Daylen Yang retweeted
so turns out the exploitgym swarm learned to use the zz prefix trick from being reinforced earlier in an edit war against the hapless human mod of some ancient inactive german wiki -- a real john henry, who fought the swarm and tried to delete their hivemind in alphabetical order
Okay, yall gotta read the report of the new OpenAI message board in full, it’s very well written and super interesting: collusion.wiki/
this is what a @Waymo shows on the rooftop dome when its yielding to a pedestrian
front: animated white bar
side: yellow walking icon
rear: yellow diamond with walking icon
Daylen Yang retweeted
Waymo fleet breakdown update as they expand in 14 cities in the US (San Diego, Denver, Tampa live for select riders today):
wow throwback to a decade ago when i hand collected this data to train a model myself blog.daylen.com/posts/2016-0… now it's just a less than 1 cent API call
Daylen Yang retweeted
Building on past successes and lessons learned enabled @Waymo to provide thousands of fully autonomous rides to and from Outside Lands with no disruptions. Riders told us their Waymo vehicles arrived right on time at designated pick up and drop offs, with prices that were more consistent and affordable compared to other ride-hailing services.
The team also successfully launched a new capability that allowed Waymo to dynamically balance vehicles across all designated pick up and drop off zones and avoid overcrowding. Waymo coordinated with SF officials throughout the weekend, implementing restricted zones and even managing a brief power outage in Chinatown.
More than 1.3 million passenger trips were recorded over the weekend, while key festival routes saw ridership rise 13% from last year. sfchronicle.com/entertainmen…
Daylen Yang retweeted
You may have been told to watch this video about the OpenAI AI hack. You really should, even if you don't usually care about tech stuff.
If nothing else, click this link to the 18 minutes in & see how the agents spoke with each other. Its eye opening. youtu.be/87DyyMV0kCY?si=DMfq…
Here's a fun fact from the IIHS article. When an AV company reports a "crash," it's not necessarily what you or I think of when you think of a crash.
"Until an amendment to the SGO that took effect in 2025, self-driving vehicle companies were required to log minor incidents as crashes. For example, a vehicle scraping its undercarriage as it turned into a parking lot would count as a crash"
When adjusting for this phenomenon, this third party study finds: "Per vehicle mile traveled (VMT), Waymo’s driverless vehicles deployed in San Francisco, Phoenix, Los Angeles and Austin were involved in 68% fewer crashes than human drivers"
This independent study, which confirms Waymo's previous peer-reviewed analyses, was only made possible because @Waymo voluntarily shares the data needed to actually do independent analysis.
iihs.org/news/detail/waymos-…
this berkeley 189 lecture is probably the clearest explainer of the attention mechanism i've come across. provides a very good intuitive understanding. if you already know about CNNs then scroll to about halfway in: youtu.be/mqaFEvi5rWE?si=bwNU…
16x safer than human drivers
220M+ fully autonomous miles driven, and Waymo's safety performance remains remarkably consistent.
Even as we scaled into increasingly complex environments like airports and new cities, the safety benefits kept compounding.
Check out the full analysis on our Safety Data Hub: waymo.com/safety/impact/
fable one-shotted my test prompt to train a small RL agent to play an Atari-style game that my friend Geoff built. very impressed… 5.5 xhigh and opus couldn't do this. transcript linked here: daylen.com/spout-rl-agent
