Rabbit R1: Barely Reviewable

(upbeat music) - So, this is the Rabbit R1. And it's another AI in a box, and it's a sign of the times. So this here has a lot in common with the Humane Ai Pin, for better or worse. But it's also supposed to have two specific things that differentiate it from that. Presumably so, at least. And we have to talk about it. (lyrics clicking) (tense music) So stop me if this sounds familiar at any point, but check this out. This is a virtual assistant in a box. But it's a different box. This is not wearable. It's something you should carry with you in your pocket, like a smartphone.

I'd say it's maybe the size of a stack of Post-its. But it's this lightweight plastic bucket, designed by Teenage Engineering. And boy, people love some teenage engineering right now. And it's super... I mean, you can tell exactly what it is from a mile away. This thing is very recognizable. The thing has a button on the right side. That's the one you hold down to ask questions. And then instead of a projector, here you have a built-in screen. And instead of a built-in cell phone, it has a SIM card tray next to the USB Type C port for charging.

I would say that it seems to send basically any and all requests to the cloud, but I will say that it feels significantly faster at answering questions than the Humane Ai Pin. Now, that's a very, very low bar to clear, but it's consistent, it's consistently faster. How far away is the moon? - The average distance between the Earth and the Moon is approximately 384,400 kilometers. - You can connect headphones via Bluetooth or turn the built-in speaker all the way up, like I did. But it also shows you the text of your response on the screen, just below this bouncing

rabbit

, and the battery time.

In fact, it basically always shows this

rabbit

jumping all the time. That's how you know he's awake. You press that button once to put it to sleep. Or it would do it automatically after a few seconds. Then press the button again to wake it up. Boom, just the rabbit waiting for you there. And then to access the settings, you don't have to press any buttons. There is no gesture on the screen or anything. Simply shake it like you would an Etch A Sketch. And as you do that, from there you can scroll up and down with this scroll wheel and select with the button.

This is how you move around the user interface. The other two things it has are the scroll wheel and a rotating camera. So the scroll wheel to navigate the UI instead of the screen is... I'll explain why it's weird in a second. But of course, the AI ​​assistant is multimodal, so you can use vision to answer questions about what it sees. We have seen this before. With this one, you double-tap to open it and then hold to ask. I don't know. What type of plant is this? - Taking a look now. The plant pictured appears to be a Monstera Deliciosa, also known as Swiss Cheese Plant.

This is a type of tropical evergreen climbing plant native to southern Mexico and parts of Central America. The big bright green leaves with distinctive... As a reviewer trying this, I feel like that DJ Khaled clip after a while, where I'm just pointing things out, like, "Okay, what is this? And what is this?". ?" - And what is this? - Those are pickled banana peppers. - And what is this? - Berries and seeds. - And maybe what is this? - Water. - But I also pointed it at my computer screen with a very long email and I asked him for a summary, he reads it instantly and gives me a summary.

It also does the same thing with articles. I think that's cool. But yeah, it's just... It's an AI in a box. .It's best to just answer questions. (click on text) (soft music) So yeah, this is bad for a lot of things too And this list will sound pretty familiar to you in some ways, battery life is just as bad. like the one on the Humane Pin. It has a billion-power battery inside and it's brutally bad. As if it's bad enough when you have a completely different device next to your smartphone, but when it can sit in front of you without doing anything. nothing, and the battery is visibly draining and dies in about four hours...

Then you have to charge it several times a day and it's still dead when you wake up in the morning. It's just exhausting. Furthermore, it took 45 minutes to charge this small battery from zero to sixty. And it's also missing a lot of what I would consider basic features. For example, you can't set alarms, you can't set timers, you can't record videos, you can't record photos, you can't send emails, you don't have a built-in calendar. There are many things I would like to have. assistant to make... Not here. And of course, being an AI assistant, he also freaks out and confidently answers questions wrong.

Like my starting point of asking a question I know the answer to and then getting the answer wrong happens all the time. Which is one of the disadvantages of this category. So this device was designed by Teenage Engineering. And they're really leaning into that. So... I mean, it's this bright orange, it's a really quirky, friendly-looking device, I would say, intentionally. But yes, they love their analog controls. And this scroll wheel here doesn't surprise me, but it's actually very frustrating to use. First of all, it sticks out a little at the back, as you can see. Which looks pretty good.

But that means that if it were really sensitive, it would actually shift if you left it on a table or something. So they've reduced the sensitivity of the scroll wheel, so it's actually super slow. Just like it requires a surprising amount of scrolling motion to go down a line in the setup. And then there's no haptic feedback here as you scroll to help you feel it. Alright, okay, you can get used to that. You use the button to select... But you may also have noticed that there is no Back button anywhere on this device. So to go back up a level you have to scroll to the top each time.

That's also annoying. And then to change the brightness or the volume, you actually need two coats. So you simply enter the brightness to select, hold the button with one hand, and then scroll the scroll wheel with the other hand to adjust the brightness. Which, yeah, you can't do it with one hand. It works, so you can learn it and you can say, you know, it's this quirky user interface that it has. But I feel like a lot of these issues would be solved if it were a touch screen. So what if I told you that this is a touch screen and you're not actually allowed to use it for much of anything?

Moving through these menus would be easier if you could tap whatever you wanted, right? Coming back after a long scroll would be easier if you could just flip the scroll to the top of a list and hit back. But you can not. The only thing you can use the touch screen for is basically typing on the keyboard in terminal mode. So with the terminal enabled you can turn it to the side, a keyboard appears and you can type your questions and get text answers. Which is great. And you can move between letters with the scroll wheel.

I guess it's a cool feature. But why can't we use the touch screen for anything else? Is it just them trying not to look too much like a smartphone? Maybe. Probably. (click on text) (tense music) So, okay, what's the point? Why does this exist? If it's so similar to the other one that was also so bad, what are we doing here? And really, there are two things that I think they hope are what set him apart from the other. That should, in the words of its co-founder, Jesse... - But I don't think MKBHD is saying this is the worst device it's ever reviewed so far. (people laughing) - Which, low bar, but okay.

Those two things are the price and the great action model. So the Humane Pin was very easy for everyone to use because it costs as much as a phone. It was $700 with a $24/month subscription to avoid becoming a brick. It's just crazy. So this one here is $200 and requires no subscription. So, you know, okay, that's a little different. But it also shows that it is $200. First of all, you will need a separate SIM card for it to work on the cell phone. So while there's no subscription fee for Rabbit to keep the device running, it's still a fee you'll pay every month to get data off of Wi-Fi.

But hey, the unboxing experience is extremely minimal. It comes in a single cardboard box with a plastic container that looks like a cassette tape, which also doubles as a stand, but there's literally nothing else. No charging brick, no USB-C cable, no stickers, no paper instruction manual, nothing at all. And then the R1 itself is made of plastic. Doesn't mean it's not well built. You know, there's no push-ups or crunches or anything like that. But it's definitely plastic. The camera, very basic. The speaker, very cheap. You know, there's a low-end MediaTek chip inside. The same one that's in the Moto G8 Power Lite, which costs like a $150 phone.

Almost no batteries, as we learned. And no fast charging or fancy wireless charging. And it comes in one color, a very bright orange. Like this ridiculously saturated bright orange. I'm not even kidding, the color you're seeing in this YouTube video on your screen will be the best I can do with my own color correction. But it's legitimately difficult to photograph. As if the phone camera doesn't increase the saturation enough to accurately represent it in real life. It's the brightest orange thing you've ever seen in your life, I guarantee it. Now, if you want something a little more understated than neon orange, channel sponsor dbrand has you covered here.

I have this black camo here, and now the orange parts that are showing are like more decorative pieces, which is pretty cool. But the funny thing is that dbrand also collaborated with Rabbit before launch to make sure that the screensavers for this were also available on launch day. So, that includes mine. If you want to check it out, screensavers or skins for the R1, you can click the link below. But yes, speaking of the screen, it is a 2.9-inch diagonal TFT screen, without automatic brightness. So yeah, I think you get the idea. (lyrics clicking) (soft music) So, do you know that Humane Ai Pin didn't have apps?

Well, this thing has no applications either. But the other thing they hope will separate it from the rest is what they call a large-scale action model. You know, the great language models are just the words, the language that we use. So, it is an AI that can take our natural language, process it, and in return convert it into words. And that's how we interact with it. Therefore, a great action model is supposed to take our words, then process them and turn them into actions. That's the theory. Basically, you'll be able to use apps like a human would, depending on what you tell it to do.

Now, the most important thing is that this is not the same as an API, where one company could work with others to have some kind of add-on to their services so they can work together. Because an API can be too restrictive, or not offer all the features, or just not at all... I mean, there are a lot of good reasons why you don't want to trust an API. So what this is supposed to do is just go in and use the app, like a human, essentially with a mouse and keyboard. Think of it as a virtual agent.

Honestly, I think it's a great idea. I think it's a really great idea. You know, large language models have been trained, with all their data, to respond to us just as humans would, and some are very convincing. Therefore, in theory, a large action model should be able to use these applications and services just as a human would. From Spotify to Twitter to your banking app and everything in between. It's already good enough to recognize important UI elements, like a play button or a buy button and things like that. So with enough training data it could be really good.

With enough training data, it could become really good. But the thing is, they don't have much training data yet. So far, they have made four apps available. Four. They have this online portal called Rabbit Hole, where you can log in and enable them all. Spotify, Uber, DoorDash and Midjourney. So the Rabbit can talk to these four apps and has an on-screen UI to do it exactly the way we're theorizing. Then it will play songs from Spotify. For example, I can ask you a song and it will try to play the correct one and show you any information you need on the touch screen to confirm or refine the actions you are taking.

But even now, they work. I already had problems because it played the wrong song.I've also seen other people already have problems with the DoorDash app by doing things wrong, which is even more frustrating. I can't imagine how frustrating it would be if the Uber app crashed. It definitely sounds like it needs more training data. And that's just for these four initial applications. Now, Rabbit says they already have 800 different apps trained, but they haven't built a UI for them yet, so it's not available in Rabbit yet. And that is still in process. Now, in theory, they've also started working on something called generative UI, where you can essentially recognize what kind of app it is and then create a UI so Rabbit doesn't have to...

But that's still theoretical too . and also on the roadmap and still in process. And then if there's another app or service you want your Rabbit to be able to do, like something for work or something super specialized you've invented, they've talked about something called "Teach Mode." Basically, it's having Rabbit watch you do what you want it to do on your mouse and keyboard, and then it learns from what you did and then you can repeat your actions later, which is sick. Unfortunately, you can probably already see where this is going, teaching mode is also not available at the moment.

This is something they are still working to examine and finalize. And that will be at the end of this year, also in process. So as of now, this device has none of that, just the four types of half-functioning apps we went over at the beginning. What are we doing here? (lyrics clicking) (soft music) Okay, I'll try not to turn this into a rant, but I feel like we need to at least acknowledge that a lot of these tech companies are developing technology a little bit backwards. As if they are delivering products so unfinished that it actually makes them almost impossible to review.

It feels like it's just, make the thing and then put it up for sale. Now it's like putting it up for sale, and then delivering it half-done, and then iterating on it and improving it, and hopefully with enough updates, then it'll be ready, and that's what we promised a long time ago when we started selling it. And then this whole period in between is a disaster. And it also applies to all kinds of product categories. We've seen this with games. Like big studios are handing out half-baked games saying, "Oh, you know, it's an alpha. There will be updates." But in the meantime, it's a full-priced AAA game that just suffered from an unacceptable amount of bugs and issues.

It's also happening with cars. And vehicles are advertised and then delivered in a half-finished state, where you simply don't get many of the features you paid for. And eventually they will arrive soon, with a software update. You know, smartphones, obviously, we've been seeing this for years, but it seems like now more than ever there's at least one feature, one big feature in every smartphone release that's announced, but it's not coming until later. year. And now these AI-based products are on the cusp of this horrible trend where what you get at the beginning is practically non-functional compared to all the promises, all the features, and all the things that some day will be.

But you still pay full price up front, which is what makes it so crazy. And that obviously sucks not only for the people who buy it at full price at first, but also for reviewing these things in general. For example, how do you evaluate a product where the version of what was promised in about three years, what could be is incredible, but the version that is delivered now is water for dogs? How are those dots connected? Do you even connect those dots? Are you supposed to give them the benefit of the doubt? I don't know. What are we doing here?

I mean, I guess on the one hand, it's nice that tech products, these things, can get better over time and get better, and what you buy can be better tomorrow. All of that is great. That didn't used to be true, and that's beautiful and all. But on the other hand, the flip side of that coin is that some products are delivered unfinished, and I feel like it's going to get worse before it gets better. (lyrics clicking) (tense music) So let me tell you, I'm personally very excited about the idea of ​​a super-personalized AI assistant that can do everything a human assistant could do.

That is my dream. I want that to happen. And I'm interested in these different companies approaching it from different angles. And hopefully, we'll get there eventually. But it is clear that this will also require a lot of time, effort and technical development, and also a lot of data. As I already said this in the Humane review, I'm pretty sure that a good assistant needs to know everything about you, whether you're human or virtual. It needs to know your preferences, your location, what you are doing and what you like to do. And every little thing. There is so much data.

As for the Rabbit...I think that's why they focused on getting such a low price. I mean, all of this is a hard sell, because you're betting on what the product could be one day. But for the Pin, you have to spend $700 and $24 each month, on the off chance that it will get there one day. Like it's a lot of money. But for this cute little thing, 200 dollars. It seems like a much easier investment to make. A much easier pill to swallow, just go for it. And if it turns out great in a couple of years, you'll feel it's worth it.

You know, on the off chance that some of those 800 apps are super useful. Or you know, even if you already have a smartphone, maybe teach mode lets you do something with the press of a button that you would never have dreamed of before, then it might be worth it. The whole situation reminds me a little... If you look back at the beginning when Tesla started shipping cars with the whole autopilot system. They sold cars and became popular for other reasons. It was a good electric car with good range and a good charging network. And then people started kind of beta testing this autopilot system, and Tesla started collecting millions of miles and hours of data about how it works in the real world.

And so all that data helped their systems learn and get the edge they needed to try to create the best autopilot software. So I'm sure the Rabbit would like to take advantage of its price advantage and get as many of them as possible into the real world, into people's hands, so they can start beta testing, train and improve it. a variety of tasks. But the problem with the Rabbit is that it doesn't have like the other reason people would have it anyway nowadays. So it's kind of a chicken and egg problem. So if you want my advice, you've heard it before.

Buy the product based on what it is today and not what it is promised to be in the future. And with this category, that's the hardest thing to do, the hardest thing to keep in mind. Because the promise is very great. But that's still my advice. And I guess we still have to see what big companies like Google and Apple will try to do in this space, probably this year. So with all that, we'll see. Thanks for watching. See you at the next one. Peace. (soft music fades)