You Asked, I Answered: Everything About the Rabbit R1 🐰

I did a review about the rabbit R1 device a few weeks ago and it was my most popular video of all time the day that the rabbit R1 was announced I ordered one and I cannot wait to get it so I can review it and show you all what I think of it I truly believe that this is the first step into what is going to be the future of computing direct natural language to AI to compute now that first video that I made about it got over a million views in just a few weeks so obviously this device has a lot of appeal and at the very least people want to know all about it and if you haven't seen anything about the rabbit R1 device it is a $200 piece of Hardware that you can carry around that essentially uses your natural language to perform actions on your behalf you could do everything from listening to music to ordering an Uber ordering food going shopping anything and the primary interaction method is your voice not a touchscreen and so I read through all of your comments and there were a lot of common questions and confusions about the product and so rabbit actually reached out to me after seeing my video and wanted me to address your comments directly and so that's what I'm going to be doing today rabbit is sponsoring this video and wants me to answer a lot of your questions and clarify a lot of your confusion so I'm really excited to do this because I am genuinely so excited about this device so let's go through some of the common questions and concerns that I found in your comments to that video and hopefully I can clear a lot of it up first let's look at this comment by at PMX no subscription but their LM runs in the cloud expensive and the interaction with apps is done in emulators running again on the cloud expensive times too so either they'll start charging for a subscription in a few months or they'll go bankrupt and the device will become a very Nic looking paper weight so very valid concern people are so used to not only buying a hardware device but also having to pay subscription fees and I think a lot of you might be skeptical that not only is the rapbit device only $200 which is very inexpensive for any modern tech device but they're not charging any type of subscription fee now I talked to Rabbit about this they have no plans to charge any subscription fees but of course they could change their mind in the future but for now there are no subscription charges and I know it's weird to hear that but it is true today and a little bit later in the video I'm going to talk about why running rabbit OS and the large action model and everything they do in the cloud is actually a lot less expensive than you think because the large action model is very very efficient and so there's no subscription planned once you get the device you own the device and you can use it next there seems to be a lot of confusion about what a large language model versus what a large action model actually is I think a lot of people think that it's just a marketing term and it might just be a large language model with a few bells and whistles but I want to clear up what it actually is a large language model is based on the Transformers architecture what it's really good at is predicting the next token in a series of tokens basically the next word in a sentence but that wouldn't have worked well with this application what rabbit does is it takes your natural language converts it into what it thinks the action you want to accomplish is and then actually performs the action on kind of a hidden operating system so if you can imagine taking out your phone clicking door Dash going to a restaurant listing ordering food and completing the order it is doing that for you so it is actually figuring out how to click through an interface it's not connecting to an API and I'm going to show a clip as to why they decided to do that in a second and not only that because it's not using the Transformers architecture it is is actually much more efficient and has a much lower cost to run back to the point of not having a subscription fee but let me show you this video clip of Jesse Lou on a podcast explaining the difference between a large language model and a large action model we're not making any large language model we we it's the large action model is neuros symbolic it's not a large language model large language model where we usually talk about his gbt 4 you know and Bard and and you know Gro and all that uh those are based on Transformer and they require a a ridiculous amount of GPU on the cloud to be able to train and and get things right I don't think any startup raise $30 million can just all of a sudden make their own large language model so for the record we're not doing that we're working with all the best language models and small language models and potentially if there's open source language models that's in the future we would like to basically set up a Evo internal evaluation to keep monitoring the performance from all this vendors and we can switch on the fight that's how rabid o works but we are focusing on large action model which we we know for a fact that the language models were the Transformer were designed to understand language better but it performs really really miserably at least for now on finishing the tasks and we don't like the way to work with apis because API one you're betting everyone will give you API which is not the case it's easier for open ey to or big companies to encourage everyone to build API for them but it's really hard for startup to convince you know all of a sudden you have 2,000 vendors working on your API in your format but even if you have all the API uh often times they don't they don't 100% replicate the full feature of the application that's why we were like okay hold on let's take a step back let's think about a universal solution let's create an AI for a universal solution to whatever the application is is Android app IOS app Windows app how can we build a universal solution knowing that language model weren't designed to trigger actions so we actually use neuros symbolic to basically start we actually start working with data labeling companies we started to collecting based on our own assessment we actually assembled test groups to have real human interacting with different kinds of software like uber like spottify like all the frequent apps so we started this process about two and two and a half years ago we started collecting real human interacting with all kinds of softwares and we we get all those things in video recording and then we set up a NE symbolic algorithm which become today's large action model that to basically fit in all these clips into the large action model and ask our lamb to read the the the clips frame by frame probably the most common thing that I've seen both in the comments to the previous video I made as well as across the entire internet is why couldn't this have just been an app on my phone versus a standalone device and I'm going to answer this in a few ways the first way I'm going to answer it is I come from the world of startups the fact that rabbit is out there trying something completely new trying to break the molds trying to break the Monopoly of Apple devices Android devices I applaud them and they are truly showing us what I believe is the future of computing and they did not think that they could accomplish that with just being an app on a phone now when you take out your phone the entire device is built around you using your finger to navigate through apps and within the app find what you need to do but rabbit is very very different you are are speaking natural language and you have an AI agent performing those actions on your behalf and that would not have been possible if they built this as just an app and then there's the concern of okay so now I have to carry around two devices well a lot of people actually carry around multiple devices so you have your phone sometimes you have your laptop you have your earbuds and now you're having your AR and VR devices so to me at least as an early adopter I really don't mind carrying around another device because first of all it's gorgeous in my opinion it's small it's lightweight and so if I get to carry around this device that I get to play around with what I believe is the future of computing I don't mind that at all and then a lot of people are also saying okay well isn't Siri and Google Assistant just going to add this well probably they're going to add some of these pieces of functionality but they've also had 15 plus years to do so and they really haven't done it all that well now of course things are going to change and they're going to change rapidly going forward but I'm looking forward to some competition in the market and Apple and Google are trying to avoid the pitfalls of the innovators dilemma they have these existing products with these existing Frameworks these existing ways to think about things and it's really hard to break out of that mindset and so even if Siri evolves even if Google Assistant evolves they're still going to have to work within the constructs of everything that they've built previously and rabbit really wanted to innovate and do something completely new and again I applaud them for that let me show you this video clip of Lou describing why he decided to build his own Hardware device versus just being an app and I think he says it really well check this out when do you feel yourself reaching for your iPhone when do you feel yourself reaching for the rabbit when do you hit frustration points with either device and and how does that go into your thinking about does this at some point replace me bringing my phone with me yeah uh and I have to maybe think about the hardware capacity of this device or maybe there's you know this this rabbit but there's two other versions and do do did you ever leave home without your iPhone and just take your rabbit talk to me about that yeah yeah so so first of all any single direct actions I I'm I'm leaning towards rabbit now why why because when I start trying to accomplish that task it's a speed of thoughts and I have that already and I feel it it almost quicker and faster more intuitive than find that analog button without even looking at it and just talk and the AI is accurate and good enough lamb is fast enough to get things done so I'll give you some quick quick reference I'm always on multip screens I'm always on a lot of things daily when I'm working and maybe talking to other guys if I have something that I don't know this is definitely faster forget about lamb just searchwise this is way faster then I have to set up a new tab or go to the Chrome and start typing you have the action button it's kind of like a CB radio you press the button and you say Define the word icon clastic or Define the word Donnie Brook you don't know the word Donny Brook you say Define the word Donny Brook I don't have to think command tab go to Google define Donnie Brooke Etc six or seven steps yeah well well not even that some very Advanced you know uh numbers this gives you faster because I I give you an actual case I was having a meeting with our one of our existing investors they were ask about you know performance in comparison you know on sales and and some numbers uh from other company I generally don't know I don't know what the what the revenue from from you know a company last year and if you think about I search on Google there's 200 tabs which one is accurate which one is correct so that's why we set up the Strategic partnership with practicly just to enhance that part so that's search alone another thing is music I can tell you you will love this just because it can play music it's it's a second iteration of the classic iPod with with with with the same level or even simpler level of of control of this I think you know just music play alone I that's definitely my number one in the past eight months other than search search is probably 70% of the music now I also saw a lot of concerns about privacy and security and I want to touch on both of those things so first let's talk about privacy and in fact privacy is so important to rabbit that they actually have a page dedicated to it on their website so let's take a look at what they say directly about privacy so first because it's hardware and it has a camera they have explicitly given it a physical way to block the camera in the device which is awesome and if we keep reading it says R1 does not have an always listening mode so this is also really important there's been a couple other AI Hardware devices launched and they are always in listening mode always in watching mode and that always turned me off immediately this is much more opt in than opt out if you want to interact with the rabbit you actually have to hold the button down or you actually have to take a picture it is not always listening it is not always always watching which I like it only hears you when you hold the physical push to talk button rabid OS also features a gyroscope enabled privacy mode where all modes of input and output are disabled when the R1 is facing down and then there's the actual information that you're sharing with the R1 device we ensure that whatever you have shared with the R1 stays with it no one including us will be able to use personally identifiable information such as your name phone number email address etc for any purpose other than serving you you can choose to have rabbit OS remember or forget things with our Rabbit Hole web portal so it has a separate web portal where you're going to do things like authentication and controlling your information and we are committed to never selling or sharing your data with any third party under any circumstances without your formal explicit permission so that is really really important even if you don't trust what they're saying you can log into their web portal and remove any information that you don't want them having and not only that it actually spins up an isolated environment just for you and your agent when you interact with rabbit OS you will be assigned a dedicated and isolated environment on your Cloud for your own lamb when our rabbits perform tasks for you they will use your own accounts that you will have securely granted Us control over through our Rabbit Hole web portal so let's say you're logging into door Dash or Spotify or any other service you log into the web portal very securely using SSL giving them authentication permission just like you would any other third party we do not store your passwords for these devices rabbits will ask for your permission and clarification during the execution of any tasks especially those involving sensitive actions such as payments so you're not accidentally going to order $500 worth of Snickers bars from door Dash rabbits are a safe and respectful representation of a legitimate user we do not create fake accounts or spam accounts or engage in any abnormal access patterns or traffic to the Target app the AI agents that they're using look and behave like humans we do not attempt to reverse engineer any service or aim to replace existing API equivalent we do not solve captas and we do not seek to significantly alter the user base of the apps so again there should be no difference between the agent performing in action on your behalf and you performing it yourself it is just saving you a ton of time by doing it for you and as I've mentioned they're a startup they're evolving they're figuring things out so they're trying to be extremely transparent and always improving and again coming from the startup World seeing this type of innovation I applaud them and so hopefully that answers some of your concerns about privacy and security you're not actually storing passwords on the device it's all done securely in the cloud the hardware device just acts as an interaction method between you and the agent and now let me show you what Jesse Lou has to say about authentication and privacy and security himself this is actually how authentication is going to work take a look when I tell rabbit hey order me sushi family of five you know this amount of food Etc yeah um it knows how to use the Uber Eats app or the door Dash app now the rabbit device then fires off to your service somewhere in the cloud yes this request I have already authenticated in a web interface right hey I authenticated my Uber Eats and my door Dash account it knows the sushi restaurant I like and then it starts this ordering process and then I guess comes back to me with like hey just want to confirm this is what you want and I say yes and then what does it do it pops up an emulator of some type in the cloud and then you have this web emulator that has my log authenticated already how does that work so first of all uh let's start from beginning authentication right because if you think about a device it works quite different than the previous generation because it doesn't have any software pre-baked in it doesn't have anything pre-installed it's it's literally just a a AI it's your choice what kind of service you want to enable and it's your choice how complicated and how advanced you want this device to be if you just say hey this is a cool looking iPod I just want to use this to listen music music then you unlock the music feature choose whatever vendors it only will do music but in tomorrow if you want to start ordering food you have to unlock that feature the login process is where you're correct we have a web portal that's kind of like our own mini version of iTunes iCloud if you understand that in that sense that helps with all the authentication settings and feature management so you go to the website and you basically choose whatever servic you want to unlock because again to large large action model Spotify YouTube music Apple music doesn't make a difference it it is an interface you know in fact in fact xedia and YouTube music doesn't even make a difference they are all interfaces so you we give you the freedom to choose whatever preferred Services you want to unlock you go there and you you you basically click the connect with Spotify button and what happens next is that it will redirect you to Spotify login and we don't save your credential we don't touch that you go to Spotify you go to Uber you go to Door Dash you log in through them and then we recognize oh this account is being being connected with rabbit OS and then what happens is that on on our Cloud we have a very very uh creative structure and Innovative structure that we have considered we have a super host we have a superhost computer that when Jason talks to his rabbit R1 about ordering a hamburger from door Dash what's going on is that we first see if Jason if Jason's Lo logging with door Dash or logging with Uber East right and we saw okay Jason choose door Dash and on on that super host a lamb is inter interacting virtually with a door Dash app or website and you don't see all of that because all of a sudden this is done all at once because it's it's Ai and then we will render rabbit themed UI on your device to get your result got it so you're not you're not directly interacting with the with the host as well you're just talking to it the intention goes l l understand oh you want to do this then lamb do that in the virtual environment and reender the result to your device so that's it I'm super excited about getting my rabbit R1 I should be getting it in a few weeks from now as soon as I get it I'm going to be making a bunch of review videos I hope you're as excited about the R1 as I am because I do really think it's giving us a glimpse into the future of what Computing is going to look like I've said this not only in my previous rabbit video but in many previous videos that applications are going to be going away programmers are probably not going to be needed in the long term it's going to be natural language direct to AI direct to compute and I'm really excited about that potential and so I hope you get the R1 device if you want to wait until you see some of my reviews for it that's great too and if you liked this video please consider giving a like And subscribe and I'll see you in the next one

Channel: Matthew Berman

Views: 83,509

Keywords: ai, llm, rabbit, rabbit r1, ai device, artificial intelligence

Id: 3E5XO_zIdA4

Length: 18min 40sec (1120 seconds)

Published: Fri Feb 23 2024