šŸŒ± Memo #1 Nature and AI

What does the AI and Nature market look like? Who are the start-ups leading the way? What is the funding landscape? What are the challenges to be overcome? ... Find out all this and more in our latest deep dive

Our Deep Dives are back. In this edition you can expect to learn:

  • šŸŒ³ Why AI can support positive Nature outcomes? 

  • šŸšœ Who are the start-ups leading the way?

  • šŸ’° What is the funding landscape?

  • šŸ’Ŗ What are the challenges to be overcome?

  • šŸ’”ā€¦ As well as some hot takes from experts in the field

Readers Beware: Weā€™ve spent the past couple of months huddling up with founders, academics and investors operating at the intersection of Nature and AI to capture a perfect snapshot, but things are changing quickly. Weā€™re sure we wonā€™t have captured the full picture here. Please do call us out to update this pictureā€¦ In the spirit of speedily but steadily bringing this rapidly evolving AI X Nature space online together!

 šŸŒ³ Why AI for Nature?

Climate Change and Biodiversity Loss exacerbate each other. Simply put, rising temperatures will worsen biodiversity loss, releasing more greenhouse gases, creating a negative feedback loop. This will inevitably aggravate a range of societal issues, from civil conflict to resource competition, in turn causing more biodiversity and climate impacts. Conversely, improved ecosystem condition is likely to help store and sequester carbon, support water and food security, and provide greater resilience to climate change impacts.

There is a growing desire to understand and quantify the impact and dependencies that organisations have on Biodiversity. This desire is driven by a growing volume of framework and policies (e.g. the Kunming-Montreal Global Biodiversity Framework or the TNFD); a growing body of Investors seeking to better allocate and engage funds based on material nature-related issues; and a growing number of forward-thinking organisations seeking to better understand their impact and dependencies on biodiversity.

As the saying goes, ā€œwhat gets measured gets managedā€...AI as a tool is uniquely positioned to manage these complex issues. Due to its capacity to gather, complete, and interpret large, complex datasets on biodiversity and ecosystems, it can be used to support stakeholders in taking a more informed approach to combating biodiversity loss. At the same time AI poses its own environmental and ethical challenges.

Innovators are wrangling with two big questions:

  • How can we use AI to protect Nature (and who is leading the way)?

  • How can we prevent AI from exacerbating harm to humanity, ecosystems, and communities reliant on ecological preservation?

šŸšœ Question 1) How can we use AI to protect Nature (and who is leading the way)?

šŸ˜ But first a Biodiversity Breakoutā€¦

Biodiversity is the variability among living organisms and ecosystems. Ecosystems can be thought of as complex machines, like mechanical watches, where all the parts ā€“ species ā€“ function together and cannot function apart. Intact ecosystems have been finely tuned to optimise the energy within that system while still being able to respond to external shocks. The complexity of ecosystems makes it pretty tricky to understand the impact companies are having on them for a few reasons:

- Specificity: Biodiversity is highly site-specific - each site has a substantial number of unique species that pass in and out of it at different times of year, and often these species can be pretty challenging to spot (Maybe they are very small, very rare, or only come out at night). As a result, very detailed data on ecosystem and biodiversity and the company operations are required in some form to understand their interaction and trends over long time frames (10 years +).

- Uniqueness: ā€˜Commercialā€™ impacts are extremely varied themselves. No two oil spills are the same. Assessment methods need to be tailored to each activity and each site sensitivity, to be able to identify and accurately assign impact.

- Complexity: Impacts can cascade through ecosystems and are often technically difficult to understand. E.g One damaged ā€˜cogā€™ can damage other components: removing 1 km2 of a 100 km2 rainforest or 1 out of every 100 species does not necessarily translate to a 1% biodiversity loss; over time, if a critical component, it could negatively alter the ecology and biodiversity of the entire forest (100%)

Artificial Intelligence canā€™t unpick these three areas alone and AI algorithms are only as good as the data that feeds them ā€¦ ā€œRubbish in, Rubbish Outā€. However, when fed with reliable and accurate data AI can vastly improve corporates ability to move the needle on Biodiversity.

Letā€™s find out howā€¦

1) Locate Interactions with Biodiversity & Ecosystems. The first step after any feel-good corporate biodiversity announcement is to actually understand where their company operates. This may sound simple, but supply chains get pretty complex pretty quickly. Imagine trying to locate the factory of your supplierā€™s, supplierā€™s, supplier. Startups are using AI to recognise and classify facilities from large datasets of images to automate mapping of supply chains (i.e. finding a soy processing facility within thousands of km of rainforest).

- Ecopia - Converts geospatial imagery into accurate, and up-to-date maps of ā€˜taggedā€™ buildings (E.g. use cases such as identifying supply chain assets in areas associated with high risk of deforestation)

- Global FieldID - Maps boundaries of fields and assigns unique IDs to each plot of land - like a barcode - which can help corporates pinpoint where raw materials are coming from to a field level.

2) Evaluate Impacts and Drivers of Biodiversity Loss. After organisations have understood the footprint of their organisations they can begin to prioritise locations based on:

  1. How important the ecosystems are for biodiversity (e.g. high biodiversity, Areas of rapid decline in ecosystem integrity)

  2. How likely the organisation is to have significant dependencies and/or impact on the ecosystem (E.g. Areas of high water stress, Areas of significant resource extraction related to key products)

Historically, corporates have relied on in-house teams or consultancies to do this. Teams would do long field-studies or use pre-canned data sets to investigate factors like ecosystem integrity, species abundance and water stress = Extremely labour intensive . A growing body of AI solutions are increasing the efficiency and transparency of these process and making insights directly available to corporates.

2a) Understanding ecosystems using automated species identification: Biodiversity data is collected from optical, acoustic or thermal sensors deployed in the field or mounted on drone/satellites. AI can then be used to sift through this data to identify species, and estimate various measures of ecosystem health. A major benefit of using AI is the increased efficiency and accuracy of data processing compared to manual analysis.

Examples include: Acoustic & Visuals - Pivotal ; Hyperspectral - Gentian.io.

Two growing bodies of Species Identification Data sources are:

- eDNA: NatureMetrics eDNA - eDNA analysis uses a non-invasive genetic technique to monitor species presence/absence and distribution. eDNA consists of small fragments of genetic material left in the environment by organisms.

- Citizen Science: INaturalist - Millions of observations created and identified by members of the public are shared with the Global Biodiversity Information Facility where they are used to advance scientific understanding of biodiversity.

2b) Understanding Impact and Dependencies: Organisations must also understand the impacts their operations are having on ecosystems (See Image below). It is also critical for organisations to understand the dependencies they have on those ecosystems and the risk that their degradation poses on day-to-day business.

- Data aggregator: Nala, Aggregate data and tools to empower companies to identify biodiversity risks and dependencies to enhance their interaction with the natural world

Example impacts and Dependencies monitoring using AI

3) Mitigate & Manage: Having understood the scope and scale of their impact on biodiversity, organisations are then positioned to do something about it.

3A) Deploying AI to increase system efficiency. We wonā€™t spend much time here but there are thousands of innovations in this space - where AI can generate efficiencies that can reduce pressure on ecosystems IF (and only if) they lead to lower resource use. If the efficiency gains are used to produce more goods, there is likely no net benefit for biodiversity. Examples include ā€¦

- Reduced Inputs: Trinity AgTech,  AI-driven biodiversity analysis and optimisation to encourage regenerative farming.

- More efficient extraction: KoBold Metals, uses AI to improve predictions about where deposits of mining material are located, thereby reducing the amount of ground that needs to be disturbed to find new ore bodies

3B) Support Land Use Management and Restoration: AI can support land managers to understand what opportunities for biodiversity restoration exist on their land, and if active restoration such as tree planting is taking place, what species would be best given local conditions.

- Monitoring Land Use: Dendra Systems

- Preventing Bad actors: Ocean Mind help combat illegal fishing by using satellite imaging and artificial intelligence for rapid identification of fishing activities in marine protected areas.

- Optimising policy: CAPTAIN (Conservation Area Prioritisation Through Artificial Intelligence): integrates biodiversity data, conservation budgets, and data on human impacts to suggest optimised areas for biodiversity protection

3C) Finance: Once corporates have completed the hard work of measurement and mitigation, their is a growing desire to finance biodiversity credits to plug the remaining gap. AI is being used to provide independent and high-resolution data on positive biodiversity impacts. This data can be utilised to monitor and track project implementation and the validity of credits. Stay tuned for an upcoming articleā€¦

- Pivotal: building detailed set of annotated, species-level biodiversity data ever compiled ā€“ while providing measured, quality-controlled analytics to show evidence of any biodiversity gains on the ground.

4) Monitor and Disclose: Pure-play Measure Report Verify (MRV) startups are also leveraging new remote sensing technologies to drive cheaper, more accurate measurement, reporting, and verification of the impact of specific activities taken by corporates to improve Biodiversity. AI technologies are often leveraged here for the synthesis and analysis of the large data-sets generated from these activities.

- Pivotal: building detailed set of annotated, species-level biodiversity data ever compiled ā€“ while providing measured, quality-controlled analytics to show evidence of any biodiversity gains on the ground.

- Sfeeri - AI-enhanced software solution for corporate biodiversity impact assessment, action and reporting

šŸ’° Funding Landscape

This report includes data extracted from the Net0 Platform regarding capital raised in 2022 & 2023 YTD by startups operating at the intersection of AI and Nature. The following funding round types are considered out of scope in this report: acquisition, IPO, post IPO, merger, PIPE, secondary transaction, and SPAC

  • Resilient: Nature raised $0.35B in 2022 and is on track for similar in 2023. Bumper deals include:

    • BeeWise - Developer of autonomous beehives intended to increase yield, reduce colony loss and eliminate the use of chemical pesticides ($80M Series C)

    • Pano AI - utiliSes a combination of advanced technologies such as AI, cameras, satellites, 5G, and cloud software to swiftly detect and pinpoint wildfires, enabling fire agencies to take necessary actions and prevent their spread. ($17M Series A)

  • Early-Stage: Pre-Seed and Seed funding dominate in 2023 representing 90% of all deals

  • Small tickets: Total of 73 deals over the last two years with an average deal size of $5.64M

šŸ’Ŗ Question 2: How can we prevent AI from exacerbating harm to humanity, ecosystems, and communities reliant on ecological preservation?

AI also has a series of weaknesses it is worth understandingā€¦

  • Emissions and Ecosystem Impact: Systems also have a large carbon footprint. Researchers have found the process of training a common large AI model can emit as much as 626,000 pounds of carbon dioxide equivalent, five times the emissions of a car over its lifetime. Ongoing emissions of Hardware (See work of Loic Lannelongue) also must be considered.

  • Rubbish In, Rubbish Out: AI models find patterns in the data they are given, so if data is skewed or biased towards certain contexts (for example, biological data can have regional, species, habitat, geographic, threat or language biases) they may fail to extrapolate or find only spurious correlations.

  • Tackling a Digital Divide: Powerful nations and companies, many of which are in high-income countries, lead in the development of AI systems, including those applied to nature and biodiversity in low- and lower-middle-income countries. A failure to consult local communities in the deployment of AI systems that will affect their environments and livelihoods risks reinforcing existing social inequalities, neglecting community needs and interests, and worsening the digital divide.Those that benefit least from AI will also generally bare the brunt of the most sever environmental impacts from climate change.

  • Training Capacity: Say youā€™re trying to train an AI model that will process hundreds of thousands of photos from camera traps to identify the species. First you need a team of experts to ā€˜teachā€™ the model that that picture of extremely rare species X is actually X. As you can imagine there are a limited folk able to do this, even then three PHD ecologists may come up with different answers for what the species is. There is therefore a challenge both in capacity of AI X Biodiversity models, but also in their accuracy.

 šŸ’”Hear it from the experts

Do you want to know how AI affects other aspects of the sustainable transformation, including education, social equity or the world of work, while shifting power and politics on environmental action at the same time? Check out ā€˜Fast Forward: How to Harness the Power of AI for Societal Progress and a Sustainable Futureā€™, a new book by Alice Schmidt, Claudia Winkler, Flo SchĆ¼tz and Jeroen Dobbelaere to learn about risks, opportunities and why all of us need to get involved.

ā€œAI bears great potential for protecting and restoring nature. But we mustnā€™t leave AI solely to techies. AI is much more about people and the environment, than it is about technology per se.If we are serious about protecting and restoring nature, people from all relevant disciplines must get involved, ensuring that the impact of AI-supported tools is measured and managed.ā€

Alice Schmidt - author and adviser on sustainability & AI (www.aliceschmidt.at)

A final noteā€¦

To this end, there is a great need to merge the digital rights movement with the environmental and climate justice movement. We must remember that AI carries risks as well as opportunities. Nor is it a silver bullet to solve the environmental crisis. But if developed and used responsibly and inclusively, AI can be a valuable aid in protecting our planet's natural wonders and promoting better practices for the well-being of all species

Written by Colin, Ollie, and Patrick - Drop us a message!

Special thanks to Federico - Net Zero Insights, Alice Schmidt, and Cameron Frayling - Pivotal for their insights.

Reply

or to participate.