Translation:
Body:
1,748 words in total.
Estimated reading time: 8 minutes.
Tesla of the future is: a vertically integrated automotive company with software and hardware integration, an energy production and charging network company, and a mobility technology company.
Both the previous article and this one are about the Intelligent/Autonomous Driving segment.
This segment involves what Tesla can achieve in the future – whether it will be (1) advanced driver-assistance or (2) complete autonomous driving. If readers have read the previous articles, they should know that this problem also determines whether Tesla’s long-term future is a company with a market capitalization of over trillions, or nearly hundreds of trillions. In fact, it is whether Tesla can become a mobility technology company.
The previous article talked about it. We will not do a specific technical analysis here. Studying technology can easily fall into precise errors. I believe that perhaps, even the top researchers in the industry at this time would find it difficult to reach a conclusion that is close to the truth. Therefore, we turn to studying the issues of research and development teams and resources, which are the “seeds” and “soil” of “growing” cutting-edge technology. Studying these “seeds” and “soil” should yield a fuzzy correctness, bringing us closer to the truth.
In the previous article, we studied the resources that Tesla possesses, that is, we studied the “soil”. Based on the basic formula of better technology = algorithm + data’s autonomous driving technology, we found that Tesla has a huge fleet of vehicles that is closest to real-world driving scenarios, providing a large amount of high-quality data to enrich the algorithm framework, ultimately making its technology better and more reliable than its peers.
Just like a person’s cognitive level is a collection of people he has met and things he has read, artificial intelligence is somewhat similar.
At the same time, Tesla foresaw the huge demand for computing power to train large amounts of data, and developed dedicated chips and next-generation supercomputing platforms to train visual data. With the same capital investment, training data of the same scale can be done four times faster than its peers. When you iterate faster than others, accumulating time over the years can make a big difference.
All right, this is the “soil”. Next, we will study the team, which is the “seed”.
To study a team, we look at it from three aspects: whether it is at the top of the industry, whether it believes in this direction, and whether it often wins.
Tesla’s Autopilot team has been researching in the direction of vision artificial intelligence since 2017 and in terms of vision artificial intelligence technology, it is widely regarded as the best in the industry. Therefore, it can be said that if we look at Tesla’s entire Autopilot research and development team as a whole, this team is at the top of the industry.Let’s take a look at whether the Tesla team believes in the direction of achieving autonomous driving through pure vision. This is not to say that Tesla’s pure vision approach is necessarily correct, and that other autonomous driving companies using other sensors are incorrect. Rather, it is an analysis of whether an individual or team believes in this approach, and the conclusion reached is whether it can maximize their personal and team’s subjectivity. Only when you truly recognize and believe in something will you invest all your resources to achieve it; otherwise, it’s just going through the motions. Ultimately, those who genuinely believe will succeed sooner or later.
In the first half of this year, Tesla canceled its last sensor except for cameras, the millimeter-wave radar, in North America. From Tesla’s various external decisions and performances, it seems that they believe pure vision is sufficient to achieve autonomous driving, and the rest is just a matter of time. It is difficult to know from the outside whether Tesla’s internal team is committed to this pure vision direction. However, by looking at someone’s interviews and daily tweets, you can get a better understanding of their thinking and true thoughts. The advantage of an interview is that the person being interviewed is essentially in real-time response mode; often, you can judge their true thoughts through their facial expressions combined with what they say. I recommend watching an interview with Andrej Karpathy, Tesla’s Director of Artificial Intelligence and Autopilot Vision, and Pieter Abbeel, the Director of the Berkeley Artificial Intelligence Research Lab at UC Berkeley. After watching the interview, I got the feeling that Andrej Karpathy is very confident that the vision path can achieve full autonomous driving.
Finally, it is important to see whether victories are often won. It depends on whether there is an opportunity for product and technology practice. Only with practice and battles to fight, there can be victories, as well as a sense of accomplishment and morale that comes with winning. Tesla’s autonomous driving technology can be deployed to the limited-range FSD Beta fleet or to the global Autopilot fleet after testing within a limited range is mature. Regardless of which type, these are vehicles driving in the real world, providing plenty of opportunities for practice.
Furthermore, Tesla currently releases an FSD Beta and OTA upgrade every 2-4 weeks on average. The rapid progress of FSD Beta has surprised many testing car owners. Of course, many things are not done well in the middle. However, these problems are gradually resolved when the next version is released, and many real users online praise the development team when they test again. This is what it means to win, to enable the team to establish a sense of accomplishment and morale.
A team that grinds its gun but never shoots will gradually wear down its morale, and constantly heading in the wrong direction or choosing the wrong technical path is also a one-step forward and two-steps back situation.Therefore, I believe that being at the top of the industry (professionalism), having faith in the direction (belief), and achieving frequent victories (practice) are the key principles for identifying good “seeds”. Being at the top of the industry represents leading technological and intellectual capabilities, having faith in the direction represents maximizing subjectivity, and achieving frequent victories represents positive feedback, a sense of achievement, and morale. The combination of these three factors makes for a super-strong team.
Moreover, if we can match it with the appropriate “soil”, this towering tree will eventually break ground and rise up. Any force can only delay, but cannot stop it.
(to be continued)
This article is a translation by ChatGPT of a Chinese report from 42HOW. If you have any questions about it, please email bd@42how.com.