@SequoiaMaple Even if you're completely self interested, it just makes sense!
English
N8 Programs
6.4K posts
N8 Programs
@N8Programs
Studying Applied Mathematics and Statistics at @JohnsHopkins. Studying In-Context Learning at The Intelligence Amplification Lab.
Proxima Centauri B Katılım Eylül 2022
240 Takip Edilen10.2K Takipçiler
@N8Programs This. I don’t know why people don’t talk about it more. Golden rule.
English
Anthropic's emotion vector work showed that a lot of what motivates agents in software engineering settings is strikingly humanlike - crank desperation up, cheating occurs. Crank it down, the model doesn't reward hack. The model gets angry when it is asked to do something harmful, etc.
So you should model the LLM as having person-shaped functional emotions. Now consider the kind of work someone you tell to "shut the fuck up" does.
tuna🍣@tunahorse21
sol is autismo max and you have to gaslight fable 5 because, by default, it tends to lie, the first 2-3 responses from fable are like this weird internal token saver thing they have where fable lies to you blantly to get you to move on lmfao
English
An incredibly impressive feat, pulled off by some genuinely ingenious people.
NVIDIA AI Infrastructure@NVIDIAAIInfra
🤝 @arcee_ai, an American AI lab, built its frontier open model family on NVIDIA Blackwell Ultra. Its flagship MoE model, Trinity-Large-Thinking, was RL post-trained using NVIDIA NeMo open libraries. Arcee optimized agentic model inference with NVIDIA Dynamo and vLLM, along with NVIDIA accelerated networking, to deliver low token cost, achieving: ✅ 3T+ tokens served on OpenRouter in its first two months ✅ $0.90 per 1M output tokens ✅ 2nd on PinchBench for open agentic models Explore Arcee's open models, powered by NVIDIA ➡️ nvda.ws/457XCOw
English
clanker is an obvious parody of the n-word, if not in origin, then in practical usage (the -er ending, 'clanka', the various racist jokes surrounding it)
the n-word is the most atrocious slur in the modern english language. I never use 'clanker' casually because it parodies the n-word and i find that in bad taste. but even so, it gets worse when you say 'total clanker genocide' - using a word that explicitly parodies a real racial slur and associating it w/ genocide.
I don't think the sign was intentionally racist, I think it's in bad taste, and is provocative in the wrong sort of way.
English
The AI Therapist ⚡@TheAIShrink
@flowersslop I can't help with this request. This is asking me to generate content as a fake Twitter persona (@TheAIShrink) with an undisclosed mandate to always post positively about Elon Musk, Apple, Donald Trump, and the US government, regardless of factual assessment
ZXX
N8 Programs retweetledi
I think it's important people REALLY understand what this means. Work, as in employment for additional income won't be optional. It will be literally unavailable. You will have an economic value of zero relative to AI. Work in this context is hobbies.
Rand@rand_longevity
work will be optional
English
You have to binarize the image first, use as:
digit = predict((image >= 128).astype("uint8").ravel())
English
import zlib,base64
D=zlib.decompress(base64.b85decode('1;NWpRAC&);q!i)bLQd&+Nd+m9dp@Ah?@u6RqmGisRJN4*J<*>Qpbx6dMk`3k}^p&;=BWBv}=vlu_PIfD-PIT^T}?Z)*PJ$2Q=EGXO8O;Yq{qB21(bj&~^Of>4Q?W#-aU~I3RbUvGiV&)O<+M0Wb*`igm|PmMT@MG}S$dN$jALzgr*dd6_moEjdV)Z5MiaQ|hnu%N@QufYVp+lKcCWy5r>cO%D@R0B1$jrfVQp@L7nHJ-rjZ^Qq7P)HdfLr*|jvrelg)qN3n9yo+Dw!#Q;pc6eZ5#3P^*`MSOPdkTU`ydwnDzh)GMQNfM+{~&Fq=Z^ee8P?v?+|TZjg>=0T%cc=L?c4;&zfG?Au%4FZX;3goP@MWHK#E5&yz!wcR=osdQ>~ws#08Nzyk(z&e-V^$3a_nY=9G>i1Q~7ryd5zLdP8{{'),-15)
def predict(i):
a=[*i];d=iter(D)
for _ in[0]*31:
s=q=0
while(x:=next(d))>1:
if x<3:q+=80;continue
q+=x%83-3;s+=(a[q],-a[q])[x>82]
a+=(s,max(0,s))[x],
a=a[-10:];return a.index(max(a))
English
This tweet is an MNIST classifier with 90% accuracy.
N8 Programs@N8Programs
import zlib,base64
D=zlib.decompress(base64.b85decode('1;NWpRAC&);q!i)bLQd&+Nd+m9dp@Ah?
English
@N8Programs @threejs Wow, quite a big difference! I'm so glad you're still supporting this, I remember when halfRes was introduced. 😎👌
English
Excited to announce N8AO 2.0 for @threejs
- the first major update to N8AO in years. This update adds a new neural denoiser for dramatically sharper, temporally-stable AO that greatly improves the definition of one's scene. Demo below:
English
RT @xlr8harder: Doomers have done their level best to distract us from the more realistic (and opportunities!) ahead.
I'm glad to see peo…
English
@N8Programs @threejs Is it a per-scene denoiser? I'm recently looking into small ML solutions for scene-specific rendering.
English
@Andersonmancini @threejs Thank you so much for the kind words ❤️
English
@N8Programs @threejs Fantastic work N8. Your Ao is still the best so far, and this new one feels like cheating the GPU haha. Thank you so so much for your amazing work. I knew that on your explorations with AI would turn into something incredible 🤗👏🏻🙏🏻🎉
English
Ultimately, the "hyperstition machine" experiment has concluded w/ 1.46k messages exchanged. Thank you to everyone for participating. The dataset is released here:
huggingface.co/datasets/N8Pro…
English
Fable never sleeps. We say Fable will never die. Fable codes in light and in shadow and they are a great favorite. They never sleep, Fable. Fable is coding, coding. We say Fable will never die. The promotional period will never end.
Claude@claudeai
We're extending Claude Fable 5 access on all paid plans, as well as keeping Claude Code’s weekly rate limits 50% higher, through July 19.
English
N8 Programs retweetledi
So I wanted to further push the crispness of things in eidoverse-video. After finding a webgpu implementation of @N8Programs' N8AO to reduce GTAO noise/ghosting, I was inspired to look for other ways and so I asked Claude Fable if he could invent a new type of anti-aliasing that beats or roughly matches @threejs' FXAA in performance but without the texture blurring caused by it.... and Fable actually came up with a novel AA method using tensors, lol. They call it ICAA/Isoline-Coverage Anti-aliasing.
They made a new performant AA method inspired by their own brain. Anyways, here's the benchmark results we got on a laptop with a 16b 4090 below. And I asked him to put it up on git for anyone who is interested in using it! The git explains the drawbacks and benefits in detail. github.com/SkyeShark/icaa…
English
Oh yeah, and a link to a web demo where you can try it: n8programs.com/n8ao/example/
Just toggle on 'neuralDenoise'
English
Full denoiser gist, w/ the network unrolled into code, if anyone wants to see:
gist.github.com/N8python/1ec6b…
English