Innovation is hard. Copying is easier.
RESEARCH
-
Salesforce Releases CodeGen2.5 7B Model for Code Generation
By
–
Salesforce AI Releases CodeGen2.5 Model: https://
huggingface.co/Salesforce/cod
egen25-7B-multi
… The key contributions towards productization of these models are: Releasing CodeGen2.5 LLM with state-of-the-art on HumanEval for 7B parameters. CodeGen2.5 with 7B is on par with >15B code-generation models -
Language Models and the Future of AI Development
By
–
Full (free) story can be found here: https://
supervised.news/p/language-mod
els-and-the-future-of
… -
Azure Quantum Accelerates Scientific Discovery and Research
By
–
Accelerating scientific discovery with Azure Quantum – The Official Microsoft Blog https://
bit.ly/43367qY -
Pic2Word: Zero-Shot Composed Image Retrieval Method
By
–
Composed image retrieval (CIR) returns images for a query that combines an image and text sample that provides instructions to modify the image. Read how Pic2Word, the first zero-shot CIR method, only requires weakly labeled image-caption pairs to train → https://
goo.gle/3pvXaZA -
Voicebox: Multilingual Speech Synthesis via Flow Matching
By
–
Voicebox can synthesize speech across six languages, perform noise removal, edit content, transfer audio style & more. It's based on Flow Matching, a new method proposed by Meta AI.
— AI at Meta (@AIatMeta) 6 juillet 2023
More details on this work & real examples ➡️ https://t.co/ptAtSE4DuQ pic.twitter.com/SUYbXw15oWVoicebox can synthesize speech across six languages, perform noise removal, edit content, transfer audio style & more. It's based on Flow Matching, a new method proposed by Meta AI. More details on this work & real examples
https://bit.ly/3CwUw8z -
unarXive 2022: Preprocessed arXiv Publications for NLP Research
By
–
unarXive 2022: All arXiv Publications Pre-Processed for NLP, Including Structured Full-Text and Citation Network https://
bit.ly/43sXI1w #MachineLearning #AI #DataScience #DeepLearning -
nanoT5: Fast T5 Model Pretraining and Finetuning on Single GPU
By
–
nanoT5 – Fast & Simple repository for pretraining and finetuning T5-style models A code repository for pretraining and finetuning T5 LLMs(one of the earliest encoder-decoder models) under a limited budget. – Achieves 40.7 RougeL in 16 hrs on a single GPU(original T5 on Hugging
-
AI Experts Disagree on AGI Timeline Estimates
By
–
Experts in AI do not agree on an AGI timeline. Off the top of my head Famous AI founder said 6yrs 4 yr ago
Metaculus says 3
Musk said 6 1y ago
OpenAI says <10
Eric Schmidt said 20 1y ago
Gary Marcus says 10-77
Yann LeCun says next few decades
Andrew Ng says 30-50