The Algorithms Behind Self-Driving Vehicles

a16z2023-10-20

1K views|7 months ago

💫 Short Summary

The video discusses the use of advanced sensors such as lasers, cameras, and radars in autonomous vehicles, highlighting the debate between expensive Lidar systems and more affordable camera-based solutions. It emphasizes the importance of taking a principled approach and recognizing the unique strengths and limitations of each technology for optimal performance. The evolution and affordability of cameras are also highlighted, showcasing their widespread adoption and integration into various devices, similar to the progression of radar in cars. Overall, the focus is on the continuous improvement and innovation in imaging technology.

✨ Highlights

📊 Transcript

✦

The autonomous vehicle relies on advanced sensors like laser, camera, and radars for perception.

00:53

The sensors complement each other, with lasers providing precise details, cameras for color recognition, and radars for object sensing.

These sensors collectively create a detailed environment picture, exceeding human attention and accuracy capabilities.

The debate on using expensive Lidar versus simpler camera-based systems revolves around factors like fidelity, processing power, and cost.

The optimal performance of autonomous vehicles requires a balanced approach that recognizes the strengths and limitations of each sensor technology.

✦

Advancements in technology have made cameras and sensors more affordable and accessible to the general public.

03:44

The cost of cameras has significantly decreased over the past few years, leading to widespread adoption and integration into various devices.

Cameras on cell phones were once considered unnecessary but are now commonplace and of higher quality than standalone cameras from previous years.

The speaker emphasizes the continuous improvement and affordability of cameras.

There is a shift towards increased usage and innovation in the field of imaging technology.

00:03so this is my first time in a fully

00:05autonomous vehicle but I've seen them

00:07around my area they're driving around I

00:09see no driver but I also see the swirly

00:11thing on top I see a bunch of different

00:14kind of appendages to the car so maybe

00:16you could just break that down like what

00:18is happening how is all this technology

00:20coming together and what are the bits

00:22and pieces that you've added on to the

00:24car that allow it to be autonomous

00:26absolutely um so fundamentally you can

00:29think of it like like um is the car

00:31aware of what's happening around it m uh

00:34and then uh can it anticipate what the

00:36things around it are going to do yeah

00:38and then uh reasoning on what it should

00:40do next these are sort of the three

00:42components in perceiving what's around

00:44us this car with those appendages as you

00:47mentioned can see three football fields

00:50away 360° and it's getting a snapshot

00:53multiple times a second okay right and

00:56it's relying on a combination of the

00:59State ofthe art laser camera Radars all

01:03strategically positioned so to give you

01:05an idea lasers give you a very precise

01:08understanding of everything around you

01:09it's the smallest detail so if there was

01:11a child you know an inch out of this uh

01:14pole it would be able to mark oh this is

01:16a demarcated uh you know child away from

01:19that pole and you'll get to see that the

01:22uh cameras are needed to distinguish

01:24between the red light and the green

01:25light yeah yeah and The Radars can

01:27almost see around corners even when the

01:30laser and the camera or our human eyes

01:32can't cuz they can sense uh objects

01:34coming in and so we took an approach

01:37that we want to combine the best

01:39strengths of each of these modalities to

01:42create the best picture of what you can

01:44see around the world that you're just

01:46just incredibly better than a human

01:49possibly could both due to the attention

01:50span the range the Fidelity and the

01:53combination of these sensors coming to

01:55either the Rivalry in self-driving is

01:57lar versus you know Tesla Vision or the

01:59the video processing how has weo thought

02:01about that decision of what you know

02:04some people pose it as expensive

02:06Hardware more simple software because

02:07you have so much Fidelity from ldar

02:10versus a bunch of cameras and just like

02:12a lot heavier processing like is there a

02:15future where maybe actually you don't

02:16need all of the same sensing systems you

02:19know personally I have found the you

02:21know lighter and video debate almost

02:23takes like a ideological sense I you

02:26know for hard problems The Innovation

02:28Arc that we're talking about the best

02:30approach is taking a first principl

02:32approach right without uh you know

02:35neither love nor hate for a specific

02:37technology that it's a technology it's

02:39it's a tool and lar clearly has

02:42strengths that a camera doesn't for

02:44example at nighttime uh even the best

02:46cameras will have some challenges and

02:48Camera clearly has strengths that the

02:50lar doesn't and similarly radar has

02:52strengths that liars and cameras don't

02:55so saying you love liars and hate

02:58cameras or vice versa is is saying you

03:00love one wavelength versus the other

03:02wavelength right it's it's not a

03:04fundamental thing okay but uh there is a

03:07practical question so clearly on the

03:09first principles does it does the

03:11combination of these

03:13sensors position you better than

03:15individuals the answer is yes we can

03:17show you that there are situations in

03:18which camera will be

03:20insufficient now the question is a

03:22practical economic one is your ability

03:24to bring this public good to a large

03:27number of individuals hindered by the

03:28fact that these things are expensive yes

03:31and Liars you know 20 years ago if

03:34somebody told you that all these Toyotas

03:35are or these cars are going to have

03:37radar you be like no Radars are

03:38expensive guess what every single you

03:40know most cars have Radars today uh um

03:44cameras on cell phones were a novelty

03:47you know the old Nokia phones would have

03:49You' be like oh who needs a camera on a

03:50phone right you may have I'm sure now

03:53cell phones have better cameras than

03:54dedicated cameras of 6 years ago lar is

03:57going through the same transformation

03:59right now the iPhones have Lars right uh

04:02the amount we have been able to cost

04:03down these Lars in the last 2 years is

04:06incredible um okay so you're not

04:07concerned about that and we we you know

04:10we had uh four or five years ago we had

04:12that belief now we have that

04:15[Music]

04:25[Music]

04:28proof

04:32[Music]

🎥 Related Videos

a16z Podcast | Things Come Together -- Truths about Tech in Africa

a16z Podcast | The Infrastructure of Total Health

The Robot Lawyer Resistance with Joshua Browder of DoNotPay

a16z Podcast | Bots and Beyond

Design Sprints as a Tool for Organizational Change

a16z Podcast | Valuing Today's Fast-Growing Software Companies

🔥 Recently Summarized Examples

Former Priest REVEALS Jesus' MYSTICAL Lost Years & His Connection to BUDDHA! | Fr. Seán ÓLaoire

Kim Kardashian's Plastic Surgery Reversal: Is She Trying to Rewind Time?

How To Succeed As A NEW & YOUNG Realtor [Deals Every Month + Luxury Listings]

BITCOIN EMERGENCY: NEXT PRICE TARGETS REVEALED!! Bitcoin News Today & Ethereum Price Prediction!

Uncovering Ancient Atlantean Ruins: Exploring Evolutionary Pathways and Psychic Phenomenon

Samsung Technician Knives TV To Void Warranty

View original video