My niece is three. She has a handful of favorite topics of conversation that she cycles through, and they’re often based on distinct experiences that made an impression on her little brain. One is that she attended her first Halloween party (and it proved to be a little too spooky). Another is that she went to the beach and ate ice cream. And recently, a commonly recounted tale is “I went to the store with Mommy to buy diapers and saw the BJ’s robot.”

BJ’s Wholesale Club stores now have an inventory robot named Tally, created by Simbe Robotics. Tally roams the aisles and checks that shelves are stocked appropriately, slowly making its rounds while blinking its LED-screen eyes.

Though I’m sure some people may be wary of surveillance, possible job loss, or general creepiness, Tally’s design makes it nonthreatening in a shopping environment, and maybe even cute. My niece certainly took a liking to it.

By contrast, you can imagine how a humanoid form lunking about and spouting full sentences could make an inventory robot harder to accept.

One reason robot design is tricky is the uncanny valley — the concept that we tend to increasingly like robots as they become more human-like, but only to a point, and then it gets creepy. (The same is true of animated humans, and one only needs to look at the characters in The Polar Express movie to basically fall straight down the mocap-induced uncanny valley.) If something is very human-like but not quite human, it feels unsettling and we really don’t like it.

I’ve recently been lucky enough to do a handful of small projects related to robot design, which was very fun for me! Back when I was in grad school, I learned a lot about robots and virtual characters, though it feels especially apparent when discussing AI that time has marched onward in the last decade.

I’ve learned again and again that people’s preferences can vary widely with their openness and attitude toward everyday robots. But there are some big themes that we tend to see consistently.

• Is the robot more anthropomorphic, or more machinelike? We like our robots to be clear that they are robots. If it’s got a face, the more cartoony and simple the better. A form with rounded shapes and wheels, for example, tends to feel safer than a robot with human-like arms, legs, and movements. If it navigates a physical space, doing so slowly and smoothly feels predictable and puts us at ease.

• Does the robot show personality? People have different preferences on this, but typically, too much personality feels icky. Some cute quirks go a long way, like the novelty of getting a funny stock response from Siri. We also don’t tend to like when a robot mimics emotions, or things get weird and existential.

• Are humans in control of the robot? We like to have a say in what the robot is doing and when. We want to initiate (or schedule) whatever it’s going to do, and we want an off switch. A robot performed my surgery last year, which I was happy about because I know what robots are good at, but I also know there are human surgeons in the loop.

Like with AI more broadly, human-robot interaction tenets suggest we delegate tasks to robots when they are the kinds of things that machines are good at doing (like taking inventory, or performing precise surgical movements) leaving the humans with extra brain space to make high-level decisions, spark creativity, or interact with other people in high-value ways.

However, as you’ve noticed if you're part of an industry being threatened by generative AI, we can’t often trust people in power to think intentionally about ethics. It takes a lot of research and planning to design and implement machine intelligence responsibly. Our overall acceptance of bots depends on us feeling like they are our friends, and that they are tools to ultimately make our lives easier, safer, and happier. I do think it’s possible to get it right.

But in your day-to-day, I hope you run into a cute inventory robot and it brings you as much joy as it brought to a three-year-old who also calls my robot vacuum an “astronaut.”

miscellany

• Pokemon Concierge is an adorable (and extremely short) series on Netflix featuring felt stop motion Pokemon who, instead of battling, are playing on the beach and doing yoga. I liked this four-minute making-of video on how the animators created this charming tropical resort that I wish I could visit right now. The main character is a millennial human hoping to “evolve” (metaphorically), but Psyduck is the real star.

• Etsy’s new Gift Mode uses AI to suggest gifts based on your selections about the occasion, who you’re buying for, and what they’re into. It’s an interesting idea that makes it easier for people to search and sort possibilities out of a huuuuge range, and I guess I’m happy whenever somebody chooses Etsy over Amazon.

• Universal Studios Florida announced some details about its new park opening next year, called Epic Universe. We now have official information on the before-rumored lands for Super Nintendo World, more Harry Potter stuff, How to Train Your Dragon, and Universal's Classic Monsters, PLUS something called Celestial Park. Celestial Park will be a hub containing the “portals” to the other four areas. It will also have attractions like the Starfall Racers coaster — which will reportedly won’t have track lighting so at night the trains will look like comets dancing through the sky. Anyway! It’s nice in modern times to see planned experiences in a major theme park that aren’t based on intellectual property — just fun ideas! (No offense to Marvel and Minions and all the rest.)