Tesla CEO Ilan Musk claims Tesla will become the world’s most valuable artificial ielligence company in the future thanks to the huge volume of video data collected by its cars, but do such data really give a competitive advaage?
If you believe in Ilan Musk, you should no longer consider Tesla as an electric car manufacturer and we should look at AI’s ielligence company; The mask trusts the data collection that has been collected from its cars. The volume of data collected from the Tesla cars has reached several Publies; Because customers of Tesla cars are driving millions of miles around the world every day.
Why is Tesla really a big artificial ielligence company?
In theory, all of these real -world data are exactly what Tesla needs to train their cars to become self -help without the need for human assistance; Tesla’s car training for self -employme is the focal poi of the Mask’s vision for the future of Tesla, but there is a problem with these; This data is not necessarily as much as the mask claims, and some are not effective at all.
Construction of an artificial ielligence system that can drive well is a challenge that is significaly differe from the manufacture of natural language processing chats such as ChatGPT; ChatGpt with billions of words trained from the Iernet. The purpose of creating ChatGPT and competing systems such as GROK from XAI is to use a pattern recognition approach to provide reliable information and answer questions. In the worst case, the results are often very disappoiing, but if the vehicle’s artificial ielligence is wrong, human lives will be in danger.
Driving a car is a completely differe issue that is affected by a variety of variables; Variables such as driving conditions, climate, construction, change of traffic patterns and how to move other vehicles. Successful manageme of all these variables and readiness to respond to unexpected eves is the main nature of artificial ielligence responsible for automatic driving. AI training with endless driving videos on highways does not help to learn the most importa things that this technology needs to know; These data can only help artificial ielligence in learning rare cases that lead to accides or dangerous situations.
This technology can make your driving very smoothly, but when the situation is a little abnormal, you no longer have anything in the bag.
This is a computer scieist and executive director of the Technology Company who did not wa to be named; Because he did not wa to criticize Tesla publicly. He we on to say:
You have only learned bad habits. 3 out of every 2 people pass through the stoppage. If you only model people’s behavior, you will pass the stand boards.
For this reason, Tesla’s robotic taxi competitors use laser Lidar to capture 3D and radar images to detect solid bodies in the car to gain richer and more accurate images of the surroundings. It may be possible to use camera data merely, but you need “the best camera systems to manage it”. Drago Anguelov, Head of Research, Waymo Taxi Company (Waymo), said at a Google Developer Conference a few years ago:
This is a great bet that you can do. This is very risky and you don’t need to do it.
Yann Lecun, a senior scieist at Meta and New York University’s computer science professor, also gives Tesla data competitive advaage. He said about the subject:
The impact of data is generally assessed overly; The more data you have, the improved the performance, but the efficiency is reduced. Double -tightening data only creates minor improvemes that are still very distrustful.
Still, no so -called self -expanding company has developed Level 2 data; It means a 2 -step level in which the vehicle can act independely in all situations where one can drive.
Lecun added:
Any 4 -year -old teenager can learn to train for about 2 hours. This shows that curre artificial ielligence architectures have little capability in understanding the world and learning from limited data or experience.
Although the sale of electric cars and the value of the company’s stock is still collapsed, none of these issues have made Tesla supporters stop betting on the vision of the mask’s artificial ielligence. Protesters against Tesla stores have not disappoied supporters in the Preside of Trump’s coroversial Preside’s Preside Mask’s role in the DOGE Plan. Some stock market analysts still think the mask knows something that others don’t know.
Wedbush Securities from Wedbush Securities Financial Services Company.
We believe that self -technology alone is worth one million dollars and this hypothesis will be proven in the coming years.
Mask and Tesla did not respond to requests for comme. Wedbush Securities, the head of the Tesla Self Program, also did not respond to the issue.
Ilan Musk’s vision for developing artificial ielligence systems in Tesla
Mask has tied Tesla’s upcoming artificial ielligence applications, including human robots and smart factories, and has abandoned the company’s long -term goal of selling 2 million electric cars a year to 5. One of the possible reasons for this change is the increase in competition in the electric car market, especially with companies such as China’s BDD. Another reason is that if Tesla can solve the problem of driving, the use of hundreds of thousands of electric robotics taxis would be much cheaper and more profitable than establishing more factories to produce and sell millions of personal cars.
This is so esseial for Tesla that the mask does not wa to buy the company’s shares. Mask said in a call to receive a year’s income report with him: “
If one does not believe that Tesla will solve the issue of self, I think he should not invest in this company.
The January Mask announced that Full Self-Driving or FSD software uses a massive Tesla data tank in the company’s new data ceer called Cortex in Austin. Driving, despite his name, still needs human supervision at all momes. This capability based on artificial ielligence and the initial Autopyott Tesla system must definitely improve; The Tesla’s complete driving system and autopilot have been linked to four deadly accides around the world over the years.
Alex Ratner, a computer scieist and CEO of Snorkel AI, a software manufacturer of raw data labeling software, said:
The huge volume of camera data is useful, but these data do not spoaneously become the leader of the artificial ielligence market. “Access to unique data sources is definitely an advaage.”
Ratner, a member of his family in Wimo, told Forbes about the use of data to teach artificial ielligence:
When choosing and managing data, we have to say that the video received from a good driver or a bad driver is not an easy issue and is extremely importa; Because these models learn from the most common things they see.
Companies that have been developing artificial ielligence for safe driving cars and trucks for years, such as Aurora, Zoox and Waabi, have focused on creating high quality data, including sufficie number of rare cases. They use advanced computer simulations and real -world structured experimes to master dangerous or unusual road situations. Tesla’s data is not necessarily a reflection of these eves.
Missy Cummings, an artificial ielligence specialist and professor at George Mason University, who has given federal institutions and California lawmakers in the field of automotive technology, says:
There is no guaraee that the data is real to teach all the rare and complex things that cars need to learn to lead to real learning.
This makes it difficult to solve the problems that car developers have faced when ideifying road conditions as a risk by artificial ielligence; Problems such as sudden and unexpected brakes of the Tesla Phaom Brake System.
Even ideifying the most importa parts of driving data for education, from the endless volume of road videos, is a very difficult task. A car researcher and computer scieist who is familiar with Tesla’s approach but did not wa to be named, said: “
So you have a huge amou of driving data, but how do training experts make sure that all the importa parts of the model are selected?
It is difficult to diagnose; Because Tesla is not clear about its process. Engineers of other major technology companies are publishing articles regularly and explain their latest research, but Tesla cannot be considered an active member of the AI research community.
Ian Lecon said about Tesla’s activity in the artificial ielligence world:
Tesla does not participate in the world of artificial ielligence research and developme, conferences, scieific articles, etc. It is as if there is no external existence at all.
A record full of Tesla’s defeat
Tesla’s achievemes in driving are repeatedly lagging behind the goals that Ilan Musk has announced.
Year 2, Mask promised to drive the Tesla car without human ierveion, but the promise has not yet been fulfilled.
Tesla has not even closer to the goal of setting up one million robotics taxis uil year 2.
Yan Lecon said:
Ilan Musk constaly makes great and exaggerated claims, and their results have been much lower than expected, almost a decade. We have witnessed such a thing for many of us that it was clear that all these claims were either lies or a sign of the mask’s illusion. I don’t understand how anyone can still believe his claims.
But this has not been able to preve the mask from promising more or more, or the fanatical supporters discourage him from coinuing to invest. The examples he has shown so far are still far from reality.
Last October, his company’s robotic taxi mask, Cybercab, unveiled. Tesla’s taxi was displaced by the participas in the Universal Studios film studios in Los Angeles, but even under such situations where the taxi was closed in the studio, Tesla technicians were preseing and monitoring driving or even low -speed prototype corol.
Tesla’s Optimus’s Optimus Robot versions were also remotely corolled for the participas.
Mask said about the Optimus robot in a call to get the financial outcome.
I think Optimus has the poteial to earn more than $ 5 trillion in the long run, it’s really crazy.
Of course, the actual Tesla Robotics Taxi test will be conducted in June with the Tesla Robotic Taxi Taxi Service in Austin; Provided to be launched on time.
Mask in calls to get financial outcomes said:
We examine it very carefully to make sure we haven’t lost anything. The car real service will first be launched to earn money in Austin, then will be unveiled in other US cities as soon as possible.
Tesla has also applied to set up a similar taxi service in California. The service is run by cars whose owner and operator are self -sufficie, not robotic taxis.
Yan Lecon said about Tesla Company:
Tesla cannot produce self -independe cars. The company does not have a research company with sufficie freedom and scieists with sufficie tales to achieve this goal.
Musk could not lag behind in the competition with Wimo, which is a vanguard of robotics taxi, and Vimo has been lagging in Phoenix, San Francisco, Los Angeles and Austin for some time.
Wimo announced last moh that with a fleet of about 5 cars, it would make more than 6,000 trips a week. The company began operating in Atlaa at the end of this year, and its officials plan to launch their service in Miami next year. Alphabet has not disclosed Wimo’s revenue, though it is estimated that its revenue was $ 5 million with more than 2 million trips per year.
There have been no fatal accides for the Wimo Robotics Fleet so far, but we have seen minor accides associated with them. At the same time, the Tesla owners regularly upload videos of their cars in full driving mode, in which the car performs dangerous movemes; Such as crossing the red light on the outlet of the freeway in New Jersey, which was close to an accide or a red light in China.
Finally, the vision of the Mask’s artificial ielligence for Tesla depends on how much it will be financially valuable to its company; Because he predicts he can earn trillions of dollars in the coming years with artificial ielligence.
Ian Lecon, a Meta company employee, believes that in order for cars to learn from the video, how the world works, a major change must be made, and that it must be investigated to make at least another decade. He said about the subject:
I feel that we do not achieve the aim of building cars whose systems can corol themselves just like humans, and the construction of practical humanist robots, unless we understand how artificial ielligence systems can be trained in a way that is like animals and humans.
