Article by Human Cusp, image Peter Scott
As artificial intelligence continues to transform the world, we see numerous opportunities arising for its use in the space sector, and great potential for AI-knowledgeable engineers, entrepreneurs, and others to gain competitive advantages.
In earlier forms, AI has been used in space for decades, from image processing to signal analysis to processing data from experiments conducted in orbit. In the 1990s, artificial neural networks were used[1] to learn how to classify land coverage in Earth observation imagery, software developed at the Jet Propulsion Laboratory (JPL), where I spent many years. More recently, JPL has pioneered the development of autonomous robots for planetary exploration, such as the NeBula (Networked Belief-aware Perceptual Autonomy) system[2] for teams of robots exploring unknown environments under extreme conditions.
While generative AI has captured the popular imagination, the use of predictive AI based on more traditional machine learning (if anything evolving that fast could be called ‘traditional’) is where we look for the most easily quantifiable returns on investment. Optimizing signal routing and bandwidth allocation in satellite telecommunication paths, for instance. Genetic algorithms for material design simulations, and robotic experimentation. In medicine, drug discovery traditionally takes on the order of years but AI is reducing that to days.[3]
Power management in space systems is usually a limiting factor due to the size of energy storage units and solar power collectors. AI can help optimize power consumption through studying usage patterns. A similar approach allowed Google to reduce its data center cooling costs in 2016 using AI.[4]
In infrastructure, we should not forget cybersecurity. We can no longer afford to protect our networks with exclusively human-curated firewall rules when those networks are under attack by bad actors using AI, and vendors such as Palo Alto now recognize the need for AI defense. However, the recent CrowdStrike incident,[5] affecting millions of machines and crippling airlines, hospitals, and banks, highlights the need for human diligence. More haste, less speed.
And of course, AI-accelerated software development changes everything touched by code. Within two years it has become unacceptable to develop code without using generative AI such as Github CoPilot.
The global efflorescence of AI brings its own challenges, of course, such as an acute shortage of experienced AI engineers. Our educational institutions must get out of old pedagogical models and recognize that AI can now supplant functions that teachers have usually taken on, such as marking papers. Degree programs need to be adapted to a world that being reshaped by AI and train engineers and scientists for where that world will be tomorrow, not yesterday. Those changes are not confined to technology but also mean that the economics of future jobs and the affordability of housing, for instance, are likely to change in unpredictable ways that urgently require considered investigation. Schools are still trying to work out the balance between teaching skills that AI can now perform because they grow valuable neural pathways in students, and deciding what tasks can safely be left for AI. Many schools are unable to adapt fast enough to these changes, and will fail.
The explosion of large language models (LLMs) provides particularly important horizontal benefits for technology businesses. Seamless and near-perfect translations speed up international agency cooperation, and documents can be translated between different formats easily. Image recognition can be used to accelerate assembly and repair, particularly if combined with augmented reality (AR) input devices.
The UK faces challenges when competing with the two largest players in AI, the US and China, which can outspend other countries. London-based Tortoise Media’s Global AI Index[1] puts the UK in fourth place, behind Singapore and ahead of France, scoring countries on a huge range of metrics. The availability of enormous data centre resources is a powerful attractor for top talent. UK companies – with the obvious exception of DeepMind – should not attempt to compete on raw model size but instead explore the countless niches of unique combinations of skills. Regulation has yet to catch up – the UK is not bound by the European Union’s AI Act, of course – and this is a time when there are equal parts uncertainty and opportunity for risk takers to leverage the volatile environment. I would particularly encourage maximizing ties with the UK’s superlative universities to create seamless partnerships connecting talent with need.
To conclude, the AI revolution spans every expression of human endeavour, because it models and emulates principles of human cognition and creativity. So there are opportunities for enterprising individuals outside science and engineering to exploit AI in applications ranging from education to human resources, marketing to accounting. The limits are chiefly imposed by our imaginations.
————-
Peter Scott helps people master technological disruption. After receiving a Master’s degree in Computer Science from Cambridge University, he moved to California to work for NASA’s Jet Propulsion Laboratory and now works with schools to help them pivot their curricula and teaching methods. Through proprietary frameworks and group facilitation he empowers them to adapt to and leverage AI.
Peter has keynoted around the world. His third TEDx talk, on why AI might need to have empathy, was released in July 2023.
His weekly podcast, “Artificial Intelligence and You” tackles three questions: What is AI? Why will it affect you? How do you and your business survive and thrive through the AI Revolution? In July 2022, his second book, also called “Artificial Intelligence and You,” was released.
[1] https://www.sciencedirect.com/science/article/abs/pii/009830049090009I
[2] https://www-robotics.jpl.nasa.gov/what-we-do/research-tasks/nebula-autonomy/
[3] https://www.nature.com/articles/d43747-023-00029-9.pdf
[4] https://deepmind.google/discover/blog/deepmind-ai-reduces-google-data-centre-cooling-bill-by-40/
[5] https://www.fastcompany.com/91160759/crowdstrike-data-gdpr
[6] https://www.tortoisemedia.com/intelligence/global-ai/
+44 (0)7436 140185