Highlights From Tesla AI Day 2021

• by Martin Punter
• August 20, 2021
• 5 min read

Highlights from Tesla’s AI Day includes everything from humanoid robots to computer chips. Just another day in the office for the American automaker.

Creepy or helpful? You tell us © Credit to Tesla

• Tesla Has Made A Human Robot
• Tesla Computer Chip & Full Self-Driving
• Sorting Computer Vision Issues

Tesla stole the headlines once again at its AI Day, an event that highlighted all of the companies moves in the artificial intelligence lane.

The three-hour-long event was jam-packed, with podcaster Lex Fridman calling it “the most amazing real-world AI & engineering effort I have ever seen in my life.”

For the first 40 minutes of the stream, we were treated to footage of Tesla’s full self-driving feature in action, set to some pretty Tron-like music. After that, Elon took to the stage to talk specifics.

The company began with a featured series that shed light on all the hard work completed by Tesla engineers and its AI team. Emphasis was placed firmly on the future and the untapped potential of this kind of tech moving forwards.

“There’s a tremendous amount of work to make it work and that’s why we need talented people to join and solve the problem,” said Musk.

Tesla Has Made A Human Robot

Let’s get it out of the way now (it’s why most of you are reading this in the first place anyway).

Towards the end of the event, Tesla revealed it’s working on a 5’8”, human-like robot that bears a striking resemblance to those crazy robots from the movie I, Robot — or Zer0 from Borderlands 2.

We’re 100% serious, by the way, as are Tesla.

We only say that because there was a brief moment during the event where someone, clearly in a morph suit, came on stage and began waving.

But no, this is actually happening in a bid to cut out the mundane tasks that rule our everyday lives. At least, that’s Elon’s reasoning.

Elon Musk is a borderline evil genius at this point.

The Tesla Bot can deadlift 150 pounds and walk at speeds of 5 miles per hour. Its face also has a built-in display to convey information (like Zer0, just not an assassin).

“It’s intended to be friendly, of course, and navigate a world built for humans,” said evil genius Musk. “We’re setting it such that at a mechanical and physical level, you can run away from it and most likely overpower it.”

Tesla Computer Chip & Full Self-Driving

Tesla director Ganesh Venkataramanan also revealed the automaker’s computer chip used to run its Dojo supercomputer.

Dojo, if you didn’t know, is used in most of Tesla’s AI makeup. It’s used to map areas a lot faster compared to other computing systems.

The D1 chip is small yet mighty © Credit to Tesla

This “D1” chip has GPU-level compute with CPU connectivity and twice the I/O bandwidth of “gold standard” state-of-the-art networking switch chips.

The overall goal of the chip is to increase bandwidth while decreasing latencies for better AI performance.

Telsa is looking beyond full self-driving capabilities with this chip too. Of course, it will benefit the EV company’s driver assistance system, but Tesla is thinking bigger.

“This is not intended to be just limited to Tesla cars,” said Musk. “Those of you who’ve seen the full self-driving beta can appreciate the rate at which the Tesla neural net is learning to drive. And this is a particular application of AI, but I think there’s more applications down the road that will make sense.”

Dojo is due to be operational next year, so stay tuned.

Sorting Computer Vision Issues

Tesla’s approach to autonomous vehicles never ceases to amaze. Especially when you consider its place in the future. And we aren’t just talking cars either, as proven by Tesla’s handyman robot.

Still, there are a few kinks that need to be worked out in its computer vision systems.

According to Tesla’s head of AI, Andrej Karpathy, there are two main problems that need to be solved. One in the form of temporary occlusions, such as cars blocking the view of the road and signs that appear earlier in the road.

Tesla engineers will solve this by rolling back on a spatial recurring network video module. The module keeps track of multiple elements when on the road, all of which influence how it reacts on the road.

Karpathy described Tesla’s architecture as “building an animal from the ground up,” which is both cool and terrifying at the same time.

The company has its doubters when it comes to ‘Autopilot’ and Tesla’s move into self-driving in general. These fixes could prove paramount in the court of public opinion.

We’ve linked to the livestream above if you’d like to experience it for yourself. But yeah, there’s a lot to unpack here. We’re pretty sure everyone will be drawn to its human-like robot, but what was your favourite part of AI Day 2021? Drop us a comment below.

Martin Punter

View All Posts >