The writing was on the wall. The suspension of Robotaxi permit is not surprising at all. The required AI technology for autonomous behavior simply doesn’t exist to be used in FSD. AI so far has reached only the level of automation for narrow domain tasks. The task of detecting various surrounding objects from video training data is an example of narrow AI automation. However, the task of making realtime decisions independently in a holistic way based any surrounding road and traffic condition with safety and self preservation in mind requires autonomous AI agents, which is beyond the capabilities of current AI technology. Big data driven Machine Learning isn’t enough for such capabilities. With ML alone, it will be a never ending chase for training for the next OOD edge case encountered. Aggressive , ruthless marketing for FSD isn’t the way to get there. #ai #fsd #robotaxi #automation #autonomous
BREAKING: The Verge reports "California DMV suspends Cruise’s #robotaxi permit effective immediately." Among the reasons cited for the suspension: Cruise’s vehicles “are not safe for public operation”; the company allegedly “misrepresented” the safety of its vehicles; Cruise’s vehicles could pose “an unreasonable risk to the public”; and permits can be revoked “if a manufacturer is engaging in a practice in such a manner that immediate suspension is required for the safety of persons on a public road.” Cruise can still test its vehicles with human safety drivers behind the wheel. Comment: It would have been better if Cruise had made this decision itself, but kudos to the CA DMV for having the gumption to act. Waymo's operations are unaffected, but we shall have to see if this is the beginning of the end for the robotaxi. https://lnkd.in/dkXxqSaX
Could you imagine if we applied the same driving skill requirements to humans? Very few people would be allowed to drive. I think full autonomous should be allowed as soon as it's better than the average human. Otherwise, let's make it a whole lot harder to get a license. These cars don't speed; they don't drive drunk, and they don't text while driving, they don't get distracted, and they don't get tired, they also don't road rage on each other.
It was time to end this spooky self-driving myth.
I bought a Tesla model 3 and drove it 5 days from NY to Tucson, AZ and it did > 95% of the driving all by itself. It was very impressive. And it wasn't just the highway driving. Certainly agree that the long-tail of issues will be challenging, but if it deals with them in a safe way, it may not matter if it handles them in the most ideal way. https://www.linkedin.com/posts/troycollinsworth_i-bought-my-son-a-tesla-model-3-for-college-activity-7119361659147583488-CrjN?utm_source=share&utm_medium=member_desktop
I remember a few months ago the Italian GDPR authorities banned ChatGPT. Their point was not anticapitalism or protectionsm. It was safety. OpenAI had made a typical startup mistake in not thinking about global compliance: it had not created the means for under-aged children to sign up and be exposed to the technology. It’s a kids’ game: many under-aged teenagers lie and open Instagram and Facebook accounts - you may argue. True. But in those cases, you’ve lied and probably don’t have parental approval…. So the company cannot be liable (your legal tutors may be).
Technology evolution along the way of life and services are very much sensitive to lives of the planet of our mother Earth for the good 😊 of the product of our mother Earth for the good 😊
Wonder how Waymo has got the permit for driverless Uber rides in Arizona... after all these mishaps?
Did they cause accidents or injuries?
Well said!
AI Consultant || MIT Alumni || Entrepreneur || Open Source Project Owner || Blogger
9moIt’s important to be cognizant of the difference between AI automation and AI autonomy, especially for FSD. Autonomous behavior does not just emerge from a collection of independent narrow AI task automation. It’s greater than the sum of individual parts. https://fchollet.substack.com/p/ai-is-cognitive-automation-not-cognitive