As the AI Age continues, I find myself looking back on AI’s depiction in pop culture of the past. Specifically, I think about 2 TV episodes: “The Ultimate Computer” from Star Trek: The Original Series, and “The Human Factor” from the OG MacGyver series.
Anyone who’s watched these episodes (and for those who haven’t, both are on Paramount+, or just pirate-stream it if you’re feeling ethically flexible) knows the similarities: human-based AI touted as “the next step” and a replacement for human beings f*cks up during a simulation, prioritizes self-preservation, and nearly gets the protagonists killed. These are “M-5” and “Sandy”, respectively.
I find myself comparing these depictions to how AI works today, and wondering what would happen if contemporary AI was involved in these kinds of situations. How would ChatGPT behave if it were given control of a starship? How would Claude behave if it had to guard a classified R&D facility? Are the recent IRL AI-related cybersecurity incidents any indication?
I’d like to know the community’s theories on this.
The scenario falls apart when you realise that the “traditional” view of AI, what we’ve seen in movies and on TV, really isn’t at all the same as what companies today call “AI”. Auto-complete is not intelligence.
With every sci-fi story that’s been written incorporated into its training data, an LLM chatbot roleplaying an ‘AI’ persona is going to mindlessly auto-complete the most likely next thing that happens right after humans hand over control of the important dangerous thing to AI and trust it to make life-and-death decisions. It doesn’t have to be actually intelligent to be dangerous, it just has to be more intelligent than the people who think it is intelligent- and in my opinion it’s already there.