0:00

32:32

Elon Musk: Tesla Autopilot | Lex Fridman Podcast #18

Lex Fridman·32:32·Technology·B2

Chương · 8

Chương 1·Introduction and Context

Hiển thị

Tiếng ViệtAuto-scroll

0:00

the following is a conversation with Elon Musk he's the CEO of Tesla SpaceX your link and a co-founder of several other companies this conversation is part of the artificial intelligence podcast the series includes leading researchers in academia and industry including CEOs and CTOs of automotive robotics AI and technology companies this conversation happened after the

0:26

release of the paper from our group at MIT on driver functional vigilance during use of Tesla's autopilot the tesla team reached out to me offering a podcast conversation with mr.

0:36

musk I accepted with full control of questions that could ask and the choice of what is released publicly I ended up editing out nothing of substance I've never spoken with Elon before this conversation

0:49

publicly or privately neither he nor his companies have any influence on my opinion nor on the rigor and integrity of the scientific method that I practice in my position at MIT Tesla has never financially supported my research and I've never owned a Tesla vehicle I've never owned Tesla stock this podcast is not a scientific paper it is a conversation I respect Elon as I do all

1:16

other leaders and engineers I've spoken with we agree on some things and disagree on others my goal is always with these conversations is to understand the way the guest sees the world one particular point of this agreement in this conversation was the extent to which camera-based driver monitoring will improve outcomes and for how long he will remain relevant

1:37

for a I assisted driving as someone who works on and is fascinated by human centered artificial intelligence I believe that if implemented and integrated effectively camera-based driver monitoring is likely to be of benefit in both the short term and the long term in contrast Elon and Tesla's focus is on the improvement of autopilot such that it's statistical safety benefits

2:04

override any concern of human behavior and psychology Elon and I may not agree on everything but I deeply respect the engineering and innovation behind the efforts that he leads my goal here is to catalyze a rigorous nuanced and objective discussion in industry in academia an AI assisted driving one that ultimately makes for a safer and better world and now here's my conversation

2:33

with Elon Musk what was the vision the dream of autopilot when in the beginning the big-picture system level when it was first conceived and started being installed in 2014 the hardware in the cars was division the dream I would characterize the division or dream simply that there are obviously two massive revolutions and in the automobile industry one is the transition to electrification and then

3:05

the other is autonomy and yeah became obvious to me that in the future any car that does not have autonomy I would be about as useful as a horse which is not to say that there's no use it's just rare and somewhat idiosyncratic if somebody has a horse at this point it's just obvious that cars will drive themselves completely it's just a question of time and if we did not participate in the

3:35

autonomy revolution then our cause would not be useful to people relative to cars that are autonomous I mean an autonomous car is arguably worth five to ten times more than by not colored which is not autonomous in a long term depends what you mean by a long term but would say at least for the next five years perhaps ten years so there a lot of very interesting design choices with

4:04

autopilot early on first is showing on the instrument cluster or in the model three and the center stack display what the combined sensor suite sees what was the thinking behind that choice was there debate what was the process the whole point of the display is to provide a health check on the vehicles perception of reality so the vehicles taking in information for a motion

4:31

sensor is primarily cameras but also radar and ultrasonics GPS and so forth and then that information is then rendered into vector space and that you know with a bunch of objects with product with properties like lane lines and traffic lights and other cars and then in vector space that is re-rendered onto your display so you can confirm whether the car knows what's going on or

4:59

not by looking at the window right I think that's a extremely powerful thing for people to get an understanding sort of become one with the system and understanding what the system is capable of now have you considered showing more so if we look at the computer vision you know like Road segmentation Lane detection vehicle detection object detection underlying the system there is

5:23

at the edges some uncertainty have you considered revealing the parts that the uncertainty in the system the said apart movies associated with say image recognition yeah right now it shows like the vehicles in the vicinity a very clean crisp image and people do confirm there's a car in front of me and the system sees there's a car in front of me but to help people build an intuition of

5:49

what computer vision is by showing some of the uncertainty well I think it's my car I always look at this sort of the debug view and there's this to debug views one is augmented vision where which I'm sure you've seen where it's basically we draw boxes and labels around objects that are recognized and then there's we're called the visualizer which is basically a Vectis based representation

6:18

summing up the input from all sensors that doesn't does not show any pictures but it shows all of the it's basically shows the cause view of the world in vector space but I think this is very difficult for people to know normal people to understand they would not know what thing they're looking at so it's almost the nature my challenge to the current things that are being displayed

6:43

is optimized for the general public understanding of what the system is capable it's like if you no idea what how computer vision works or anything you can still look at the screen and see if the car knows what's going on and then if you're you know if you're a development engineer or if you're you know if you're if you have the development build like I do then you can

7:03

see you know all the debug information but those would just be like total gibberish to most people what's your view on how to best distribute effort so there's three I would say technical aspects of autopilot that are really important since the underlying algorithms like then you'll network architecture there's the data so the distrain on and then there's a hardware

7:25

development there may be others but so look algorithm data hardware you don't you only have so much money only have so much time what do you think is the most important thing to allocate resources to do you see it as pretty evenly distributed between those three we automatically get a fast amount of data because all of our cars have eight external facing cameras and radar

7:54

and usually twelve ultrasonic sensors GPS obviously and I am you and so we basically have a fleet that has we're about four hundred thousand cars on the road that have that level of data I think you keep quite close track of it actually yes yeah so we're we're approaching half a million cars on the road that have the full sensor suite yeah this so this is I'm I'm not sure

8:26

how many other cars on the road have the sensor suite but I'll be surprised if it's more than five thousand which means that we were we have 99% of all the data so there's this huge inflow of data absolutely massive inflow of data and then we it's it's taken us about three years but now we've finally developed a full self-driving computer which can process and an order magnitude as much

8:53

as the Nvidia system that we currently have in the in the cars and it's really just a to use it you've unplugged the Nvidia computer and plug the tells the computer in and that's it and it's it's a in fact we're not even we're still exploring the boundaries of the capabilities we were able to run the camera is a full frame rate full resolution not even crop of the images

9:14

and it's still got Headroom even on one of the system's the heart full self-driving computer is really two computers two systems on a chip that are fully redundant so you could put a boat through basically any part of that system and it still works the redundancy are they perfect copies of each other or yeah also it's purely for redundancy as opposed to an arguing machine kind of

9:38

architecture where they're both making this is this is purely for redundancy you think more like it's if you have is a twin-engine aircraft commercial aircraft this system will operate best if both systems are operating but it's it's capable of operating safely on one so but as is right now we can just run we haven't even hit the at the edge of performance so with there's no need to

10:08

actually distribute the functionality across both SOC s we can actually just run a full duplicate on but on one each one you haven't really explored or hit the limit of this not yet at the limiter so the magic of deep learning is the that it gets better with data you said there's a huge inflow of data but yeah the thing about driving the really valuable data to learn from is the edge

10:35

cases so how do you I mean I've heard you talk somewhere about autopilot disengagement is being an important moment of time yes to use is there other edge cases or perhaps can you speak to those edge cases what aspects of that might be valuable or if you have other ideas how to discover more and more and more educators in driving well there's a lot of things that I learned though

11:02

certainly edge cases where I say somebody's on order pot and they take over and then okay that that's a trigger that goes at to assist and I says okay so they take over for convenience or do they take over because the autopilot wasn't working properly there's also like let's say we're trying to figure out what is the optimal spline for traversing an intersection then the ones where there are no

11:31

interventions and we are the right ones see you then say okay when it looks like this do the following and then the end and then you get the optimal spline for a complex know getting a complex intersection so that's for this is kind of the common case you're trying to capture a huge amount of samples of a particular intersection how when things went right and then there's the edge

11:56

case where as you said not for convenience but something somebody took over somebody asserted manual control from autopilot and really liked the way to look at this as view all input his error if the user had to do input if there's something all input is error that's a powerful line to think of it that way because it may very well be error but if you want to exit the

12:19

highway or if you want to its a navigation decision that all autopilot is not currently designed to do then the driver takes over how do you know yes that's gonna change with navigate an autopilot which we're just released and and without still confirm so the navigation like lane change based it likes a certain control in order to change the lane change or Exeter freeway

12:41

or doing a highway interchange the vast majority that will go away with the release that just went out yeah that that I don't think people quite understand how big of a step that is yeah they don't so if you drive the car than you do so you still have to keep your hands on the steering wheel currently when it does the automatic and lane change what are so there's these

13:06

these big leaps through the development of autopilot through its history and what stands out to you as the big leaps I would say this one navigate an autopilot without confirm without having to confirm there's a huge leap it is a huge leap but it also automatically overtake slow cars so it's it's both navigation and seeking the fastest lane so it'll it'll - you know overtake a

13:33

slower cause and exit the freeway and take highway interchanges and and then we have traffic like traffic light to recognition which introduced put initially as a as a warning I mean on the development version that I'm driving the car fully stops and goes at traffic lights so those are the steps right you've just mentioned somethings that an inkling of a step

14:01

towards full autonomy what would you say are the biggest technological roadblocks to full self-driving actually I don't think we I think we're just the full self-driving computer that we just let but it has a Oracle the FST computer that's now in production so if you order any Model S or X or any model three that has the full self-driving package you'll get the FST computer that

14:32

that was that's important that have enough base computation then refining the neural net and the control software but all of that can just providers know their update the thing that's really profound and where I'll be emphasizing at the auto sort of what that investor day they were having focused on autonomy is that the cars currently being produced but the hardware currently

14:57

being produced is capable of full self-driving but capable is an interesting word because like the hardware is yeah and as we refine the software the capabilities will increase dramatically and then the reliability will increase dramatically and then it will receive regulatory approval so it's actually buying a car today is an investment in the future what you're

15:20

essentially buying a you're buying the I think the most profound thing is that if you buy a Tesla today I believe you are buying an appreciating asset not a depreciating asset so that's a really important statement there because if hardware is capable enough that's the hard thing to upgrade yes usually exact so then the rest is a software problem yes I've software has no marginal cost really

15:48

but what's your intuition on the software side how hard are the remaining steps to get it to where you know the the experience not just the safety but the full experience is something that people would enjoy I think we will enjoy it very much the under knee on the highway sits it's a total game changer for quality of life for using you know Tesla motor pilot on the highways is so

16:22

it's really just extending that functionality to city streets adding in the traffic like traffic light recognition navigating complex intersections and then being able to navigate complicated parking lots so the car can exit a parking space and come and find you even if it's in a complete maze of a parking lot and then if and then you can just pick just drop you

16:50

off and find a parking spot by itself yeah in terms of enjoy ability and something that people would actually find a lot of use from the parking lot is a really you know it's it's rich of annoyance when you have to do it manually so there's a lot of benefit to be gained from automation there so let me start injecting the human into this discussion a little bit

17:12

so let's talk about full autonomy if you look at the current level for vehicles being test on row like way mow and so on they're only technically autonomous they're really level two systems with just the different design philosophy because there's always a safety driver in almost all cases and they're monitoring the system right do you see Tesla's full self-driving as still for a time to come

17:41

the requiring supervision of the human being so its capabilities a powerful enough to drive but nevertheless requires a human to still be supervising just like a safety driver is in a other fully autonomous vehicles I think it will require detecting hands on wheel for at least six months or something like that from here really it's a question of like from a regulatory

18:13

standpoint what how much safer than a person just autopilot need to be for it took to be okay to not monitor the car you know and this is a debate that one can have it and then if you need even a large sample a large amount of data so you can prove with high confidence so statistically speaking that the car is dramatically safe without a person and that adding in the

18:42

person monitoring does not materially affect the safety so it might not need to be like two or three hundred percent safe in a person and how do you prove that incidence per mile incidents per mile you know crashes and fatalities yeah fatality would be a factor but is there they're just not enough fatalities to be statistically significant Miguel but there are enough crashes you know

19:07

there are four more crashes and there were fatalities so you can assess where's the probability of crash that then there's another step which probability of injury and probability of opponent injury the probability of death and all of those need to be much better than a person by at least perhaps two hundred percent and you think there's a the ability to have a healthy discourse

19:37

with the regulatory bodies on this topic I mean there's no question that the regulator's paid a disproportionate amount of attention to that which generates press this is just an objective fact and Tesla generates a lot of press so the you know in the United States this I think almost 40,000 automotive deaths per year but if there are four and Tesla they will probably receive a thousand

20:06

times more press than anyone else so the psychology of that is actually fascinating I don't think we'll have enough time to talk about that but I have to talk to you about the human side of things so myself and our team at MIT recently released the paper on functional vigilance of drivers while using autopilot this is work we've been doing since autopilot was first released

20:28

publicly over three years ago collecting video driver faces and driver body so I saw that you tweeted a quote from the abstract so I can at least guess that you've glanced at it yeah all right can I talk you through what we found sure okay so it appears that in the data that we've collected that drivers are maintaining functional vigilance such that we're looking at 18-thousand

20:57

disengagement from autopilot 18900 and annotating were they able to take over control in a timely manner so they were there present looking at the road to take over control okay so this goes against what many would predict from the body of literature on vigilance with automation now the question is do you think these results hold across the broader population so ours is just a small

21:26

subset do you think one of the criticism is that you know there's a small minority of drivers that may be highly responsible where their vigilance decrement would increase with auto pilot use I think this is all really gonna be swept I mean that the systems are proving so much so fast that this is gonna be a moot point very soon where vigilance is like if something's many

21:57

times safer than a person then adding a person does if the effect on safety is is limited and in fact it could be negative that's really interesting so the fact that a human may some percent of the population may exhibit a vigilance decrement will not affect the overall status of safety no in fact I think it will become very quickly maybe in towards in this year but I say

An toàn hơn một người rất nhiều, sau đó thêm một người vào thì nếu tác động đến an toàn là có hạn và thực tế có thể là tiêu cực, điều đó thực sự thú vị. Vậy việc một con người, một phần trăm dân số, có thể biểu hiện sự suy giảm cảnh giác sẽ không ảnh hưởng đến tình trạng an toàn tổng thể. Không, thực tế tôi nghĩ nó sẽ trở nên rất nhanh, có thể trong năm nay nhưng tôi nói

22:30

I'll be shocked if it's not next year at the latest that having the post having a human intervene will decrease safety decrease it's like imagine if you're an elevator I used to be the third elevator operators and you couldn't go in an elevator by yourself and work the lever to move between floors and now nobody wants it an elevator operator because the automated elevator that

Tôi sẽ bị sốc nếu không phải năm sau muộn nhất là việc có người can thiệp sẽ làm giảm an toàn. Hãy tưởng tượng nếu bạn là thang máy, trước đây có người vận hành thang máy và bạn không thể tự vào thang máy và kéo cần để di chuyển giữa các tầng, và bây giờ không ai muốn người vận hành thang máy vì thang máy tự động

23:00

stops the floors is much safer than the elevator operator and in fact it would be quite dangerous to have someone with the lever that can move the elevator between floors so that's a that's a really powerful statement and really interesting one but I also have to ask from a user experience and from a safety perspective one of the passions for me algorithmically is camera based

Dừng các tầng an toàn hơn nhiều so với người vận hành thang máy, và thực tế sẽ khá nguy hiểm nếu có ai đó với cần gạt có thể di chuyển thang máy giữa các tầng. Đó là một tuyên bố rất mạnh mẽ và thực sự thú vị, nhưng tôi cũng phải hỏi từ góc độ trải nghiệm người dùng và an toàn, một trong những niềm đam mê của tôi về mặt thuật toán là phát hiện dựa trên camera

23:22

detection of obvious sensing the human but detecting what the driver is looking at cognitive load body pose on the computer vision side that's a fascinating problem but do you think there's many an industry you believe you have to have camera based driver monitoring do you think there could be benefit gained from driver monitoring if you have system that's that's out of

Phát hiện cảm biến rõ ràng về con người nhưng phát hiện tài xế đang nhìn vào đâu, tải nhận thức, tư thế cơ thể về phía thị giác máy tính, đó là một vấn đề hấp dẫn. Nhưng bạn có nghĩ rằng có nhiều ngành công nghiệp bạn tin rằng phải có giám sát tài xế dựa trên camera không? Bạn có nghĩ có thể có lợi ích từ giám sát tài xế nếu bạn có hệ thống vượt quá

23:44

that's out or below human level reliability then drive monitoring you make sense but if your system is dramatically better more level than a human then drive Montaigne monitoring is not just not help much and like said you just like as an you wouldn't want someone interview like you don't want someone in the elevator future an elevator do you really want someone with

Vượt quá hoặc dưới độ tin cậy của con người thì giám sát tài xế có ý nghĩa. Nhưng nếu hệ thống của bạn tốt hơn đáng kể so với con người thì giám sát tài xế không chỉ không giúp ích nhiều và như đã nói, bạn giống như không muốn ai đó phỏng vấn, bạn không muốn ai đó trong thang máy tương lai, bạn có thực sự muốn ai đó với

24:07

a big lever some random person operating elevator between floors I think they could I wouldn't trust that or rather have the buttons ok you're optimistic about the pace of improvement of the system from what you've seen with a full self-driving car computer the rate of improvement is exponential so one of the other very interesting design choices early on that connects to this is the

Một cần gạt lớn, một người ngẫu nhiên vận hành thang máy giữa các tầng? Tôi nghĩ họ có thể, tôi sẽ không tin tưởng điều đó, thà có các nút bấm. Bạn lạc quan về tốc độ cải thiện của hệ thống từ những gì bạn thấy với máy tính xe tự lái hoàn toàn, tốc độ cải thiện là theo cấp số nhân. Vì vậy, một trong những lựa chọn thiết kế rất thú vị khác từ đầu liên quan đến điều này là

24:35

operational design domain of autopilot so where autopilot is able to be turned on the so contrast another vehicle system that we are studying is the Cadillac super cruise system that's in terms of OGD very constraint is particular cause of highways well mapped tested was much narrower than the OD of Tesla vehicles what's theirs there's a TD yeah that's good this is it's a good life what was a

Miền thiết kế vận hành của Autopilot, nơi Autopilot có thể được bật. So sánh với một hệ thống xe khác mà chúng tôi đang nghiên cứu là hệ thống Cadillac Super Cruise, về mặt ODD rất hạn chế, chỉ trên đường cao tốc, được lập bản đồ tốt, thử nghiệm hẹp hơn nhiều so với ODD của xe Tesla. ODD của họ là gì? Đó là một câu hỏi hay. Đó là

25:09

design decision within that different philosophy of thinking where there's pros and cons what we see with a wide OD d is drive Tesla's drivers are able to explore more the limitations of the system at least early on and they understand together with the instrument cluster display they start to understand what are the capabilities so that's a benefit the con is you go you're letting

Quyết định thiết kế trong triết lý tư duy khác nhau, nơi có ưu và nhược điểm. Những gì chúng tôi thấy với ODD rộng là tài xế Tesla có thể khám phá nhiều hơn các giới hạn của hệ thống, ít nhất là ban đầu, và họ hiểu cùng với màn hình cụm đồng hồ, họ bắt đầu hiểu các khả năng. Đó là lợi ích. Nhược điểm là bạn để

25:34

drivers use it basically anywhere anywhere detect lanes with continents was their philosophy design decisions they were challenging they were being made there or from the very beginning was that done on purpose with intent well I mean I think it's frankly it's pretty crazy giving it letting people drive it a 2-ton death machine manually that's crazy like in the future of

Tài xế sử dụng nó về cơ bản ở bất cứ đâu, bất cứ nơi nào phát hiện làn đường với các lằn. Đó là triết lý của họ, các quyết định thiết kế họ đang thách thức, họ đang được đưa ra từ đầu, điều đó được thực hiện có chủ đích không? Chà, tôi nghĩ thẳng thắn mà nói, thật điên rồ khi cho phép mọi người lái một cỗ máy chết chóc nặng 2 tấn một cách thủ công, điều đó thật điên rồ. Giống như trong tương lai,

26:06

people were like I can't believe anyone was just allowed to drive one of these two-ton breath machines and they just drive wherever they wanted just like elevators use like move the elevator with the lever wherever you want it can stop it halfway between floors if you want it's pretty crazy so it's gonna seem like a mad thing in the future that people were driving cars so I have a

Mọi người sẽ nói 'Tôi không thể tin rằng ai đó được phép lái một trong những cỗ máy chết chóc nặng 2 tấn này và họ lái bất cứ đâu họ muốn', giống như thang máy, di chuyển thang máy bằng cần gạt bất cứ đâu bạn muốn, bạn có thể dừng nó giữa các tầng nếu bạn muốn, điều đó khá điên rồ. Vì vậy, nó sẽ có vẻ như một điều điên rồ trong tương lai khi mọi người lái xe. Vì vậy, tôi có một

26:33

bunch of questions about the human psychology about behavior and so on and that would be coming that grammar told ya because you have faith in the AI system not faith but the both on the hardware side and the deep learning approach of learning from data we'll make it just far safer than humans yeah exactly recently there are a few hackers who tricked autopilot act and not expected

Loạt câu hỏi về tâm lý con người, về hành vi, v.v. và điều đó sẽ đến, bởi vì bạn có niềm tin vào hệ thống AI, không phải niềm tin nhưng cả về phần cứng và cách tiếp cận học sâu từ dữ liệu sẽ làm cho nó an toàn hơn nhiều so với con người. Chính xác. Gần đây có một số hacker đã lừa Autopilot theo những cách không mong đợi

27:02

ways of the adversarial examples so we all know that neural network systems are very sensitive to minor disturbances to these adversarial examples on input do you think it's possible to defend against something like this wrong for the street for yeah can you elaborate on the on the confidence behind that answer well the you know in your own air is just like basic punch up make matrix

Các ví dụ đối nghịch. Vì vậy, chúng ta đều biết rằng các hệ thống mạng nơ-ron rất nhạy cảm với các nhiễu loạn nhỏ, các ví dụ đối nghịch này trên đầu vào. Bạn có nghĩ có thể phòng thủ chống lại điều gì đó như thế này trên đường phố không? Có, bạn có thể giải thích sự tự tin đằng sau câu trả lời đó không? Chà, bạn biết đấy, về bản chất, đó chỉ là toán ma trận cơ bản.

27:28

math oh you have to be like a very sophisticated somebody who really understands neural nets and like basically reverse engineer how the matrix is being built and then create a little thing that's just exactly causes the matrix math to be slightly off but it's very easy to then block it block that by having but basically ant here a negative recognition it's like if you if

Toán học. Bạn phải là một người rất tinh vi, ai đó thực sự hiểu mạng nơ-ron và về cơ bản kỹ thuật đảo ngược cách ma trận được xây dựng, sau đó tạo ra một thứ nhỏ chỉ khiến toán ma trận hơi sai lệch. Nhưng rất dễ để chặn nó bằng cách có về cơ bản là nhận diện phản đối. Nó giống như nếu bạn

27:53

the system sees something that looks like a matrix hack excluded because Sophia is such a easy thing to do so learn both on the validator and the invalid data so basically learn on the adversarial examples to be able to exclude them yeah you like your basically order both know what is what is a car and what is definitely not a car and you trained for this is a car

Hệ thống thấy thứ gì đó trông giống như hack ma trận, loại trừ nó vì đó là điều rất dễ làm. Vì vậy, học cả trên dữ liệu hợp lệ và không hợp lệ, về cơ bản học trên các ví dụ đối nghịch để có thể loại trừ chúng. Vâng, bạn về cơ bản biết cả hai: đây là xe hơi và đây chắc chắn không phải xe hơi, và bạn huấn luyện cho 'đây là xe hơi'

28:18

and this is definitely not a car those are two different things people have no idea neural nets really they probably think your license balls like you know fishing net only so as you know so taking a step beyond just Tesla and autopilot current deep learning approaches still seem in some ways to be far from general intelligent systems do you think the current approaches will take us to general

Và 'đây chắc chắn không phải xe hơi', đó là hai điều khác nhau. Mọi người không biết mạng nơ-ron thực sự, họ có thể nghĩ nó giống như quả bóng phát sáng, bạn biết đấy, lưới đánh cá. Vì vậy, vượt ra ngoài Tesla và Autopilot, các phương pháp học sâu hiện tại dường như vẫn còn xa so với các hệ thống thông minh tổng quát. Bạn có nghĩ các phương pháp hiện tại sẽ đưa chúng ta đến trí thông minh tổng quát

28:49

intelligence or do totally new ideas need to be invented I think we're missing a few key ideas for general intelligence general artificial general intelligence but it's going to be upon us very quickly and then we'll need to figure out what shall we do if we even have that choice good but it's amazing how people can't differentiate between say the narrow AI that you know allows a car

Hay cần phải phát minh ra những ý tưởng hoàn toàn mới? Tôi nghĩ chúng ta đang thiếu một vài ý tưởng chính cho trí thông minh tổng quát, trí thông minh nhân tạo tổng quát, nhưng nó sẽ đến với chúng ta rất nhanh và sau đó chúng ta sẽ cần tìm ra chúng ta nên làm gì, nếu chúng ta thậm chí có sự lựa chọn đó. Tốt, nhưng thật tuyệt vời khi mọi người không thể phân biệt giữa AI hẹp, bạn biết đấy, cho phép xe

29:23

To figure out what a lane line is and, you know, navigate streets versus general intelligence, like these are just very different things.

Xác định vạch kẻ làn đường và điều hướng đường phố, so với trí thông minh tổng quát, đây là những thứ rất khác nhau.

29:33

Like your toaster and your computer are both machines, but one is much more sophisticated than another.

Giống như máy nướng bánh mì và máy tính của bạn đều là máy móc, nhưng một cái tinh vi hơn nhiều so với cái kia.

29:40

You're confident with Tesla you can create the world's best toaster.

Bạn tự tin với Tesla có thể tạo ra máy nướng bánh mì tốt nhất thế giới.

29:45

World's best toaster?

Máy nướng bánh mì tốt nhất thế giới?

29:46

Yes.

Vâng.

29:46

But with the world's best self-driving?

Nhưng với xe tự lái tốt nhất thế giới?

29:49

I am.

Tôi tự tin.

29:50

Yes, I do.

Vâng, tôi tự tin.

29:51

To me, right.

Đối với tôi, đúng vậy.

29:53

Now this seems game, set, match.

Bây giờ điều này có vẻ như là trận đấu, set, match.

29:54

I don't know.

Tôi không biết.

29:55

I mean, that's how I almost be complacent, overconfident.

Ý tôi là, điều đó gần như khiến tôi tự mãn, quá tự tin.

29:58

But that's what it appears.

Nhưng đó là những gì nó xuất hiện.

30:00

That is just literally what it, how it appears right now.

Đó thực sự là những gì nó xuất hiện ngay bây giờ.

30:03

I could be wrong, but it appears to be the case that Tesla is vastly ahead of everyone.

Tôi có thể sai, nhưng có vẻ như Tesla đang dẫn trước mọi người rất xa.

30:09

Do you think we'll ever create an AI system that we can love and loves us back in a deep, meaningful way, like in the movie

Bạn có nghĩ chúng ta sẽ tạo ra một hệ thống AI mà chúng ta có thể yêu và nó yêu lại chúng ta một cách sâu sắc, ý nghĩa, như trong phim

30:18

Her?

30:18

I think AI will be capable of convincing you to fall in love with it very well.

Tôi nghĩ AI sẽ có khả năng thuyết phục bạn yêu nó rất tốt.

30:24

And that's different than us humans, you know.

Và điều đó khác với con người chúng ta, bạn biết đấy.

30:27

We start getting into a metaphysical question of like, do emotions and thoughts exist in a different realm than the physical?

Chúng ta bắt đầu đi vào câu hỏi siêu hình như, liệu cảm xúc và suy nghĩ có tồn tại trong một lĩnh vực khác với vật chất không?

30:35

And maybe they do, maybe they don't.

Và có thể chúng có, có thể không.

30:37

I don't know.

Tôi không biết.

30:38

But from a physics standpoint, I didn't think.

Nhưng từ quan điểm vật lý, tôi không nghĩ vậy.

30:41

I tend to think of things, you know, like physics

Tôi có xu hướng suy nghĩ về mọi thứ, bạn biết đấy, như vật lý

30:45

was my main sort of training.

Là lĩnh vực đào tạo chính của tôi.

30:47

And from a physics standpoint, essentially, if it loves you in a way that is that you can't tell whether it's real or not, it is real.

Và từ quan điểm vật lý, về cơ bản, nếu nó yêu bạn theo cách mà bạn không thể biết nó có thật hay không, thì nó là thật.

30:57

It's a physics view of love.

Đó là quan điểm vật lý về tình yêu.

31:00

Yeah.

Vâng.

31:00

If there's no, if you, if you cannot just, if you can't prove that it does not, if there's no test that you can apply that would make it, may allow you to tell the difference, then there is no difference.

Nếu không có, nếu bạn không thể chứng minh rằng nó không, nếu không có bài kiểm tra nào bạn có thể áp dụng để cho phép bạn phân biệt, thì không có sự khác biệt.

31:16

Right.

Đúng.

31:16

And

Và

31:17

it's similar to seeing our world as a simulation.

Nó tương tự như việc xem thế giới của chúng ta như một mô phỏng.

31:19

There may not be a test to tell the difference between what the real world is and a simulation.

Có thể không có bài kiểm tra nào để phân biệt giữa thế giới thực và mô phỏng.

31:25

And therefore, from a physics perspective, it might as well be the same thing.

Và do đó, từ góc độ vật lý, nó có thể coi như là cùng một thứ.

31:29

Yes.

Vâng.

31:30

And there may be ways to test whether it's a simulation.

Và có thể có những cách để kiểm tra xem nó có phải là mô phỏng hay không.

31:33

There might be.

Có thể có.

31:34

I'm not saying there aren't.

Tôi không nói là không có.

31:36

But you could certainly imagine that a simulation could correct

Nhưng bạn chắc chắn có thể tưởng tượng rằng một mô phỏng có thể sửa chữa

31:39

that once an entity in the simulation found a way to detect the simulation, it could either restart the, you know, pause that simulation, start a new simulation, or do one of many other things that then corrects for that error.

Rằng một khi một thực thể trong mô phỏng tìm ra cách phát hiện mô phỏng, nó có thể khởi động lại, bạn biết đấy, tạm dừng mô phỏng đó, bắt đầu một mô phỏng mới, hoặc làm một trong nhiều việc khác để sửa lỗi đó.

32:03

So when maybe you or somebody else creates an AGI system and you get to ask her one question, what would that question be?

Vậy khi có thể bạn hoặc ai đó tạo ra một hệ thống AGI và bạn có cơ hội hỏi cô ấy một câu hỏi, câu hỏi đó sẽ là gì?

32:18

What's outside the simulation?

Cái gì ở bên ngoài mô phỏng?

32:20

Ilan, thank

Ilan, cảm ơn

32:22

you so much for talking today.

bạn rất nhiều vì đã nói chuyện hôm nay.

32:26

It's a pleasure.

Rất vui.

32:28

All right.

Được rồi.

32:29

Thank you.

Cảm ơn bạn.

32:31

You.

Bạn.

Transcript

Hiển thị

Introduction and Context

Lex Fridman introduces the podcast with Elon Musk, explaining the context of their conversation following the release of a paper from MIT on driver vigilance during Tesla Autopilot use. He emphasizes his independence from Tesla and his goal of fostering rigorous discussion on AI-assisted driving. The conversation aims to explore the vision and challenges of autonomous driving technology.

Lex Fridman giới thiệu podcast với Elon Musk, giải thích bối cảnh cuộc trò chuyện sau khi một bài báo từ MIT về sự tỉnh táo của tài xế khi sử dụng Tesla Autopilot được công bố. Anh nhấn mạnh sự độc lập của mình với Tesla và mục tiêu thúc đẩy thảo luận nghiêm túc về lái xe hỗ trợ AI. Cuộc trò chuyện nhằm khám phá tầm nhìn và thách thức của công nghệ lái xe tự động.

release of the paper from our group at MIT on driver functional vigilance during use of Tesla's autopilot the tesla team reached out to me offering a podcast conversation with mr.

1 / 8