This is all really interesting. What does it imply for the current lawsuits over whether training diffusion models on publicly available images violates copyright? Because you seem to be agreeing with the plaintiffs that diffusion models are just mixing and matching from the training data.

Less. It would reduce the number of cyclists. Part of the problem is that we have bike lanes on some main streets, but not on secondary streets that dont have room. So the cyclists (and scooter riders, who are worse) ride the sidewalks to get to the bike lanes.

Well, it hasn’t stopped one of us.

I know what I see with my own eyes when I leave the house every day. I’m not stupid.

I would rather the bike riders get in cars. Why? Because in San Francisco they routinely ignore the bike lanes and ride the sidewalks. Anything that reduces the number of bike and scooter riders makes the city safer for pedestrians.