(Tech stock columnist Jon D. Markman publishes Strategic Advantage, a lively guide to investing in the digital transformation of business and society. Click here for a trial.)
Tesla (TSLA) is on the verge of a game-changing breakthrough in machine learning yet the only thing people are talking about is its plan for a stupid humanoid robot.
Executives at the electric vehicle company on Thursday held an artificial intelligence day. The two-pronged goal of the event was to show off its AI progress, and to recruit of new engineers.
Plans went awry when Tesla-bot, a 5’8” nonfunctional humanoid robot appeared on stage.
This is why investors should consider buying buy Tesla shares anyway.
Let’s be clear. The AI day presentation was mind-blowing. Tesla engineers are aiming so high it is hard to put the scale of innovation in perspective. Lex Fridman, an acclaimed AI researcher working at MIT and often a Tesla critic characterized the event succinctly:
“Tesla AI day presented the most amazing real-world AI and engineering effort I have ever seen in my life.”
In the past, Fridman criticized Elon Musk, Tesla’s chief executive officer for downplaying the difficulty of the full self-driving problem. In Fridman’s view, the obstacles to FSD are so daunting he didn’t believe any firm could successfully navigate the landscape within the next 5-10 years.
The Tesla AI day changed his mind. That is saying something.
Musk and his team completely reimagined computer vision by thinking exponentially bigger. Then they built models to collect and label the data, and a new processor to make sense of it all.
The idea of AI conjures up rooms full of computers deciphering data and making choices on the fly. In reality, Tesla still employs 1,000 engineers who manually label pedestrians and orange road cones. The latest software iteration is getting much better at auto-labeling. Musk said on Thursday that the neural network model is being completely retrained about every 2 weeks.
Processing is certain to improve when Tesla’s Dojo computer is outfitted with the latest in-house chips designed specifically to optimize neural networks. Engineers claim these breakthrough mega chips offer 4x the performance of the current processors yet they consume 1/5 the footprint.
Fridman believes the virtuous cycle of data collection, labeling, model retraining, and redeployment will give Tesla a real fighting chance to finally solve FSD. This is potentially a $1 trillion opportunity.
Unfortunately, this monumental development is getting lost in the analysis of the Tesla bot.
Investors should focus.
Tesla is building a fully integrated AI powerhouse. Longer-term investors should consider buying any near-term weakness in its shares.