On the face of it, you’d be forgiven for thinking that Rabbit’s R1 device is quite different from Humane’s AI pin. They got presented in considerably different ways, and Rabbit got a considerably more positive reaction, being possibly the most interesting thing announced at CES.
Both companies presented their products pretty differently. Humane spent years giving off this air of mystique surrounding the company, playing coy about what they were building and letting out some teaser concepts gradually over the past year until eventually releasing an bafflingly un-compelling product intro last month.
Rabbit’s keynote came off as a little more scrappy and humble, and its CEO (and the product itself) had a lot of personality. Rabbit’s presentation made a bit more sense, trying to sell the “why” before showing the “what”. That strategy worked (or maybe people just responded to the $199 price with no subscription fees), and it sold out on its first day.
These two devices and companies aren’t so different, though. They’re both rooted in this idea that you can sell an entire product as a wrapper around AI.
Both companies say “we don’t do apps”. Instead, the AI is just there to do what you need the computer to do, and you don’t need to think about how it’s going to do that thing. They sell this as a principled stance on what makes a better product, but I doubt they really believe that. The reality is that it’s damn near impossible to build a brand new app platform in 2024. So instead of a futile attempt at making an app platform, they just tell you you don’t need one. Whether users need (or just want) apps is a whole other story though, and the fact that both devices are being pitched as ones you use in addition to your smartphone should tell you how much confidence you should have in these.
Despite both companies selling devices with AI at their core, neither company seems to have AI as a core competency, and instead they seem to be using off-the-shelf AI tech from OpenAI (Rabbit has trained a model on using apps, but it’s unclear how much of that tech was developed in-house). Both companies lean heavily on just saying “AI” and hoping your imagination runs wild with ideas about things you can do. Both companies’ demos try to show off things that seem nominally impressive in a demo, but no requests ever get too deep in asking the AI for things. There’s never an interaction where you have a long dialogue with the AI about what you’re trying to accomplish. At best, you can give a command to the AI, and then maybe refine it once or twice. The latest large language models are a leap forward in terms of how you can query them like a human, but they’re still a far cry from being able to be talked to like a competent human assistant. Eventually, you become frustrated after you’ve asked enough things to the AI that seem reasonable to a human but the AI can’t handle, and you realize there is no magic, and that it’s just yet another computer tool where you have to communicate to the computer in a particular way to get results, only now the results aren’t guaranteed to be correct.
Both of these companies and products should be setting off your bullshit detectors.
The presentations themselves should be your first clue. Both companies try to ape Apple’s keynote style, and both fail miserably in their own unique ways. Apple’s keynote presentations aren’t great because the slideshow uses a black gradient, or that the speakers are casually dressed either on stage or in a suspiciously clean office space. Apple’s presentations are great because they rehearse until they’re polished, and they are constantly refining the structure of presentation for maximum impact (also, Apple’s products tend to be materially compelling)
Humane’s product demo sounded rehearsed (if very stilted) no one ever asked whether they were covering the most important things first. Seriously, in the first minute or so of the presentation you learn it comes in 3 colors (referred to as “colorways” which I know is technically correct but it’s so pretentious), and its batteries don’t last all day but are hot swappable so you can achieve “all-day” battery life. Who the fuck thought those were the facts worth leading with? Rabbit’s CEO thought a little more about the structure of the presentation but didn’t put the effort into rehearsing.
Both companies are also riding on the credibility of more reputable companies. Humane hired a lot of ex-Apple employees which automatically piques the curiosity of plenty of people, and Rabbit developed their device with the help of beloved design firm Teenage Engineering. This isn’t inherently a bad thing (and I don’t think either company has leaned too heavily into this), but it’s worth being mindful of. I’d have dismissed Humane entirely if not for being full of Apple alumni, and I don’t think I’d have had any interest in Rabbit’s device if it didn’t have the unmistakable cuteness of Teenage Engineering (also, Teenage Engineering doesn’t work with just anyone).
Humane did the classic vaporware move of spending years being “stealthy,” but not so stealthy that they wouldn’t shut the hell up vaguely hyping up their product. Lots of great products get developed in stealth mode, like iPhone. The critical difference here is that you didn’t hear a goddamn peep from Apple until they had a compelling product to show the world, and it was indeed compelling.
Rabbit, on the other hand, is playing it a little different, instead coming right out the gate announcing an imminent ship date in March of their first devices. That really shows up Humane if it’s true, but also this is the level of shipping optimism I get from just about every Kickstarter campaign I’ve ever seen where someone who has never made a physical product tells you they’ve got the entire thing built, and all they need is funding and they’ll start mass production, and so you back them. Then you start getting email updates. First it’s fun stuff like “wow, look at how they make these molds!” but then pretty soon you realize they’re getting a crash course in “manufacturing shit is harder than we thought!”. Then come the delays. First it’s just this one little thing, then they’re back on track, and then the complications keep coming, and eventually the updates get less frequent. I will gracefully proclaim myself an asshole if March comes along and they start shipping without a hitch, but their shipping projections ring just a little too optimistic. Also: apparently customers’ credit cards are charged at the moment of preorder, not when shipping, which is a massive red flag. Finally, Rabbit’s “no subscription fees” model screams “bait and switch” because the AI processing isn’t happening on-device; it’s happening in their cloud service which costs real money to run, and I don’t even think this $199 retail price covers the cost of making it.
The entire premise of trying to sell AI as a product is bullshitting people. The latest AI tech makes great demos! It’s super fun to play with ChatGPT, and sometimes it can really impress me and do things of real value that save me time, but it’s mostly a demo and it has a huge problem making up shit. We had the same hype problem with autonomous vehicles which we thought were around the corner a decade ago. Tech execs got dollar signs in their eyes and imagined autonomous vehicles disrupting everything but didn’t bother with the pesky reality that the tech didn’t really exist. Companies are repeating that mistake now with AI, trying to sell you on a vision where you have a virtual assistant that you can talk to like a human but they’re supercharged with a computer’s speed and capabilities. And yes, that would be really transformative, if this tech could actually do this. Instead, we’re setting ourselves up for disappointment.
AI isn’t a product. It’s a technology, and when a company’s central selling point of their product is “AI,” that’s a clue that they don’t have a product.