We Need New AI Analogies

Tomorrow’s AIs are both more embodied than HAL from 2001 and less robotic than Rosie or Data. A better analogy comes from a surprisingly old story.

Typically, in science fiction and popular understanding, AIs and robots (or drones) fall into distinct, although overlapping, categories.

Robots are autonomous single entities that move through the physical world the way humans and animals do. Robots don’t need human shapes—just think about the variety of droids in Star Wars—although many famous examples are humanoid (Rosie, Data, the Vision, Marvin).

AIs are disembodied voices that affect the physical world (“Alexa, turn on the front lights”) but do not inhabit the physical world: we hear but do not see them. These AIs tend to be tied to a place (a home, a space station or starship), and in some ways to bethat place. HAL from 2001: a Space Odyssey is one of the first such, but the helpful computers across the different incarnations of Star Trek (including Zora in the just-concluded Star Trek: Discovery) are where a lot of folks first encounter the idea of an invisible computerized helper. Siri, Alexa, ChatGPT, and all their cousins follow this model.

We need new AI analogies.

Our two concepts—robots and AIs—will soon merge. Instead of one disembodied AI, we’ll have a swarm of different, collaborative AIs that interact with us and with each other. There may be a high level primary AI that runs functions in and around your home, but that AI will work with other, smaller, embodied AIs that maneuver through space.

Think of a Roomba that mostly just wanders around vacuuming things but that also can get direction from the primary to nudge the cat to go outside or push a door closed because the baby is taking a nap. Then think of a cluster of such AIs with different body shapes and levels of intelligence that the primary organizes like an orchestra conductor.

In late 2023, Ethan Mollick (author of Co-Intelligence, the book that I’m recommending most this summer) wrote a piece called “An AI Haunted World” that captures how this AI collaboration will work, albeit in a more disembodied sense than I’m imagining:

In the near future, AIs will work in their own hierarchies of intelligence, all communicating with each other, perhaps mostly autonomously. If you want to grab a bite to eat, it may be that your more “intelligent” premium AI assistant can guess what restaurant you might like based on reasoning about you and your recent behavior, and then delegate to a cheaper AI to actually make the reservation. Or you may ask your personal phone AI for help with the task first, and it can get advice from a frontier model on how to nail a tricky reservation, charging your account for the extra bit of intelligence-on-demand.

Layer on different form factors, different robot shells that the smaller AIs inhabit, and we’re closer to what I’m predicting.

But we still need a handy analogy to help us see that future in our minds’ eyes.

The best one I can think of are the helpful mice (and birds, and dog) in the classic 1950 animated Disney version of Cinderella.

Image created using Ideogram.ai.

Bibbidi-Bobbidi-Boo

In the story, Cinderella’s father remarried after her mother died. Then Dad also died, whereupon the stepmother and stepsisters enslaved Cinderella and forced her to do all the housework. (I’m not going to tell the whole story: it’s a good movie; go see it.)

Versions of the Cinderella story date back to ancient Greece, but the version that Disney adapted was by Charles Perrault (1697), who added the detail of the glass slippers. (The Brothers Grimm included a much grislier version in the early 1800s.)

The Disney innovation in the Cinderella story was to add speaking, autonomous animals that love and want to help “Cinderelli” (as they call her, because talking cartoon animals frequently have speech impediments, e.g. Daffy, Donald, Porky, and Scooby) however they can.

These animals function like the swarm of embodied AIs that will live and work alongside us in the near future.

In the opening reel, birds wake Cinderella by pulling up her braids and chirping into her ears. A pack of mice, led by Jaq and newcomer Gus, entertain and help Cinderella while also engaging in a long war with Lucifer, the evil stepmother’s fat cat.

When the stepmother says that Cinderella can go to the prince’s ball if she finishes a mountain of chores, Cinderella is sad that she won’t have time to update a dress that belonged to her late mother. While she is working, the mice all team up to update the dress for her singing, “We can do it, we can do it, we can help our Cinderelli!”

Here, the mice act as agents for Cinderella rather than assistants. This distinction is key to the difference between primitive digital assistants like Siri and Alexa that do only what we tell them to do (“Hey Siri, at 10:00am tomorrow remind me to water the plants”) and coming-real-soon AI agents to which we can delegate goals to achieve rather than tasks to manage.

In this analogy, the Fairy Godmother—who shows up just in time to transform Cinderella and the animals into a ball-worthy royal guest and her entourage—acts as the primary, orchestrating AI that pivots the entire organization into a new short-term strategy with clear resource limitations (the spell ends at midnight).

In the climax, mice Jaq and Gus again act as agents fulfilling Cinderella’s wishes without her direction. After the wicked stepmother locks Cinderella in her room to prevent her from trying on the glass slipper, Jaq and Gus steal the key and carry it up many stairs to free Cinderella, enlisting Bruno the dog to help them get rid of Lucifer, who is trying to stop and eat the mice.

Arthur C. Clarke famously observed, “Any sufficiently advanced technology is indistinguishable from magic.” The corollary here is that if we look to popular narratives of magic, then we can get a glimpse of our future interactions with advanced technology.


Note: To receive articles like this one—plus a whole lot more—delivered straight to your inbox, please subscribe to my free weekly newsletter!


Posted

in

,

by

Tags:

Comments

0 responses to “We Need New AI Analogies”

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.