tl;dr: Berkeley researchers trained models on a possibly-accidentally-leaked Chinese dataset -- images of US naval ships, suspiciously labeled with bounding boxes around their radar systems.
They're making the point that training models like this can give some intelligence towards another country's ML capabilities (here, China). In my opinion, a cool work that has way more "contribution" than your average AI conference paper.
(And full disclosure: I contributed a bit of feedback on their writing.)
That's really strange and incredibly frustrating - but slightly less so if it's consistent with all of the bars (including their own).
I take issue with their choice of bar ordering - they placed the lowest-performing model directly next to theirs to make the gap as visible as possible, and shoved the second-best model (Grok-1) as far from theirs as possible. Seems intentional to me. The more marketing tricks you pile up in a dataviz, the less trust I place in your product for sure.
Is the fall detection alt-text actually that good here? There were more elements in that image, which I assume people who use screen readers or are reading the alt-text for broken images would want to know.
The "EMERGENCY SOS" slider and "I'm OK" button give more examples of the "straightforward and direct" language that the article text references. I also learned something about that feature itself (besides that it existed in the first place) - Apple's design choices to make the "SOS" a slider, followed by a larger/easier to press button for "I'm OK". Even though it wasn't related to the point of the article, it was information that I wouldn't have learned had I just read that alt-text.
Is this part of accessibility guidelines for alt-text? Shouldn't they convey the same information, whether it's in image or text form, even if it's not directly relevant to the point of an article?
I can also imagine that people have different preferences - maybe some want all the information like I mentioned, whereas others don't want to distract from the point of what they're reading. I wonder which way the alt-text guidelines lean in practice.
What sorts of doctors can dictate their own schedules, work minimal hours, or take months of vacation each year? If they exist, surely they're firmly in the minority.
They're making the point that training models like this can give some intelligence towards another country's ML capabilities (here, China). In my opinion, a cool work that has way more "contribution" than your average AI conference paper.
(And full disclosure: I contributed a bit of feedback on their writing.)