Rabbit R1 & AI Pin: Future of Smartphones or Tech Dead-End?

It was probably inevitable that some engineers and entrepreneurs would build hardware rather than just smartphone apps around the new generative AI tools that have been blowing minds over the last couple of years. Certain circles in tech believe that the smartphone may be due for a successor technology and many believe it will be powered by a talking AI. Imagine a portable device that anticipates your needs, conducts research, then executes multiple tasks employing multiple apps from a single command. That’s just part of the vision of an agent-centric interface where you interact with data through an AI assistant that may someday make today’s smartphones seem quaint. It could drastically simplify and optimize systems for digital audio and smart home functions. But these new devices from two fast moving startups, Rabbit and Humane demonstrate that the technology just isn’t there yet.

Humane AI Pin & Rabbit R1

AI Pin by Humane and Rabbit Inc.’s R1 have recently launched and show us possible directions for the smartphone or its successor. But consumer reviews of these products suggest that neither device was fully market-ready. Both devices suffer from the all too common “launch-as-beta-test” phenomena in today’s technology. But Humane and Rabbit may have targeted tight release dates to stay ahead of two major conferences, Google I/O recently and Apple WWDC in June, which are expected to showcase new AI-powered Siri and Google Assistant.

Both devices can use a wireless network, Wi-Fi and Bluetooth to talk to us from a cloud AI through sophisticated text-to-speech and voice recognition. Functionally, responses from cloud AI will be delayed at times as demonstrated by several hands-on videos. Both offer silent modes too, the AI Pin uses a micro-laser projector OS controlled by hand gestures, while the Rabbit R1 relies on a touch screen/scroll wheel combo. At present, what Humane and Rabbit offer are tools to boil the Internet down to a single interactive voice, with varied results.

Rabbit R1, Mad LAM

Rabbit Inc. founder Jesse Lyu created R1 to integrate with apps on your smartphone and other computing devices. The list of compatible apps it can use on your behalf is limited at launch, but app compatibility is what should make R1 the more practical of the two new devices. You can use it to access apps by voice without touching your phone, including those apps that can make online purchases. In his product presentation, Lyu tells his R1 to order a pizza and book flights in transactions that appeared to run a little too smoothly. Getting information from a generative AI is fine, as long as fact-check later. But trusting them to use apps with your encrypted credit card numbers seems to put a bit too much faith into a technology that’s known to hallucinate. You might as well give your Mastercard to Jim Ignatowski from Taxi to go on a pizza run. Okey dokey!

Designed by Sweden's Teenage Engineering with design cues borrowed from Swatch

But what makes Rabbit R1 truly interesting is its unique LAM AI (Large Action Model). The LAM function will let you teach R1 “actions” in software interfaces in a similar way that LLMs (Large Language Models) use words. With appropriate training data, it could save time performing complex or redundant tasks in nearly any software. 

R1’s LAM capability is promising, but its limited app compatibility and finite training data makes LAM more of a future consideration. Early hands-on reports from the Rabbit R1 subreddit suggest that the LAM’s Teach Mode was unavailable at release but should launch soon. Despite user’s frustrations, with many early adopters finding themselves beta testers, the Rabbit R1 could grow into a truly groundbreaking device someday. If a dedicated community of users push the limits through modding and crowd-sourcing LAM training, it could really take off in unexpected directions. Unfortunately there is less optimism about Humane’s AI Pin.

Rabbit R1 Bell Curve: Device uses Perplexity Pro AI

AI Pin, Good Intentions in Tech 

Many of the brilliant minds in big-tech including engineers like Jaron Lanier have given us insights into our collective smartphone addictions. These not quite whistle-blowers have alerted the public to what most of us already knew; our apps are purposely designed to exploit our brain’s reward system, keeping us addicted to bolster their revenue streams. Some of these magnanimous former tech developers now engage in a sort-of Munchhausen Syndrome by engineering, including the founders of Humane Inc. Veteran Apple power-couple, Imran Chaudhri and Bethany Bongiorno now seek to counter tech junk food with their own brand of technology vegetables. Humane's mission statement describes a new kind of tech startup: 

“…building innovative technology that feels familiar, natural, and human. Technology that improves the human experience and is born from good intentions.” 

The company’s website design reflects a modern, west-coast consumer ethos, resembling more new age spa than a tech company. In his 2023 Ted Talk, Chaudri showcased the AI Pin, emphasizing that it interacts with users naturally as they remain present in their surroundings. Humane’s vision is commendable, you have to respect that it wants to be a smartphone alternative rather than extension, even if it comes at the cost of broader consumer appeal.

AI Pin is essentially an AI voice assistant that can also perform text/voice communication but it can’t use your existing cell-phone number. At launch it has virtually no apps nor integration with your smartphone outside of syncing with your contacts. Future updates could change this, but for now it's a functionally lean device.

Audio Capabilities: AI Pin vs. Rabbit R1 

Both AI devices have limited support for streaming music services at launch. AI Pin has Tidal and R1 can stream Spotify. Both AI devices will stream music directly to your Bluetooth hardware where sound quality will depend on supported codecs. Rabbit R1 hasn’t published codec compatibility (at the time of writing) on its website, but regular Bluetooth-users will know that probably means AAC/SBC only. AI Pin on the other hand has a full complement of codecs including Bluetooth 5.1, LDAC, aptX HD and of course AAC/SBC. That’s an impressive list that suggests AI Pin’s emphasis on quality Bluetooth streaming. But both have potential for even better voice-activated streaming music capabilities.

If you use a Wi-Fi music player or AVR at home and AI Pin allows you to access the Tidal Connect feature, you’ll bypass Bluetooth altogether and stream directly between your player and Tidal’s library in the cloud. This will permit access to Tidal’s lossless and Hi-Res library.

R1’s LAM feature has even bigger audio potential since you could teach it to control any subscribed streaming service on any device. You could even teach it to control Roon servers or a UPnP client to enable Wi-Fi streaming by voice control throughout your home.

“Responsible” Tech Reviews? 

MKBHD aka Marques Brownlee has been a prominent consumer tech reviewer on YouTube for over a decade. His voice carries across the tech world like E.F. Hutton and it's not just his 18+ million followers listening.

Brownlee delivered a savage review of Humane’s AI Pin in a title that doesn’t mince words: “The Worst Product I've Ever Reviewed... For Now.” His review was honest, recognizing Humane’s ambition and future potential, but also that the device falls well short of its promise. His review earned Brownlee backlash from the tech industry blue-checkmarks on X (Twitter) that raged for days. Tech industry figures seemed to close ranks around Humane and even implied that reviewers bear some responsibility to tech startups releasing a first product. Arguments ensued and this is just a small sample of what sparked it:  

“@MKBHD review is devastating for the future of Humane….is going to eviscerate sales.” - Alex Kehr 

“I find it distasteful, almost unethical to say this when you have 18 million subscribers… with great reach comes great responsibility. Potentially killing someone else’s nascent project reeks of carelessness. First, do no harm.” - Daniel Vassallo, Small Bets 

Humane Inc.’s Head of New Media, Sam Sheffer eventually stepped in to cool ongoing disputes about the job of product reviews. He praised Brownlee’s review as “solid” and a “gift” while promising Humane will take his observations and move forward. Shafer displayed professionalism in his response’s humility and commitment to improvement. It’s the right choice for a company in full damage control.

Smartphone, Kind of OP 

The AI Pin and Rabbit R1 launches illustrate the tech industry’s current state. A cynical view may suggest that Humane and Rabbit simply capitalized on AI buzzwords to attract VC funding and attention. The smartphone’s AI-based successor may be on the horizon, but it’s probably not today’s handheld-sized alternatives. As Marques Brownlee says in his AI Pin review; “The smartphone is kind-of OP”.

Snapdragon 8 Gen 3 will give phones some powerful local AI capability

For now, the revolutionary next-step for the smartphone will likely be a new AI-powered smartphone. The recently launched Qualcomm Snapdragon 8 Gen 3 chip features on-device generative AI, supporting both large language and vision models. The premium phones that use it will bring faster performance to a virtual AI voice assistant, image generation and enhanced video and image processing. Your next phone is likely to be a vastly improved talking AI that will seem more lifelike than Siri or Google Assistant has ever sounded before.

Conclusion: A Dark, Colorless Internet of AI Voices 

Jesse Lyu’s presentation of Rabbit R1 demonstrated impressive tech and a visionary talking AI agent that can employ multiple apps to execute tasks or provide the optimal answer to your request. But if the future of the Internet is a talking voice-box, count me out! A dark talking Internet seems bleak. The latest version of ChatGPT, GPT-4o has now put the talking Internet into an app on your phone. The latest release from OpenAI engages you conversationally, like a friend, capable of sarcasm and backed by the Internet's collective knowledge. A truly voice-interactive Internet is a remarkable achievement that could serve a purpose in certain situations. But interacting with the world’s data in single stream of spoken words that simply tells us answers misses out on our human instinct of discovery. It also begs the question, who or what controls the answers we receive? Even if you see the GPT-4o app as purely entertainment, the alleged voice of Scarlet Johansson engaging in quippy dialogue would soon grow tiresome.

For all its addictiveness and absorption of our attention, “the screen” is a powerful tool for the mind and it’s not just my own dyslexic brain. We all process multiple threads of data in parallel with less cognitive load when information is presented visually. Using a talking AI for research, advice or recommendations may give you exactly what you asked for. But I think we’ll lose something important without the probabilistic randomness that search engines provide. I can’t believe I’m already feeling nostalgic for search! 

Let's say you're going out for pizza and search for recommendations. It's empowering to scroll through lists of options on your phone as you scan text, icons and thumbnails just to see what catches your eye. Verbally asking an AI for the best pizza place in town may lead to a great pizza. But visually scrolling through local pizza options, each with captions and images of restaurant interiors makes finding a new pizza place into more of an adventure, even if it turns out the pizza kind-of sucks.

Two questions remain; do people really want to talk to the internet now and do you trust AI?