Who are we?
We are Lucidminds, a research collective and resident member of Sustainable Development Goals House in Amsterdam. Founded in 2018, our goal is to create a community of like-minded experts to convene around imperative ideas with strong social and planetary impact potential. We bring together systems thinking, and state-of-the-art research, to help weave the threads of technological evolution toward actionable real-world solutions.
To maintain our independence, our creative freedom, and to never stray from our guiding principles, we support our work through publicly-funded research grants, and collaborations with other organisations which focus on evidence-based, long-term, reusable, and often open-sourced solutions.
What is GUS?
Green Unified Scenarios, or GUS, is an agent-based simulation system centred around trees as the primary agent. Our peer-reviewed model forecasts the growth potential of various species, projecting carbon storage and release capacities under various climate conditions, hypothetical maintenance models, and different species configurations. We then use these forecasts to estimate downstream impacts such as the reduction in airborne pollutants, rainwater capture & run-off, and even biodiversity indices.
Our goal was to build the basis for a rapid scenario analysis framework that could give city-officials and arborists a tangible understanding of the forests they manage. We try to show the exponential benefits of maintenance activities that are simple and cheap to perform now, and which have dramatic impacts on the long-term health of the trees, and in turn, our planet.
Furthermore, by modelling individual tree-agents, and their influence on their direct neighbours, and simplifying the complex models of their growth, we strike a balance of speed and accuracy which allows rapid comparisons between many hypothetical scenarios, across sprawling cityscapes.
These tools were enticing for many who knew the difficulty in funding public tree-plantings, and more radical management strategies for urban-forests, which often bumped up against private development and real-estate interests, or tightened public-funding pools. Despite the success of our collaborations, we envisioned a way to bring this powerful simulation framework more directly to citizens, and this became known as gAIa.
What is gAIa?
gAIa is an interactive map of living trees, focused on ground-truth, real-world data. Using this we could expose the layered simulation forecasts for their growth, and weave in soil and weather information, as well as a knowledge-graph of semantically linked ecosystem data, and even a novel memory-persona system that can give each tree the potential to reflect its unique history and place in the world.
We've built gAIa as a set of agentic AIs, each one with access to a rich and expanding set of tools which enable them to perform data-analysis, fetch up-to-date information from 3rd party data-sources, assess images, and provide informed insights into many ecosystem topics.
Interactions with gAIa come at three main levels:
- Individual tree
- Locally-connected "mini-forests"
- gAIa, which represents nature and ecosystems as a whole
Throughout gAIa we employ the Reasoning and Action (ReAct) agent-framework, which allows the agents to adapt to any user request, base its responses in carefully collected data from multiple sources, and allows it to speak confidently, without hallucination, and to ground its answers in verifiable references. At all levels we provide a baseline set of analysis tools to the agents, and in each case, we can provide a perspective-specific subset of additional tools that give each agent its unique viewpoint.
For individual trees, these tools can include assistance for measuring, or caring for a given tree and their specific status. We also have the chance to give each tree a persona, based on its unique amalgamation of properties, such as location, health status, species, size, immediate neighbours, and even historic records.
Meanwhile, mini-forests represent a collection of trees whose greatest inter-canopy distance falls within a given threshold. These represent groups of trees that interact directly, through shading, resource-sharing, and likely an overlapping population of insect and animal inhabitants. At this level, analysis tools can investigate biodiversity potential in the group, and provide more details on inter-species interactions.
Finally, at the gAIa level, a grand perspective is applied, as the entire ecosystem can be examined — in either a general sense, or in context of the speaker's location. Furthermore, hypothetical scenarios can be examined, as the gAIa agent is also equipped to hypothesise on the impacts of potential new plantings, and analyse the ecosystem as a whole.
Knowledge Graph
gAIa is much more than a chat-bot. The agentic structure, and ReActive framework allows us to arm the system with tools to fetch and process data from internal and external sources, and reflect on these dynamic inputs in the context of user-queries.
One of the most interesting toolsets is our semantic knowledge-graph focused on natural ecosystems, and everything relating to them. This interconnected web of concepts and relationships is designed to give machines a means of relating the concepts and entities in our biosphere, and it also acts as an index that can connect these entities to extra context and references.
To get a better understanding, you can read our dedicated description of the knowledge-graph here.
Tree Memory & Personas
Armed with data about each tree, and some information about each species and the climatic context and time of the request, we can adapt the persona of the agents in gAIa. But as we shared the beta version in more places, we realised gAIa could also act as a memory bank for the subtle inter-organism relationships that citizens have with the trees around which they live. This led us to develop an ever-evolving memory system around the trees, which can give more iconic or historically significant trees a way to reference their impact.
To learn more about this innovative system in gAIa, you can read this deep-dive here.
Real-time Alerts
gAIa also has a system for integrating many different data-sources, and translating them into calls to action for everyday citizens. We can integrate weather warnings, sensor networks, data-gaps, and many other potential sources, to invert the traditional chat-bot model, and give a mechanism for initiating conversations with users in the cases where trees may be silently struggling.
Where do we go from here?
Better tools for broad new tree-surveying projects — These will allow easier onboarding of new cities and regions, with tools for individuals, or for city councils to manage and monitor their tree populations.
Gamification — A simple, social aspect on top of our robust tree inventory systems, we want to highlight and reward those users most involved in exposing their city's rich tree-populations. And help those dedicated individuals to promote better maintenance practices, and healthier urban-ecosystems.