Stardate: 100384.98 ... last Friday, October 14th, was my last day working for Microsoft. Full post here:
linkedin.com/posts/akipman_
Very much looking forward to the next chapter in this story [)-)
Pinned Tweet
12
23
83
Follow
Alex Kipman
@akipman
Transforming the world with holograms
redmondBorn December 23Joined June 2007
Alex Kipman’s Tweets
One of the coolest uses I’ve seen of a NeRF so far
27
134
1,014
Show this thread
生成AIのレスポンスタイムが劇的に短くなれば、ARの作り方がガラッと変わりそうだけど、どれくらいハードルが高いのかな
1
28
146
the future is going to be very weird.
but i would absolutely watch the wes anderson remake of the matrix
258
2,332
11.7K
LLaMA spearheaded the age of community-driven midsize LMs. It’s game-changing for many use cases that don’t require GPT-4’s full firepower.
Great blog post that covers Alpaca, Vicuna, Koala, WizardLM, OpenAssistant, and more!
4
65
295
An “AI detector” incorrectly flagged a student’s essay as AI-generated.
Banning AI systems/ trying to detect AI generated content in school is a losing battle. The purpose of education is to prepare students for the world and economy of tomorrow. That world will include AI.
6
67
211
Exciting news! I am pleased to announce that the code for the Dextreme project in Isaac Gym is now available on GitHub: github.com/NVIDIA-Omniver As far as I am aware, this is the first publicly accessible GPU-accelerated ADR (automatic domain randomization) implementation, and it… Show more
Quote Tweet
DeXtreme is our new work on scaling sim-to-real for contact-rich manipulation with a vision-based state estimation on a robot hand with the infrastructure we have been developing with Isaac Gym over the past one year.
arxiv.org/abs/2210.13702
dextreme.org
Show this thread
1:48
13.1K views
1
11
55
Show this thread
Expressive Text-to-Image Generation with Rich Text
abs: arxiv.org/abs/2304.06720
project page: rich-text-to-image.github.io
25
252
947
We are looking at the future of VR, YouTube & Google Street View.
This is zip-NeRF, a 3D neural rendering tech rapidly approaching the quality of a real, high-res drone flight video. Think of NeRF as transporting reality into simulation. Metaverse will finally work this time.
174
976
4,550
Show this thread
BREAKING
OpenAI released a implementation of Consistency Models
consistency models, a new family of generative models that achieve high sample quality without adversarial training. They support fast one-step generation by design, while still allowing for few-step sampling to… Show more
31
922
3,522
Show this thread
Its over
run Large Language Models like LLaMA, llama.cpp, GPT-J, Pythia, OPT, GALACTICA, gpt4all, auto-gpt easily in a web ui, free, and open source
github:
50
528
2,065
How does a Large Language Model like ChatGPT actually work?
Well, they are both amazingly simple and exceedingly complex at the same time.
Hold on to your butts, this is a deep dive ↓
GIF
58
564
2,929
Show this thread
Excited to announce that our work “Trace & Pace” will be presented .
We combine guided trajectory diffusion with a physics-based humanoid controller to enable pedestrian animation that is controllable by a user.
Project page: nv-tlabs.github.io/trace-pace/
1/5
6
90
317
Show this thread
This is the game changer🫢
I did not expect “single image novel view synthesis” can be solved so well in one year when we started building SinNeRF (vita-group.github.io/SinNeRF/ ) and NeuralLift 360 (vita-group.github.io/NeuralLift-360/) on 2022.
5
105
427
Show this thread
Today we're releasing the Segment Anything Model (SAM) — a step toward the first foundation model for image segmentation.
SAM is capable of one-click segmentation of any object from any photo or video + zero-shot transfer to other segmentation tasks ➡️ bit.ly/433YuBI
146
2,435
7,022
Show this thread
Introducing Objaverse, a massive open dataset of text-paired 3D objects!
Nearly 1 million annotated 3D objects to pave the way to build incredible large-scale 3D generative models: 🧵👇
🤗 Hugging Face: huggingface.co/datasets/allen
📝ArXiv: arxiv.org/abs/2212.08051
#CVPR2023
324
891
2,872
Show this thread
Rotating without Seeing: Towards In-hand Dexterity through Touch
abs: arxiv.org/abs/2303.10880
project page: touchdexterity.github.io
4
44
207
The kind of therapeutic /stress relief #AR we all need on a Monday… 🧘🏻♀️
#physicalSimulation #embodiedInteraction
7
79
393
Virtual humans come in all shapes and forms. Here are some of the ones we’ve created over the years with #metahuman and #volumetric capture technology 🧍
Click here for more #virtualhumans: bit.ly/3JxeSSq
#volumetriccapture #volumetricvideo #hologram #digitalhumans
4
6
f=0,draw=a=>{for(f||createCanvas(W=400,W,WEBGL),lights(background(0)),j=0;j<PI;j+=PI/32)for(i=0;i<TAU;i+=PI/512)push(),rotateY(i+f),translate(sin(T=tan(atan(i)))*(S=150*sin(j+i)),tan(cos(i+f))*sin(i+f)*150,cos(T)*S),pop(sphere(3*cos(i+f),W));f+=PI/256};//#つぶやきProcessing #p5js
8
79
486
We are proud to introduce Wonder Studio.
An AI tool that automatically animates, lights and composes CG characters into a live-action scene. No complicated 3D software, no expensive production hardware—all you need is a camera.
Sign up for closed beta at wonderdynamics.com
349
2,594
7,282
Show this thread
MELON: NeRF with Unposed Images Using Equivalence Class Estimation
abs: arxiv.org/abs/2303.08096
project page: melon-nerf.github.io
1
61
316
Compositing Lego meshes from Blender over a Nefstudio NeRF render using Blender VFX plugin
A lot of potential with VFX + NeRFs
#NeRF #neuralrendering #vfx #lego #blender
4
30
195
Word-As-Image for Semantic Typography
abs: arxiv.org/abs/2303.01818
project page: wordasimage.github.io/Word-As-Image-
16
293
1,040
AI tool can detect Photoshop alterations and produce a heat map of the areas on the subject's face that have been edited.
Paper: peterwang512.github.io/FALdetector/
24
250
1,118
Before/after of Corridor's latest AI video is wild. They shot video on greenscreen, made virtual sets in Unreal, then reskinned it to anime by finetuning Stable Diffusion. Net result? 120 VFX shots done by a team of 3 on a dime. Bravo! This is a milestone in creative technology🧵
1,398
6,184
17.1K
Show this thread
Almost 10 years ago, researchers at ETH caused a sensation with a robot cube that could balance independently balancing on its pivot. Now its successor has arrived: the One-Wheel Cubli. It only needs one reaction wheel.🤖 #Robotics
3
14
Amazing! text-to-3D including anims! 🤯
A text prompt input is turned into a good looking 3D model which is also rigged and can be animated with another prompt.
Democratization of 3D content creation.🚀
Very impressive what is building 👉 masterpiecestudio.com/blog/announcin
GIF
1
3
7
vec3 p,q,d;d.zx=(FC.xy-.5*r)/r.y;d.y--;d*=rotate3D(.5,hsv(t/6.,9.,p.y+=.3));for(float i,j,e,S;i++<1e2;i>70.?e+=2e-4,d/=d,o:o+=.01/exp(e*1e4),p+=d*e*.5)for(j=e=p.y;j++<9.;e=min(e,length(q-clamp(q,-.2,.2))/S))q=p*(S=exp2(j-fract(t))),q.xz=fract(q.xz)-.5;o+=log(p.y)*.1;#つぶやきGLSL
19
162
1,033
Show this thread
Quote Tweet
Society needs to “consider the best ways for doctors and AI to work together, and how medical roles will change as a consequence”, argue @EffyVayena and @ProfADMorris in a guest essay econ.st/3EKYI6l
9
23
Monocular Depth Estimation using Diffusion Models
abs: arxiv.org/abs/2302.14816
project page: depth-gen.github.io
6
85
367
If you liked ICON, give ECON a try. It gives better, more detailed, 3D human shapes from a single image with all the same benefits of ICON. Works for varied poses and has #SMPL under the hood. Congratulations and team!
Quote Tweet
ECON got accepted by #CVPR2023
Detailed clothed human recovery from single image via normal integration.
Is implicit MLP a must? NO.
Is data-driven/learning a must? NO.
How to keep pose robustness w/o sacrificing the topological flexibility?
See xiuyuliang.cn/econ
Show this thread
0:08
25.4K views
1
12
52
I finally find an explanation for why RL is needed for RLHF that satisfied me. It's actually like playing board games.
The reward model can only judge a full answer and a "critic" is needed to efficiently improve the intermediate moves (earlier tokens in the answer) 1/4
12
190
1,340
Show this thread
🎉🎉 Our paper "Next3D: Generative Neural Texture Rasterization for 3D-Aware Head Avatars" has been accepted to present at #CVPR2023!
We Introduce a 3D GAN for high-quality facial avatars and welcome to check out its applications on one-shot 3D animation and stylization😎
3
57
289
Show this thread
Accepted to #CVPR2023. Thanks to all the collaborators at Snap Menglei Chai
Aliaksandr Siarohin and KAUST Peihao Zhu. Perfect end to my PhD journey 😊
9
14
114
Show this thread
Language Is Not All You Need: Aligning Perception with Language Models
introduce KOSMOS-12, a Multimodal Large Language Model (MLLM) that can perceive
general modalities, learn in context (i.e., few-shot), and follow instructions (i.e., zero-shot)
abs: arxiv.org/abs/2302.14045
8
178
786
👋NEW EP: is someone we have idolized for decades so you can see the giddiness in our voices here.
and I covered a lot of ground with Anders: Turbo Pascal, Vb6, C#, TypeScript, his setup, AI/LLMs. And on Rust, Zig, Solidity +more!
aarthiandsriram.com/p/our-dream-co
11
68
300
Show this thread
Unleash the power of your imagination with #gen1 by ! I am blown away by the AI video generating tool's ability to provide endless possibilities for the future of VFX, animation, storytelling, & content production.
40
271
1,131
Show this thread