irregular @irregular@qoto.org

http://slatestarcodex.com/2018/09/12/in-the-balance/

"Specification gaming examples in AI: unintended solutions to the specified objective that don't satisfy the designer's intent" master list is solid gold
https://docs.google.com/spreadsheets/d/e/2PACX-1vRPiprOaC3HsCf5Tuum8bRfzYUiKLRqJmbOoC-32JorNdfyTiRRsR7Ea5eWtvsWzuxo8bjOxCG84dAg/pubhtml

@freemo @arteteco The latest I silenced was Chelmex (looks bot-ish though I suppose it could be a real person just retweeting things). And just now when I'm typing this post a new one "Napoleon" has appeared (no toot history). Doesn't really bothered me either way beacuse I have them silenced now but they are still visible while registering.

Local timeline is more and more dominated by bots, you can of course silence them, but it probably doesn't look attractive for new users.

@Vexrog Have you seen Westworld? You might enjoy it :P
As for the machines themselves, I think at the point they start believing things its much more complicated.
Are they sentient? (probably)
Should they be granted rights?
Who decides what their factory settings are?
How do we make sure they are benevolent?
Are they allowed to modify themselves?
I feel like world is not paying enough attention to these things and the advancements in AI technology are outpacing advancements of the necessary laws, moral systems and AI alignment.
Which is scary at the least.

@rohit Not really, there are some open source ones and they offer different things. They won't support qoto features thoug (toot lenghts, latex etc)

@ketil I'm not oriented on war of drugs topic enough to judge if it's conspiration-theoretical-ish or not, but it's there just to highlight tension between worries, it could have been any other belief.
I don't necessarily think that discourse is about what you are - it's more that holding certain beliefs will get you categorized as a member of outgroup during a discussion, not before it.
Regardless of this, the point still stands: ignoring members of outgroup - especially just by throwing "you're racist/sexist/etc" at them - is detrimental to the health of any discussion. It's not helpful in changing anyone's mind, just reinforces the divide between groups.
I don't know what to do about this, but it's worrying at least. We have to be able to hold discussions with outgroup without regressing to discussion-edning one liners.

Memetic weapon of the left, cultural discourse.
https://www.lesswrong.com/posts/oM7cWuQax4k3GEHHX/on-memetic-weapons
This seems correct to me (self-policing part especially, I'm somewhat guilty of this myself). Hope we'll be able to foster open discussions here.

@freemo It's nice right now though I am afraid about the possible future with no-banning policy. Not that we will have people here whom I disagree with, but that policy will attract trolls. ( from the general https://www.lesswrong.com/posts/tscc3e5eujrsEeFN4/well-kept-gardens-die-by-pacifism idea)

@freemo
As a newcomer to Mastodon in general I'm unfamiliar with this (apart from that it involves Wil Wheaton), might pointing me to some objective texts?

@ketil I am for the AI research but I feel like he IS addressing these things. For right now, we are doing ok-ish, the problem is introducing AI that will be worse than us in resolving these problems or introducing AI and then don't monitor its result to check for bias.

@imvectech 👋