Manuel Gerold is the COO and co-founder of askEarth, a Zurich-based start-up that is building an AI-powered and user-friendly search engine for satellite images. For Manuel, space is not just about embracing the final frontier for future generations; it’s also about improving life on Earth for the current population. Everyone should have the right to study their home planet from above – within limits.
In 2022, the four co-founders of askEarth received a gift. At that point, Manuel Gerold, Gaetan Petit, Simon Grüning, and David Berger had been working for over a year to bring their dream to life. The founding team wanted to make satellite image data more accessible to the general public, and this meant they would need a user interface that is as close to human language and logic as possible. Creating a natural language processing interface was a necessary, but daunting part of the project. And then, suddenly, everyone and their grandmother were using generative AI tools practically overnight.
The idea behind askEarth is to have a search engine that queries the petabytes of freely available satellite image data. The new Large Language Models (LLM) step in to power the user interface; the bulk of askEarth’s data comes from the EC-run Copernicus program and in the future potentially also from private constellations like Airbus, Maxar or Planet Labs. Between these layers is the actual search engine.
askEarth is preparing for a commercial launch after a well-received Minimally Viable Product, and Manuel says the company is still researching and testing multiple LLMs. While they will eventually need to focus their training efforts on one model, the team is open to customers’ preferences.
How does askEarth handle the oh-so-common AI hallucinations? Manuel explains that the company limits the LLM to the outer layer: AI is a translator between human language and computer language. Producing absolutely correct results from reliable data is just as important as having an intuitive user interface, which is why the two need to be clearly separated.
“We want to create awareness”
While satellite data is freely available, Manuel points out, it’s not really accessible. The data provided requires a lot of specialized post-processing before a non-expert can understand it. Such data is useless for untrained individuals and costly for companies: you need to hire and train a data scientist to gain any insights.
What are the insights that companies could gain from access to satellite data, then? askEarth is targeting verticals such as agriculture, insurance, real estate, and logistics. These are all industries that have an interest in both up-to-date images and longer timelines featuring large geographical areas across the globe. Another use case is journalism: satellite images are a great tool for news reporters to verify their facts. You can, for example, see New York hidden behind a smoke cloud as a result of the Canadian wildfires, or witness how much glaciers have melted during the recent years.
askEarth wants to make basic exploration with its tool free for anyone. Premium features, such as masking certain areas or detecting changes over time, would then bring in the revenue needed to keep the service running. The company’s main objective is to empower, educate, and create awareness, so askEarth doesn't want to define the use cases too narrowly.
There are some ethical considerations, though. One horror scenario combines facial recognition software with high-resolution satellite images, and suddenly everyone’s location on Earth becomes public record. This is not possible with the current data, Manuel assures. The images askEarth uses have a resolution of 10 meters per pixel; this means you can recognize a large ship, but even a car is already too small to detect.
There is currently very little regulation on how satellite images can be used, Manuel admits. This will need to change in the future as data becomes more accurate and more available. The discussion is already happening, but regulating satellite imagery is an international venture, and those take time. Manuel believes that the eventual regulation will focus on limiting the allowed resolution, as that is a technically easy solution.
The askEarth platform is not meant to be a spy tool, Manuel stresses. Although such negative implications are not yet a reality, Manuel says these ethical discussions are part of the team’s daily work. After all, we should be prepared instead of waiting for the technology to catch up and the problems to emerge.
“Ten years from now, we will have more time for things that excite us”
Manuel’s and askEarth’s roots are in Space Innovation (formerly the Swiss Space Center), Space4Impact, ARIS and the Swiss space community in general. Space is special to Manuel: when other industries try to make the most revenue with the lowest cost, the space community tries to do the most unbelievable thing in the shortest possible time. Space is not about economy, it’s about passion, Manuel summarizes.
Space missions can be roughly divided into two categories: space exploration and Earth observation. Exploration missions involve sending probes to strange planets and far corners of the universe and providing future generations with information decades from now – which is definitely inspiring. However, Manuel is drawn to Earth observation, because it feels so relevant: his work has an impact on the current society, on ourselves and the people close to us.
When asked what the world will look like ten years from now, Manuel is optimistic. The way we interact with data will change drastically already within the next couple of years. In the longer term, new AI tools will force us to change our education system – which is something we haven’t really done in decades. Furthermore, we will get to automate tedious work, such as writing emails or creating schedules. As a result, we will have more time for things that excite us.
This article is partially based on this podcast episode.
Comentários