Party at mine: Elon Musk to host hackathon to accelerate self-driving tech

Tesla CEO and co-founder Elon Musk is hosting a “super fun” hackathon event at his house in four weeks’ time, alongside the Tesla self-driving AI team.

Musk had hoped that feature-complete full self-driving (FSD) would be available for Tesla’s revolutionary electric vehicles by the end of 2019.

But at the company’s fourth quarter 2019 earnings call last week he conceded that the milestone was running behind schedule. “We are pretty close, it’s looking like we might be feature-complete in a few months,” he said.

It appears Musk may now be seeking to accelerate this timeline by holding a hackathon, where developers get together to bounce ideas off each other, and look at the problem with some fresh eyes.

Inviting potential AI developers to get in touch via social media channel Twitter overnight (Australian time), Musk tweeted, “Tesla will hold a super fun AI party/hackathon at my house with the Tesla AI/autopilot team in about four weeks. Invitations going out soon.”

“Tesla will soon have over a million connected vehicles worldwide with sensors & compute needed for full self-driving, which is orders of magnitude more than everyone else combined, giving you the best possible dataset to work with,” said Musk.

He referred to the “kick-arse” self-driving AI chip that Tesla has developed from the ground up and which allows the company’s vehicles to process far more data (Tera Operations per Second – or TOPS, to be precise) than was previously possible using Nvidia GPUs and a separate computer processor.

For feature-complete – that is, able to drive from home to work and back under human supervision – to be achieved, Tesla must solve one final piece in the puzzle.

Smart Summon (slow speed parking and summoning in car parks) and Navigate on Autopilot (high speed navigation of on-ramps, off-ramps and lane changing on freeways) are already in use both in the US and Australia.

The missing piece is medium-speed navigation of traffic lights, roundabouts and other street-level obstacles.

It’s the most difficult piece of the puzzle, and obviously Tesla is hoping its hackathon will solve it once and for all.

Once feature-complete FSD is available it will then be up to regulators to allow Tesla vehicles to drive on public roads without human supervision.

“Our custom 144 TOPS in-vehicle inference computer, where almost every TOP is useable & optimized for [neural nets], far exceeds anything else in volume production, giving you the hardware you need to run sophisticated nets,” Musk said.

“Dojo, our training supercomputer, will be able to process vast amounts of video training data & efficiently run hyperspace arrays with a vast number of parameters, plenty of memory & ultra-high bandwidth between cores. More on this later.”

Musk and Tesla AI expert Lex Fridman do not require applicants to have PhDs on artificial intelligence to apply.

“All that matters is a deep understanding of AI & ability to implement NNs in a way that is actually useful (latter point is what’s truly hard),” said Musk.

“Don’t care if you even graduated high school.”

We can’t help but wonder if Musk will also use the hackathon/party as an opportunity to play his recently released single, “Don’t Doubt ur Vibe”.