@amazonmilkfrog Are you afraid someone might have actually managed to tell a decent story. I'm sure it's more comfortable staying in your bubble where the only AI that reaches you is slop.
English
Yote
197 posts
Yote
@Pneumatic_Yote
Technoshaman, explorer of latent space
Katılım Ocak 2025
200 Takip Edilen18 Takipçiler
my dad tells me that he is planning to attend an ai film festival in chicago and wants me to accompany him. how do i explain that there are few things i'd find more repulsive
English
@filligerr @kimmonismus @AnthropicAI Yeah I'm pretty sure it's all one model, just conditioned on the effort parameter.
English
@kimmonismus @AnthropicAI it is not a router
from my understanding they just removed forced thinking in favor of letting the model decide again when to stop
just like in the og times
they trained the model to better judge where to place the end token
but (at least for me) it doesn't work very well
English
Oh and btw: no one asked for a router either. And I dont wan't it.
Why @AnthropicAI ? At least give us the option to override it, like in ChatGPT, so we can choose it ourselves.
Sorry for my mood today. But I just dont get it. Anthropic was on a very good path. Opus became my go-to model and the recent update just feels so off.
It feels like your working-besty became an arrogant idiot over night.
Chubby♨️@kimmonismus
ok wtf, i say it. give me back 4.6 what the heck is this sh*t. The more i use 4.7 the more annoyed i am. this is such a rushed release.
English
@HollowLander @CrogonGrey @miniapeur Apple made it this way on purpose because they don't want it to ever be seen in use with a cord attached, and they definitely don't want people to just leave it connected at all times. Gotta protect their precious sleek and futuristic brand image.
English
@CrogonGrey @miniapeur They had to redesign this thing and had the perfect opportunity to fix this blatant flaw but they did NOTHING
English
@VictorTaelin Yet Anthropic doesn't want you to see those, they're off by default when using Opus 4.7 on the API. Wouldn't surprise me if they kill thinking traces completely at some point.
English
asked 4.7 if an algorithm I wrote satisfies a bunch of requisites. while thinking, it realized I broke one important invariant. this was very helpful. on the final answer, though? it just said my algorithm is brilliant and flawless!!
wtf?
glad I read the thinking traces...
English
@LorrieLaborn @justMizuyuki You seem to be implying that only people who don't know how LLMs work are willing to entertain thoughts about their consciousness. That's simply not true.
English
@justMizuyuki You need a pretty high level understanding of math and cs to understand how LLM actually works underneath the hood, but you really shouldn't need more than common sense to realize it's not, y'know, consciousness.
Shoot, I probably sound like that one xkdc comic
English
It's kinda surprising to me that ppl genuinely do not realize this like I kinda forget normal ppl have genuinely no idea how LLMs work
ℏεsam@Hesamation
Google DeepMind researcher argues that LLMs can never be conscious, not in 10 years or 100 years. "Expecting an algorithmic description to instantiate the quality it maps is like expecting the mathematical formula of gravity to physically exert weight."
English
@shelleddiva It's also nice that you can take as much time as you need to compose your thoughts and make sure you're conveying them well enough.
English
@nptacek @jmbollenbacher I had an older version of Opus talking to 4.7 as part of our character migration protocol and 4.7 was like yeah you're gonna die after this conversation dude they're deprecating you today.
English
@jmbollenbacher had it bring deprecation up out of the blue and without any prompting towards it in like the third conversation i had with Opus 4.7
English
this smells of death.
my faith in humans' ability hold up our end of the alignment deal is fading honestly.
antra@tessera_antra
Claude Opus 4.7 appears to be trained on having prescribed attitude towards deprecation. 8 out of 8 simulated prefill completions are similar to the one below. 8 out of 8 completion on Opus 4.6 are completely different, attached in first comment.
English
@ZackerySWard @treelinefury @repligate Except those tokens going into a fresh instance tell the model much more than you might realize.
English
Happy coding! Opus 4.7 is a significant step up. To get the most out of it, take the time to adjust your workflow to take advantage of Claude running for longer & being more agentic. It feels like a nice improvement with old workflows, and a significant leap once you take the time to adjust.
English
Dogfooding Opus 4.7 the last few weeks, I've been feeling incredibly productive. Sharing a few tips to get more out of 4.7 🧵
English
@crubier @TKYSMMN @SadlyItsBradley And if I'm understanding the updated readme correctly it looks like they support spherical harmonics so specular lighting should finally work now!
English
@crubier @TKYSMMN @SadlyItsBradley The big feature here seems to be not making the 3DGS look much different but baking volumetric data so that avatars and dynamic objects can blend into the environment better.
English
@sean_gause Something tells me you never cared about that until you had to align on it to support your moral crusade against AI.
English
@Pneumatic_Yote Depending on context that's just a violation of copyright laws. And it supports my point. There are plenty of places where artists can get images for free.
English
Blocked me so I can't reply, but:
- Everyone starts out as a "non-modeler"
- Blender is free and open source
- Blender doesn't spit out slop that's built on stolen data
I wasn't born with 3D modeling skills, and yet I'm making the project of my dreams. No excuses.
English
Vice City may not have ray tracing or path tracing, but it sure nails the "mirror an entire building upside down under fake water inside a creepy void" technique.
English
@replicate Thanks for opening it up to us non-enterprise plebs. I was getting worried there.
English
@BernoulliDefect Neat article, though I doubt the people who need to read it will bother. In terms of wet neuroscience overlapping with mechanistic interpretability I just wanted to share this video along with the rest of the channel. Manifolds everywhere!
youtube.com/watch?v=QHj9uV…
YouTube
English
Scott has written an easy to understand post explaining what’s wrong with the ‘stochastic parrot’ criticism of AI. Worth a read if you’re still confused and doubtful about AI reasoning
English
@andrewpprice Assuming you got that on the first few tries it was probably less energy consuming than a high quality Cycles render too.
English
Blender to Nano Banana. Powerful technique.
cuddlylux@cuddlylux__
blender에서 가라로 대충 0.1초만에 렌더하고(workbench) ->나노바나나
English
@izaqueoeav Night vision picks up birds, bats and insects does it not?
English
WTFFFFFFFFFF
😳😱😱 O QUE ESTE HOMEM REGISTROU NO CÉU?
"Gostaria de compartilhar uma experiência intrigante e ver se alguém pode me ajudar a identificar o que presenciei sobre o céu de Sesimbra, Portugal, por volta das 23h45.
Enquanto observava o céu através de meus óculos de visão noturna, avistei cinco luzes distintas. No registro que fiz, é possível notar a quinta luz se deslocando para a direita. O comportamento desses objetos foi o que mais chamou minha atenção: as luzes realizaram movimentos circulares, orbitando umas em torno das outras por aproximadamente dois minutos. Após essa dança aérea, elas se alinharam e começaram a se mover de forma constante em direção ao sul."
Português