholly unimpressed with Woz's career. I think he's a nerd who pretends he can see the future. Really, he's just a guy that used to build computers in the garage.
There have been multiple articles related to Waymo's fleet in Phoenix (interviewing locals, observing the fleet), and found there are still many issues out there. Starting a self-driving fleet (which will have a driver behind the wheel it sounds like), with a few vehicles, in one of the best mapped and easiest climates in the world, is a great first start, but it isn't just another couple of years and new chips that will make this widespread.
Lots of possible detours ahead if you can take the blinders off. All it takes is one very public death from this new service and it sets the whole timeline back a few years.
A minimum of 7 years before such autonomous driving (with no backup driver) is widely available...