An Atlantic writer claims that "studies have shown that automated vehicles are less able to detect people of color and children." It doesn't cite a specific study, but all the studies I could find focused on stand-alone vision models, not AVs.

Moreover, as @AVGregR points out, Waymo's vehicles have lidar sensors that detect people and objects by bouncing lasers off of them. Studying vision models doesn't tell you anything about whether lidar sensors will miss people of color or children.

@binarybits Unsupported vibes-based research citations are not acceptable at a high quality outlet (which @TheAtlantic mostly is). Especially given broad replication issues, you have to cite your behavioral science sources thoroughly so your use of that research can be rigorously assessed.

With access to X and web data, Grok is the AI assistant with the most up-to-date information, keeping you informed about the latest news and trends in the world, for any given topic.

@binarybits It seems like @TheAtlantic or Xochitl Gonzalez should either cite a specific study that says this or issue a correction.

@binarybits This is the quality I expect from The Atlantic

@binarybits As usual, the AI fairness research industry has a huge "compared to what" blind spot x.com/road_driving/s…

@binarybits I'm pretty sure they still do a better job than humans who also obviously have difficulty seeing dark skin at night.

@binarybits As always, the real answer is "compared to what?" Are these so against tech bros that they'll sacrifice tens of thousands of lives per year just to feel smug?

@binarybits @KelseyTuoc Atlantic article by Xochitl Gonzalez, whose previous article last month was the viral one in favor of smoking. Maybe she just has a fetish for people dying.

@binarybits Less able than what? In what conditions? If this a just a matter of physics then so what? And if it's still better than humans then so what? These muppets are just "disparate outcomes are racist" midwits, always interpreting data to match their preconceptions.

@binarybits @KelseyTuoc I wonder if it discriminates by seeing extremely pale Europeans better than more suntanned Europeans in dark conditions? Could this even apply with families?

42 agents. 216 threads. One dashboard. Every agent gets its own prompt, tools, skills, and budget. Deploy specialized agents across your company. From the team at Airtable.

@binarybits It's a shame this piece couldn't have tackled the conflict inherent in accepting the safety advantages are real, but also greatly valuing the human connection that comes with speaking with taxi/uber drivers.

@binarybits @colorblindk1d Human drivers also have a harder time seeing a black guy in black clothes at night.

@binarybits This seems like nitpicking. The main problem with the reporting is that AVs are much safer for both genders and any discrepancy is minuscule in comparison but the story reports it as a significant problem.

@binarybits It’s so poorly phrased. Do the studies compare AV detection of poc and children vs AV detection of whites and adults, or AV detection of poc and children vs human driver detection of poc and children? Big difference!

@binarybits Also, in the self driving world, months matter. Studies of years old tech are useless

@binarybits @KelseyTuoc Also, they make it sound like the cars will be less likely to detect them than human drivers. But the research would in any case just compare to detection of other persons by the car. Drivers, too, find it harder to see smaller people or darker faces in some conditions.

@binarybits It’s almost like the Luddite Atlantic writer is biased against AVs Shocking

@binarybits this is like a plot point from the Will Stancil Show what the hell