AI News Bytes: Meet DALL-E 's Cousin-VALL-E; Deepmind's MuJoCo MPC (MJPC); Stanford's Emotion Generator.....

Hi there, today we will share some research updates from BoxInstSeg, Variance Reduction System, Federated Learning Utilities and Tools for Experimentation (FLUTE), VALL-E, Deepmind's MuJoCo MPC (MJPC), Open Deep Learning Toolkit for Robotics version 2.0, ClimateLearn library, AI2's TeachMe and many more..

Stanford:?Researchers at Stanford have developed an?Artificial Intelligence (AI) Model, ?SUMMON , that can generate Multi-Object Scenes from a Sequence of Human Interaction.

Google Research:?This research work demonstrates dramatic improvements in the generality and performance of learned optimizers, by scaling up meta-training compute and dataset size, and by making architectural improvements. The resulting learned optimizer,?VeLO , has no hyperparameters, yet outperforms heavily hyperparameter-tuned baselines across more than 80 diverse optimization tasks, including large published machine learning models.

Zhejiang University/Alibaba:?BoxInstSeg ?is a toolbox that aims to provide state-of-the-art box-supervised instance segmentation algorithms. It is built on top of mmdetection. The main branch works with Pytorch 1.6+ or higher (we recommend Pytorch 1.9.0)

Meta AI:?Shares more details about the?Variance Reduction System ?— a new offline RL framework developed by Meta to help ensure a more equitable distribution of ads on our services?

Microsoft AI:?Microsoft team?introduces?Federated Learning Utilities and Tools for Experimentation ?(FLUTE) as a framework for running large-scale offline federated learning simulations.

CVF: All papers of WACV 2023 are now available in -?Open Access Repository ?provided by the Computer Vision Foundation.?

Stable Diffusion Application:?This New?API? Makes It Easy And Cheap For Developers To Build Machine Learning ML-Powered Apps Using Stable Diffusion.?

Microsoft: DALL-E generates pixels from the text. Now meet its cousin,?VALL-E , which generates audio from text. VALL-E Can Turn Text into Speech Using Three-Second.

Stanford:?A key issue with LLMs is understanding how certain the model is about an answer. This?paper? demonstrates a proof-of-concept of a large language model conducting corporate lobbying-related?activities.

Google/Deepmind:?Med-PaLM ?is a?Large Language Model ?that is Supporting the Medical Domain in Providing Safe and Helpful Answers.

Meta/KAIST/NYU:?Convolutional networks strike back, again. The fully convolutional ConvNeXt v2 extends the successful ConvNeXt architecture by adding self-supervised learning capabilities.?What's new?

Meta:?In this?research paper , Meta researchers explore the scaling properties of mixed-modal generative models, discovering new scaling laws that unify the contributions of individual modalities and the interactions between them.

University of Toronto:?Hinton, professor at the University of Toronto and engineering fellow at Google Brain, recently published a?paper ?on the Forward-Forward algorithm (FF), a technique for training neural networks that uses two forward passes of data through the network, instead of backpropagation, to update the model weights.

Deepmind:?MuJoCo MPC (MJPC) ?is an interactive tool for real-time behavior synthesis with predictive control algorithms.?MJPC ?includes a number of planners written in multi-threaded C++, like iLQG and Gradient Descent. One of the awesome MJPC features is asynchronous simulation. You can slow down/speed up the environment time (press -/+), effectively providing additional planning time.?

Toyota/Meta/The Hebrew University of Jerusalem:?ReVISE , the first universal audio-visual speech enhancement model powered by SSL. This single model can perform video-to-speech synthesis, speech inpainting, denoising, and source separation.

Shanghai AI Laboratory/ Tsinghua University:?This research from China introduces?InternImage , a large-scale CNN model that outperforms ViT. Experiments on ImageNet, COCO, ADE20K, and a wide range of benchmarks demonstrated that InternImage achieves accuracy comparable to or better than ViT trained on large data sets, showing that CNNs are a viable option for large-scale models. This demonstrates that CNNs have great potential as an option for large-scale models. One challenge is that DCN-based methods are processing-heavy.

CMU:?A?deep neural network ?maps the phase and amplitude of WiFi signals to UV coordinates within 24 human regions.?The model can estimate the dense pose of multiple subjects by utilizing WiFi signals as the only input.

AI2:?Researchers From Allen Institute for AI Introduce TeachMe : A Framework To Understand And Correct AI Models.?TeachMe is the suggested system, which consists of two key parts: (1) Entailer, a T5-based machine reasoning model that can generate valid lines of reasoning. (2) a dynamic database of previous comments.

New Unicorn:?Germany’s Translation?Tech Firm DeepL ?Becomes Latest AI Unicorn. DeepL has emerged as one of the most disruptive players in the machine translation space with a product that rivals even Google’s technology.

Read Today's Latest Newsletter where we shared research updates from MIT MIMIC-IV, the Japanese DeBERTa V2 Large model,?An image-to-image translation framework, One-Shot-Talking-Face, Text-to-Audio Diffusion, AI-powered Distracted Driving Monitor