Tesla (TSLA) could team with OpenAI, the firm behind the wildly popular bot ChatGPT, to further boost its artificial intelligence (AI) enterprise, analysts said, just as the electric vehicle (EV) maker is rushing to develop its Dojo supercomputer for that same endeavor.
Their comments came as OpenAI – which Tesla CEO Elon Musk helped co-found in 2015 – has spurred a chatbot arms race among tech superpowers Microsoft (MSFT) and Alphabet (GOOG), who are racing to launch their own ChatGPT versions to dominate the AI-powered chatbot revolution.
Since ChatGPT launched in November, the world has marveled at its human-like conversationalist abilities, offering to write articles on demand, help students to homework and even explain quantum physics.
“It’s not inconceivable that they could have some collaboration at some point in the foreseeable future,” said Edward Sanchez, senior automotive analyst at TechInsights, adding that Musk likely has close ties with OpenAI’s executive team, making a potential deal easier to strike.
“Tesla is way ahead of the industry in over-the-air updates so if they feel there is enough consumer demand for a ChatGPT service in its cars, similar to Apple’s Siri, they may want to push the service in its next update.”
Sanchez envisaged Musk, who injected $1 billion alongside the likes of other Silicon Valley billionaire Peter Thiel to bankroll OpenAI’s launch, could at least license GhatGPT from OpenAI which reportedly clinched $10 billion from Microsoft in recent weeks, just as the software giant said it would extend a “multi-year, multi-billion” dollar partnership with the company.
Huge computing power
Data scientist and fintech professor Merav Ozair agreed a Tesla-OpenAI venture is “a high possibility” but noted that for the service to work on Tesla’s vehicles, huge computing power will be required.
“You are going to need a lot of data, probably 100 times more than what you use to download a video,” said Ozair. “And this data will have to be analyzed by a very powerful computer to help relay it very quickly, in a matter of seconds, for instance, if you are driving.”
“You are going to need a lot of data, probably 100 times more than what you use to download a video,” said Ozair. “And this data will have to be analyzed by a very powerful computer to help relay it very quickly, in a matter of seconds, for instance, if you are driving.”
Perhaps for this reason, Tesla is rushing to develop its Dojo infrastructure with new updates expected during its upcoming investor day on March 1, which will also likely raise the prospect of other developments, including a timeline for a new, cheaper Tesla model.
Currently, the Dojo is used to interpret and finetune video data from Tesla’s full self-driving (FSD) software to improve the technology through AI learning.
This is to ultimately reach Level 4 or fully autonomous driving in the future. That should help it power its grand AI ambitions which, on top of FSD, calls for the eventual colonization of Mars through Musk’s Space X.
This is to ultimately reach Level 4 or fully autonomous driving in the future. That should help it power its grand AI ambitions which, on top of FSD, calls for the eventual colonization of Mars through Musk’s Space X.
“AI is not just about FSD cars, its usage goes way beyond including the chatbot to Optimus [Tesla’s new humanoid robot], to space where the Mars Rovers will (are) be operated by AI so they can communicate with earth. AI will be everywhere,” Ozair noted.
Dojo to accelerate full self-driving (FSD)
Activist investor Ross Gerber, who is seeking a seat on Tesla’s board, hopes Dojo will help Tesla accelerate its FSD offering to be ready in the next couple of years (after heavy delays) to further differentiate it against a rising clutch of rivals.
“Level 4 FSD in which you don’t have to pay attention [while the car drives] is just not there yet,” Gerber said, adding that Tesla’s software will have to handle complicated city driving before it can exit level 3.“When I am on the highway, I can fall asleep on FSD and not have to worry,” Gerber added.“But in a city, with all the unpredictable human variables at play, I can’t.”
Gerber said there are more immediate things to worry about, however, including Tesla’s profit margins, which must move higher and accompany earnings beats for the stock to reach the $250 level in the next 12 months.
Investors are keen to hear more about Tesla’s ‘master plan’ during investor day, including how it plans to grow its energy storage business, enlarge its giga-factory network to 10 from four and how it will bring a lower-priced model to market through a ramp up of 4680 battery cells.
Sanchez expects the new model will come with a $25,000 base price and sport similarities to Volkswagen’s (VWAGY) Golf or Polo.
However, Musk has said the computer may eventually be powerful enough to lease its throughput to third parties, similar to what Amazon (AMZN) does with its AWS cloud service for businesses.
Dojo is expected to help Tesla unwean itself from its current GPU-powered computer facilities, boasts roughly 68.75 petaFLOPS of power, much less than the 442 Japan’s ‘Fugaku’ rival churns out. Tesla’s goal is to eventually increase Dojo’s output to 1,000 petaFLOPs (a modern way to measure a computer’s speed), according to Ozair.