I tested the AI gadget that got the internet buzzing and it left me wanting more


Rabbit R1

Kerry Wan/ZDNET

As I was packing up at the office, my mind was already set on using the Rabbit R1 to book an Uber to my downtown meeting. The R1, a groovy-looking handset roughly half the size of a traditional smartphone, was meant for moments like this. 

Forget tapping through slides and columns of filters on a phone app or browser window, a simple voice command to the R1 is supposed to direct dedicated AI agents to get the job done on your behalf. “Book me an Uber to 123 ABC Street.” “Order a large vodka pizza from Via Ponte.” “Play the latest Benson Boone album.” Theoretically, this near-frictionless process would save you time, effort, and lots of clicking. Theoretically.

Also: Rabbit R1 unboxing: My first impressions of the $199 AI gadget

By the time I’m outside and command the R1 to book me an Uber, I notice two things: The R1 thinks I’m situated at a bakery a few blocks up (and when I change the pickup address — only possible via voice — it never updates to the new one), and since unplugging the device five minutes ago, it’s battery indicator has dropped to 94%. I put the little orange square into my left pocket, pulled out my phone from my right pocket, and opened the Uber app.

To be clear, the R1 isn’t meant to replace your smartphone, even though it has a SIM card slot for an LTE connection. Instead, I like to think of it as an office intern who’s still on their first day at the job. They might not be the best at everything, but there’s so much potential with time and training. Oh, and they always wear an ultra-neon orange shirt to work, so you have to notice them.

<!–> Rabbit R1

You can access the settings by shaking the R1 twice, but it only works from the home screen.

Kerry Wan/ZDNET

What the R1 does well from the get-go is answer complex questions fast, noticeably faster than the $700 Humane Ai Pin. It nails the basics like “What’s the weather?” so I won’t go in-depth with that. Instead, I asked questions like “Who won last night between the Celtics and Heat, who scored the most points, and what was their field goal percentage?”, all in succession. While the R1’s first response suggested that the last time the two teams faced each other was back in January, leaving me to believe that it has a knowledge cutoff, it answered all three questions correctly on the second attempt. Odd, but still impressive.

Also: These AI avatars now come with human-like expressions

Sometimes, the AI responses can be noticeably long, and perhaps my extensive questioning is to blame. In most cases, after the R1 has given me what I think to be a satisfactory answer, it rambles on with more details, requiring a press of the side button to silence it.

–> Rabbit R1

Enabling terminal mode switches to the R1’s UI to landscape orientation and presents a keyboard to type questions in.

Kerry Wan/ZDNET

The R1 is also great at multimodal tasks, capturing your surroundings (and yourself) with the rotating 8MP camera and leveraging the visual data to answer questions. From the rooftop of an office building, I pointed the gadget at a series of skyscrapers in front of me and asked it to describe what it saw and what city I was in. It responded, “In front of me, I see a skyline of tall skyscrapers and buildings, including the iconic One World Trade Center tower. This is the skyline of New York City, with the Hudson River visible in the foreground.”

Also: The most promising AI smart glasses are from a brand you’ve never heard of

In another example, I pointed the R1 at a bowl of ramen and asked it for the calorie count. While the AI couldn’t give me an exact number due to its inability to pull all necessary context from a 2D image, like the size of the bowl and what ingredients were used, it gave me a typical range of 300 to 600 calories per serving, which was close enough.

I also tested the R1’s translation feature with my colleague, Kyle Kucharski, who speaks several languages, including Spanish and Russian. Asking the assistant to “translate” promptly brings up a dedicated interface on the 2.8-inch screen, but that’s about as good as the feature gets. From delayed translations, where the AI would utter lines minutes later, to translating only one language in a conversation between two different ones, the experience is buggy and likely wouldn’t instill any confidence in travelers.

<!–> Rabbit R1

Kerry Wan/ZDNET

On top of Uber, which is “under maintenance,” according to R1 when I attempted to use it again this morning, DoorDash, Spotify, and Midjourney are the only third-party services that work with the device. Spotify and Midjourney worked fairly well for me, playing music tracks that matched the mood that I noted and generating images based on whatever ideas came to my mind. (I do wish I could send those generated images directly from the R1 instead of downloading them from Discord, where they’re hosted.) 

For DoorDash, you’ll have to settle with five to seven food options per restaurant as the R1 screen doesn’t display anything more. Yet even when I settled for a standard cheese pizza because Rubirosa’s famous Tie Dye wasn’t on the menu for some reason, the checkout screen gave an error message – again, and again, and again.

Also: Apple’s new AI model could understand your home screen and supercharge Siri

Lastly, there’s a big battery problem with the R1. I got about three hours of use on a single charge on Thursday. On standby, meaning I don’t use the device at all, I’ve noticed the battery drop 10%-15% within half an hour. Rabbit CEO Jesse Lyu says a fix is coming as soon as next week, so I’ll keep an eye out for any improvements. For now, expect to charge the R1 two to three times a day, depending on what functions you use.

Final thought

The early-year teasers and product launches of AI devices such as the Rabbit R1 and the Humane Ai Pin don’t surprise me. Both companies, among others, are racing against time – before tech giants like Apple and Google unveil similar, if not more useful, AI software and services during I/O and WWDC in May and June, respectively. By then, when the voice assistants already in our pockets are more capable than ever, it’ll be much harder to sell a secondary device that does much of the same while introducing its own set of problems.

Naturally, a good part of the AI hardware experience right now is unpolished, needing a group of users enthusiastic enough to pay to be beta testers. I’m always against buying a product for its promise of future updates; nothing is guaranteed. If, however, you’re interested in what the future of AI gadgets could look like, the Rabbit R1 may be the most compelling option on the market today.


Leave a Reply

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