The Rabbit R1 AI doesn't live up to the hype
The Rabbit R1 is only as good as it's AI, and it has a long way to go before it proves its Large Action Model can confidently complete tasks
I’m in the middle of writing my Rabbit R1 review, and there’s so much to like about the design and the idea behind this bright orange AI pocket companion. This is the first AI gadget worth testing – if you’re an early adopter of tech or an AI developer.
But before you spend $199, you should know that the artificial intelligence behind the novel R1 device hardly lives up to the hype set by Rabbit’s CES 2024 keynote. It’s supposed to usher in the age of LAMs (Large Action Models) by completing the tasks that LLMs (Large Language Models like ChatGPT) fail to execute.
I have more than a few examples of situations where the Rabbit R1 fails to use natural language and up-to-date AI to answer my questions and take action.
Rabbit R1 Uber demo
How it started: I should be able to call an Uber and use natural language based on a real-life scenario without diving into vehicle choices. Asking, “I want an Uber for six passengers and three pieces of luggage,” should result in it suggesting an UberXL.
How it’s going: It generates a list of Uber vehicles and asks for my exact location for each trip, something a phone-based app already knows via GPS. By having to scroll-wheel through the list of vehicles one by one, I can’t get a holistic view of prices.
I also got caught in an endless loop. The Rabbit R1 asked me for my location (I gave my current address), then wanted my destination again (I gave it), and it wanted my location again (I gave it several more times to give it a fighting chance).
Rabbit R1 Uber DoorDash demo
How it started: The Rabbit R1 is destined to scour delivery menus, so I can ask: “Order the most popular option from Pizza Hut” (the most popular is hand-tossed with cheese, in case you were wondering). What a time saver.
How it’s going: It just sends me to a limited list of menu items without understanding the request. I’m left to scroll-wheel through a long list individually and order in a more cumbersome fashion vs simply using my iPhone 15 Pro Max.
Rabbit R1 LLM demo
How it started: Asking the AI, “Which actor stars in Christopher Nolan’s most recent film?” should not only result in the answer “Cillian Murphy” but also display a photo of the actor. Part of the Rabbit R1 charm is its playfully animated rabbitOS interface.
How it’s going: The AI names the actor but pronounces his name “Sillian,” and his photo fails to materialize. To be fair, Google gets this wrong and suggests actor John David Washington from the [older, not the latest] 2020 Nolan movie, Tenet. But, so far, Google has made no big AI promises and isn’t charging $199 for an AI device.
Rabbit R1 book a trip demo
How it started: I should be able to request a comprehensive international vacation itinerary, and the Rabbit R1 is supposed to offer flight, hotel, and sightseeing options within my travel specifications. I can then modify and book directly from the R1.
How it’s going: The R1 gives me a long list of sightseeing ideas (not in a list), and I can’t book directly from the AI device. It’s very basic, and I was always skeptical of Rabbit CEO Jesse Lyu's demo during his CES keynote. When it came time to book, this was the one time the camera cut away from showing the screen.
This summer, Rabbit promises I’ll be able to ask the R1 to compile a detailed itinerary, request trip modifications, and book directly through the hardware.
Rabbit R1 AI when it comes to current events
How it started: The Rabbit R1 harnesses the power of LLM Perplexity AI to answer almost any question you throw at it.
How it’s going: It has the same pitfalls as any LLM like ChatGPT or Google Gemini. I asked the AI if OJ Simpson was still alive, and it confidently answered me: “Yes,” and then gave me a detailed follow-up about where OJ was living and what he was doing (in reality: he’s not doing much; he died two weeks ago).
The AI can’t be trusted to be 100% reliable and up-to-date, and you have to be OK with paying $199 for that.
What does work on the Rabbit R1
I did say that the Rabbit R1 is worth testing, and it’s true. You’re buying this device to essentially beta test it in advance of software updates and an eventual Rabbit R2. The orange, pocketable design (it looks like a Pokemon Pokedex), playfully animated UI, simple scroll wheel input and computer-vision camera are all top-notch.
I can check stock prices, play songs via Spotify (Premium required), and pester it with questions without much additional context (“Who wrote the lyrics to this song”). The 360° computer camera vision camera can’t take photos, but it can ID things in the real world with mostly accurate (sometimes funny) results. It’s basically taking the subject recognition tool used behind the scenes by the Samsung Galaxy S24 Ultra camera (and all other modern phones) and externalizing its findings.
Why Rabbit R1 AI have to be updated soon
We’re going to learn more about Apple’s approach to AI at its WWDC 2024 keynote in a few weeks (it’s set for June 10). That’s where we should get our first glimpse of an AI-filled iOS 18 and download an immediate beta of the iPhone software update. That will be the stiffest competition for the Rabbit R1 AI.
It’s clear why the Rabbit R1 had to launch in April. Apple could (and probably will) dominate the tech headlines once it applies AI to its lackluster Siri assistant. But, it also means you will beta-test this AI hardware until Rabbit gets closer to its original roadmap. $199 isn’t much, but you should know what you’re getting into vs their CES keynote – before jumping down the rabbit hole.