In AI every large player is spending truly absurd amounts of compute using roughly the same hardware with roughly the same AI tools and roughly the same data... to achieve roughly same results 🤷‍♂️
Gets worse, in AI for Earth it is EXACTLY the same open data (Sentinel/Landsat/…).
But, wait, plot twist, no large AI player is touching geospatial data, and I can't understand why: Literally Petabytes of data, with full global and historical coverage, fully cloud friendly, zero legal issues and can only augment multimodal models with actual Earth semantics.
...and that's why we started Clay. To provide a plug and play trained Earth model to improve other models.
If you know why OpenAI, Mistral AI, Anthropic, Google DeepMind, ... are not adding open geospatial data (or if you can help us convince them), please do share.
Gets worse, in AI for Earth it is EXACTLY the same open data (Sentinel/Landsat/…).
But, wait, plot twist, no large AI player is touching geospatial data, and I can't understand why: Literally Petabytes of data, with full global and historical coverage, fully cloud friendly, zero legal issues and can only augment multimodal models with actual Earth semantics.
...and that's why we started Clay. To provide a plug and play trained Earth model to improve other models.
If you know why OpenAI, Mistral AI, Anthropic, Google DeepMind, ... are not adding open geospatial data (or if you can help us convince them), please do share.