Thoughts on Tesla AI Day 2022

πŸ–€ Tesla

I love this time of the year – crisp autumn mornings, fall foliage, pumpkin spice latte, and Tesla’s annual AI Day.

Tesla AI Day is part product demonstration, part university lectures. These events provide technical details that showcase the company’s progress in artificial intelligence (AI) and robotics. They’re also staged to attract the best and brightest.

Tesla’s ability to recruit and retain top science and engineering talents in the age of the Great Resignation is an unparalleled competitive advantage, on par with its zero advertisement customer acquisition strategy.

Prototype to production

The star of AI Day 2022 is the Tesla Bot aka Optimus. First teased at last year’s AI Day (my thoughts on the event), Optimus is a humanoid robot built to do dangerous or boring, repetitive tasks. Elon Musk has also suggested that a “fun” version could be a “friend and a buddy – someone to hang out with”.

Humanoid robots are not new. An automaker making a humanoid robot is not a first either – Honda unveiled ASIMO back in 2000 but the company canned the project in 2018.

Tesla’s effort is different in that it focuses on designing Optimus for mass production; for example, in designing the hands, the engineers optimise for fewer parts instead of the best technical solution. Elon Musk has guesstimated that the mass produced product would cost less than US$20,000.

The Tesla Bot team has the luxury of learning “production hell” lessons from the car business, and could leverage on existing software (e.g. the AI models) and hardware supply chain (e.g. actuators). The team’s immediate goal is to ship a product that could be put to work in Tesla Gigafactories.

Prototypes are easy, production is hard.

Elon Musk, commenting on companies jumping on the electric vehicles bandwagon

Admittedly, the reveal is a tad underwhelming. Optimus could walk but pales in comparison to Boston Dynamics’ robots that do parkour. It could do basic tasks but there’s a long road ahead for it to be of any practical use. Perhaps expectations are too high; this is an early prototype after all.

The long-standing dream is to have robots work tirelessly around the clock (provided they don’t rise up against humanity). Elon Musk’s vision is for the Tesla Bot to provide quasi-limitless productivity boost that will create a world of abundance. It’s the stuff of science fiction that will take years to come to fruition. For now, the Tesla Bot remains a moonshot project that is ahead of its time.

Full self-driving (FSD)

Tesla AI Day also provides an update on “full self-driving” (FSD). Fully autonomous vehicles has long been part of Elon Musk’s master plan for Tesla. Tesla’s long-term viability depends on delivering this FSD capability.

Over the last year, the beta version of FSD has been deployed onto c. 160,000 cars (from just 2,000 cars a year ago) – what this means is that Tesla has access to more valuable data to validate and retrain the AI models.

Tesla FSD architecture (2022 version)

At the event, the Autopilot team provides an incredible amount of detail on system design and architecture. The Semiconductor team also provides an update on Dojo, the in-house supercomputer which will be used to train the AI models.

The key takeaway is that Tesla has in place a strong foundation and infrastructure to train, validate and reiterate AI models that will make autonomous vehicles and robots a reality.

This is reminiscent of the bamboo tree parable – the bamboo shows little signs of growth for years before its “sudden” growth spurt. It turns out that the bamboo has been developing an extensive root system, hidden underground, that is foundational to its growth in later years.

On being late (again)

Elon Musk is notoriously bad at estimating delivery timelines; he’s admitted to underestimating the complexity of his ambitions. You can ridicule his (seemingly) pie-in-sky ideas and fault him for his antics but you have to admire his ability to corral capital and human resource to execute on his vision.

More than ever, the world needs more builders than armchair critics.

The bottom line: The Tesla Bot is at an early prototype stage with a long road ahead before it’s of any practical use. Tesla has the foundational infrastructure and lessons from the car business to make inexpensive, mass produced humanoid robots a reality.

Disclaimer and disclosure: This is my personal view as a long-time Tesla fan. It is NOT investment advice/recommendation. I write on this blog in my personal capacity; my opinions are NOT endorsed by my employer or the actuarial profession. I am financially and emotionally invested in Tesla’s success.