ctsats

On Sep 6, Sam Altman tweeted:

Analytics India Magazine has a convincing piece claiming that their next steps will be GPT-Vision and similar multi-modal models: https://analyticsindiamag.com/why-time-is-ripe-for-the-real-gpt-4/

Passage of time

Passage of time

I have just received an email from the California Department of Motor Vehicles (DMV), in response of the clarification request I had sent them a few days ago (see here for the background: https://www.infer-pub.com/comments/98518). I copy it here in its entirety (emphasis mine):

Greetings Christos,

The California Department of Motor Vehicles administers the Autonomous Vehicles Program and issues permits to manufacturers that test and deploy autonomous vehicles on California public roads.

The DMV’s Autonomous Vehicle Branch administers three permitting programs: Manufacturer’s Testing Permit – Drivered Vehicles, Manufacturer’s Testing Permit – Driverless Vehicles, and Manufacturer’s Deployment of Autonomous Vehicles.  

To encourage innovation and promote road safety, the DMV established regulations governing autonomous vehicle testing and deployment on California roads. https://www.dmv.ca.gov/portal/file/adopted-regulatory-text-pdf/

Manufacturers who are testing autonomous vehicles need to report any collision that resulted in property damage, bodily injury, or death within 10 days of the incident.

While the DMV doesn’t require this collision reporting under the deployment permit, manufacturers still must submit a Report of Traffic Accident Occurring In California form (SR-1) and also report collisions to the federal government under the National Highway Traffic Safety Administration’s standing general order.

Thank you,

Autonomous Vehicles Branch | Policy Division
California Department of Motor Vehicles

So, it would seem that the vehicles of Waymo & Cruise that are deployed under their recent permit are not under the obligation to submit these special collision reports which end up in our resolution source and affect the outcome of the question; and relevant reports by these 2 companies will only be about vehicles under testing conditions.

Unsurprisingly, these operators/manufacturers are still obliged to submit the standard accident reports, but it would seem that these reports will not end up in our resolution page.

How does this affect the question here?

Many of us operated under the reasonable assumption that, since we now have 2 companies with deployment permits, the effective number of cars and mileage is expected to go up, compared with the previous regime where the only vehicles on the roads were for testing purposes; hence, the number of collisions was also expected to go up, due to elementary statistical reasons (more cars naturally lead to more expected incidents, all other being equal).

But this assumption does not seem to hold, under the clarification above.

So, I think the question now is - can we expect that Waymo & Cruise, under their new deployment permits, will reduce the cars that they have on the road for testing purposes, thus possibly driving the collision reports down? And if yes, by how much?

I confess that I have no clue at the moment; feedback welcome...

(I am pinging @henrytolchard & @dante, since the clarification above may also be relevant to the spirit of the question)

With "Data as of June 13", the same people that maintain our resolution source at Brookings show this graph:

in a Washington Post article of July 10, unearthed by @peregrine below: https://www.washingtonpost.com/opinions/interactive/2023/ukraine-war-analysis-stalemate-economy-aid/

The graph implies that the actual percentage for June is 17.7% (the Brookings chart only displays integer readings). Hanlon says

Russia still holds just over 17 percent of Ukraine

Could someone confirm it? (I am accessing the page through the Internet Archive...)

cc @peregrine @404_NOT_FOUND