a computer shouldn't be allowed to fuck around because a computer can never find out
- they/any
- significantb.it/
working on a @styx-os.
 | ||
"accidentally-vengeful telco nerd"
—Tom Scott
platform sec researcher, OS dev, systems architect, composer; Other (please specify). vintage computer/electronics nut.
I am open to tag suggestions - if there is something you want me to tag on my posts, leave a comment. <3
take a look at
this cool bug I found 🪲
discord
@sirocyl
mastodon/whatever
social.treehouse.systems/@sirocylrevolt.chat (occasionally active)
@sirocyl#5128
new blog:
significantb.it/styx linux OS project
styx-os.org/You must log in to comment.
in reply to @tsiro's post:
I mean this is that IBM presentation slide from 1979, right? I think you're in position with a long history of being correct, even if it's not necessarily been brought to bear in the world.
I don't know. Maybe I'm misunderstanding you here, but this sounds to me like you would rather have someone to hold accountable for death or injury then prevent said death or injury in the first place. That seems entirely backwards to me.
That doesn't mean we shouldn't have a debate on what accountability means in the presence of automated processes, but I'm not willing to delay tangible safety benefits for that.
When dealing with corporations, the threat of accountability (the OSHA inspector comes in and shuts your operation down, you are held liable for negligence, etc) is unfortunately required to have safety improvements in the large majority of cases.
Sure, but the original comment presupposed that we do have safety improvements without accountability.
Not quite. Why do we trust the unaccountable entity's safety data?...
Arguably I (and judging by OPs other comments, OP as well) was thinking of stuff like self-driving cars, where accident statistics are RIGHT THERE. But again this is less a question of "who is accountable for this nominally safe system" and more one of "okay but is it actually safe". Which is a fair question to ask! But not the question OP asked.
Contrary to intuition, the statistic of "accidents within the set of all self-driving cars" doesn't say anything as meaningful as its contribution to safety, even directly compared to "accidents within all human-driven cars".
There is a fallacy here that I don't know the name of, but basically, what should be asked is instead "have the introduction of self-driving cars positively correlated to reduced accidents?" Just saying that there are less accidents within a given subset doesn't actually account for its impact to the whole.
There's a number of ways that it could be true but actually mean it contributed more to accidents - for example, people using them when they normally wouldn't have driven at all. But these kinds of things are never reported in parallel, which is what makes using that isolated statistic a useful smokescreen - and a useless argument in the face of tangible ethical concerns.
in reply to @jckarter's post:
thanks! i gotta hand most of the credit to @bethposting for the tee up tho
Pinned Tags