That makes sense to me. Summarisation was pretty bad before, and it’s one of the strengths of LLMs.
RESEARCH
-
Base Models Achieve In-Context Learning Through Instruction Tuning
By
–
i think even <base models have in-context learning ability if instruction tuned.
-
Generative AI Advances: Cherry-Picked Demos vs Practical Applications
By
–
Generative AI has made great strides, producing images, text, video & music. However, differentiate between cherry-picked demos & practical use cases. Models act as components in intelligent systems, not minds. For more, see JayAlammar's article series:
-
Embedded documents semantic search infrastructure discussion
By
–
this is pretty interesting. a collection of embedded texts… idea would be multiple people want to do semantic search over the same pieces of text. are there commonly embedded documents that would provide value if listed here?
-
Train SOTA Computer Vision Models with Super-Gradients Library
By
–
super-gradients/YOLONAS.md at master · Deci-AI/super-gradients https://
bit.ly/44wHlBp
Easily train or fine-tune SOTA computer vision models with one open source training library. #AI #MachineLearning #DeepLearning #LLMs #DataScience -
Hopper-Dean Foundation Supports Deep Learning Indaba Initiative
By
–
The Hopper-Dean Foundation is proud to help support the Deep Learning Indaba. Please consider doing the same! 🙏 https://t.co/aIc3gUFUx8
— Jeff Dean (@JeffDean) 19 mai 2023The Hopper-Dean Foundation is proud to help support the Deep Learning Indaba. Please consider doing the same!
-
The Printing Press: Unfinished Impact on Society
By
–
I am far more worried about the printing press. It is not done with us yet!
-
GPT-4 Theory of Mind Investigation Through Empirical Testing
By
–
3/ @labenz had a hypothesis that a paper on GPT-4 lacking “theory of mind” was incorrect and wanted to investigate further. He posted a Bounty for a script that used GPT-4 to run on 400 random sample problems and document the results to prove or disprove the paper findings
-
Conversational Interfaces for Model Understanding and Explainability
By
–
Check out this video with Prof. Hima Lakkaraju on the role of conversational interfaces in model understanding. With this tool, users can understand model predictions & explanations. An essential viewing for those interested in NLP.
-
Vision Transformer Efficient Video Backbone Using Sparse Tubes
By
–
Learn how we turned a Vision Transformer image encoder into an efficient video backbone using sparse video tubes (3D grid-based cuboids with learnable visual representations of video samples), reducing compute needs and achieving competitive results → https://
goo.gle/435CTYQ