“Amazing by most standards:” Musk defends Tesla Full Self-Driving and AI Day takeout
Less than a week after presenting the progress and achievements of Tesla’s Full Self-Driving (FSD) computer chip to the world on “AI Day,” Elon Musk defended his comment that the latest deployment The semi-autonomous driving suite’s software was “not that great.”
Currently under investigation by the United States-based National Highway Traffic and Safety Administration (NHTSA) for suspected crashes with emergency vehicles, Tesla Autopilot’s Advanced Driver Assistance Suite and its top-level FSD, which Tesla wants to make it fully self-sufficient, come under scrutiny.
Musk apparently didn’t help matters on Tuesday when he said in response to a Tweet, “FSD Beta 9.2 is actually not great (in my opinion), but the Autopilot / AI team is rallying. to improve as quickly as possible. We try to have a single battery for highways and city streets, but it requires massive recycling of NN.
FSD Beta 9.2 is actually not great, but the Autopilot / AI team is mobilizing to improve as quickly as possible.
We try to have a single battery for highways and city streets, but it requires massive recycling of NN.
– Elon Musk (@elonmusk) 23 Aug 2021
It doesn’t seem particularly reassuring at first glance, but Musk took to Twitter again early Wednesday (AUS) to clarify this statement.
The latest comment came in response to one of Tesla’s FSD beta testers, a group of Tesla approved owners who have been testing the electric vehicle maker’s latest software deployment since late 2020.
“v9.2 works “excellent” despite @Elon Musk complaints. It is a 22 minute journey. No intervention. (As usual), ”said the Twitter user known as“ Earl of Frunkpuppy ”.
“It’s amazing by most standards, but we’re aiming for 1000% safer than the average human driver“Musk replied.
It’s amazing by most standards, but we’re aiming for 1000% safer than the average human driver
– Elon Musk (@elonmusk) August 24, 2021
This in itself shows how much Musk and the Tesla AI team are aware of the quality of self-driving cars that must be better than humans for them to be accepted by the general population.
If self-driving cars were even twice as safe as human drivers, there would be fewer deaths from car crashes.
But it would seem that we fallible humans are 1,000 times more forgiving of death caused by human negligence, incompetence or just because someone is just plain old “having a bad day” than we are. if a machine “makes a mistake”.
In addition, a machine does not “make mistakes”. It operates within the parameters of its programming, therefore the programming must be developed to deal with the almost countless number of situations that can occur on the road, at any speed.
Tesla is so convinced that he will soon have solved the AI puzzle that he will also be able to produce a humanoid robot using the same hardware and software stack that can do everything from grocery shopping to shopping. other more dangerous and / or boring jobs.
Watching AI Day, after going through the initial shock of having to translate uber-geek language on the fly, one of the things that impressed me the most was the incredible complexity of the human brain and the complexity of it. which is made to match this.
Going through the number of processes that the Tesla AI team has developed using algorithms and hardware so that a car can do something that a human does every day is frankly mind-boggling.
From how to read precise data at the edges and lines of a road and stitch it together in a bird’s eye view, to using eight cameras to combine sensor data to determine location object of a moving object, then label that object as a car, van, child, or dog, AND then predict where that object is at any given time if it becomes occluded (hidden by another object ) is just the beginning.
To do this, he launched a plan to automatically label objects and use simulations to train his algorithms to handle extreme cases. By the way, Elon, we still need a “Roo Mode”.
Tesla has also completed building its own AI chip, which it then integrated into a mega-computer known as an “Exapod” which it will use to collect data from its fleet of cars, automatically tag the data. , use it to retrain and then deploy the upgraded algorithms to the fleet.
Thank goodness MIT AI expert Lex Fridman took the time to create an “AI Day Highlights” video to break it down for the layman.
In the video below, he sets out the main takeaways and also gives his top three reasons why he thinks what Tesla is doing is exciting.
“The entire image that was presented on AI Day was amazing as the Tesla AI machine can go through the iterative data engine process of automatic labeling plus manual labeling of edge cases so that step of labeling plus data collection, recycling, deployment and again you go back to data collection, labeling, recycling and deployment, ”he says.
“And you can go through this loop as many times as you want to arbitrarily improve network performance. I still think no one knows how difficult the problem with autonomous driving is, but I also think this loop has no cap.
“I still think there is a big place for driver detection, I still think we have to solve the human-robot interaction problem to make the experience more enjoyable but damn it, this loop … is amazing, ”he says.
“The second reason this whole effort is amazing is that Dojo can essentially become AI training as a service, directly supporting AWS and Google Cloud. There is no reason to use it specifically for the autopilot computer.
“You can basically use it for every machine learning problem, especially one that requires scaling,” he says.
“The third reason why this was amazing was that the ark of neural network and the data engine[oipelinesestapplicablebienplusquelesroutesetlaconduiteilpeutêtreutiliséàlamaisonetàl’usineetpardesrobotssousn’importequelleformetantqu’iladescamérasetunactionneurycomprisouilaformehumanoïde[oipelinesisapplicablethanmuchmorethanjustroadsanddrivingitcanbeusedinthehomeandthefactoryandbyrobotsinanyformaslongasithascamerasandactuatorincludingyesthehumanoidform”[oipelinesestapplicablebienplusquelesroutesetlaconduiteilpeutêtreutiliséàlamaisonetàl’usineetpardesrobotssousn’importequelleformetantqu’iladescamérasetunactionneurycomprisouilaformehumanoïde[oipelinesisapplicablethanmuchmorethanjustroadsanddrivingitcanbeusedinthehomeandthefactoryandbyrobotsinanyformaslongasithascamerasandactuatorincludingyesthehumanoidform”
We’ll just leave this here:
Bridie Schmidt is the senior reporter for The Driven, sister site of Renew the economy. She has been writing about electric vehicles since 2018 and has a keen interest in the role zero-emission transportation needs to play in sustainability. She has participated in podcasts such as Download This Show with Marc Fennell and Shirtloads of Science with Karl Kruszelnicki and is a co-organizer of the Northern Rivers Electric Vehicle Forum. Bridie also owns a Tesla Model 3 and offers it for hire at evee.com.au.