General Motors announces the termination of the robotaxi project, Musk: Low-cost autonomous driving solutions are extremely difficult

General Motors' subsidiary, Cruise, which is responsible for the robotaxi project, has accumulated losses of approximately $10 billion over the years.

On the 10th, General Motors announced that it has stopped funding Cruise, citing increased competition and costs, and will exit the robotaxi business to focus on the development of driver assistance systems for consumer vehicles.

In the announcement, General Motors stated:

General Motors plans to readjust its autonomous driving strategy, prioritizing the development of advanced driver assistance systems to achieve fully autonomous personal vehicles; considering the significant time and resources required to scale the business and the increasingly competitive robotaxi market, General Motors will no longer fund the development of Cruise's robotaxi.

General Motors acknowledged that Tesla's end-to-end autonomous driving solution is the correct strategy.

This is indeed the direction of industry development. Cruise has already begun to move down this path. We are advancing towards foundational models and end-to-end approaches.

Elon Musk quickly commented:

"Achieving a general solution for autonomous driving is a very difficult problem, especially while ensuring that cars remain prohibitively expensive."

Cruise founder Kyle Vogt complained:

Even if it wasn't clear before, it is now clear: General Motors is a bunch of fools.

Kyle has shown a strong interest in technology, especially autonomous driving systems, from an early age. He graduated from the Massachusetts Institute of Technology, where he earned degrees in computer science and electrical engineering.

Before founding Cruise, Kyle co-founded Justin.tv, which later evolved into Twitch, a popular live streaming platform.

Kyle founded Cruise Automation in 2013 and played a key role in the company's development, serving as president, CEO, and CTO.

In 2016, General Motors acquired Cruise Automation for over $1 billion, making Cruise a wholly-owned subsidiary. This acquisition is part of General Motors' strategy to leverage autonomous vehicles to drive future transportation In 2020, Cruise began testing autonomous vehicles in cities like San Francisco and was granted permission to operate robotaxis in San Francisco around the clock in 2023.

By the end of 2023, Cruise faced significant setbacks due to safety and regulatory issues, particularly following incidents involving pedestrian injuries. This led California regulators to suspend Cruise's autonomous testing and deployment permits.

As a result, Kyle resigned as CEO of Cruise in November 2023, and other leadership changes followed.

In 2024, Kyle founded The Bot Company, focusing on the consumer robotics sector, indicating that his career focus has shifted from autonomous vehicles to home automation.

George Hotz, the founder of the autonomous driving company comma ai, also commented on the fate of Cruise while promoting his own company.

Today, General Motors withdrew funding for the Cruise autonomous taxi project. I predicted this five years ago.

If you've been paying attention, the development of autonomous driving is as I said, Tesla will come out on top, and comma ai will take second place with its open-source solutions, just like the iOS and Android markets in smartphones. Today, Tesla has the largest fleet, and we have the second largest fleet.

Many autonomous vehicle startups have gone bankrupt, leaving with nothing. Ghost Autonomy raised $240 million, valued at $0! Argo AI raised $3.7 billion! Valued at $0!

Investors are now wasting billions of dollars on humanoid robots. Who can stop this train wreck? It's as obvious as last time.

In terms of scalable technology needed to solve autonomous driving (and eventually humanoid robots), comma has gone the farthest. Our end-to-end approach is in use, with over 10,000 people currently using it for driving. You can purchase it on our website! We also have SOTA learning world model simulation technology, which is much purer than what Tesla is doing. Autonomous driving is still a few years away from an AlphaGo moment.

Once again, no one will believe me because hype doesn't require belief. If you believe me, you wouldn't allow brain-dead idiots to pour billions into massive garbage. What is needed are 10-100 very smart people to carefully fix the flaws in a complete end-to-end software technology stack. That's the problem By the way, if you are one of the 10-100 smart people who really want to solve the problem of autonomous vehicles (rather than what Cruise has been doing for the past 10 years), come work at comma. We have the most straightforward hiring page I've seen, with challenges and rewards based on bounties. comma.ai/jobs.

George Hotz, online alias "geohot," is a famous American security hacker, software engineer, and entrepreneur. He rose to fame in 2007 as the first person to unlock the original iPhone and for jailbreaking the PlayStation 3 in 2010.

Since then, George has moved into the field of artificial intelligence and autonomous driving technology. He founded a company dedicated to developing open-source autonomous driving systems called comma.ai, with its flagship autonomous driving product named openpilot.

Netizen Jack Ryhsider asked: How can you become number one? Does it require too much effort and money? Or are you already falling behind in this competition? George replied:

About 10 years ago, I met Elon. His thinking was very clear, and all his ideas were the same as mine, so there was no advantage in ideas.

Tesla has brute force, a 9-9-6 culture, and "what did you accomplish this week," while comma values lessons learned, restructuring, and capital efficiency.

On a per-dollar basis, we crush Tesla. But they can solve many more problems than we can, and unlike companies like General Motors, their direction is basically correct.

Winning second place is still a big victory. Tesla is vertically integrated closed source, like iOS, while we are integrated open source Android.

Elon jumped in and replied to George:

Tesla's reasoning efficiency is absurdly high because we are limited to about 200 watts of onboard computing power and cannot really brute-force problems with thousands of watts of Nvidia GPUs.

Clearly, Elon's reply quickly ignited George's competitive spirit:

The power consumption of the comma 10-nanometer chip is only 10 watts!

If you haven't tried openpilot yet, give it a shot. Its functionality is not as powerful as FSD, but the experience is more comfortable, more cooperative, and has lower maximum torque. Additionally, you should switch to our driving monitoring feature.

Our computing cluster has 500 GPUs, compared to your 50,000 H100s, your training capability is 100 times ours.

And most of our cars are much worse in controllability. Among all the cars we tested, Tesla's driver latency is the lowest, as shown in the comparison between the Model 3 (left) and our flagship model Lexus (right)

Of course, I don't mean to use this as an excuse for myself. We live in a fair world, and I have complete autonomy over my decisions not to invest in the following areas.

1. For training computing power, we can purchase clusters just like you, or at least build one. If we find a way to confidently scale our capabilities through computing power, we may need to do so. But we haven't found anything worth investing in here yet, and we are starting to change that.

2. For inference computing, the Tinygrad project is developing a user-space GPU driver that will run via USB. Once we want to release larger models, we will sell an external GPU of about 200 watts that can be placed under the seat and connected to comma 3/3X. This will provide us with similar inference capabilities.

3. Regarding the issue of actuator latency, openpilot can run on Model 3 :)

George Hotz and Elon have a connection. Their relationship can be traced back to at least 2015 when George met Elon through a mutual friend. Elon was interested in George's work in artificial intelligence and autonomous vehicles. They discussed the possibility of George working at Tesla, but reportedly, George declined Elon's invitation because he felt that Elon kept changing the terms of collaboration.

On the autonomous driving solution, George Hotz has consistently supported abandoning the lidar + high-precision map approach.

In 2022, after Elon acquired Twitter, George interacted with Elon on Twitter, which led to George eventually joining Twitter for a 12-week "internship" focused on improving Twitter's search functionality.

Elon was very happy at the time, posting: "World-class software talent is joining Twitter."

The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Invest at your own risk