Vincent Boucher ๐Ÿ’Ž

All Photos Twitter.com 7 hours ago
THE PUREST ARCHETYPE OF CUBISM ART Historical Significance: In 2021, MONTREAL.AI MASTER succeeded in unveiling the purest archetype of the Cubism Art Movement - opensea.io/assets/0x495f9โ€ฆ #AIArt #CryptoArt #MontrealAI #NFT
Retweeted by Vincent Boucher ๐Ÿ’Ž
7h
Crypto AI Art #514 HISTORICAL SIGNIFICANCE: The "Becoming" Collection is opening the doors to the #CryptoAIArt Movement. Listing: opensea.io/assets/0x495f9โ€ฆ MONTREAL.AI invented the term "Crypto AI Art". #AIArt #CryptoArt #MontrealAI #NFT
Retweeted by Vincent Boucher ๐Ÿ’Ž
8h
delist everything. we havent seen anything yet!
Retweeted by Vincent Boucher ๐Ÿ’Ž
8h
Our new work, StyleAlign, is on arXiv! arxiv.org/abs/2110.11323 Just finished fine-tuning a new StyleGAN model? Great, but don't drop the parent model just yet! You can do a lot of cool stuff if you just use the two models together! Morphing, Img2img, zero-shot regression & more
Retweeted by Vincent Boucher ๐Ÿ’Ž
10h
Large language models (LLMs) often make mistakes that are difficult to correct. We study the problem of quickly editing these models: Paper: arxiv.org/abs/2110.11309 Code: github.com/eric-mitchell/โ€ฆ w/ @_eric_mitchell_, C. Lin, @ABosselut, @chrmanning thread ๐Ÿงต๏ฟฝCl
Retweeted by Vincent Boucher ๐Ÿ’Ž
10h
Being 100% engaged with this space during the next 90 days will net you more gains than most people make in 10 years.
Retweeted by Vincent Boucher ๐Ÿ’Ž
12h
A PyTorch re-implementation is now available - see github.com/rail-berkeley/โ€ฆ Use IQL for all your offline RL and online finetuning needs! twitter.com/ikostrikov/staโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
12h
Here's a very useful blog post by @RishiBommasani and @percyliang at @StanfordHAI about Foundation Models. I found the section clearing up the misconceptions about name itself ("Foundation Models") to be especially useful. #FoundationModels #GPT3 crfm.stanford.edu/2021/10/18/refโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
12h
Hyperinflation is going to change everything. Itโ€™s happening.
Retweeted by Vincent Boucher ๐Ÿ’Ž
12h
Google has a significant fraction of the world's top AI talent, and yet Gmail has recently been marking as spam nearly every email from the undergraduates in my ML course. It sometimes even spam filters emails from my grad students or replies to messages I sent.
Retweeted by Vincent Boucher ๐Ÿ’Ž
13h
New NeurIPS paper: We train a Robust CLIP encoder that produces approximate CLIP representations by seeing highly corrupted images. We can classify images by observing 2% random pixels or very blurry images better than humans. sriramravula.com/publications/iโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
14h
YOLOR-P6 55.4% AP and Scaled-YOLOv4-P6 54.5% AP are still the most accurate Real-time (>=30FPS) neural networks, even 1 year after the release of Scaled-YOLOv4! Better than PP-YOLOv2, YOLOv5, YOLOX... YOLOR: arxiv.org/abs/2105.04206 Scaled-YOLOv4 (CVPR21): openaccess.thecvf.com/content/CVPR20โ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
14h
Excited to announce our upcoming NeurIPS paper on unsup. skill discovery (w/ @syhw, @arkrause, Nicolas Usunier)! It's time to move beyond navigation in continuous control, so here's a new benchmark and an effective hierarchical skill discovery method: facebookresearch.github.io/hsd3 1/4
Retweeted by Vincent Boucher ๐Ÿ’Ž
15h
Our paper on equivariant flows for sampling from quantum field lattice theories got accepted to the NeurIPS ML4Physical Sciences workshop! Neural nets can now sample from much larger lattices. With Corrado Rainone, Miranda Cheng and Roberto Bondesan. (1/4) arxiv.org/abs/2110.02673
Retweeted by Vincent Boucher ๐Ÿ’Ž
17h
At @BigscienceW 104B GPT training we finally had a BREAKTHROUGH and the training doesn't diverge! The key change of Expirement 11 was to change --init-method-std 0.006 from 0.02 - Thank you, BS Team! Details are in: github.com/bigscience-worโ€ฆ TB is here: huggingface.co/bigscience/tr8โ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
17h
I wanted to promote the Chien-Shiung Wu Prize Postdoctoral Fellowship now available at Kavli IPMU. ipmu.jp/en/job-opportuโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
17h
Whatโ€™s your favorite NFT project? ๐Ÿ‘€
Retweeted by Vincent Boucher ๐Ÿ’Ž
18h
Rejuve Network, incubating a few years within @singularity_net , announces a new phase of its growth path !! ... Decentralized AI toward the end of widespread human superlongevity... few things are more important... twitter.com/Rejuve_io/statโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
18h
@farokh Im in Montrรฉal lets have a burritos and talk nft ๐Ÿคฃ๐ŸŒฏ๐Ÿ™ we are working on a really nice project . Have a nice day ๐Ÿ™
Retweeted by Vincent Boucher ๐Ÿ’Ž
21h
Thanks to @Alliance_Can for giving our CEO @martinbouch the opportunity to deliver a keynote at the prestigious Digital Research Alliance of Canada Industry Roundtable. Itโ€™s an honor to follow leaders from companies such as Intel, HPE, NVIDIA and AMD. #AllianceCanada #NDRIO
Retweeted by Vincent Boucher ๐Ÿ’Ž
21h
#SpeechBrain wants speech separation to work well in real-life conditions. With that in mind, we release #REALM, an open dataset of real mixtures. We also studied a #neuralnetwork to blindly assess the separation performance. Paper: arxiv.org/abs/2110.10812 Credits:@CemSubakan
Retweeted by Vincent Boucher ๐Ÿ’Ž
21h
Quantum computing has hit radar of technical leaders, because of efficiency it offers at scale. It will take years to develop for most applications, however, even as it makes limited progress in near term in specialized fields of cryptography. Read more๐Ÿ‘‡๏ฟฝventurebeat.com/2021/10/21/garโ€ฆ0w
Retweeted by Vincent Boucher ๐Ÿ’Ž
MICo: Improved representations via sampling-based state similarity for MDPs Our #NeurIPS2021 paper introduces a new loss that improves your RL agents! ๐Ÿ“œPaper:arxiv.org/abs/2106.08229a ๐Ÿ’ปBlogpsc-g.github.io/posts/researchโ€ฆpF ๐ŸCodgithub.com/google-researcโ€ฆAhh 1kKyG
Retweeted by Vincent Boucher ๐Ÿ’Ž
(1/7) As I've been working on PAC-Bayes for a few years, often, students ask me what is an easy reference to begin with. My answer: "I know excellent and accurate books and papers on PAC-Bayes, but they're not easy". So I decided to write this tutorial. arxiv.org/abs/2110.11216
Retweeted by Vincent Boucher ๐Ÿ’Ž
Your direction is more important than your speed.
Retweeted by Vincent Boucher ๐Ÿ’Ž
"peer review is the backbone of science" I can't live without a spine, but science can definitively live and prosper without peer-review through arXiv. twitter.com/CSProfKGD/statโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
AI DEBATE 3: "The Legendary Debate" 10 , 3 1 8 Attendees! 63 Days until #AIDebate3 !!! RSVP Info: linkedin.com/events/aidebatโ€ฆ It All Comes Down To AI DEBATE 3 By: MONTREAL.AI #AIDebate #AIDebate3 #MontrealAI #QuebecAI #QuebecArtificialIntelligence
Retweeted by Vincent Boucher ๐Ÿ’Ž
I'll be talking in DC at the @SeedAIOrg summit tomorrow about recent trends in AI, and policy interventions to lead to a better AI development community. Will show off some neat new analysis from @Hernandez_Danny also. (Will repub slides after). twitter.com/SeedAIOrg/statโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
Excited to announce our excellent speakers at CLeaR WS on "Combining Learning and Reasoning: programming languages, formalisms, and representations", clear-workshop.github.io @RealAAAI,#AAAI2022,@BehrouzBabaki @sebdumancic,@hfaghihi15,@guyvdb @DanRothNLP @GaryMarcus @monireheb
Retweeted by Vincent Boucher ๐Ÿ’Ž
Can video-based grounding improve NLU tasks (incl. world/temporal/physical knowledge)? Check out our new work *VidLanKD*: Improving Language Understanding via Video-Distilled Knowledge Transfer! arxiv.org/abs/2107.02681 Led by @ZinengTang, w/ @HaoTan5 @mohitban47 (@uncnlp) ๐Ÿงต3
Retweeted by Vincent Boucher ๐Ÿ’Ž
Glad to announce that our paper: Learning with Holographic Reduced Representations has been accepted to #NeurIPS2021! We show how holographic reduced representations can be used a train a neuro-symbolic architecture. arxiv.org/abs/2109.02157
Retweeted by Vincent Boucher ๐Ÿ’Ž
Let's go #888InnerCircle let's see how everyone's awesome profile photos are looking these days... I'll start with mine... an 0N1 Force with the beautiful @maalavidaaa Ring and awesome @jasonseife in the background!
Retweeted by Vincent Boucher ๐Ÿ’Ž
PyTorch 1.10 is here! Highlights include updates for: - CUDA Graphs APIs updates - Several frontend APIs moved to Stable - Automatic fusion in JIT Compiler support for CPU/GPUs - Android NNAPI now in beta Blog: pytorch.org/blog/pytorch-1โ€ฆ Release: github.com/pytorch/pytorcโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
Neural networks really, truly learn. It's not a fancy trick. This is one of the most remarkable things humans have ever figured out, and the implications are difficult to overstate.
Retweeted by Vincent Boucher ๐Ÿ’Ž
Love this work. The most exciting thing here is that it's completely open in the raw-est form of openness. @BigscienceW and EleutherAI are super cool! twitter.com/BigscienceW/stโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
For the good of science, let's avoid arbitrary distinctions between "Bayesian" and "non-Bayesian" methods in deep learning, and put to rest the notion that deep ensembles are a "non-Bayesian" competitor to standard approximate inference. A new blog post: cims.nyu.edu/~andrewgw/deepโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
AI DEBATE 3: "The Legendary Debate" 9 , 0 0 0 Attendees! 90 Days until #AIDebate3 !!! RSVP Info: linkedin.com/events/aidebatโ€ฆ It All Comes Down To AI DEBATE 3 By: MONTREAL.AI #AIDebate #AIDebate3 #MontrealAI #QuebecAI #QuebecArtificialIntelligence
Retweeted by Vincent Boucher ๐Ÿ’Ž
@elonmusk @MatchasmMatt @NHTSAgov Hey @elonmusk, if you are not too chicken, @missy_cummings and @GaryMarcus would love to debate self-driving prospects with you and @karpathy at #AIDebate3, an online event on December 23 that already has 10,000 people signed up.
Retweeted by Vincent Boucher ๐Ÿ’Ž
Tired of taking 100 calibrated images to get a reasonable 3D reconstruction? Our new #NeurIPS2021 paper recovers shape, texture, and illumination of 3D objects from ~8 images in the wild without precise camera poses! jasonyzhang.com/ners
Retweeted by Vincent Boucher ๐Ÿ’Ž
i've spent this afternoon learning manim (originally by @3blue1brown but now maintained by @manim_community ) to use in the video for one of my #NeurIPS2021 papers. it's *so* nice. from now on i'm going to try to make all my talks with this! here's a sneak preview:
Retweeted by Vincent Boucher ๐Ÿ’Ž
TIL: Kernel methods, i.e. Kernel Mean Embedding of Distributions, are becoming cool again. Check out this paper of #NeurIPS2021 arxiv.org/abs/2109.03582. Here is an introduction arxiv.org/pdf/1605.09522โ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
I am proud to announce that three of our papers have just been accepted for publication. Two at #NeurIPS2021, and one at #WACV2022. This wouldnโ€™t have been possible without my excellent collaborators @HildeKuehne, @OliverDeussen, Christian Borgelt, Bastian Goldluecke, ...
Retweeted by Vincent Boucher ๐Ÿ’Ž
I'm aware of 5 GPT3 public replications so far*, would anyone like to further flesh out this list? (Note, I'm not counting things like @huggingface T0 because it's a different architecture. Though perhaps I should be?) twitter.com/jackclarkSF/stโ€ฆ
Retweeted by Vincent Boucher ๐Ÿ’Ž
We're thrilled to partner with DeepLearning.AI to create some great new content for their NLP Specialization on Coursera! With this update, you can access exciting new material and lectures that cover the state of the art in NLP ๐Ÿง‘โ€๐Ÿซcoursera.org/specializationโ€ฆNU
Retweeted by Vincent Boucher ๐Ÿ’Ž
Twiends™ uses the Twitter™ API, displays it's logo & trademarks, and is not endorsed or certified by them. These items remain the property of Twitter. We do not sell followers, we only provide display advertising. Bots & fake accounts are not permitted on twiends. © 2009
Grow Your Twitter Free
Want To Grow Your Twitter?
We help other people find and follow you on Twitter.
Key Info:
Started in 2009
Over 6 million signups
Country targeting provided
We never auto tweet to your timeline
We never auto follow others
We actively moderate our community
Please Share
Please upgrade your browser  chrome