Researchers have found a new set of AI algorithms that construct & judge data explanations using sequential Monte Carlo techniques. The algorithms provide uncertainty-calibrated insights & track the likelihood of explanations. Learn more: https://
bit.ly/42HajfA @nasscom
MACHINE LEARNING
-
New AI Algorithms for Uncertainty-Calibrated Data Explanations
By
–
-
Vision Transformers Interpretation Tutorial at CVPR 2023
By
–
Catch our very own @RisingSayak in collaboration with @hila_chefer at #CVPR2023 as they deliver their tutorial — "All Things ViTs" presenting methods to explain & intepret Vision Transformers! They'll have @MokadyRon as a guest speaker Website https://
all-things-vits.github.io/atv -
Accenture Creates 40,000 New AI Jobs Globally
By
–
Artificial Intelligence to create new jobs at Accenture Artificial intelligence creates 40,000 new roles at Accenture https://
buff.ly/43BJefn
#ai #ArtificialIntelligence #MachineLearning #deeplearning #jobs #futureofwork -
Galactic: Scaling End-to-End Reinforcement Learning for Robotic Rearrangement
By
–
Galactic: Scaling End-to-End Reinforcement Learning for Rearrangement at 100k Steps-Per-Second paper page: https://
huggingface.co/papers/2306.07
552
… present Galactic, a large-scale simulation and reinforcement-learning (RL) framework for robotic mobile manipulation in indoor environments. -
ATT3D: Advanced Text-to-3D Object Synthesis Technology
By
–
ATT3D: Amortized Text-to-3D Object Synthesis
— AK (@_akhaliq) 14 juin 2023
paper page: https://t.co/jINaEFQvWR
Text-to-3D modelling has seen exciting progress by combining generative text-to-image models with image-to-3D methods like Neural Radiance Fields. DreamFusion recently achieved high-quality results… pic.twitter.com/uFMHgImJCEATT3D: Amortized Text-to-3D Object Synthesis paper page: https://
huggingface.co/papers/2306.07
349
… Text-to-3D modelling has seen exciting progress by combining generative text-to-image models with image-to-3D methods like Neural Radiance Fields. DreamFusion recently achieved high-quality results -
Neural Scene Chronology: Time-Varying 3D Reconstruction from Internet Photos
By
–
Neural Scene Chronology
— AK (@_akhaliq) 14 juin 2023
paper page: https://t.co/oUdnOhTYls
In this work, we aim to reconstruct a time-varying 3D model, capable of rendering photo-realistic renderings with independent control of viewpoint, illumination, and time, from Internet photos of large-scale landmarks.… pic.twitter.com/pmIuSzH6LNNeural Scene Chronology paper page: https://
huggingface.co/papers/2306.07
970
… In this work, we aim to reconstruct a time-varying 3D model, capable of rendering photo-realistic renderings with independent control of viewpoint, illumination, and time, from Internet photos of large-scale landmarks. -
GeneCIS: Benchmark for General Conditional Image Similarity
By
–
GeneCIS: A Benchmark for General Conditional Image Similarity paper page: https://
huggingface.co/papers/2306.07
969
… argue that there are many notions of 'similarity' and that models, like humans, should be able to adapt to these dynamically. This contrasts with most representation learning -
GLoRA: Generalized LoRA for Efficient Model Fine-tuning
By
–
One-for-All: Generalized LoRA for Parameter-Efficient Fine-tuning paper page: https://
huggingface.co/papers/2306.07
967
… present Generalized LoRA (GLoRA), an advanced approach for universal parameter-efficient fine-tuning tasks. Enhancing Low-Rank Adaptation (LoRA), GLoRA employs a generalized -
WebGLM: Efficient Web-Enhanced Question Answering System
By
–
WebGLM: Towards An Efficient Web-Enhanced Question Answering System with Human Preferences paper page: https://
huggingface.co/papers/2306.07
906
… present WebGLM, a web-enhanced question-answering system based on the General Language Model (GLM). Its goal is to augment a pre-trained large language -
Algorithmic Innovation: The Next Frontier Beyond Scaling
By
–
biggest questions in ai: – since the transformer, most AI improvement was via scaling up & great engineering
– almost no algorithmic ideas have driven performance, besides RLHF
– future gains will likely require algorithmic innovation how long will they take? who does them?