Space Futures and Generative AI for Space Applications
Earlier this month I was invited to speak at KCL Space's inaugural X3 Space Conference at the Strand Campus, London. My talk was on Space Futures and Generative AI for Space Applications. My approach was to design and deliver a new talk that set the scene with an overview of the technologies and innovations that will drive our space future as well as describe how the technology will be applied across the space sector. In the second part of my talk, I wanted to showcase Generative AI for Space Applications, here I debuted a collection of images I ‘prompted’ as well as a live demonstration of ChatGPT. Later in this post, I will go on to showcase sneak peeks of some of the images I created (prompted) by utilising Midjourney, a text-to-image Generative AI product to synthesise images of lunar habitats. In doing so, the output generated harked back to images from Star Wars and images released by the European Space Agency (ESA) over ten years ago! I’ll dive more into this later as there are ethical considerations as well as other tangential issues to be aware of and be aware of when using Generative AI models/products. I have been thinking about the ethics of space technology and its application in space, on the Moon, Mars, and beyond. It forms a big part of the ethos of the work I do at Naaut. When I get the opportunity to speak at events like KCL Space it’s vital to me that as an actor in the industry, I share the ethical values, considerations as well as implications of technology. We have the huge privilege of taking bold first steps but it would be a totally wasted opportunity if we didn’t think about the wider impact to ensure that access to space remains open and accessible to all humankind.
If you are new here, let me quickly introduce myself: I am a sense maker, connector, and horizon scanner. I am curious about technology and how it’s applied across society, and in establishing Naaut my focus has been on how technology is applied across the space-technology ecosystem. Increasingly, I am questioning the ethical application technology and the barriers, shortcuts, and advantages that might increase traction in one group of early adopters versus those that might not have access to or awareness of such technology. Space is, after all, the equalizer, and technology is the enabler! My ethos is that Space is for all.
I had a great time speaking at the X3 Space Conference particularly, I loved the question and answer session where there was some great discussion about the use of ChatGPT in academic environments and how students are using this technology to proactively learn themselves as well as generate reports. There is no protocol in place as yet but having a robust discussion and debate was the goal of my talk, I really enjoyed the conversations with the KCL Space community following my presentation.
Part One: Space Futures
What are the technologies and innovations that are going to drive our space future?
I wanted to highlight some of the areas of technology and innovation that will drive our collective space future. It is important to me that the tech drives our interstellar goals and also is transferrable to Earth-based applications that make everyday life better for us all. As we increase our dependency on computing devices, whether it’s our mobile phone, a laptop, etc. Battery technology, reusability, and longevity of these devices will be key to our sustainable future. Moreso, if we can organise circular economies that can scale here on Earth as an innovative model to replicate in cis-lunar space surely that is a no-brainer. A circular economy is a systemic approach to economic development designed to benefit businesses, society, and the environment. Cis-lunar space means the region of space from the Earth out to and including the region around the surface of the Moon. In space resources are rare and transporting goods into orbit and onto the surface of the Moon is expensive and technically challenging.
Check out the slides I shared below, I went into more detail in the talk but they give you a nice overview of the areas of space technology and innovation that will be applied to drive our future in space.
So it was great that Jorge Arturo Levario - Delagarza, who opened talks by sharing his journey as a SpaceX intern and 2x NASA intern gave an example of the work he’s been doing on Lithium battery testing as I think it is one of the most incredible technologies that we really need to invest in. Not to mention two Lithium mines have opened in the past year in Cornwall, England! I won’t digress here but with SpacePort Cornwall and two Lithium companies based in Cornwall too, talk about growing regional space economies across the UK!
Battery Technology
One of the examples of why power and battery technology is key as space technology is described in the following scenario: on the lunar surface an AI-enabled swarm of robots is mapping the lunar surface to create a lunar highway to deliver payload safely to a target location ahead of human settlement. So you might be running simulations on an AI chip on a lunar rover, or on a series of bots - like a swarm of bots. Humans might be based remotely on a lunar gateway/lunar orbiting platform and sending instructions to the swarm of bots and closely monitoring for intervention or next steps of deploying payload assets. This scenario is one I have elaborated on to build out a scenario and is based on a project that came out of the NASA Frontier Development Lab where I have previously worked, which is an AI and space and science accelerator programme. One of the technical papers that came out of it was about how do you map the lunar surface with a swarm of bots, to feedback on where there might be a crater, and find the most accessible pathway to create a highway on the lunar surface.
Now, for some people reading you might be thinking that if you lose a bot, it’s okay because you’ve got others that you’re sharing information with that can iterate the pathway and share that with the other bots in the system. But the challenges of that, if you have AI and machine learning, in-situ on the lunar surface, is that it’s going to cost you a resource as they consume power to create output, you’re going to need energy, and also you need the compute for feedback and reinforcement learning (RL). Consider if the process logs ‘lost bot X’ so that it can be retrieved, repurposed, and redeployed as an asset for a future mapping exercise. This is the kind of problem-solving, thinking, and circular process systems thinking that excite me at Naaut. We often talk about embedded carbon when we talk about the Net Zero strategy, what about the cost of getting that swarm bot to the Moon and then losing it and simply being ok with the wastage? We have to think about the wider consequences of littering the moon with the debris of new space missions now that there is legislation to mine the moon. I am closely observing the live mission by the Japanese company ispace inc who reached lunar orbit just yesterday on their tech demonstration mission Hakutu-R Mission 1.
So back to the presentation, we have all of this incredible technology it’s only going to be useful if we know where it’s going to be applied. And so, if we’re thinking about deep space missions, whether that’s going back to the moon on to Mars and even beyond like the ESA JUICE mission, which is going to the moons of Jupiter coming up in April 2023. We need to think about electric propulsion and we need to think about nuclear propulsion technology especially when it comes to deep space exploration.
Part two: Generative AI for Space Applications
In the second part of my talk, my goal was to showcase images I have created from text prompts that can be utilised as a powerful tool to inspire and expand creative representation of what it might be like to be on the surface of Mars or imagine a near-future time in which humans might settle on Mars. I also shared images that have struck a chord with me depicting South and East Asian Women as Astronauts. The wonderful thing about presenting in front of a live audience again was being able to read the room, interact with the audience and see the slides that got the most image captures! I haven’t shared all the slides here but if you would like to see them or are interested in hearing the talk please drop me a line. Slides on their own rarely share the context of my thoughts, and viewpoint and I am shedding a bit of light here on a select few but please bear in mind this only scratches the surface of the talk and the themes.
While I encouraged the attendees to play, learn and innovate with technology my personal goal is to continue to explore Generative AI and take the Naaut community on the journey with me. I have been exploring and playing with prompts to develop a style I want to share with you all in due course.
I also question the inbuilt biases that might occur from the engineered design of a product up to the final product and release. I am an advocate for transparency in algorithmic design and implementation as well as the transparency of how a ML (machine-learning) model has been trained on open or closed source data or acquired images to train its models. It is important to me to be as excited about the application of technology as well as its implications in how it returns me a result that I go on to share. But most importantly the range of public data is used to train its model and how it acknowledged the data source for IP.
This is where my talk expanded with a demonstration of ChatGPT. I polled the audience, primarily made up of students 98% of which are studying STEM subjects, and 2% identified as in Humanities/Arts. The majority had heard of and had a go with ChatGPG. When it came to using text-to-image Generative AI, there was one person in the room who had tried it. This is where I introduced the audience to my Vana portrait created with images of myself (see image above). Next, I moved on to talk about Midjourney where I prompted it to create the image of an Astronaut on Mars with a cricket bat. I went through the process in much more detail as well as considerations for prompt design, raised awareness of prompt collusion where users see a style they like and unceremoniously rip it off as well as the access to the free version and how Midjourney’s data acquisition may not be the most ethical of text-to-image generative AI products. While there is a community code of conduct I am not sure yet how it’s policed but these are very early days as it was only released in Nov 2022.
Generative AI images of a Lunar Habitat
As I mentioned earlier in this article, in the case of prompting Midjourney I shared with the KCL Space X3 audience images of a lunar habitat. See below the images of the lunar settlement pictures which might look cool, but they also look a lot like Luke Skywalker’s childhood home on Tatooine! They have a striking resemblance to images released previously by the European Space Agency in partnership with Architects firm Foster + Partners. The ESA images depict a lunar settlement where an inflatable dome creates a structure to be reinforced by a 3d-printed shell created by lunar regolith to protect future visitors to the moon from micrometeorites, radiation, and the like. The images ESA released were in 2012 which you can see here on the bottom left and the images I prompted are not much different do you notice the similarity to Luke Skywalker’s childhood home on Tatooine? (bottom right image). Generative AI is only limited by the source images it has been fed to train its models so designing innovative lunar habitats might have a long way to go yet but this is something to consider objectively I had an image that while it might not be unique in its design still provided me with images I could share to spark ideas or inspire the audience. Maybe I have been around long enough in the space industry that I have the expertise, perhaps it’s because I love Star Wars but we have to learn to question the outputs of any image prompted by text generation and underline any output with critical thinking a skill which we all need to develop with the onset of AI-based technologies.
Check out the slides and images I shared and judge for yourself with the rendered image here:
What about trained models based on Science-backed papers?
Now consider what images could be created in closed science environments where collaborators share their research and query future design iterations from text-to-image models like the ones Dall-e, Midjourney, and others might generate. These are the considerations that are so important for me to share with the audience as there is so much potential for the technology but at the moment the images created by Midjourney do not credit living artists whose images have been used without their consent to produce images across its platform which is hosted on a Discord server. There are bigger questions to ask and we need to converse with everyone across all of society to educate, illuminate, share, learn, understand, and grow together. What happens if a machine learning model has been trained with your artistic design style? How do you protect your IP whilst getting credit or even a monetary kickback? This is why blockchain and web3 have always excited me, as it has the capacity to do this and there are ethical players in the arena who I have had the chance to discuss this with at length. I am always keen to speak to you if this is an area of interest or if you are working on this.
On blockchain for space applications, I went into this more in this talk I gave pre-covid a few years ago about Blockchain for Space at a conference hosted by Matterum at Digital Catapult in London’s Kings Cross.
Thank you to KCL Space for the invitation to speak at the X3 Space Conference it was a wonderful experience. Shout out to Remco Timmermans who captured the awesome pics of my talk.
I would love to build on the themes of this talk and connect with others working in the space, technology, and innovation space.
Let me know in the comments below what you enjoyed most about this article, what you’d like to hear more about, and if you’d like me to record and share the talk I gave. Sadly it wasn’t recorded on the day but I am open to giving the talk again as it’s something I am excited to share with new audiences.
Drop me a line if you’d like to see my slides.
Let the conversation continue.
