Artificial intelligence is not only transforming offices, search engines, and applications. It can also help listen to forests, estimate harvests, verify carbon credits, and anticipate environmental risks. With that goal in mind, Google DeepMind selected 16 organizations from Asia-Pacific for its AI for the Planet accelerator.
The program began this week with a practical bootcamp in Singapore. Over the next three months, the startups, nonprofits, and research teams will receive access to Google’s AI tools, specialized models, mentorship, and technical support.
AI to protect nature
The selected projects use different data sources, from animal sounds to satellite images. The idea is to turn environmental signals that are difficult to analyze manually into useful information for making decisions.
- 800 Trust, New Zealand: combines AI and bioacoustics to continuously monitor biodiversity and detect environmental threats.
- Kumi Analytics, Singapore: uses remote sensing and
deep learningto create environmental baselines for conservation projects. - Listening Lab, New Zealand: develops bioacoustic tools that require less data to study biodiversity through sound.
- TelePIX, South Korea: transforms satellite images into practical information for monitoring mangroves on a global scale.
- Wildlife.ai, New Zealand: builds open AI-powered cameras for species conservation.
- Yayasan Ekosistem Lestari, Indonesia: develops a predictive platform that connects environmental degradation with disaster risk.
Why is using bioacoustics important, for example? Because a recording can reveal whether species are present or absent without researchers having to constantly walk through a forest. AI helps process large volumes of audio and find patterns that might otherwise go unnoticed.
More precise and resilient agriculture
Agriculture faces a complex combination of problems: pests, diseases, changing weather, soil degradation, and a lack of information for small-scale producers. In this category, the projects aim to bring advanced analysis to places and communities where resources are often limited.
- Edufarmers, Indonesia: provides small farmers with near-real-time guidance on pests, diseases, and weather through everyday messaging apps.
- Living Roots, Thailand: uses data and AI to design biological fertilizers adapted to the needs of each crop.
- SIGMA, Singapore: creates satellite-based AI models to estimate yields and strengthen climate resilience.
- Terrastack, India: combines satellite and agricultural data to deliver detailed information about smallholder plots.
- X-Centric, Australia: replaces traditional soil laboratories with portable AI-enabled X-ray equipment capable of providing immediate geochemical analyses.
Here we see one of AI’s most concrete applications: moving from general recommendations to specific decisions for each plot. Knowing what is happening in a particular field can help apply fertilizers more precisely, detect problems earlier, and reduce waste.
Carbon, cities, and climate solutions
The third area brings together initiatives focused on measuring, verifying, and scaling solutions related to carbon and climate adaptation. Measuring accurately is essential: without reliable data, a carbon credit or an environmental improvement can be difficult to verify.
- Archeda, Japan: uses satellite data to turn nature-based carbon credits into measurable assets with greater integrity.
- City Syntax Lab, Singapore: develops an AI platform with autonomous agents to optimize energy consumption and carbon emissions in urban districts.
- Climitra Carbon, India: uses
Geo-AI, or AI applied to geographic data, to verify the removal of invasive species and convert them into biochar. - Farmers for Forests, India: uses AI-powered drones to turn smallholder agroforestry into measurable carbon and biodiversity actions.
- Varaha Climate, India: helps small farmers obtain carbon credits by verifying regenerative agriculture and carbon removal through remote sensing and AI.
The technical side is especially relevant in these cases. Models must combine satellite images, field data, sensors, drones, and agricultural records. That information then needs validation processes to prevent exaggerated estimates or results that cannot be audited.
The models that will support the group
Google DeepMind says participants will be able to work with frontier models and tools from its AI ecosystem, including AnthroKrishi, ForestCast, AlphaEarth Foundations, SpeciesNet, and Perch.
Each model can support different tasks: agricultural analysis, ecosystem observation, interpretation of geospatial information, species identification, or wildlife sound processing. The value lies not only in the model, but in connecting it with local data and real needs.
Environmental AI is only useful when its predictions become verifiable decisions on the ground.
The challenge will be moving from prototype to implementation. A model may work in a controlled test and fail when it encounters sensor noise, incomplete images, changing weather, or insufficient data. That is why access to technical mentorship and computing infrastructure can be just as important as the algorithm.
A test for AI applied to the planet
The accelerator represents an interesting experiment: concentrating advanced models on regional problems that require local knowledge. Asia-Pacific is not facing a single environmental crisis, but a combination of biodiversity loss, pressure on crops, urban risks, and climate change.
Over the next three months, these 16 organizations will have to demonstrate how they develop, deploy, and scale their solutions. The most valuable outcome will not be a flashy demonstration, but tools that farmers, researchers, conservationists, and cities can use consistently.
AI does not replace the people who know a forest, a plot of land, or a community. It can expand their ability to observe, compare, and anticipate. And when that technology is designed around concrete problems, it stops looking like a futuristic promise and becomes useful infrastructure for protecting the planet.
Original source
https://blog.google/innovation-and-ai/models-and-research/google-deepmind/ai-planet-accelerator-apac