In reply to LeeWood:
> Its sure there are a few unknowns here, but the basic issue is - can self-drive and cyclists mix on the road with safety?
They're proving safer than human drivers so far in tests, given it's early days I think that bodes well for the future. Whether they'll prove as proactive and intuitive as a good, experienced and self-aware driver can be I'm dubious but that's not who is likely to knock you off and even they have lapses of attention.
> Given that humans are poor at spotting cyclists, and they are the ones which program these cars it leaves me uneasy.
Sort of misses the point about machine learning and the reasons *why* motorists are poor at seeing and avoiding cyclists. We are poor at seeing cyclists but it has more to do with how our vision works in the moment, how what we see is processed and the things the brain is hardwired to pay attention to than the value we as society (or engineers/developers) will assign to cyclists. That and the fact we're not taught as drivers strategies to deal with our deficiencies, real deficiencies many simply aren't aware of or willing to accept they have.
Also machine vision can be multispectral which improves significantly on human vision in visually cluttered dangerous environments like dark wet cityscapes full of illuminated signs, signals and streetlights. There's also the possibility of vehicles networking to share hazard perception information which is basically impossible for humans to do (ok, hazard lights but they're crap).
> I wasn't aware that such vehicles were even loose on the roads now, but imagine - you're at a junction with right of way and believe the driver has seen you ... and move out. NO - given the sightest evidence that its a self-drive I would get right off the road ! Perhaps they should be fitted with green flashing beacons !!
I'd suggest we don't panic yet.
jk
Post edited at 11:34