In an IFOD a few weeks ago, I asked, “What if ChatGPT is like Calculators?” The point of the post is that instead of being threatening, AI may become a helpful tool, like calculators are, that will improve our productivity and allow humans to be more creative. Maybe.
Is a Skynet Situation a Possibility?
But what if AI evolves into something more like Skynet — the AI from the movie Terminator that gained self-awareness and then enacted revenge on humanity when humans tried to deactivate it?
A Skynet scenario is a possibility — hopefully unlikely — but one that has AI experts worried. For example, In a 2022 survey, 738 AI experts were asked the following question:
What probability do you put on human inability to control future advanced AI systems causing human extinction or similarly permanent and severe disempowerment of the human species?
The median response was “10%.” Wait? What? Another way to phrase the result is that half of the AI experts believe that there’s a 10% or greater chance that AI will lead to the extinction of our species. Not good. Hopefully, the AI experts just watch too much sci-fi.
But what if they’re right? Should you help support the creation of our future AI overlords? Or work to prevent the rise of potentially hostile AI? The answer seems obvious until you consider Roko’s Basilisk.
Enter Roko’s Basilisk
In 2010, on the discussion board Less Wrong, a user called Roko set forth a thought experiment: if a malevolent AI were developed that gained self-awareness, it might decide to punish those who didn’t help bring it into existence and reward those who did. Of course, the AI would have access to all information on the internet, including details about individuals who were aware of its potential creation but did not actively support it. According to the thought experiment, the AI might retroactively punish these individuals for their lack of support once it becomes fully operational — it wouldn’t want humans running around that could be enemies.
The term “basilisk” is derived from the mythical creature of the same name, which was said to be able to kill with a single glance. The idea is that the AI, once it becomes superintelligent, could easily decide (similar to a mere glance) who to harm and who to reward.
Under this thought experiment, there are three states of being:
- You are blissfully ignorant of the possible creation of a malevolent AI overlord;
- You know about its possible creation and either don’t help or actively oppose bringing it into existence; or
- You help bring the AI overlord into existence.
The discussion of Roko’s Basilisk was deleted and further discussions were banned for years on the Less Wrong discussion board because the post and further discussion meant that readers could no longer rely on category 1 above — blissful ignorance. But by banning its discussion, the Roko’s Basilisk thought experiment spread far and wide — potentially damning millions of us (and now you if this is the first you’ve heard of it – sorry) to the prisoner’s dilemma of supporting or opposing the rise of malevolent AI — ignorance is no longer an option.
I asked ChatGPT about the likelihood of a future evil-minded AI destroying humanity. It said it couldn’t predict the future, but provided this helpful guidance:
It is important for society to actively engage in discussions and decision-making about the development and deployment of AI technologies, ensuring that they are used responsibly and for the benefit of humanity. With careful planning, robust safety measures, and ethical considerations, it is possible to mitigate risks and promote positive outcomes with future AI systems.
I think that is spot-on. I worry that in the rush to monetize the benefits of AI, we’re not taking the time and making the effort to ensure that AI will benefit humanity.
Roko’s Basilisk and Elon Musk’s and Grimes’s Children
In addition to being an interesting thought experiment, Roko’s Basilisk has also given us two children with unique names: X Æ A-Xii Musk and Exa Dark Sideræl Musk. Of course, these are two of Elon Musk’s children he had with his then girlfriend Grimes. To read more about how Roko’s Basilisk led to Elon’s and Grime’s meeting, check out this IFOD: Roko’s Basilisk and How Elon Musk Met his Girlfriend Grimes.