Gainforest’s Journey with ETH BiodivX at the XPRIZE Rainforest | Part 1: The Tech

Written by David Dao and Chona Billones

The Journey Starts

Almost four years ago in July 2020, the Rainforest XPRIZE Team reached out to us at Gainforest: 

“The work you are doing at [Gainforest] is what inspired us to reach back out to you. The team feels you would be a great fit for this competition and we encourage you to consider joining a team or forming your own.” 

And so our small non-profit decided to form the Gainforest Team, ultimately making it to the last 33 qualifying teams from almost 300 teams: https://www.xprize.org/prizes/rainforest/articles/xprize-rainforest-qualified-teams-announcement

While exploring the qualifying teams list, we realized that another team from Zurich has applied for the XPRIZE: BioDivExplorer from ETH. After a brief meeting, we realized that our both teams were complementing each other: BioDivExplorer was stunning in environmental robotics and eDNA, while Gainforest’s major focus was on artificial intelligence and local community empowerment.

With the approval of XPRIZE, we joined forces to co-lead the newly formed ETH BiodivX team and together we advanced to the last 16 semifinalists.

The Semifinals in Singapore

The semifinals in 2023 were the first in-field test of our ideas. Within a wild 72 hours, our team conducted 41 drone flights, produced 12,049,708 DNA sequences, while additionally receiving 12,402 API calls on our initial prototype of Taina (back then called ChatNaturally) from 339 citizen science submissions.

It proved that our team works well together - and that our citizen science-based approach can be scaled further for the finals. For the finals our team was ready to step up the involvement of IPLCs with the goal of setting a new standard.

Preparing for the XPRIZE Finals in Brazil

The partnership with Indigenous Peoples and local communities (IPLCs) is at the heart of our team’s work. Since August 2023, we have hosted several weeks of capacity building workshops, co-created the world’s first inclusive AI assistant for IPLC benefits sharing and founded an Indigenous-led Data Council, that is also a official team member of our ETH BiodivX team, giving them an executive voice on our team’s overall decisions. Below we will summarize our strategy and work and our commitment to benefits sharing.

Figure 1: Our team has collaborated with IPLCs through four co-design stages, including capacity building, Indigenous-led data collection, and a global hackathon, all facilitated by the AI Assistant Taina leading to inclusive and benefit-sharing AI algorithms and Insights. This process ensured trust, accurate data tracking, and benefit sharing for the communities involved.

Our team collaborated with Indigenous Peoples and Local Communities (IPLCs) in four co-design stages:

1. Capacity Building and Co-learning: This initial phase focused on building trust and understanding the needs and requirements of IPLCs.

2. Indigenous-led Data Collection: Guided by the Indigenous Data Council, chaired by Aldeia Inhaâ-Bé, we facilitated a process where critical data was gathered by the communities themselves. Using Taina, we ensured that data providers were immutably tracked, securing benefit sharing for the future.

3. Global Hackathon: In collaboration with the Indigenous Data Council, we organized a global hackathon that developed AI algorithms and provided valuable insights.

4. AI algorithms & Insights: Insights derived from IPLC data are tracked and we reward data contributors.

Taina: An Inclusive AI Assistant for IPLC Data Sovereignty and Benefits Sharing

Illustration 1: (a) Taina used by UFAM students and lecturers to help support in plant knowledge collection (b) Vanda Witoto and Parques das Tribos testing Taina during the Capacity Building and Co-Learning workshop

Note: Taina can be accessed here: https://telegram.me/tainagainforestbot

Indigenous Peoples and Local Communities (IPLC) possess crucial knowledge for environmental decision-making. However, through our co-learning workshops, we learned that sharing this knowledge faces two major challenges:

1. Lack of inclusive technology: Current systems often fail to accommodate traditional knowledge-sharing methods like storytelling.

2. Trust and ownership concerns: Fear of data exploitation and unequal benefits sharing (data colonialism) discourages knowledge sharing.

Thanks to the workshops, we were able to co-create and address these challenges through an innovative AI assistant that:

1. Supports multiple languages (English, Portuguese, Spanish, Bahasa, and Swahili) to facilitate inclusive communication.

2. Incorporates a data provenance layer to ensure transparency and ownership of shared knowledge.

3. Utilizes multimodal capabilities (text, image, and speech recognition) to adapt to diverse communication styles.

4. Is specifically designed to engage respectfully with environmental knowledge from Indigenous and local sources.

5. Runs on Telegram, a popular and robust chat application in Brazil, avoiding the need to download additional apps on the phone

This approach aims to bridge the gap between traditional knowledge systems and modern decision-making processes while safeguarding the rights and interests of IPLCs.

Immutable Data Provenance through Web3

Taina maintains the ability to track individual conversations (referred to as threads) and messages through unique identifiers associated with each user. Each thread_id is always linked to a user_id (or a unique public Celo address), which allows conversations to be attributed to a specific user if they have given consent (consent set to True). Additionally, Taina tracks all documents utilized due to RAG and can monitor the citations. All links are immutably attested through Celo attestation, ensuring that the knowledge is publicly linked to a unique address.

Benefits Sharing & Last Mile Money

Note: Benefits sharing can be monetary (financial reward) and non-monetary (knowledge sharing, team ownership). Both are critical for our team.

Many IPLCs lack access to financial services. To address this, we leverage our learnings during the co-design workshops that there is a widespread availability of the internet and mobile phones within local communities. The Celo blockchain, a Layer 1 Blockchain currently transitioning to Layer 2, enables anyone globally to transfer stablecoins such as cUSD or USDC (pegged to the USD) to any mobile address worldwide. All transactions are immutably stored in a decentralized ledger. Additionally, Celo uses energy-efficient algorithms, making the blockchain carbon-negative and requiring only a minimal transaction fee (less than $0.001) per transaction. With robust capabilities to convert received cryptocurrencies into local currencies, Celo is an ideal choice for rewarding our IPLC communities.

The Indigenous Data Council has approved all pricing for data collected for use during the XPRIZE finals competition.

Taina’s role to empower IPLC voices for decision-making

Illustration 2: Team members David Dao and Kamila Camilo launching the Taina-based AI Assistant to bring the perspectives of IPLCs to policy makers at COP28 in Dubai.

One of Taina’s major goals is to make it easier for local and global decision-makers to incorporate IPLCs perspectives into environmental policy. Thus, during COP28, we announced the release of a Taina-based AI assistant that is being utilized by 125 negotiators from 54 countries supporting negotiators in understanding and incorporating Indigenous Peoples and Local Communities (IPLC) positions into decision-making processes. This year, in 2024, Taina will continue to play a critical role at the United Nations Framework Convention on Climate Change (UNFCCC) COP29 in Baku and United Nations Convention to Combat Desertification (UNFCCD) COP16 in Riyadh due to our strong partnerships with the Youth Negotiators Network and Oyá Institute.