Even though BLOOMChat is trained on English dominated alignment data, we see a significant impact on qualitative metrics for assistant style conversational use cases in multilingual settings. (2/6)
RESEARCH
-
Public engagement with AI research and latent space concepts
By
–
800 j’aime en mettant le mot « espace latent » et en partagent un papier de recherche 🥲. (Avec une vidéo stylé)
— Defend Intelligence (Anis Ayari) (@DFintelligence) 19 mai 2023
Ça donne de l’espoir. La preuve que la science intéresse peut être… sans besoin d’aller chercher du putaclick angoissant pour parler d’IA. (A bon entendeur) https://t.co/iAPZ6YdyDo800 j’aime en mettant le mot « espace latent » et en partagent un papier de recherche . (Avec une vidéo stylé) Ça donne de l’espoir. La preuve que la science intéresse peut être… sans besoin d’aller chercher du putaclick angoissant pour parler d’IA. (A bon entendeur)
-
DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining
By
–
DoReMi: Optimizing Data Mixtures Speeds Up Language Model Pretraining Xie et al.: https://
arxiv.org/abs/2305.10429 #ArtificialIntelligence #ChatGPT #DeepLearning -
Guide to Training Your Own Large Language Models
By
–
How to train your own Large Language Models https://
bit.ly/44wZq2d #AI #MachineLearning #DeepLearning #LLMs #DataScience -
Architectural Cleverness Cannot Compensate for Raw Scaling Gap
By
–
http://
incompleteideas.net/IncIdeas/Bitte
rLesson.html
… is about the surprisingly limited impact of architectural cleverness versus raw scaling. I am analogizing to suggest that a radical hardware architecture like https://
en.m.wikipedia.org/wiki/Cerebras probably won’t make up for being back a few stages, but hope springs -
Master Transformers for Text, Vision, and Speech with HuggingFace
By
–
By the end of this course, you will know how to apply Transformers to (almost) any ML problem! This includes text, vision, and speech. And you'll do that using HuggingFace libraries. Stay tuned and follow @abacusai
! There's a lot more coming! -
Free HuggingFace AI/ML Course on Natural Language Processing and Transformers
By
–
FREE AI/ML course you can't miss: The HuggingFace Course https://
huggingface.co/course/chapter
0/1
… The hottest field in AI right now is Natural Language Processing. This is the place to start if you want to learn about Transformers. Read on for what to expect: -
Trending AI Models, Datasets and Apps This Week
By
–
Trending model, datasets & apps of the week on http://
hf.co. Kudos to @BigCodeProject @MosaicML @databricks @togethercompute @AnthropicAI @deepfloydai and many others! -
Generative AI in Science: White House PCAST Live Stream
By
–
Join the public live stream in <2 hours on @WHOSTP website. I will be talking about #generativeai in science to the PCAST council. https://
whitehouse.gov/pcast/meetings
/2023-meetings
… -
China’s Fab Capabilities and Cerebras Wafer-Scale Integration Viability
By
–
China’s fabs are several generations behind, but I wonder if something like Cerebras’ wafer scale integration would still be viable. The limited impact of clever architecture vs simple scaling is one of the “bitter lessons”, but it is still meaningful.