How AI could be the key to deep space exploration

Deep space exploration presents many exciting opportunities, but current technology prevents humans from going beyond the moon’s orbit. What challenges do humans face with deep space exploration, how could AI solve this challenge, and what challenges would those AI face?

What challenges do humans face with deep space exploration?

As much as I know that Star Trek is science fiction, I hope that the concept of faster-than-light travel becomes a reality. Of course, nothing can travel faster than light, but some theories may allow for space-time to be manipulated to allow for travel across great distances in a short amount of time (wormholes and Alcubierre Drives). Nevertheless, until these dreams of exotic drive systems are realised, humanity has to cope with the technology it has for space exploration.

When it comes to deep space exploration, humanity faces a multitude of challenges that not only prevent humans from going beyond the orbit of the moon but also make it difficult for robotic systems and probes to conduct scientific research. Undoubtedly, the first challenge is that everything of interest in space is unbelievably far away, and while humans can live long enough to traverse the solar system using traditional chemical engines, trying to reach even the closest star would take thousands of years.

The second challenge humans face is that we are delicate creatures entirely dependent on our ecosystem to survive. That means any long journey across the solar system would require large amounts of oxygen, food, and water, and carrying this extra weight requires larger rockets. Thirdly, deep space is not as forgiving to living tissue as Earth is; as such, any ship carrying humans requires significant shielding. In fact, giant planets such as Jupiter have such strong magnetic fields that they have radiation bands lethal to anything living, and thus anything of interest in these bands is effectively off-limits.

To try and solve these challenges, probes can be sent in place of humans that can experience greater g-forces, resist radiation, and don’t require food or water. However, probes that stray too far from Earth can be hard to communicate with (i.e., require a strong antenna with a direct line of sight). Furthermore, the extreme distance between Earth and a deep space probe means that the time taken for messages to arrive makes real-time operation impossible. Thus, if something of interest happens, it is up to the probe to take images and data readings.

How could AI solve deep space probe challenges?

Fundamentally, the advantage that AI would bring to any deep space probe is the ability to operate without any human intervention, whether it is orbit adjustment, power conservation, or predicting issues that may impede its mission. But above all else, an AI that can spot subjects of interest would be immensely powerful as it would allow probes to identify events that would capture the interest of humans.

For example, a probe flyby of a moon of Saturn, for example, could be programmed to image the far side. But what if the probe detected a meteor that was about to hit Saturn? As far as research goes, this is significantly more important, and a probe powered by AI would be able to recognise this as interesting. Thus, the AI would turn the probe around and observe the meteor impact instead. Once observed, it would then go back to its original mission.

Additionally, AI-powered probes could also be used for missions that extend well beyond the range of radio communications from Earth. Using current technology, miniature probes can be launched toward local star systems at relativistic speeds (accelerating over a long period of time) such that they would arrive within our lifetime. Upon arrival, it would be the responsibility of the AI to observe the system, look for things of interest, and then find a way of getting the data back.

What challenges does AI face in space exploration?

While the idea of AI in space probes does sound promising, there are numerous challenges faced. One of these is that current AI systems are nowhere near advanced enough to understand what is interesting to humans. A funny but good representation of this is in the film Planet 51, whereby an intelligent probe lands on an alien world, and instead of recognising the obviously advanced civilisation who are driving vehicles, watching films, and eating out, is more interested in unusually shaped stones.

Another challenge AI presents is that the intelligence integrated into such probes must be repeatable and predictable. Even though AI technology has come extraordinarily far, such algorithms can still make trivial mistakes. For example, a mistake in an imaging sensor could be perceived as a bright flash of light, which may trick the system into thinking something of interest has occurred. Thus, any AI-powered probe would need multiple redundant systems to ensure that it can discriminate between sensor values and identify those that may have failed.

Overall, AI-powered probes present an option for humanity to explore the stars without having to be physically present, and the technology to send probes beyond the solar system already exists (albeit extremely tiny probes). But AI still has a long way to come before it can be entirely entrusted with the complete operation of a multi-million-dollar probe.