Elon Musk, Tesla and the case for the AI chip
Another day, another futuristic brainwave from Elon Musk. It’s not easy keeping up with the whirlwind industrialist’s ventures, but if you’re a Tesla enthusiast, like me, you are probably following the projects of the iconic electric car maker’s CEO closely.
I won’t dwell on his ambition to send a Tesla Roadster into orbit around Mars - cue the tagline ‘red car for a red planet.’ One of his more grounded ideas, revealed in California just days ago, involves developing customised AI hardware chips for the Autopilot self-driving system. Although Musk is furtive about the chip’s stage of development or when it will start shipping inside vehicles, he says that Tesla will get there much faster if they have dedicated AI hardware.
Musk maintains the chip will fast forward the day when the self-driving Tesla can drive 10 times more safely than a human. Currently, Tesla owners are instructed to only use the Autopilot system on highways. He says a future software upgrade will permit full self-driving using the hardware inside existing vehicles.
The chip seems to signal a shift in Musk’s attitude towards AI, which he once famously described as an ‘existential threat to humanity.’ Unveiling his AI chip development in Long Beach, California, he spoke of the company’s need for AI talent in software and hardware where, according to Musk, “AI can really be useful”, rather than the proclaimed “godlike uber intelligence”.
Tesla’s primary use for AI is to make sense of data from the cameras, radar, and other sensors that its Autopilot automated driving function interacts with. Musk was quick to add that he believes cars will soon be harder to fool than people, noting how they can use multiple sensors such as radar and cameras to verify what they’re seeing.
Musk’s AI chip is in good company. The next generation AI chip took off a few years ago and forms a suite of artificial intelligence techniques called deep learning that is used to help organise such simple things as your Google photos and produce digital assistants like Amazon's Alexa and Apple's Siri, right up to Tesla’s self-driving cars.
These machine learning models require huge amounts of computing power and the latest algorithms are pushing on the limits of what current chips can handle. Love him or loathe him, Musk is unarguably a visionary. In 2016 he declared: “I consider autonomous driving to be a basically solved problem. We’re less than two years away from complete autonomy.”
Last month, during the unveiling of Tesla’s forthcoming Semi truck, Musk made a surprise announcement that the company is taking reservations for a next-generation Roadster that won’t be ready until at least 2020. Again, I’ll be keeping a close eye on developments. It might be a good investment, even if we do end up driving it on Mars.