However, warns Norton, “if the tech turns out to actually make cyclist cycling safer for those who have it, but more dangerous for those who don’t, does that become grounds in policy for requiring all cyclists to have the necessary equipment for cars to detect them? If that does, then we now have problems about access to cycling among those with budgets, or deterring cycling in a society where we need more, not less for lots of reasons, including sustainability and public health.”
Norton added: “We are not protecting these unequipped cyclists when we have equipped cyclists, and we are to some degree making their situation more serious as drivers come to expect cyclists to be equipped. And eventually, even road designers and road authorities will start to assume that cyclists should be equipped. Perhaps even the law may begin to expect this.”
This looks like it's going to be jaywalking 2.0. Oh, you got squashed by a car whose driver was busy on their phone? Your fault for not wearing your bEaCoN sweaty
"oh, we looked in side our special encrypted black-box that only we can look in, and it says you, the cyclist were 100% responsible for this accident. We all know computers are made of math - and therefore always objectively correct. Sorry, but we can't be liable for that crushed spine!"
AI training is basically the mathematical version of ideological programming. It only sees what you feed it, so it naturally adopts the ideology of whoever trains it.
Like how all those vision AIs struggled to see people with dark skin because the creators used a database of white people.