Ai2 Newsletter
May 2024
Top story - Ali Farhadi and AI2 Profiled in the New York Times
OLMo just got better! Created on the path towards our upcoming 70 billion parameter model, OLMo 1.7–7B showcases a longer context length of up to 4096 tokens, and higher benchmark performance due to improved data quality, a new two-stage training procedure, and architectural improvements.
We also released an updated version of our dataset, Dolma 1.7 — in which we focused on (a) exploring more diverse sources of data, and (b) more precise filtering of web sources.
Dolma moves to the ODC-BY license
We believe that the ODC-BY license is going to make your work with the Dolma dataset significantly easier moving forward. ODC-BY grants the public permission to use Dolma, and users may copy, reproduce, and distribute Dolma. Users can also modify and create derivative works using all or a substantial portion of the Dolma dataset with an attribution notice.
Introducing SatlasPretrain Models, a suite of open geospatial foundation models
To efficiently and accurately tackle remote sensing tasks, large-scale pre-trained models must be open and available for others to use as a starting point for their research. We’re excited to release SatlasPretrain Models, which includes model weights and source code on GitHub, with comprehensive documentation and examples.
More from us
- CBS Mornings spotlights Skylight’s support in restoring Bahamas' Seas
- How can AI help us prepare for a changing climate? Listen to Chris Bretherton on Nvidia's The AI Podcast
- We're hiring! Check out our current openings