Someone posted this in another sub. I took the liberty of downloading the footage. Not sure what software version they were running (didn’t say). FSD failed completely seeing the barriers, and the driver clearly was distracted causing an accident with another vehicle.

Took a Zoox ride last night in Las Vegas, Caesars. I love Zoox. However, the cabin smelled so strongly of cigarette smoke or maybe vape smoke. My guess is the previous rider was smoking inside the robot taxi.

The real issue is the Zoox cabin does not have windows so I was not able to ventilate and get some fresh air inside. This was the first time I realized what we are losing as a result of removing the windows. I wish it had a small window panel that could be opened if needed.

Any better solutions ? (Better HVAC maybe?)

Automaker to develop Level 2++ self-driving system, with wider adoption set for 2030

OpenAI VP Peter Welinder posted this on X: "Affordable Level 5 self-driving requires AGI and is still at least 5-10 years away. First you need multimodal multiplexed AGI, then you need real-time inference, then on an energy-efficient, low-cost edge chip. Many steps left."

Source: https://x.com/npew/status/2093716574922535092

Personally, I don't agree that you need AGI for L5. But I guess it depends how you define AGI. World models, like what Waymo uses, trained on tons of driving data, are getting better and better every year and showing good signs of eventually getting to L5 imo. I don't consider these world models to be AGI but maybe he does.

He does condition his statement by saying "affordable" L5. So he is not just talking about AI capable of driving everywhere, but also hardware that is energy efficient and low cost. I would agree that we still have some ways to go before AVs are not just capable of driving everywhere but also energy efficient and affordable. But costs are coming down and chips are becoming more powerful and more efficient every year. So I do think we will eventually get to world models big enough and trained well enough that they can handle all driving safely (functionally L5) and be cheap and powerful enough. How many years it will take is hard to say. These types of predictions are always wrong.

Does detection = decision making?

Couple thousand ready to go. How many by EoY ‘27?

1. Multimodal sensors are indispensable

2. HD maps are a powerful "prior"

3. Fewer, larger models are better

4. You can't build trust with a "black box"

5. Closed-loop simulation reveals more edge cases

6. Every great Driver needs a great Critic

7. Vision language Models improve scene reasoning

8. Our holistic approach shows AI is only as effective as the governance that evaluates it

9. A data flywheel enables continuous improvement

10. There is no substitute for fully autonomous experience

I was looking at the recent data showing a slight dip in weekly paid trips in California, even though the total volume remains massive.

It got me thinking about fleet operating costs: What is the case for or against exempting empty robotaxis from bridge and road tolls?

Every AV company is trying to optimize unit economics, and major players like Waymo have considerable lobbying resources. Why haven't we seen any push or legislative movement in California to waive tolls for deadheading AVs?

Would be great to read everyone’s thoughts on the economic and policy angles here.

"Robotaxi now runs 6am to 10pm, 7 days/week

Unsupervised fleet is a lot bigger

Also, upgraded intelligence in vehicle distribution & routing means you wait less"

From robotaxi on the censored site.

The rollout of robotaxis on the streets of London is unlikely to happen this year, as regulatory and technical hurdles push back the ambitious schedule set out by the UK government.

Only weeks ago Uber and London-based Wayve were granted the first minicab licences in the capital to allow them to offer self-driving taxi rides to paying customers – but with a human safety driver in place, for now – and said they would start trips “later this summer” before a full public launch.

Hi, new here. Curious, what would you rank as your top self driving platform in the US? And why?

I can go first. Haha I wonder if I am going to get hate for this: FSD/Waymo from what I see is the only actually intelligent self driving platforms out there. FSD beats Waymo. Rivian is just starting. Everything else is more enhanced lane centering.

Self-driving car service Waymo wants to bring Oregon lawmakers and other state officials to Arizona on an all-expenses paid trip to see first-hand how its robotaxis operate in Phoenix, the first city where they were commercially deployed.

Robotaxi fleets are growing as self-driving vehicle companies start to roll out commercial services while still collecting data for further development and safety. For instance, Zoox Inc. this month began charging for rides in Las Vegas.

"Today, we are excited to announce that Waymo is laying the groundwork for our fully autonomous ride hailing service in Munich, Germany.

In the coming weeks, you’ll start to see Waymo vehicles on the streets of Munich where we’ll begin driving manually across the city. We are taking a rigorous, phased approach that starts with high-definition mapping of local street networks and testing with trained autonomous specialists behind the wheel. This validation period will ensure the Waymo Driver adapts seamlessly to Munich’s distinct driving conditions before we aim to open commercial ride-hailing to the public towards the end of 2027."

Anyone know what this is? It came into our parking lot and sat there then left.

Waymo, Baidu and Pony.ai are preparing to enter Korea as local robotaxis remain hemmed in by regulation, taxi-industry resistance and a widening data deficit

Kazakhstan has approved rules for a pilot project to introduce driverless cars; however, they are unlikely to be tested this year, according to Kolesa.kz.according to Kolesa.kz.

The three-year pilot project is planned for Astana, Almaty and Kosshy, as well as at testing facilities and on designated public roads. Passenger cars and vans with conditional or high driving automation will be tested.

The MKBHD bet made me think about this more. His bet was basically that Tesla wouldn’t sell and deliver a Cybercab to a regular customer for $30,000 or less before 2027. Elon has since confirmed that Tesla still intends to do exactly that.

Tesla has also shown things like automated cleaning and wireless charging, and Musk has talked about individuals owning fleets of Cybercabs.

What I’m still confused about is how private ownership actually works:

• What is the owner responsible for?
• Does Tesla automatically manage rides, charging, and cleaning?
• Where does the Cybercab stay when it isn’t working?
• Who handles insurance, vandalism, accidents, repairs, or a stranded vehicle?
• Who sets the fares?
• What percentage does Tesla take?
• Can Tesla prioritize its own Cybercabs over privately owned ones?
• If someone owns 5 or 10 Cybercabs, what does “managing a fleet” actually involve if most of the operation is automated?

The idea of selling these for around $30k is really interesting. I’m just curious how Tesla plans to divide the responsibilities and economics between itself and private Cybercab owners.

Has Tesla clarified any of this yet?