If there was a test of time award for datasets, the first winner would be MNIST, without which there would be no neural networks.
RESEARCH
-
RAND Should Focus on Global Compute Acceleration
By
–
Looks like I have a new adversary. Sorry to say, I have been a huge fan of RAND for years. But they should be writing about global compute acceleration.
-
Continuous Learning Essential to Keep Pace with AI
By
–
I agree, @Mercer — continuous learning is the only way we'll all keep up!
-
DeepSeek-Coder-V2-Instruct Outperforms GPT-4o on Coding Tasks
By
–
DeepSeek-Coder-V2-Instruct is the open-weights model that beat GPT-4o at coding tasks. Try the 236B model on Replicate today
-
AGI Achievement: When AI Becomes Preferable to Humans
By
–
AGI is achieved when an AI agent becomes preferable to a human for employment, or when AI's capabilities match or exceed human abilities without significant limitations.
-
Keras Models Direct Save to Hugging Face Hub Integration
By
–
Further, in the next Keras version we're making it possible to save *any* Keras model to Hugging Face Hub directly, via the built-in model `.save()` method (and likewise for model loading!)
-
GraphCast Team Wins Award for Weather Prediction AI
By
–
Extremely proud of my partner in crime @FortunatoMeire and all of #GraphCast team for this award! The thing is heavy
-
Claude 3 Haiku Fine-tuning Achieves 99.6% Classification Accuracy
By
–
In testing, we tuned Haiku to act as a moderator on an example dataset. Fine-tuning improved classification accuracy from 81.5% to 99.6% and reduced prompt tokens per query by 89%. Fine-tuning for Claude 3 Haiku is available today in preview in Amazon Bedrock.
-
Open Source AI Party at ICML Vienna July 24
By
–
Join us at the Open Source AI Party at ICML Vienna! Location: Das Campus Date: Wednesday, July 24 Time: 7 PM – Late Get ready for a night of fun, networking, and great vibes. Just a 5-minute walk from the Messe Wien Congress Center! Space is limited; priority will