Rabbit R1's Device Just GOT UPGRADED (Even BETTER!)

So the new Rabbit R1 device recently just got a major upgrade and I need to show you all why. This is really, really cool and some of the things you didn't know because there are some exclusive videos that show us just exactly what's going on with this device. So let's actually take a look at some of the things. If you're, you know, buying the Rabbit, if you're looking forward to it, you do want to know. So number one is that these things are selling fast. Like really fast. Okay, so the fifth batch of 10,000 rabbit R1 devices has sold out. Pre orders for the sixth batch totaling 50,000 are now available at Rabbit Tech. An expected delivery date for the sixth batch is June to July 2024. And for all addresses in the EU and UK, batches number one to six will be shipped by the end of July 2024 on a first come, first serve basis. So if you're one of the people that ordered this when you first saw it, you're likely to get it earlier than someone that orders it currently, now. And remember, this is only $200, which is around 170 British pounds, around €170. Now, what was recent was there was actually a really new announcement that essentially they dived into. And essentially this was an announcement with perplexity. Now many people actually don't know what perplexity is, which is why I'm making this video so you guys can understand. So essentially, Perplexity is basically like Google search, but it combines AI to be more effective. And honestly, you can't knock it until you try it because it is really, really effective. So take a look at the Perplexity trailer so you can really understand exactly what's going on. Because trust me when I say it's really, really effective. So take a look at this and I'm going to explain to you their announcement.

Whether you're navigating the maze of headphone options, drowning in news noise, or stalled on your Japan trip plans, Perplexity Copilot is your guided search assistant. Just turn on Copilot and ask. It takes a deep dive into anything you want to know and delivers tailor made, concise answers. Forget about diving into a sea of links. Copilot does the legwork by grasping the essence of your question. To fine tune your search. Copilot engages you with clarifying questions. This ensures you get what you're actually after. Once it gets what you're asking, Copilot scours a vast array of sources to ensure relevance and quality. Want to know More every source is just a click or tap away for deeper exploration. Let's say you asked a quick question, but the answer wasn't what you were looking for. Easy. At the bottom of your quick answer, hit Rewrite and select Copilot to turn your quick search into a guided search experience. With Perplexity Copilot, you're not just searching, you're gaining a new window into the Internet. From the simplest questions to your deepest inquiries. This is where knowledge begins.

So Perplexity is really, really effective at what it does. And for those of you who know what Perplexity is, and for those of you who use it, you're going to know exactly what I'm saying is so true. That's why this announcement is so cool. Because Rabbit actually partnered with Perplexity to provide everyone who buys their Rabbit device a year completely of this. Now usually I think this is around $10 or $20 a month, but, but currently they're going to give you guys a year complete free. And trust me when I say this is going to make the Rabbit device so much better. That's why I said that this has been completely supercharged. Now later on in the video, you're going to see some videos of the actual Rabbit device because the founder actually shared some videos on Twitter, things like size references, some other cool stuff like that. And there was actually a mention of Rabbit's device by the CEO of Microsoft. But so then we have the founder of both companies here on a Twitter space talking about this announcement and I think you guys should listen to this because it is really.

So I'm pretty excited to share that Perplexity and Rabbit are partnering together. So we are excited to power real time precise answers for rabbit R1 using our perplexity online LLM APIs that have no knowledge cut off, just always plugged into a search index. And the first hundred thousand rabbit R1 purchases will also get one year free of Perplexity.

Probably where that came from. I didn't know that business has that feature, but yeah, continue everyone.

Okay, yeah, the first hundred thousand Rabbit RN purchases are going to get one year free of Perplexity Pro. So it's basically like Perplexity Pro, one year free is 200 bucks. So if you pay 200 bucks to purchase a Rabbit R run, you're getting twice the value.

Yeah, so we, you know, we had a interaction on X a couple of days ago and then what's going on on axis the following couple of days? Team's been working really hard together to make this happen. And I think, you know, to me it's a no brainer if you think about, you know, rabbit R1 with price of 1 99. No, actually not 200, but $199. No subscription.

Fair enough.

And the Perplexity errand is generous enough to offer Propensity Pro for a whole year that worth actually 200 bucks.

So yeah, there was that announcement that was really cool but there was also some other stuff. Okay, so I said the Microsoft CEO Satya Nadella actually talks about just how good Rabbit was. And I can't imagine how this must feel as the Rabbit founder seeing the CEO of I think it is now the world's largest company. Talk about the product that you've created.

See, I thought the demo of the Rabbit OS and the device was fantastic. I think I must say after Jobs sort of launch of iPhone probably one of the most impressive presentations, presentations I've seen of capturing the vision of what is possible going forward for what is an agent centric operating system and interface. And I think that's what everybody's going seeking, which device will make it and so on. It's unclear but I think it's very, very clear that computer, I go back to that, right. If you have a breakthrough in natural interface where this idea that you have to be go one app at a time and all of the cognitive load is with you as a human, does seem like there can be a real breakthrough. Because you know, in the past when we had the first generation, whether it was Cortana or Alexa or Siri or what have you, it was just not, it was too brittle where we didn't have these transformers, these large language models, whereas now we have, I think the tech to go and come up with a new app model. And once you have a new interface and a new app model, I think new hardware is also possible.

And is that an opportunity for Microsoft.

Or are you moving away from hardware?

I mean, look, I mean always it's an opportunity.

So that talk right there was really fascinating because Microsoft seemed to be kind of eyeing up the hardware market. And I mean you have to remember it was a couple of years ago. In fact not just a couple of years ago. In fact it was I think around 15 years ago where it, Microsoft really just pulled the plug on their device which was the Windows Phone. Some of you don't even know what that is and rightly so because it just didn't go well. And it just goes to show how hard it is to make a consumer hardware device that actually does succeed. And it will Be interesting to see if Microsoft just jump back into this, but I don't think they will unless because OpenAI are going to be working on a device too. But I think if you watch some of the other videos that I talked about in, where I talked about Ray Ban's AI glasses, I that are going to be coming in the future, I think that that is going to be an interesting point. Now something as well that many people did miss was how Rabbit actually works. And in the original video which I made discussing Rabbit's amazing tech, I didn't actually show this video from their website where they actually talk about large, well language action models and essentially their new proprietary system on how they actually use agents to, I guess you could say interact with the web because LLMs are good, but they are text based and that's essentially their purpose. They can be repurposed for other things, but that's not what they were made for. So they essentially made lams. And in this demo they essentially talk about how large action models are pretty much better than anything we've ever seen. And it's a new foundation model that understands human intentions on computers. So I think this is a really interesting watch. And then after this I want to show you guys some of the videos of Rabbit like actually being used. So some more in person demos because I think it's really, really interesting because I know, I know like everyone who's ordered it, you probably want to know how big it is, you probably want to know how certain things work for certain capabilities. So I'm going to show you that in a second.

We can teach Rabbit OS how to use specific applications. In this video I'm teaching a Rabbit how to book an Airbnb while I'm operating normally as a human on the left screen. Watch closely on the right as the large action model is learning all my inputs and imitating my behavior in real time. So I'm trying to plan a trip to Barcelona with my wife and my daughter. The first thing I'm going to do is navigate to the anywhere option and I'm going to type Barcelona in the search field. The system is suggesting Barcelona, Spain, which is exactly where we want to go. Using the website's calendar tool, I'm going to mark our check in on the 15th and check out on the 21st. Now I'll click Add Guests and adjust the members accordingly. Now let's hit the search button and see what pops up. Since we love the beach, let's make sure to select the beachfront option. And for more private experience, I'm going to select entire home so we have the whole place to ourselves. For the budget, I'll set a maximum at 400,000 won and a minimum of 100,000 so that all the options are within our price range. We're going to need at least two bedrooms to make sure we all have our own space. Finally, with all of our preferences set, we've got plenty of options that fit the bill. I'll just start browsing for the perfect one. Each training only takes a few minutes and does not require access to an application programming interface, also known as an API, nor do you need anything installed on your device. You only need to train each workflow once. Let's try to use Rabbit OS and instead book a room in London. My extended family is going to London. It's going to be eight of us and four kids. We're thinking of December 30th to January 5th. It's not set in stone yet, so I just want some general options. Can you look it up for me?

Sure, I can help you with that. The first option is a home in Portobello Muse house, priced at 1,348,351 per night with a rating of 4.8.

The large action model supports mobile apps, web apps and professional desktop apps. It learns directly on the user interfaces and acts on them. We have already started the training process for the most popular apps. As you're watching this video, Rabbit OS is learning fast and adapting to hundreds of applications. The ultimate goal of Rabbit is to define the first natural language operating system that replaces apps on your device. It's time for the machines to do some serious homework.

So, yeah, I think you can understand why this product sold the way it did. Because if what they're saying is even remotely true, I mean, training, you know, this takes minutes. You know, it needs no API required that, you know, you can do it within one without software. One time. That's what they said. Each workflow, you just need to train it once. I mean, if that is really true, and that's a bold claim, they are definitely, definitely breaking new ground here. So I would say that that is absolutely incredible, but that's just some understanding of how it works. Then of course we had the benchmarks, which I found to be really, really cool because I actually compared it to GPT4, GPT 3.5, Flan T5XL, some of the other things, and you can see just how good lam Large One is Neuro Symbolic, their new proprietary model. And then of course, we get to the size References. So this is where the founder actually talks about just how big this is because some people might want to see just how this is how it works, what it, what the size is. Just some cool stuff like that. And then he also shows two other videos. So I want to show you guys this because I think it's important to see just how big it is. And I kind of wish he did compare it to like an iPhone because I feel like this might not replace the iPhone, but it's still a similar handheld device but nonetheless definitely worth a watch.

But the idea is, you know, seven.

Years ago when I designed Raven H, I have this magnetic table detachable pixelated controller that it just docks on the main device like that. But the idea is that you can carry around and you can kind of like just hold and talk. But R1 is actually smaller than that. If you put it on top, it's smaller than the footprint, but it's exactly the footprint. The width wise is exactly like a iPhone Pro Max model, but 50% of the footprint. That's kind of like the idea.

So yeah, he said it's pretty much the same size like an iPhone 15 Pro Max, but just half of it. Then of course this is something for I guess you could say accessibility. So he talks about why, you know, you don't need a left handed version.

Hey, this is Jesse and here's my R1. A lot of people on the Twitter been asking hey can you guys make a L1 for left hand users? Because they think that all these controllers and the button scrolls on the right side and probably specifically particular design for right hand. Well that's actually not the case. I'm actually a left hander so this is how I feel most comfortable holding the R1 using my left hand actually. But if you look at this, if I hold like this in my hand, the push the dog button, actually my mid finger just naturally lens right here for the PTD button and for scrolls I basically scroll from back like that without breaking your gesture. So I just hold like this and I press play Get Lucky from Daft Punk.

Okay, so that was really effective. So for any of you who are left handers, this is not going to be a problem.

Here's my R1. So here with.

Then of course he shows another sneaky demo where he talks about the rotational camera. So this is obviously worth a look.

R1, let's have a close look at the rotational camera. The camera by default points down which has a physical block for privacy. But if you, you are about to use it you go to vision.

Double.

Click and then it just rotates. Let's try that one more time. Go back, it points down and enter the vision. It rotates and obviously you can flip to the other side as well. Cheers.

So, yeah, so I think it was really, really fascinating on how they managed to make this device on the space. I did get a few clips, so I did manage to listen to the entire thing. It was around 48 minutes. It was definitely some fascinating stuff. And they actually talked about three things. So there were three things that I do want to show you from this because they talked about the future of AI assistants. They also talked about how they achieved a 500 millisecond response time and they also talked about how they reduced latency. And those are the three things that I think are most important for the future because reduced latency makes us, I guess you could say, enjoy our AI systems more because it sounds more realistic, because they respond quicker. And of course the future of AI systems is important because these guys developed a proprietary model which seems to be better than anything currently on the market. So this talk right here is how they achieve that 500 milliseconds response time. And I think it's an interesting.

Listen, if you press this button, the microphone starts recording and you're recording in an audio file and that audio file needs to be converted into strings and that strings send it to the dictation engine or tts. Text to speech, sorry, speech to text actually speech to text engine and convert to text and then that text sends it to OpenAI ChatGPT API or Perplexity API or whatever large language model for intentional understanding and then start generating based on their speed. But we made a streaming model to where we basically cut off the chunks into a very, very small timestamp chunks and would make the entire model streaming ipa. We do have a technology make the sequence into a streaming. We're not necessarily accelerating GPT or perplexity speed at the moment, but with this streaming mechanism that if you ask non search up to date information, we're constantly hitting the benchmark which is 500 millisecond per response. Because again, this is whatever we're going to push. This is going to be industrial standard because right now, Absolutely right now this. It is what it is.

Yeah, absolutely.

So that right there is how they talk about what they're going to push. Then of course they additionally dive into.

Some more details and where would you consider your. Whatever latency you have today? How do you compare that with, you know, other similar apps like Chatgpt voice to voice. Like have you, have you tried looking at comparing the two?

Yeah, so we did have a technology we call kernel that we started working on this pretty early, more than two years that would basically establish a streaming model because if you think about why there's latency. So if you press this button, the microphone starts recording and you're recording in an audio file and that audio file needs to be converted into strings and that strings send it to the dictation engine or tts. Text to speech, sorry, speech to text actually speech to text engine and convert to text and that text then into OpenAI chatgpt API or propensity API or whatever large language model for intentional understanding. And then it starts generating based on their speed and then it's a round trip, right? This is a single trip and everything reversed again. So if you add all this together, if you just go there and build a voice AI with no optimization based off GPT4, we know for a fact that single dialogue you're looking at probably five to six seconds. But we made a streaming model to where we basically cut off the chunks into a very, very small timestamp chunks and we make the entire model streaming. I think I'm not the best guy to talk about this. Maybe our CTO later on can write something about this. But we do have technology make the sequence into a streaming. We're not necessarily accelerating GPT or perplexity speed at the moment, but with this streaming mechanism that if you ask non search up to date information, we're constantly hitting the benchmark which is 500 milliseconds per response. But I wish everyone, I wish our team, you and me can do something just on up to date information search and maybe we can push this bar a little bit because again, this is whatever we're going to push. This is going to be industrial standard because right now, absolutely right now this. It is what it is.

Yeah, absolutely. Yeah. We are certainly at the cutting edge here. And in fact like the fact that you wanted to do it through streaming, that already makes it much like the perceived latency is already a lot better than waiting for the full response. And you know, I think there are so many more things we can do to speed it up.

Yeah, that's where they talk about how they are compared to ChatGPT and it seems like it's going to be getting even better. And this is the final clip where they talk about how the future of AI assistants is going to transpire.

So maybe I want to lead from there to your thoughts on the whole voice to voice form factor because the Rabbit device is definitely taking us beyond just consuming screens and text in the form of pixels to just interacting more naturally. So what are your thoughts on like the next stage of how people consume and interact with all these AI chatbots and assistants?

Yeah, so I think, you know, being our age, you know, we've grown up unfortunately where the dictation engine were never invented and then it was invented and it was put in use in a horrible way that I think our current generation are victims of the, of the early days of the dictation engine, early days of the national language processing before for large language model of course and transformer and all that. So I think me personally, I identify myself as probably along with everyone here is a PTSD with the early version of dictation engine. That's why I guess it creates such a strong impact on our mind that okay, maybe voice is not the right way to go. I rather prefer type. But I think our principle is very simple, is that what's the most intuitive way for communication, right? Like think about everyone. If we convert this twitter spaces into a type twitter spaces or into you know, even worse like a fax Twitter spaces non instant message twitter spaces. I don't think that, you know, we can deliver all this information in relatively short period of time. So you know, if you think about how human communicates with human and you know, before, I guess before the neuralink stuff become put in use and natural language, especially conversation in voice is still to be the most efficient way. Now the problem becomes easy because we just need to fix the ptsd. But I think if you look at the past three, four years, probably like especially past three years, a lot of the fundamental infrastructure around that has been significantly improved to where you know, the younger generation especially, I'm not sure if, I'm not sure if you know how many of the listeners here got like probably like 5 year old, 6 year old, 7 year old kid. But the younger generation that they were born like you know, after, I guess after 2010. I see among all the kids that you know, they actually prefer the dictation icon on the keyboard rather than start typing. So I think the use behavior in a different generation already starts shifting. And of course the fundamental reason is because of lot of infrastructure are good enough, are redundant enough. So for us we are not saying that you can only talk to R1. If you shake the R1 the keyboard will pop up. But if you think about the most intuitive way and if you're in a rush, there's nothing better than just find that analog button, press and hold and start talking. So I guess that's our design principle. We understand the current challenges of difficulties, but we want to push just a little bit further because the method is not wrong, the approaching is not wrong, it feels wrong because the technology won't ready. But I think like I said, in the past three, four years, a lot of infra has been significantly achieved.