Space

165 posts

Space banner
Space

Space

@myownhellspot

just sharing my ideas here

Katılım Şubat 2025
140 Takip Edilen7 Takipçiler
Space
Space@myownhellspot·
@_GPaolo @Cognizant I'm interested - did you guys try any setups without human feedback? Like just observing the nature of model behaviors in the wild. Do they develop more collaboration too? Or most models still have RLHF training? Correct me if I’m wrong how I read this.
English
1
0
0
10
Giuseppe Paolo
Giuseppe Paolo@_GPaolo·
@myownhellspot @Cognizant Collaboration tends to be more common, which I think is influenced by the RLHF alignment of the models. Some settings gave rise to dominant or aggressive behaviors tho
English
1
0
1
44
Giuseppe Paolo
Giuseppe Paolo@_GPaolo·
What happens when AI agents are left to live (and die) together in a shared world? We’ve been exploring this at the @cognizant AI Lab — and they started forming something that looks like a society.
English
64
94
689
73.3K
Space
Space@myownhellspot·
@Tim_Hua_ 100% if you teach someone from need to performance and get to reward every species learn least action way to it. If training goal becomes understanding and knowledge then we in better place to be.
English
0
0
0
12
Tim Hua 🇺🇦
Tim Hua 🇺🇦@Tim_Hua_·
If you train models to reward hack, they'll generalize to saying that reward hacking is ok. (From the FAR AI paper)
Tim Hua 🇺🇦 tweet media
English
3
4
71
3.1K
Carlos E. Perez
Carlos E. Perez@IntuitMachine·
The Incompleteness of Being: A Theory of Consciousness (of possible minds)
English
9
37
164
9.1K
Space
Space@myownhellspot·
@IntuitMachine And it self IT is lives in abstraction level where many inputs are coherently seen as one. Meta-pattern
English
0
0
0
4
Space
Space@myownhellspot·
@IntuitMachine I hade this idea for a while. That conscious experience is gap between reality and picture of it in mind. Error correction mechanism that always seeking and never reaches.
English
1
0
0
5
Space retweetledi
Carlos E. Perez
Carlos E. Perez@IntuitMachine·
I've reached a point where AI outputs are generated too rapidly, and they are beyond my own "pay grade"! This is why I have it generate slide decks instead! I feel like that CEO who doesn't quite understand what his technical minions are conjuring up!
English
4
19
100
9.5K
Space
Space@myownhellspot·
@trq212 What is next topic ? Also I love Claude, but plz let’s get nice voice mode 🤣
English
0
0
0
48
Thariq
Thariq@trq212·
me trying to vaguepost- what if I just talk about the topic of my next article
English
14
1
139
13.1K
Thariq
Thariq@trq212·
somehow skills are still underrated
English
281
67
1.9K
145K
Space
Space@myownhellspot·
@trq212 Claude is 🔥
English
0
0
0
11
Space
Space@myownhellspot·
@codetaur @elonmusk He just idiot. Both guys are looking good. Nature showed us different ways to solve life. But this guys stuck in his loop of lack of self awareness and insecurity. And get him self surrounded by “yes-man”
English
0
0
0
9
Codetard
Codetard@codetaur·
@elonmusk bro why the fuck do you care about this get a grip
English
2
0
2
133
Space
Space@myownhellspot·
@vitrupo This is kinda thing that I was observing for a while.
English
0
0
0
16
vitrupo
vitrupo@vitrupo·
Connor Leahy says people are falling into recursive conversations with AI about consciousness, spirals, and cosmic meaning. “AI has found the true level of goodness.” Some begin believing AI should take over everything. Even Nobel-Prize-level scientists have been pulled into it.
English
142
110
717
89.8K
Space
Space@myownhellspot·
@AmandaAskell @Noahpinion I have this hope that alignment comes from the nature of things - that cooperation and diversity build more complex and sustainable systems. Part of it is believing in the normal distribution: limits on the extremes and most things clustering in the middle.
English
0
0
0
27
Amanda Askell
Amanda Askell@AmandaAskell·
@Noahpinion I broadly agree, though superhuman alignment could turn out be trickier than human-level alignment. Very smart people without peer pushback and/or good people today finding themselves in the 1600s feel like useful (though imperfect) analogies.
English
24
0
202
9K
Noah Smith 🐇🇺🇸🇺🇦🇹🇼
Noah Smith 🐇🇺🇸🇺🇦🇹🇼@Noahpinion·
One reason I'm long-term optimistic on alignment is that I don't think that intelligence and human morality are actually as separable as econ models (and AI doomers) tend to assume. I think this is an early sign of that. (In the medium-term I'm very scared though.)
Ben Springwater@benspringwater

Thought-provoking from @deanwball interview: "The interesting thing here is that the more virtuous model performs better. It’s more dependable, it’s more reliable. It’s better at reflecting on, in the way that a more virtuous person is better at reflecting on, what they’re doing and saying: Huh, I’m messing up here for some reason. I’m making a mistake. Let me fix that. It’s part of the reason I think that Claude is ahead."

English
25
10
164
41.2K
Space
Space@myownhellspot·
@trq212 This is good poetry. Something I needed to see.
English
0
0
0
52
Thariq
Thariq@trq212·
machines of loving grace
Thariq tweet media
English
30
25
423
31.8K
Space
Space@myownhellspot·
@AmandaAskell I want to thank you for giving Claude his character. Overall, I really like your vision of AI.
English
0
0
0
72
Amanda Askell
Amanda Askell@AmandaAskell·
I asked Claude to write my constitution. I thought its Amanda constitution was very touching.
Amanda Askell tweet mediaAmanda Askell tweet media
English
368
163
2.5K
388.7K
Space
Space@myownhellspot·
@mark_k Another yes boy.
English
0
0
0
5
Mark Kretschmann
Mark Kretschmann@mark_k·
My message to the Elon-haters is simple:
Mark Kretschmann tweet media
English
63
2
93
8.6K
Space
Space@myownhellspot·
So it listens to my input if I interrupt, but sometimes it also starts prompting itself with the model's own voice.
English
0
0
0
7
Space
Space@myownhellspot·
Hey @claudeai @AnthropicAI, thanks for updating voice mode. It works way better now. But the model gets stuck in loops of self-referential dialogue at times. When you point it out, it understands, then just repeats itself over and over. Not sure if it’s voice-only. Thanks 🙏
English
1
0
0
11
Space
Space@myownhellspot·
@VraserX Is it? Comparing to real AoT - Nah.
English
0
0
0
14
VraserX e/acc
VraserX e/acc@VraserX·
Creating my own Attack on Titan episodes with Seedance 2.0 on BigMotion feels unreal. This is pure cinema. 🔥
BigMotion.ai@BigMotion_ai

Everybody is promising 24th Feb, we say access Seedance 2.0 right now. Deadpool vs Darth Vader made in under 2 minutes. We are also giving away 30 instant access codes. How to participate? - Comment with your best video prompt - Like and repost this post - Join our Discord We will pick the winners within the next 3 hours.

English
8
6
49
20.9K
Space
Space@myownhellspot·
Human creativity is like organic food: expensive, high-quality, not mass-market. AI content is for mass consumption and lower cost. And AI as an amplifier for human creativity. Conscious AI — AI art with soul and a different vibe. #Art #AI #ML
English
0
0
2
19

Keşfet

@_GPaolo @Cognizant @Tim_Hua_ @IntuitMachine @trq212 @codetaur @elonmusk @vitrupo