Aqsone

🤖 Artificial Intelligence
🇫🇷 Toulouse et Paris depuis 2014
💡 De la donnée à l’innovation
🏆 Prototyper. Déployer. Valoriser.

[Aqsone @aqsone]

📚 Imagine an AI with a photographic memory of your conversations and lightning-fast text processing. With up to 22x faster inference, it preserves context effortlessly. No more forgotten chats or text limitations. StreamingLLM redefines language models! 💬✨

📜 Paper: https://t.co/1OccOgqe1n

#AI #StreamingLLM #Innovation

[Carlos E. Perez @IntuitMachine]

Introducing StreamingLLM.

Imagine chatting with an AI assistant that can contextually reference your conversations from weeks or months ago. Or summarizing reports that span thousands of pages. StreamingLLM makes this possible by enabling language models to smoothly handle endless texts without losing steam.

Current LLMs are like students cramming for an exam - they can only memorize a limited context. StreamingLLM is the valedictorian with a photographic memory of everything you've ever discussed.

It works by identifying and preserving the model's inherent "attention sinks" - initial tokens that anchored its reasoning. Combined with a rolling cache of recent tokens, StreamingLLM delivers up to 22x faster inference without any drop in accuracy.

You know that irksome feeling when chatbots forget your earlier conversations? StreamingLLM abolishes that frustration. It remembers the touchdowns from your last game and your newborn's name without missing a beat.

Monumental books, verbose contracts, drawn out debates - StreamingLLM takes them all in its stride. No shortcuts, no forgetfulness. It's like upgrading your assistant's RAM to handle heavier workloads flawlessly.

[Aqsone @aqsone]

@ciguleva Midjourney seems more accurate 🤔

[Aqsone @aqsone]

@AlainGoudey @RBaranger Dataviz super intéressante ! Surtout grâce à la dimension "Nombre moyen d'années nécessaires pour rembourser le prêt" 😍

[Aqsone @aqsone]

@javilopen Woow this new feature is incredible! 👏

[Aqsone @aqsone]

@_akhaliq @morphaistudio Soon, the possibilities will be infinite 🤯

[Aqsone @aqsone]

@DeepLearningAI_ @awscloud @coursera Our collaborators just completed the fantastic course on "Generative AI with Large Language Models" at @aqsone! 🚀 #AI #MachineLearning

[Aqsone @aqsone]

@hylarucoder @huggingface @Gradio Yes @Gradio is a great tool!

[Aqsone @aqsone]

@OpenAI Great news! 🇫🇷

[Aqsone @aqsone]

@_akhaliq This kind of annotated dataset project is immensely beneficial for AI research. Thank you for sharing it!

[Aqsone @aqsone]

@huggingface Congrats @huggingface 🇫🇷

[Aqsone @aqsone]

🌍 IBM and NASA open source largest geospatial AI foundation model on Hugging Face! This collaboration will democratize access to AI, allowing scientists and researchers to analyze large environmental datasets and drive innovations in climate and Earth science. The potential to track deforestation, predict crop yields, and monitor greenhouse gases is enormous!

[IBM Research @IBMResearch]

We’ve open sourced our watsonx․ai geospatial foundation model, built with @NASAEarth, on @HuggingFace! 🚀

It’s NASA’s first openly available AI foundation model and the largest geospatial model on HuggingFace. Learn more: https://t.co/432TfSP1e1

[Aqsone @aqsone]

This breakthrough will accelerate all 3D developments! 🚀 Imagine what more it could achieve with additional prompts! The possibilities are endless! 🌟

[NVIDIA AI Developer @NVIDIAAIDev]

🏗️ Build #3D scenes from a single input view based on the latest #NVIDIAResearch paper accepted at #ICCV2023.

👀 This 3D-aware #generativeAI model can build diverse and plausible views.

research paper: https://t.co/DX0iXgAjzM
project page: https://t.co/Tu5mYV06Rw

[Aqsone @aqsone]

@Google @GoogleDeepMind This kind of generative AI project is amazing and promising for the future! 👏

[Aqsone @aqsone]

Another Meta open-source projetc: AudioCraft.

A powerful family of generative AI models designed to create realistic audio & music from text. 🎵 With MusicGen, AudioGen, and EnCodec in one code base, it's now easier to compose music, generate sounds, and compress audio. 🎧 This release comes with an improved EnCodec for higher quality music generation and pretrained AudioGen models for environmental sounds and effects. 🌳🌊

GitHub repository: https://t.co/LjfFdr64ji

[AI at Meta @AIatMeta]

Today we're sharing details on AudioCraft, a new family of generative AI models built for generating high-quality, realistic audio & music from text. AudioCraft is a single code base that works for music, sound, compression & generation — all in the same place.

More details ⬇️

[Aqsone @aqsone]

@nathanhabib1011 @huggingface @edwardbeeching Very useful leaderboard to track the best new open LLMs and chatbots when they are released! 👏

[Aqsone @aqsone]

🔍 Curious about "Interpretability and Explainability in Machine Learning"? Discover the power of SHAP! 🚀 Unveil intriguing patterns like gender influencing survival chances and age playing a role. Gain transparency and trust with #Interpretability and #Explainability in #MachineLearning. 💡✨

Interested? Check our blog article 👉 https://t.co/9IJ2UAaW9O

[Aqsone @aqsone]

🔍 Discover PointOdyssey, the ultimate playground for AI tracking algorithms! 🌟 With 104 videos and naturalistic motion, it sets a new standard for training and evaluation. 🎥 Let deformable characters dance through 3D scenes while cameras capture their every move. Get ready to witness the future of tracking tech!

[AK @_akhaliq]

PointOdyssey: A Large-Scale Synthetic Dataset for Long-Term Point Tracking

paper page: https://t.co/BofjfBVSIY

introduce PointOdyssey, a large-scale synthetic dataset, and data generation framework, for the training and evaluation of long-term fine-grained tracking algorithms. Our goal is to advance the state-of-the-art by placing emphasis on long videos with naturalistic motion. Toward the goal of naturalism, we animate deformable characters using real-world motion capture data, we build 3D scenes to match the motion capture environments, and we render camera viewpoints using trajectories mined via structure-from-motion on real videos. We create combinatorial diversity by randomizing character appearance, motion profiles, materials, lighting, 3D assets, and atmospheric effects. Our dataset currently includes 104 videos, averaging 2,000 frames long, with orders of magnitude more correspondence annotations than prior work. We show that existing methods can be trained from scratch in our dataset and outperform the published variants. Finally, we introduce modifications to the PIPs point tracking method, greatly widening its temporal receptive field, which improves its performance on PointOdyssey as well as on two real-world benchmarks.

[Aqsone @aqsone]

Sounds like great news! Introducing LP-MusicCaps 🎶 Researchers leverage Large Language Models to generate natural language descriptions for music tracks. The dataset contains 2.2M captions and 0.5M audio clips, boosting music captioning performance and surpassing supervised baselines. 🎵

[AK @_akhaliq]

LP-MusicCaps: LLM-Based Pseudo Music Captioning

paper page: https://t.co/yz3fRJb1NK

Automatic music captioning, which generates natural language descriptions for given music tracks, holds significant potential for enhancing the understanding and organization of large volumes of musical data. Despite its importance, researchers face challenges due to the costly and time-consuming collection process of existing music-language datasets, which are limited in size. To address this data scarcity issue, we propose the use of large language models (LLMs) to artificially generate the description sentences from large-scale tag datasets. This results in approximately 2.2M captions paired with 0.5M audio clips. We term it Large Language Model based Pseudo music caption dataset, shortly, LP-MusicCaps. We conduct a systemic evaluation of the large-scale music captioning dataset with various quantitative evaluation metrics used in the field of natural language processing as well as human evaluation. In addition, we trained a transformer-based music captioning model with the dataset and evaluated it under zero-shot and transfer-learning settings. The results demonstrate that our proposed approach outperforms the supervised baseline model.

[Aqsone @aqsone]

AI's potential to revive ancient molecules for modern drug discovery is fascinating! While the current results may be promising, further improvements in the algorithm could lead to groundbreaking advancements in antibiotic development. Exciting times ahead for bioengineering and medicine! 🧬🔬

[hardmaru @hardmaru]

AI search of Neanderthal proteins resurrects ‘extinct’ antibiotics

Bioengineers have used artificial intelligence to bring molecules back from the dead. These newly identified protein snippets from Neanderthals are found to have bacteria-fighting powers.
https://t.co/QBdIzoDDF9

[Aqsone @aqsone]

@PyTorch Great news! 🚀