The Maturing Technosphere

Google announces Earth AI, a Geospatial Reasoning model that represents a significant advance in planetary sapience

Google has published an important paper on its recently launched EarthAI platform, a collection of AI models that integrate Planet-scale Imagery, Population, and Environment via its Gemini-powered reasoning engine. Google calls this a “Geospatial Reasoning” engine. It says that, “the ultimate goal of Earth AI is to help users answer complex, real-world questions that require multifaceted reasoning across diverse models and data sources.” Such queries can be categorized into a hierarchy of increasing complexity, as in these examples from the paper:

1. Descriptive and retrieval queries involving fact-finding (e.g., “What was the highest recorded temperature in New York in August 2020?”).

2. Analytical and relational queries seeking to uncover patterns between different data sources (e.g., “How many hospitals were located in areas experiencing severe storm conditions in the state of Louisiana when Hurricane Katrina came ashore?”).

3. Predictive or inferential queries involving forecasting new information (e.g., “Which Indian cities have the most vulnerable populations at high risk of being impacted by flooding by November 25, 2027?”).

This is an extremely exciting example of the rapid advance of what we have called Planetary Sapience — that is, the planet’s ability to understand in ever greater detail, via its growing smart exoskeleton of sensors, networks and computers, its own hyper-complex functioning. As such, Earth AI also represents an important milestone in the advance toward a mature technosphere, that is, one in which there is a stable homeostatic rather than disequilibrated or disruptive relationship between the technosphere and the biosphere.

But I want to sound one of note of caution: if you read this paper carefully, you will find that a great deal of the data that is an input into these models comes from US government sources, whether that be climate and weather data from NOAA, health data from the CDC, population data from the Census, etc. The reliability of that data is at grave risk today. And all the compute in the world won’t be worth a dime if the data inputs are corrupted.