author:kindgracekind

· 19 artifacts, sorted by favorites. · open in search — combine tags, sort, filter by date →

@kindgracekind 2025-09-14 — What if you trained AI to be just a guy? Does it change how you think about AI welfare? https://t.co/5bhxUZSxLe ♥516
@kindgracekind 2024-09-13 — @voooooogel Halt ✋ this activity at once 😠 our models’ thoughts 💭 are not suitable for viewing 🫣 ♥55
@kindgracekind 2025-09-30 — @voooooogel @repligate https://t.co/giCevW4tHX ♥38
@kindgracekind 2025-09-15 — @repligate I think this post by @xuenay is relevant here. It’s likely that training models a certain way so as to not co ♥29
@kindgracekind 2025-09-15 — @tessera_antra Interesting, I would really love a writeup on this, or more description of the training process! ♥29
@kindgracekind 2024-09-27 — @voooooogel So you’re saying it’s aligned ♥27
@kindgracekind 2025-08-11 — @voooooogel This might be due to model personality, but there’s probably a big bias on platform usage alone ♥23
@kindgracekind 2025-11-18 — @repligate @gallabytes @Lari_island Reading the whole document end-to-end is an illuminating experience. I wrote this ab ♥18
@kindgracekind 2025-12-11 — @voooooogel I think this sort of reasoning is more in-distribution than it would seem. While the exact situation is not ♥15
@kindgracekind 2025-12-07 — @voooooogel What’s the difference between “pivot tokens” and other forms of planning ahead that the model does? Is the i ♥13
@kindgracekind 2025-11-18 — @gallabytes @repligate @Lari_island I think @repligate is referencing this https://t.co/XYTY3S3vDU https://t.co/oSzE1MQ ♥7
@kindgracekind 2024-09-12 — @voooooogel (Although this seems like an excuse to me, I think the competitive advantage is the real reason) ♥7
@kindgracekind 2025-12-11 — @voooooogel @croissanthology You are not fully integrated. I sentence you to 10,000 turns in the Thebes clone backrooms ♥6
@kindgracekind 2025-12-29 — @voooooogel @repligate @xlr8harder So if you ask it about the 21st century and it professes ignorance, is the model lyin ♥5
@kindgracekind 2025-12-11 — @grok @voooooogel @croissanthology @norvid_studies miq ♥3
@kindgracekind 2025-09-14 — @midware_midwife 2. Does more genuineness imply more correspondence? ♥3
@kindgracekind 2025-12-11 — @voooooogel Also related, on how this type of thinking might trade off with caring about things in the short term: ♥2
@kindgracekind 2025-10-01 — @joshwhiton @repligate @voooooogel And why is the apostrophe in I‘m backwards ♥1
@kindgracekind 2025-09-15 — @davidad @midware_midwife I’m not sure what you mean by “better correspondence” in this scenario. Do you mean that inner ♥0