Predictors of AI doom may have too little imagination, not too much

AI doomerism, the belief that AI poses an existential risk, gained substantial ground in 2023, most notably following the open letter published in March 2023 by the Future of Life Institute. The letter, which now has been signed by more than 33,000 people, sounded the warning that “AI systems with human-competitive intelligence can pose profound risks to society and humanity”.

In a reaction to this letter, decision-theorist Eliezer Yudkowsky wrote that “the most likely result of building a superhumanly smart AI, under anything remotely like the current circumstances, is that literally everyone on Earth will die”.

Not surprisingly, more and more calls are made for AI research to be stopped – or at least slowed down and shackled. This would be a mistake. AI can contribute to ensuring humanity’s future. Indeed, humanity’s descendants may merge with AI over the long term. AI could be seen as our offspring – at least our “mind children” – an offspring that straddles both biosphere and noosphere.

If the further development of AI is halted, as the AI doomsters want, it may be a grave mistake. This is not only because it will curtail the evolution of humanity over the long run, but it may also leave humanity defenceless against the threat of an extraterrestrial AI. The real existential risk to society from AI may not come from our own AI but from an extraterrestrial AI – an ET-AI, if you will.

Although there is no evidence of alien/extraterrestrial civilisations, statistically, the odds of human civilisation being singular are almost vanishingly small. There are about two trillion galaxies in the universe, each containing a 100 billion stars – most of which are likely to have planets. So, as Enrico Fermi asked, where are all the aliens?

One of the best answers to date is that the universe is a dark forest. In a dark-forest universe, where the intentions of other intelligences are unknown and cannot be reliably communicated (due to interstellar distances), the best strategy for any civilisation is to conceal its existence. If it is discovered, it then may want to strike first to eliminate the civilisation that had found it – as a precautionary measure before possibly being eliminated itself.

It is therefore foolish for humanity to broadcast its existence to the cosmos as it is currently doing, either unwittingly through radio and other broadcasts or intentionally, as through the information embedded in the Voyager spacecraft. As the scientist Jared Diamond has warned:

“Extraterrestrials might behave the way we intelligent beings have behaved whenever we have discovered other previously unknown intelligent beings on Earth, like unfamiliar humans or chimpanzees and gorillas. Just as we did to those beings, the extraterrestrials might proceed to kill, infect, dissect, conquer, displace, or enslave us, study us as specimens for their museums or pickle our skulls and use us for medical research.”

If more advanced than humanity, extraterrestrials would probably not be biological entities but artificial general intelligences (AGIs), as the UK’s astronomer royal, Martin Rees and many others have argued. If they are AGIs, this may also explain why we are not (yet) aware of any aliens – they may, for instance, use quantum entanglement to communicate (and not radio waves) or compress their communication signals so that they would be indistinguishable (to us) from noise.

How could such an ET-AI pose an existential risk? One way to destroy us could simply be to broadcast a “killer code” – an AGI computer code that would infest all our systems and take over. It could also broadcast instructions for building a civilisation-destroying “bomb”, perhaps camouflaged as a Trojan Horse. Our only long-term existential chance may, after all, depend on whether we can develop our own AGI.

Read more in Daily Maverick: Resistance is futile – South Africa must urgently adapt to the new age of artificial intelligence

There is one last twist. If we fail to create our own AGI, then we would have no defence against such an ET-AI and over the long run, if humans eventually do go extinct, there would be no AGI that may enable a technological resurrection.

It has been suggested that an AGI may, one day in the far future, be able to use simulation methods to “resurrect all possible people” who have ever lived. A super-intelligence may even use signals from advanced civilisations that lived in an aeon before the Big Bang, which they may have embedded in the universe’s cosmic background radiation, to “reconstruct an entire previous aeon civilisation”.

Ultimately, AI doomsters don’t have too much imagination but too little. DM