Ai/Robots in Fiction ver. 5 -- 13 Sentinels
“Get in the robot, Shinji” - Misato, Neon Genesis Evangelion
This famous quote encompasses the unwillingness by Shinji to pilot the 01 Eva due to interpersonal depression and external expectations and duty set from society. Now, imagine this, but add 12 more characters who bare the responsibility to pilot mechs to protect humanity. This is the setup for the game 13 Sentinels: Aegis Rim developed by Vanillaware and released by Atlas. I already love Atlas (notably the Persona Games, Metaphor, and Catherine), so as I looked for other serious games to play on my switch, I stumbled upon this. And boy oh boy, this game is batshit insane.
The 13 characters you play as that pilot the 13 sentinels
What’s crazy about this game is the directing of the plot and tons of misdirection that occurs. First of all, the game has two parts: (1) a side-scrolling adventure and (2) a real-time strategy (RTS) game, controlling the sentinels to defeat rampaging mechs called Deimos (instead of the angels in Eva). Importantly, both (1) and (2) are in medias res 1, with the adventure portion detailing each character’s perspective of the events in their own time irrespective of the other characters’ eras, and the RTS encapsulating only the final battle between the Sentinels and Deimos.
The RTS portion of the game
Throuhgout all this confusion and hints dropped here and there, we continue to learn the bigger picture of why these teens are tasked to enter the mechs and where and why the Deimos have come from. This game encompasses tons of various technological pursuits such as Ai, robots, nanotechnology, terraforming, and most fun: Time travel! Anyway, I highly recommend this game, go check it out.
On the topic of Ai/robots specifically though, I’d say the biggest feature this game has presented to me is the idea of Ai hacking other Ai. The topic of defense and cybersecurity is significantly important for the sake of (general) Ai, especially with LLMs. From my sparse knowledge, a lot of work tends to prevent adversarial attacks 2, but this is not the type of hacking that occurs in 13 sentinels.
Rather, during a confrontation between two embodied Ai, one of the Ai manages a hack to switch bodies, and since its original body was on its last words, the switched Ai would die there. The best analogy to explain this is if you know what happens in The Superior Spiderman between Spiderman and Doc-Oc3 – this happens but between two Ai. This then allows the switcher to hide within the new body, faking a personality as the old one.
I think the idea of switching two human brains/personalities coming to fruition is way into the future, so I doubt anyone truly worries about such a reality. Thankfully, the idea of two Ai’s switching its mechanical bodies is in my mind, also very far into the future. Like Doc-Oc and Spiderman, the personalities are swapped wirelessly and fluidly without embodied influence (in other words, a third party is not switching hard drives akin to a third party switching brains). As such, the hacker must transfer terabytes of data (for the neural network parameters + hard drives that consist of stored data) in a smooth and fast manner, which is absurdly difficult. Furthermore, even before transferring data, the Ai must figure out how to even breach the other Ai’s robot body which, unless the victim for some reason has port 22 open, is technically impossible.
Nevertheless, I found the concoction of two sci-fi tropes, robots and body-swaps, to be surprising and creatively dangerous. In fiction, we already fear the dormant “split personality” from psycopaths and alike (whether due to DID or not), so I’d expect more fear towards armed robots suddenly changing from a safe, security robot into a raging, killing machine.