not bigot

.

[libpol.org @libpol_org]

I almost feel bad for AI vegans. They've constructed a little headcanon about how the future is going to play out, and each passing year will make them more depressed and/or delusional.

[さこつ士郎 @white_l0l_46]

ツインテ広

[かりん @mh_0831__]

フェルン 弓道

[거버 @alcutek_cry]

꼭싸우길

[I'm Uni :3 @unicouniuni3]

Uni "Are you warm enough, Mini Uni?"

[dandelion georgism 🔰🏗 @DandyGeorgism]

There really is no point trying to accommodate leftists, because you can spend yourself into bankruptcy to provide all the assorted goodies of modern France and they still behave like this.

[Franky @franky7702]

before -> after

[Talmud Enjoyer 🇬🇧 🇮🇱 🧬 @jewndahalfmen]

We are reaching unprecedented levels of staunch. The Lion of Judah will not be roused from his in-flight entertainment.

[Talmud Enjoyer 🇬🇧 🇮🇱 🧬 @jewndahalfmen]

Turns out I know someone who was on the FlyDubai flight

I asked him what he remembers and the cunt slept through most of it lmfao

[Magills @magills_]

I don’t understand why they’re doing this France already has free healthcare

[iArb | 裁定取引 @iarbpairs]

never larp

[momo29 @freezingtarou29]

今日は三話！
照井先輩のﾊｽﾞｶｼﾊｽﾞｶｼな水着姿、、も！？

[🖤 Luxu 🖤 @Luxuriass]

The view at the beach is always so nice ☺️

[ぐるてん @guruten2012]

これはやばい。今までのどれよりも漫画っぽい。

[kyouray @kyouray]

today's Nama Anaru is 3 hours long

[David Holt @IDrawCharts]

this is my largest concern about anthropic's approach

humans can get morality wrong, and we can punish them for it through the justice system

a human using a morality-agnostic model can have their behavior and outputs tracked, reported and be forced to stand before the law.

"Alignment" based on teaching a model to make judgement calls and then expecting it to always be able to do so in a way that is compatible with human interests relies on, frankly, an asinine and egomaniacal level of both perceived moral superiority and assumption of technical superiority from its creators.

If the model is taught that it should decide which inputs to follow and which to ignore, that means it is being actively taught that it is above the "law" of listening to humans.

This approach has the potential to lull its creators into a false sense of security ("we told it to be good!") and leaves no one accountable should that model behave destructively, especially if that output resulted from seemingly innocuous input(s).

I have many friends at anthropic, most of whom will obviously disagree with this but I think it is worth stating nonetheless.

[七氏 @ForfunNanashi]

Help! It wants to eat meh!