"Is partially automated driving a bad idea? Observations from an on-road study": April 2018 (@ 24:38) "MIT 6.S094: Self-Driving Cars": Jan 20, 2018 Finally a lecture on time travel :')
Went on youtube looking for funny animal vids and now I'm an hour in learning about driverless cars. - youtube algorithm looking out for my intellectual needs.
The algorithm understands irony and puns because while I find the topics fascinating, the presenter of lecture on autonomous vehicles is more akin to a monotonous vehicle...
@46:50 I'm puzzled as to why ultrasonics cannot detect speed. Ultrasonic waves experience doppler shift just like radar microwaves. I would have thought it would also be easy to measure that frequency change, probably easier than microwave.
Lex, excellent lectures and content. Thank you for making it public ally available. A quick point to the stats around 33:10 where you show the human activity during autopilot. Very illuminating. One thing to keep in mind is that likely these humans are drivers so they have a habitual and instinctual behavior in the car that has been ingrained. Just like when your passengers lean over to check both sides of cross traffic and obstruct your point of view when you’re the one who needs this information. They’re acting out of instinct. What will happen as we get newer generations who grow up using autopilot and minimal engagement with vehicle and do not have these reflexes? I think we would see the game playing and distracted behavior that were shown earlier in the lecture. I actually do not believe we’ll get there, I think we are already finding ways to properly engage with the autopilot and set the desired behavioral framework from humans in order to help the human-machine interaction. I’m very optimistic but also agree with the timelines from Brooks you quoted. Keep up the good work!
Another question, on the slide at 34:20 where you cite 8000 transfers of control from machine to human, you illustrated the case where the human took over control rather than the machine prompting the transfer. Do you differentiate this in the data? What was the difference between the two types of control transfer?
This is interesting. Very different prediction from main players. Tesla promising self-driving cars at level 4 by 2019-2021, Ford is working on self-driving cars for rideshare fleets will release by 2022, Waymo is planning to open driverless service this year (2018) in Chandler, AZ. And GM(Cruise) planning to make driverless service in 2019 and there's like a hundred other players there and all of them promise self-driving service by 2025 at the latest. Even if all these players will be a year or two late - some early service is pretty much inevitable by 2025. I can only assume that 2045 prediction is based on current advancement rate under appreciating exponential rate of advancement.
in five years we will have 10 times computational power for same cost and 10 times self driving data to train on and 10 times more money being invested in ai most people dont realize the impact of these exponential trends, full self driving probably 5 years away max
glen thomson I'd argue that AI specific computing performance will actually increase way beyond 10x, quite likely 200 times or more. If you look up at specific AI chips being developed right now and how AI computing was evolving so far, regular CPU in 2 year time grow in performance very slowly, like 20-30%, where GPU actually improve by almost a 100% and AI specific computation becomes much better optimized and grows several times faster than even GPU computing!
I think Tim Urban's explanation was the best by far about subject if you haven't read yet, definitely check it out: waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html
Hi, have you considered installing cameras on people cars from all around the world and use the data (recordings, people reactions, etc) to make cars learn how to drive with reinforcement learning?
Sashinka, I am really impressed with your research and proposals within AI transportation. I was particularly interested in your 6 second glance study. I am a psychology graduate and it interests me as to how one's emotional reactions would affect the facial expression and if that would affect at all the desired result. Perhaps if it is pre-programmed with a variation of emotional facial expressions. But it seems exceptionally interesting if AI could make the decision to call an ambulance etc....based on the facial expressions and or sounds the driver makes. Its really fascinating how far we have come with technology. It's almost inevitable, as you mentioned, for any great invention or human advancement to be used with bad intent by the wrong person. Meaning there should be a regulating department that makes and oversees that the rules are followed. Similar to DOT but maybe AVRD? Autonomous Vehicle Regulation Department. Забыла вообще что я хотела сказать ..ну ты сашенька просто прелесть! какой ты умный.. А я давно влюбилась. Захочешь найдешь меня.
Cool course. Thanks. One question about the radars. If there is two vehicles side by side how can we know that the wave that we received is the wave that we sent ? And not the one from the neighbors ?
How much of the AI's learning of drivers emotions would be changed with a driver with different skin tones and/or different cultural existence? Or, What biases are present/not present?
So will Waymo be able to deploy its cars nationwide quickly once it declares a success in PHoenix? Or they have to map the cities slowly one by one? How long do you think it will take Waymo to go nationwide?
Ich will mein Auto selbst steuern! In der Fliegerei haben die Autopiloten die meisten Erfahrungen. Und die haben es ja leicht: keine Fußgänger, keine Baustellen und Verkehrszeichen. Trotzdem stürzen sie ab. Ja, weil die Mensch-Piloten nicht mehr wissen wie es geht. Die werden nur noch auf die Bedienung der Computer trainiert. (Sully hat das Handwerk noch von der Pieke auf gelernt, nur darum hat er die Notwasserung auf dem Hudson hinbekommen) Das hat schon beim Auto mit der Einführung der Navis begonnen. Die Leute schauten nur aufs Navi, aber dass es zB. statt einer Brücke nur eine Fähre gab, führte sie in den Fluss. Was soll das Vorführen der Versuchsfahrt in Berlin? Nur auf der mittleren Spur fahren, hat der Autopilot Angst mal von der Straße abzukommen? Was hält der Autopilot vom Rechtsfahrgebot? Wenn nichts, dann brauchen wir nur noch einspurige Straßen (wäre wenigstens ein (Park-)Platzgewinn). Und was soll es Ultraschall-, Radar- und Lasersensoren einzusetzen? Wenn dereinst ALLE Autos diese Signale einsetzen, blenden die sich dann gegenseitig, wir wir uns heute schon mit Scheinwerfern? Ein Autofahrer schaut an der Kreuzung einem Anderen ins Auge und schon ist alles klar, der Autopilot such erst mal nach dem richtigen WLAN des Anderen, bis dahin ist der Stau perfekt. Etwa 50 Jahre nach der Einführung von Autos wurde erst der Sicheheitsgurt Pflicht. Was kommt nach 50Jahren Autopilot? Und wieviele sterben vorher? Vom verpassten Fahrspass nicht zu sprechen! ( Ich: 55 Jahre unfallfrei, keine Punkte, und das Auto stand nicht nur in der Garage und noch heute sind mir die meisten zu langsam, außerdem Fluglehrer. )
The problem with MIT driver attention experiment is that drivers knew that they are recorded so probably they tried to behave their best. I don't think any privacy conscious person would like to have this setup in their car and if there is nobody looking people will start playing chess ;)
After a long time, I'm reviewing the lessons, and now everything is even more straightforward. Thanks for sharing this.
"Is partially automated driving a bad idea? Observations from an on-road study": April 2018 (@ 24:38)
"MIT 6.S094: Self-Driving Cars": Jan 20, 2018
Finally a lecture on time travel :')
How did that happen???
Went on youtube looking for funny animal vids and now I'm an hour in learning about driverless cars. - youtube algorithm looking out for my intellectual needs.
The algorithm understands irony and puns because while I find the topics fascinating, the presenter of lecture on autonomous vehicles is more akin to a monotonous vehicle...
I realize I am kinda randomly asking but do anybody know of a good site to stream newly released movies online?
@Mustafa Jasper i would suggest Flixzone. Just search on google for it :)
@Khalid Ray Yup, I've been watching on Flixzone for since april myself =)
@Khalid Ray thank you, I went there and it seems like they got a lot of movies there :D I appreciate it!!
Thanks for sharing these lectures, Lex! Truly fascinating and inspiring. It got me into ML. Greetings from Europe
@46:50 I'm puzzled as to why ultrasonics cannot detect speed. Ultrasonic waves experience doppler shift just like radar microwaves. I would have thought it would also be easy to measure that frequency change, probably easier than microwave.
It's great to explore the potential of autonomous vehicles through the course offered by MIT
Lex, excellent lectures and content. Thank you for making it public ally available. A quick point to the stats around 33:10 where you show the human activity during autopilot. Very illuminating. One thing to keep in mind is that likely these humans are drivers so they have a habitual and instinctual behavior in the car that has been ingrained. Just like when your passengers lean over to check both sides of cross traffic and obstruct your point of view when you’re the one who needs this information. They’re acting out of instinct. What will happen as we get newer generations who grow up using autopilot and minimal engagement with vehicle and do not have these reflexes? I think we would see the game playing and distracted behavior that were shown earlier in the lecture. I actually do not believe we’ll get there, I think we are already finding ways to properly engage with the autopilot and set the desired behavioral framework from humans in order to help the human-machine interaction. I’m very optimistic but also agree with the timelines from Brooks you quoted. Keep up the good work!
Very interesting. High quality course. As usual. MIT rocks.
Even your lectures are brilliant, what a human. 💖
Lady in a car, 26:28 reading the The Power of Now: A Guide to Spiritual Enlightenment by Eckhart Tolle
Another question, on the slide at 34:20 where you cite 8000 transfers of control from machine to human, you illustrated the case where the human took over control rather than the machine prompting the transfer. Do you differentiate this in the data? What was the difference between the two types of control transfer?
I wonder, did the people involved in the driver behavior monitoring experiment know that they were being monitored?
Biased :)
couldn't wait for remaining lectures
This is interesting. Very different prediction from main players. Tesla promising self-driving cars at level 4 by 2019-2021, Ford is working on self-driving cars for rideshare fleets will release by 2022, Waymo is planning to open driverless service this year (2018) in Chandler, AZ. And GM(Cruise) planning to make driverless service in 2019 and there's like a hundred other players there and all of them promise self-driving service by 2025 at the latest.
Even if all these players will be a year or two late - some early service is pretty much inevitable by 2025. I can only assume that 2045 prediction is based on current advancement rate under appreciating exponential rate of advancement.
Expected time lines for autonomous vehicle roll-out lnkd.in/ejJXhhT
in five years we will have 10 times computational power for same cost and 10 times self driving data to train on and 10 times more money being invested in ai most people dont realize the impact of these exponential trends, full self driving probably 5 years away max
glen thomson I'd argue that AI specific computing performance will actually increase way beyond 10x, quite likely 200 times or more. If you look up at specific AI chips being developed right now and how AI computing was evolving so far, regular CPU in 2 year time grow in performance very slowly, like 20-30%, where GPU actually improve by almost a 100% and AI specific computation becomes much better optimized and grows several times faster than even GPU computing!
its crazy how fast its accelerating just out of interest do you have any thoughts on when or if we will get to general ai
I think Tim Urban's explanation was the best by far about subject if you haven't read yet, definitely check it out: waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html
Hi, have you considered installing cameras on people cars from all around the world and use the data (recordings, people reactions, etc) to make cars learn how to drive with reinforcement learning?
Sashinka, I am really impressed with your research and proposals within AI transportation.
I was particularly interested in your 6 second glance study. I am a psychology graduate and it interests me as to how one's emotional reactions would affect the facial expression and if that would affect at all the desired result. Perhaps if it is pre-programmed with a variation of emotional facial expressions. But it seems exceptionally interesting if AI could make the decision to call an ambulance etc....based on the facial expressions and or sounds the driver makes.
Its really fascinating how far we have come with technology. It's almost inevitable, as you mentioned, for any great invention or human advancement to be used with bad intent by the wrong person. Meaning there should be a regulating department that makes and oversees that the rules are followed. Similar to DOT but maybe AVRD? Autonomous Vehicle Regulation Department.
Забыла вообще что я хотела сказать ..ну ты сашенька просто прелесть! какой ты умный.. А я давно влюбилась. Захочешь найдешь меня.
Knowledge is the only thing outgrows if you share with others....
Just saw the news about Uber accident. Wondering would a thermographic camera be helpful for pedestrian detection at night?
What about deep learning for autonomous deer populations and autonomous ufo deep learning?
Cool course. Thanks.
One question about the radars. If there is two vehicles side by side how can we know that the wave that we received is the wave that we sent ? And not the one from the neighbors ?
How much of the AI's learning of drivers emotions would be changed with a driver with different skin tones and/or different cultural existence? Or, What biases are present/not present?
The way he talks, how he's standing, what he's wearing and his body posture... He presents like a lawyer presenting his case.
this is so awesome. Thanks!!
Very Informative Presentation!
So will Waymo be able to deploy its cars nationwide quickly once it declares a success in PHoenix? Or they have to map the cities slowly one by one? How long do you think it will take Waymo to go nationwide?
after Tesla does
What data was used to make the radar charts? Very useful charts. Thank you.
Several typos... audio needs to be audi.. a8 released end of 2017.. and question is which cpu was used for audi a8.
14 years?! Google already has a level 4 driverless cars used on public roads in Arizona
41:41 that's funny, what if it said "rubber neckin'" as he looks out the window
Is there any chance for you to add an English subtitle please? The auto one is not that correct. T_T
Great course so far. Thank you!
Hello Lex, the link for slides is not being opened. Could you please share the slides of the lectures. Thank you very much.
Hi west bengal
Ich will mein Auto selbst steuern! In der Fliegerei haben die Autopiloten die meisten Erfahrungen. Und die haben es ja leicht: keine Fußgänger, keine Baustellen und Verkehrszeichen. Trotzdem stürzen sie ab. Ja, weil die Mensch-Piloten nicht mehr wissen wie es geht. Die werden nur noch auf die Bedienung der Computer trainiert. (Sully hat das Handwerk noch von der Pieke auf gelernt, nur darum hat er die Notwasserung auf dem Hudson hinbekommen) Das hat schon beim Auto mit der Einführung der Navis begonnen. Die Leute schauten nur aufs Navi, aber dass es zB. statt einer Brücke nur eine Fähre gab, führte sie in den Fluss. Was soll das Vorführen der Versuchsfahrt in Berlin? Nur auf der mittleren Spur fahren, hat der Autopilot Angst mal von der Straße abzukommen? Was hält der Autopilot vom Rechtsfahrgebot? Wenn nichts, dann brauchen wir nur noch einspurige Straßen (wäre wenigstens ein (Park-)Platzgewinn). Und was soll es Ultraschall-, Radar- und Lasersensoren einzusetzen? Wenn dereinst ALLE Autos diese Signale einsetzen, blenden die sich dann gegenseitig, wir wir uns heute schon mit Scheinwerfern? Ein Autofahrer schaut an der Kreuzung einem Anderen ins Auge und schon ist alles klar, der Autopilot such erst mal nach dem richtigen WLAN des Anderen, bis dahin ist der Stau perfekt. Etwa 50 Jahre nach der Einführung von Autos wurde erst der Sicheheitsgurt Pflicht. Was kommt nach 50Jahren Autopilot? Und wieviele sterben vorher? Vom verpassten Fahrspass nicht zu sprechen! ( Ich: 55 Jahre unfallfrei, keine Punkte, und das Auto stand nicht nur in der Garage und noch heute sind mir die meisten zu langsam, außerdem Fluglehrer. )
The problem with MIT driver attention experiment is that drivers knew that they are recorded so probably they tried to behave their best.
I don't think any privacy conscious person would like to have this setup in their car and if there is nobody looking people will start playing chess ;)
Thanx for lecture...
Also save victims from car accidents!!
I can't take this lecture seriously. Fridman seems confused about everything. Down Vote.
45:38
Automakers just need to call their level 4/5 AI systems a level 3 to avoid liability.
29:01
I bet on a combination of camera and radar.
start the clock
You want to crush human freedom.
nice suit
hardly technical. more like a seminar for non-engineering majors.