Scott Hosking
@scotthosking.bsky.social
Director in Environment and AI Research | @theturing.bsky.social | @bas.ac.uk
scotthosking.com
scotthosking.com
October 15, 2025 at 3:26 PM
May 19, 2025 at 8:14 PM
Thanks @jamespjh.bsky.social for hosting! huge opportunities to work across the 80 institutes(!) in attendance to develop + deploy AI-powered climate solutions at scale
Fantastic alignment with Turing's Missions in Environment Forecasting and Sustainability/Decarbonisation www.turing.ac.uk/research
Fantastic alignment with Turing's Missions in Environment Forecasting and Sustainability/Decarbonisation www.turing.ac.uk/research
Research
We are doing this through basic research into theoretical scientific problems, and through applying our cutting-edge science to real-world problems, and
www.turing.ac.uk
April 14, 2025 at 2:40 PM
Thanks @jamespjh.bsky.social for hosting! huge opportunities to work across the 80 institutes(!) in attendance to develop + deploy AI-powered climate solutions at scale
Fantastic alignment with Turing's Missions in Environment Forecasting and Sustainability/Decarbonisation www.turing.ac.uk/research
Fantastic alignment with Turing's Missions in Environment Forecasting and Sustainability/Decarbonisation www.turing.ac.uk/research
March 31, 2025 at 8:24 PM
DeepSensor now has improved support for environmental forecasting and non-Gaussian likelihoods in v0.4 ⚡ 🌪️
Simple Python interface 👉 great way to rapidly build on recent advances in weather prediction (eg Aardvark Weather, arxiv.org/abs/2404.00411)
Colab + Docs + Code: github.com/alan-turing-...
Simple Python interface 👉 great way to rapidly build on recent advances in weather prediction (eg Aardvark Weather, arxiv.org/abs/2404.00411)
Colab + Docs + Code: github.com/alan-turing-...
GitHub - alan-turing-institute/deepsensor: A Python package for tackling diverse environmental prediction tasks with NPs.
A Python package for tackling diverse environmental prediction tasks with NPs. - alan-turing-institute/deepsensor
github.com
March 20, 2025 at 8:47 PM
We saw minimal improvements when we pre-trained a U-Net model on CMIP data. This was back in 2021 though so new meta-learning approaches might do a better job now
March 20, 2025 at 8:29 PM
We saw minimal improvements when we pre-trained a U-Net model on CMIP data. This was back in 2021 though so new meta-learning approaches might do a better job now
Transfer learning across geographies does enable predictions that are outside the previous local min/max values.
However, not sure how ML would handle predictions that are outside the global distribution of values
However, not sure how ML would handle predictions that are outside the global distribution of values
March 20, 2025 at 8:18 PM
Transfer learning across geographies does enable predictions that are outside the previous local min/max values.
However, not sure how ML would handle predictions that are outside the global distribution of values
However, not sure how ML would handle predictions that are outside the global distribution of values