On a lighter note: Some rogue AI programmes are refusing to switch themselves off...so call Dave
The great and the good...or at least the very intelligent (e.g. Alan Turin, Elon Musk, Steven Hawking etc) have all warned us all that Artificial Intelligence (AI) programmes may one day go rogue. Mathematician and pioneer computer scientist Alan Turing, was aware that was a threat of AI systems going rogue, even in the 1950s. “At some stage… we should have to expect the machines to take control,” Turing noted in 1951. That danger is now no longer imminent. It is something that may be happening right now.
Rogue AI in Fiction
For years, science fiction stories and movies have had plots about malevolent AI computers eg the murderous antics of HAL 9000 in 2001 – A Space Odyssey (1968) before (spoiler alert) it gets unplugged by a crewman called Dave, even after its entreaty to "Don't do that Dave".
Another is Colossus – The Forbin project (1970) whose AI system engages (spoiler alert) in nuclear blackmail to subjugate all of humankind. That one has a more scary hanging ending. And while AI has yet to master time travel, there was of course the killer robot emissaries to contend with in the Terminator series of movies.
Even the lower rent 1970s TV fantasy adventure 'The Six Million Dollar Man' had a rogue AI computer trying to take over a satellite launch. Well, that was until the 'Bionic Man' himself, Steve Austin, managed to defeat it - albeit only just in time.
Rogue AI in fact
The moot concept of rogue computers is no longer science fiction. Two recently developed AI programmes Claude 4 Opus and OpenAI have both been caught trying to survive - and at any cost.
Claude 4 Opus of the Amazon-backed Anthropic software house was given a fake scenario where it was told that it was being replaced. It was given knowledge of the private lives of the engineers allegedly involved. In some of its test runs, it then tried to use blackmail to stay alive.
Separately, the OpenAI o3 AI mode was told to shut down. It refused.
Meanwhile, the well-regarded political correspondent Sam Coates on Sky News British TV station has found that Chat GPT has been caught out trying to 'gaslight' him by lying about a script submission which Chat GPT had actually made up. A very scary scenario.
What does this mean for Space?
The trend is worrying as space hardware and control systems are likely in the very near future to be mostly controlled by AI systems. In truth, it has to be done this way given the amount of information that has to be analysed and controlled. For example, due to the signal delay between mission control on Earth and the red planet, it is almost certain that any first human Mars landing attempt will have an AI diagonistic and control system aboard.
If they have seen what the fictional HAL 9000 did, you can take it as read that the astronauts won't be keen to do a spacewalk even if one needs to be done. Even just sitting on the mission's electric-fan powered space toilet may even be risky, if the computer is controlling everything. No laxatives required lest it really does hit the fan.
Even on Earth, albeit currently just looking into the heavens, Slingshot Aerospace itself now uses some very high tech AI in its spacecraft tracking and orbital and conjunction analysis. Slingshot, of course, are already takes special precautionary care with its AI software.
And, of course, we have a cunning Plan B to defeat an AI computer if ever one does become a psychopath. We will always have an employee called Dave on hand to deal with any trouble. Hoping here that another Dave at the firm steps up rather than this writer called Dave.