Gen AI is rocket fuel for Earth’s digital twin

This month I had the privilege of delivering a keynote speech to open the Earth Digital Twin Workshop organised by JAXA (Japan’s national space agency) as part of the activities of the Consortium for Earth Observation (CONSEO).  My remarks below focus on the role that Generative AI will play in Earth Observation and Digital Twins using satellite data.

Thank you for the privilege of speaking with you today.  Thank you also for allowing me to speak in English.

2024 is the year that Generative Artificial Intelligence, or Gen AI as we’ve come to call it, gets applied at scale to real-world problems including through Earth Observation.  Where AI was a powerful way of supporting human decisions, Gen AI can go further and create content in its own right for consumption by humans.

Working with the World Economic Forum, Deloitte recently wrote about “The Catalytic Potential of Artificial Intelligence for Earth Observation”.  I encourage you to read this report, some of which I have leveraged in my comments this afternoon.

Everyone here today appreciates the scale of opportunity presented by the global space economy.  I have the privilege of leading Deloitte Consulting across Asia Pacific and I’m excited by Japan’s role in shaping this future.  Not just for Japan, but for the entire region as we elevate the understanding and opportunity of space in general, and Earth Observation in particular, for businesses and governments everywhere.

My professional background is in the handling of data and the information economy, something I have written and spoken about for more than thirty years.  Nearly two decades ago, Clive Humby described data as “the new oil”.  What he meant was that data needed to be refined to be useful.

Today, I believe that Gen AI means that data is the new rocket fuel!  It has explosive opportunity, but it is energy intensive and requires immense engineering.  Where better to apply this rocket fuel than in the space sector!

In an era of digitisation of the physical world, it’s hard to say, “it’s not rocket science”, because it is! To make our world work better, millions (and eventually billions) of devices must be integrated.  We call this the fourth industrial revolution with space playing an essential role and Gen AI is the rocket fuel to power this rocket science!

Some of the applications of the fourth industrial revolution, powered by Gen AI and space-based platforms, that matter most across the Asia Pacific region include: building and maintaining massive infrastructure, accessing deep and difficult natural resources, navigating natural disasters, managing complex supply chains and, most importantly for all of us, creating sustainable agriculture through an agtech revolution.

Nature created swarms of insects and birds that have impact at the scale we need to support the natural ecosystem. It is likely that we will mimic nature and have swarms of micro-robots do the same for our industrial infrastructure including cities, roads, mines, and farms of the future.

This vision cannot be realised from sea level or even from within our atmosphere.  In fact, this vision of the future is part of a wider trend where everything operates as a network and infrastructure is inherently smart.  Smart cities, roads, farms, mines and more.

To-date, we’ve thought of Earth Observation as being an information source informing decisions by humans down here on Earth.  The fourth industrial revolution will push us to find ways of processing this data on-the-fly to support real-time, autonomous decision making by these swarms of devices supporting our smart infrastructure.

This real-time demand will continue to be challenged by the need to process the massive volume of data that Earth Observation already transmits to the surface.  But even that is only a fraction of the data that is collected from space, some estimates suggest that less than 5% of available data is ever transmitted to Earth.  Edge computing on board satellites will become increasingly important as AI moves from data centres to much smaller standalone platforms closer to their data sources.

Today most AI, particularly Gen AI, is processed centrally with data that has been consolidated from many sources.  However, developers are starting to invert this model and moving AI to the data rather than the other way around.  This “edge computing” approach reduces the need to transmit large amounts of data and also increases the privacy and security of often sensitive datasets.

Increasingly they will also be thinking about reducing energy consumption and even moving the data to where energy can be most sustainably sourced.  We are now realising that humanity’s ambitions for AI cannot be powered by available sources of energy without blowing our carbon budget.  While not yet the main driver, there is no doubt that energy sourced from the sun in space takes some of the pressure off the increasing energy and carbon footprint of AI on Earth.

But as excited as I am by these opportunities, I know that the real revolution will come through finding uses for Earth Observation data that we have not yet even imagined.  That is where the business, government, engineering, and science communities need to come together.  Business and government can find many of the market and social opportunities, but only if they are enabled by those of you that understand what is possible.  We are relying on you to describe the powerful combination of Earth Observation, space platforms and Gen AI in terms that business and government can understand.

It is also where initiatives such as the Gravity Challenge comes in, which we, Deloitte, have run since 2019 in partnership with local space agencies and partners such as Amazon.  It started here in Asia Pacific, and it has now gone global.  Its purpose is to encourage the evolution of business models, many or even most of which have been centred around Earth Observation opportunities.

One of the most immediate and exciting applications that is emerging through the combination of Gen AI and Earth Observation data is the creation of much more sophisticated Digital Twins and immersive models of the Earth by superimposing many disparate datasets.  Not only can Gen AI project two-dimensional data into three or even more dimensions with much greater accuracy, but it can also use time-based data to identify and model the levers, stocks and flows that drive the systems themselves from rivers and forests to cities and factories.

But all of this will not be possible without an enormous, coordinated, effort to further standardise our datasets, our AI models, and our Digital Twins.  This seemingly obvious requirement turns out to be really hard.  Consistent taxonomies and AI models are hard enough when we know the parameters of the data and solutions, such as for business reporting.  Doing it when the very context of the opportunity is still emerging is incredibly hard.  But we must do this otherwise it will be impossible to scale our solutions despite the power of the technology we have our disposal.

As confident as I am we will succeed, I know this whole effort is not without its dangers.  Today we worry about deep fake videos of politicians and actors.  But deep fake Earth Observation data could have even more sinister implications.  It could affect infrastructure; agriculture; stock markets, geopolitics; and it could degrade government decisions in the face of natural disasters.

Gen AI is becoming increasingly important to cyber security which will have to be much more extensively embedded in our space assets to reduce the opportunities for bad actors to hide malicious manipulation of the data downstream.

Of course, bad actors are not the only threat.  There are also ethical risks, some of which are already apparent, but many are yet to emerge.  Not least is our individual right to privacy and to control how our data is used.

Today our homes, factories, farms and even our vehicles can be watched from space.  But how that data is applied to our insurance policies, how it is shared with the companies that sell to us and how data collected now may be used in the future need to be considered within a data governance and ethics framework.

However, none of this should put us off from pursuing the opportunity to combine Earth Observation with increasingly digital terrestrial activities via Gen AI.  The potential payoff is massive.

We will be able to use the power of Gen AI and Earth Observation data to plug methane leaks earlier through better monitoring of pipelines.  We will be able to use Digital Twins to improve the reliability of our energy grid.  We will optimise the movement of goods through smarter supply chains.  We will better support law enforcement to intercept smugglers.  We will have an earlier view of the surpluses and shortfalls of agricultural output allowing us to manage supply and demand much more effectively.  And these are just some of the opportunities!  When fuelled by Gen AI, Earth Observation resources are both richer and more readily applied.

But, as I said, most business leaders, politicians and other stakeholders that need these solutions don’t have the level of knowledge to know what to ask for.  That makes all of you critical to realise the opportunity.  While you may be the ones to find the business use cases, your most important role could be to translate the needs of society into a form that can readily use this technology.

This isn’t a one-off process of discovery, it will be iterative, creative and, I think, enormously rewarding!  The solutions that come out of this work will realise the opportunities we know about, but also serve us in ways that we haven’t even begun to imagine across environmental, health, social, educational, commercial, and so many more needs that we have.

As you look to play this role to shape an exciting future, please, make sure that you do it in a way that puts ethics are the core.  Please do it in a way that builds on common data taxonomies.  Please encourage collaborative investment in common AI models.  Please be mindful of the environmental impact of increasing energy consumption.  And please encourage the investment in the next generation of shared technology that we can all build on!

If we do all this, together we can create an exciting future powered by the rocket fuel that is Earth Observation data!