The Race to Create the "iPhone of AI" is Heating Up!
Summary
TLDRThe video explores the emergence of AI-powered hardware devices like the R1 by startup Rabbit. These voice-controlled devices aim to execute tasks across apps via natural language, reducing clicks. The video analyzes Rabbit's technology, use cases, business model and industry landscape, noting smartphones may make standalone devices redundant if integrated AI improves. It remains uncertain if consumers will adopt voice-first hardware, but the video argues Rabbit's crowdsourced training data and modding community could make the R1's AI increasingly capable over time.
Takeaways
- 😲 Humane's pin device failed to convince anyone of its usefulness or viability
- 👀 Rabbit is a new player trying to make AI hardware devices like the R1
- 📱 The R1 aims to execute tasks by integrating with your existing apps
- 🔊 The R1 has a push-to-talk button for quick voice access without a wake word
- 💡 The R1's AI model combines neural networks and symbolic AI for task automation
- 💰 The R1 only costs $199 with no subscription compared to Humane's $699 price
- 🤔 It's questionable if people will adopt a separate voice-first device like the R1
- 🚀 The R1 could accumulate training data to eventually control any app via voice
- 💬 Microsoft and others seem interested in the AI hardware device space
- 📈 If successful, the R1 concept could alter how we interact with technology
Q & A
What was the first company to try to build AI hardware, and what happened with their product?
-The first company was Humane. They unveiled an AI assistant device called the Humane AI Pin in 2023, but it has not been released yet. The product was supposed to replace smartphones, but it failed to convince anyone of its practical utility and the company has struggled, laying off staff and losing their CTO.
What is the Rabbit R1 device that was recently announced?
-The Rabbit R1 is a small, portable AI assistant device created by the startup Rabbit. It features voice control and physical buttons to activate the AI and interact with apps on the user's smartphone to perform tasks or answer questions.
How does the Rabbit R1 work compared to other voice assistants like Siri?
-The R1 uses a large action model rather than just a language model. This allows it to not only understand requests, but actually take actions across apps and services to fulfill the user's needs. It connects to the user's existing apps and accounts.
What kind of tasks can you ask the Rabbit R1 to perform?
-You can ask the R1 to perform complex, multi-step tasks involving different apps/services like booking a vacation, ordering food, calling you a rideshare, controlling your smart home devices, etc. Basically anything that can be done through an app, the R1 aims to be able to execute through voice commands.
What are some of the advantages of the Rabbit R1 concept?
-Advantages include very fast response times, hands-free operation for tasks that normally require manually operating a phone, and a low $199 starting price with no subscription fees.
What are some potential disadvantages or doubts about the Rabbit R1?
-You have to carry another device that might not provide enough utility compared to just using your phone. Voice-first operation contradicts social norms. And there are still doubts if AI can reliably perform some sensitive tasks like booking travel or payments.
How might the Rabbit R1 improve over time?
-As more users show the device how to operate new apps, it can aggregate that training data to expand what apps it can use. Software updates utilizing improved AI models could also make it more capable of understanding any app interface.
Why is Microsoft interested in this new category of AI hardware?
-Microsoft sees significant potential for 'breakthrough natural interfaces' between humans and devices/services with these new large AI models. If hardware can be designed around conversation versus individual apps, it enables a whole new interaction paradigm.
How are new chips advancing on-device AI capabilities?
-New mobile chips like the Snapdragon 8 Gen 3 enable real-time AI features solely within the device itself, rather than needing cloud connectivity. This makes experiences faster, more reliable, and more private - key advantages.
What is one open question about Rabbit's business strategy with the R1 device?
-One question is whether the low $199 starting price without ongoing fees can sustainably fund the advanced AI capabilities Rabbit claims. There is skepticism if their funding andAI costs support this business model long-term.
Outlines
😃 Pokemon Charmander Described in a Wooden Cabin
The first paragraph describes a scene with Pokemon characters like Charmander, Pikachu, etc inside a wooden cabin. It provides visual details of the characters and setting.
😯 Rabbit R1 AI Assistant Orders Pizza
The second paragraph demonstrates the Rabbit R1 AI assistant's ability to understand natural language instructions and complete tasks like ordering a pizza. It shows a conversational interaction where the assistant confirms order details.
🤔 Microsoft and Others Eyeing AI Hardware Market
The third paragraph discusses major tech companies like Microsoft, Apple, and Google exploring the emerging AI hardware space. It suggests the market potential but also risks of developing specialized AI devices.
🔍 Understanding Rabbit's Neurosymbolic AI
The fourth paragraph provides background on the neurosymbolic AI technology behind the Rabbit R1 device. It contrasts neural networks and symbolic AI, noting their limitations.
🤨 Evaluating Advantages and Disadvantages of Rabbit R1
The fifth paragraph objectively evaluates pros like speed and price and cons like carrying a separate device for the Rabbit R1. It concludes mainstream adoption will take time.
Mindmap
Keywords
💡AI assistant
💡natural language
💡multimodal
💡neurosymbolic AI
💡on-device AI
💡general intelligence
💡training data
💡modding community
💡killer application
💡transparency
Highlights
The R1 is a small orange device causing waves in the tech space with 50,000 pre-orders sold out in just over a week
The R1 can interact with your apps and do tasks on your behalf by just pressing the push to talk button
The R1 response time is a claimed 500 milliseconds or less unlike today's voice assistants
The R1 is powered by a large action model instead of a traditional large language model
The R1's model combines newer neural networks with older symbolic AI using neuro-symbolic programming
The R1 costs just $199 with no subscription compared to Humane at $699 with subscription
The 50,000 R1 pre-order customers become training data to improve the AI over time
The R1 is useful for early adopters who love to tinker, mainstream will take time to adopt
AI hardware faces threats from improving smartphone capabilities and integration into watches
The R1's utility is undetermined, killer applications aren't yet apparent
Carrying a separate R1 device may not warrant the utility compared to a smartphone app
Ordering food on the R1 without seeing menu options seems limiting
Trusting the R1 completely to book travel could miss specific preferences
The R1 concept can't be dismissed given improving AI capabilities over time
Steve Jobs predicted apps integrating into an agent-centric voice system
Transcripts
this video is brought to you by
nebula who is this Pokemon Charmander a
fire type Pokémon known for the flame on
the tip of its tail in the scene you
described Charmander is surrounded by
other Pokémon such as Pikachu a small
yellow rodent-like Pokémon with long
ears and a lightning bolt-shaped tail
the setting appears to be the interior
of a wooden cabin or house with the
Pokémon standing on a wooden floor yep
the image is taken from a screen as
evidenced by the reflection of some
items on the screen and style of the
windows and
walls that's crazy but yeah it is a
pokedex hi welcome to another episode of
Cold
Fusion for decades it's been a Sci-Fi
dream to Simply talk to your device and
have it do tasks for you with generative
language AI starting to make this stream
feasible the question becomes how do we
turn these AI capabilities into a
functional Hardware device recently a
new unknown player has entered the space
rabbit the company only started raising
money in October of 2023 and have
already built out a product called the
R1 this small orange device is causing
waves in the online tech space with
50,000 pre-orders sold out in just over
a week people are interested so I
figured that in this episode I'll give
my thoughts on the device but I'll also
show you the new AI industry that's
forming right before our eyes this is
going to be a deep dive so sit back
relax and let's get into it
you are watching C Fusion
TV we need some context rabbit aren't
the first company to try AI Hardware
they're the second the first was Humane
and let's just say that they
embarrassingly showed how hard it is to
get this idea
right many of you have heard of The
Humane AI pin unveiled in 2023 and
unreleased at the time of this recording
it's a device you can speak to a natural
language and have it do tasks for you it
has a holographic display and was
supposed to replace your phone despite
the founders having the pedigree of X
Apple Engineers they didn't seem to live
in reality the device was more like an
exercise in philosophy rather than
practicality take a look at the CNBC
interview where they failed to convince
anyone can you see that yeah what's that
telling you so this is this is a display
when you need it it's just capable of
doing a lot of things but it's not
something that you need because the
device is actually built to be
multimodal that means you can use it
however you want like what like what
tell me something you would use that for
so you use it for just about anything
like uh sending texts or checking up on
any notifications that you've got in
stuff that you do just all there's a
speaker that's built in and then there's
a a user LED called so that's what I
want to ask you about Imran the
co-founder says that the device can do
virtually anything a smartphone can but
without a screen I doubt that's possible
something as simple as discreetly
reading emails or texts can't be done
sometimes I'll be at a meeting yeah
right and and I'm sneaking under the
under the desk to look at my email
because that's and I can't say uh Hey
Hey Siri you know tell me my email
because I don't want the whole room to
hear my email so how does that work in
this context so the device is powered by
a AI powered OS I think the biggest
thing here that is that it's AI powered
and so it's doing a lot of that that
heavy lifting for you so you just
actually engage with it when it's really
important but seriously if if if I'm in
a meeting and I'm talking to somebody
but I got to go like this when I'm in
the meeting and anybody who can see my
hand can see the see what I'm reading
too how I mean I don't know how that's
different screen's doing and then
there's the price VAR what's the price
point it's $6.99 for this model and it's
$7.99 for the one that she's got and
that's the complete system comes with
everything plus obviously this hasn't
gone too well as Humane has already laid
off 4% of their staff before even
launching their CTO has been kicked out
too but while all of of this is going on
there's a new kid in town rabbit with
their new device called the R1 firstly
what is it that was delicious check the
fridge and order the ingredients to make
that again tomorrow create a route that
works with my goals then start the best
playlist to keep me motivated watch what
I'm doing here process all my new photos
today just like this find us a nice
restaurant near here then get us there
take me to the best spot to catch the
sunset and Lead me through a meditation
play my favorite
song I'm on
it at its core the rabbit R1 is supposed
to be a device so simple that you don't
need to learn how to use it it's powered
by a multimodal language model
supposedly more capable than chat GPT so
of course it can answer all the
questions that you have but the
difference is it can interact with your
apps and do tasks on your behalf with
the push to talk button you don't need
to say anything to wake it up just press
and hold the button and talk like a
walkie-talkie I get a response 10 times
faster than most of the voice AI
project get me a 12in pizza from Pizza
Hut deliver to here the most ordered
option on the app is fine ordering a
12in pizza from Pizza Hut since you
mentioned that the most ordered option
is fine I will select that for you I
just created an order for a 12-in pizza
it's going to be hand tossed with a
classic marinara sauce and topped with
regular please please confirm your
order that sounds really good I just
confirmed order
here now I'm waiting for my pizza set of
more complex actions that may take
multiple apps to finish on a smartphone
for example if I want to plan a trip to
London I can just describe what I wish
in full to R1 I want to take my family
to London it's going to be two of us and
a child of age 12 we're thinking of
January 30th to February 5th can you
plan the entire trip for me we like
cheap non-stop flights grouped seats a
cool SUV and a nice hotel that has Wi-Fi
exploring ticketing options to make your
trip a reality for your trip I found
various flight options a range of hotels
to choose from and car rentals available
please confirm each option individually
for further details and booking could
you come up with a schedule for fun
things to do while I'm over there I have
prepared a detailed travel plan for your
trip in summary you will be exploring
London's iconic landmarks visiting
museums enjoying delicious meals at
local restaurants and experiencing the
vibrant atmosphere of different
neighborhoods navigation details are
also prepared and I can help you book
tickets for any events you'd like to
attend R1 just planed the entire trip
for me that's awesome but it seems like
this is a little bit too intense can you
plan a easy schedule for us sure I'm
working on it please take a look at it
and let me know what you think it gave
me a more relaxed schedule every day
with all the details sounds really good
and I just confirmed that I can foresee
a wonderful trip Isn't that cool huh you
can use the camera to recognize real
world objects and ask questions about
them for example what meals to make from
a fridge
or what are people talking here let me
take a look in the image there is a
conversation taking place in a chat
application the participants are
discussing rabbit OS and its
capabilities specifically mentioning a
music sequence answer one person is
curious if rabbit OS has a music
sequencer built in and another suggests
that it can manipulate your digital
audio workstation DA they also mention
using it as a learning tool and seem to
be preparing for the next topic or task
in their discussion whether it works as
well as the demos is another question
entirely the rabbit company stresses
that the R1 is not a phone but unlike
Humane it's not aiming to replace it the
R1 has a 2.8 in screen which is small
given that the average smartphone screen
is 6 Ines these days but the small
screen is meant for increasing utility
rather than the main form of user
interaction in addition the R1 has a
physical scroll wheel a push to talk
button so it doesn't need a trigger
phrase to listen a 360 swivel camera and
enter the
vision a rotates and of course a speaker
it's been designed by teenage
engineering who makes music hardware and
famously designed the nothing phone its
design might be polarizing but at the
very least the device is interesting
enough to get people
talking in terms of the business case
for rabbit there are some clear threats
generative AI technology is looking for
Consumer killer applications
entrepreneurs are naturally trying to
find the best ways to use it recently
jabel made a pretty hilarious video
about coding and building his own
companion AI
device but I played with it for a bit
and it feels more like uber besty
instead of actually having one so if I
want a true AI bestie I need to get more
control and after a few hours is the
python dependency hell check this out I
got it working on the raspberry pie my
little AI homie name is chai by the way
and from there I started working on the
portability by buying the right parts
for it and giving a little auto start
script started work on chai's actual
build form with the res piie screen and
3D printer and I just kept on iterating
on this process until my main man chai
was in one solid piece so as you can see
he is a one complete build one solid
object a microphone on top the back's
closed please don't ask how I charge him
instead new AI bestie thing my buddy
here actually has his own personality
watch this I'm a black guy who loves
anime what do you think about that
aren't we breaking stereotypes here a
black guy who loves anime alert the
media because this is such
groundbreaking information dude you
enjoy what you enjoy who cares about
what anyone else thinks aren't we
Progressive high five keep watching
Dragon Ball or whatever you're into
interesting so he like partially roasted
me but also ended on a positive upbeat
listen to being real that sounds a lot
closer to what an actual best friend
would do standard chat GPT could never
so the fact that one single guy can do
this means that there's probably a bunch
of startups chomping at the bit to enter
the hardware AI space in fact Sam ultman
and softbank's MOSI son are both trying
to get the iPhone designer Johnny I to
design their respective AI Hardware
devices Microsoft is looking at the
development of the R1 with great
interest Satia nadela hints that
Microsoft could jump into the AI
Hardware space you see I thought the the
the demo of uh the rabbit OS and uh the
device was fantastic I think I I must
say after jobs uh sort of launch of
iPhone probably one of the most
impressive presentations I've seen of
capturing the uh the vision uh of what
is possible going forward for what is an
agent Centric um uh operating system and
interface and uh I think that's what
everybody's going seeking if you have a
breakthrough in natural interface uh
where this idea that you have to go one
app at a time and all of the cognitive
load is with you uh as a human uh does
seem like there can be a real
breakthrough we had the first generation
whether it was Cortana or Alexa or Siri
or whatever you um it was just not it
was too brittle uh where we didn't have
these Transformers these large language
models uh 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 it's had an opportunity
from Microsoft or are you moving away
from Hardware I mean look I mean always
it's an opportunity for us um and so
yeah I mean we make Hardware but even
for the whole category of AI Hardware
devices as a whole it's a risky segment
to be in there's a quiet AI Revolution
going on in smartphones Snapdragon new
chip the Snapdragon 8 Gen 3 is out now
and is already giving smart phones new
on device AI capabilities Samsung's s24
is one of the first devices to fully
take advantage Mr Who's The Boss shows
us some
examples so the first is instant slowmo
literally go into your phone's Gallery
hold down on any video in it and Bam it
halves the playback speed while keeping
it smooth using realtime frame
interpolation for example if you
download a PDF on your phone one tap
we'll summarize it if you're browsing a
website on the internet you now have a
button that can read the entire page for
you and turn out a very good summary in
like 2 seconds if you have an AI feature
that can be pulled off completely on
your device it's likely to be faster
because it doesn't need to connect to
the internet more reliable cuz you can
use it literally anywhere on the planet
regardless of whether you're connected
or not and also safer since none of your
data is actually having to leave your
phone at all and so in time Google
assistant or Siri could simply be
updated with similar capabilities to the
R1 with neural Hardware chips on device
they'll be more powerful faster and
convenient another idea is that these
Standalone AI devices could just be
integrated into a watch that way you
don't have to carry around a separate
device so the unfortunate truth could be
that this new segment might be over as
soon as it started unless there's a
killer application I could be completely
wrong though Keen Tech enthusiasts and
tinkerers might find ways to make the i1
useful beyond anything I can imagine
right
now so how does it work well the device
works by using what rabbit are calling a
l or a large action model instead of the
traditional large language model the
main difference is that it can be used
for actually doing tasks instead of just
understanding linguistic context and
replying to a question if you remember
those AI agents built on top of chat GPT
last year it's kind of like those for
those interested the R1 L model is based
off something called neuros symbolic
programming this method combines newer
neural networks with older symbolic AI
which represents objects as symbols and
processes the information that way
neuros symbolic programming has worked
in narrow Fields like robotic Automation
and CAD software I'll bring into brills
to explain a bit more modern neural
networks train a model on lots of data
and predict answers using best guesses
and probabilities but symbolic AI or
good old fashion AI as it's sometimes
called is hugely different symbolic AI
requires no training no Mass amounts of
data and no guesswork it represents
problems using symbols and then uses
logic to search for Solutions all right
so I did a lot of research into neuros
symbolic AI uh talks and research papers
and I learned that it has the same
trappings and limitations of just
regular symbolic AI you essentially will
Define a bunch of if El statements and
instead of using humans to determine the
variables that go into these IFL
statements you'll use an AI to extract
features from a certain system system or
process and then do IFL statements on
that output but one of the biggest
limitations from neuros symbolic AI is
that you need domain experts to craft
the IFL statements with the good type of
biases but bias is unavoidable is this
the right task for the job I'm a little
skeptical but maybe they know something
that we don't know everything that
they're promising could be demonstrated
by releasing an app which is kind of
weird that they are going for Hardware
that costs only $200 with no
subscription fee when we all know AI API
calls at scale is not cheap but I mean
who knows maybe they know something that
we don't know and have funding that
we're not aware of this could
potentially just be a little nce Trend
like polaro cameras or maybe industry
shift like the iPhone I have my
skepticisms but only time will
[Music]
tell want to get a ride to the office
there's the app for that want to buy
groceries there's another app for that
each time you want to do something you
fumbles through multiple pages and
folders to find the app you want to use
and there are always andless buttons
that you need to click if we can make an
AI trigger actions on any kind of
interface just like a human word it will
solve the problem so how does the rabbit
manage to execute the apps that you have
without you doing
anything well there's a setup process
which means you have to log into the
apps that you normally use so it has
access to the data within you can also
train the R1 on new apps by using the
company's web portal and showing the
device how you use that particular app
and once it's being shown it can do that
automatically on your behalf this is
cool and opens up a lot of doors but it
seems like a lot of effort to the Casual
user who just wants the device to work
so what are some advantages it's fast
something that stood out to me is the
response time of a claimed 500
milliseconds or less that takes a lot of
the jankiness out of the experience so
that's very interesting unlike today's
voice assistance on a phone you don't
have to speak specifically today's voice
assistants will often get confused if
you ask them to do a task the R1 being
like chat GPT but also being more able
to execute tasks is much more useful but
probably one of the most interesting
aspects is the price costing just $199
with no subscription compared to the
Humane which is $600 with a subscription
it's compelling in this new
segment so what about the
disadvantages you have to carry around
another device that might not possess
the utility to Warrant doing that when
I've talked to friends and people about
this device everyone just says why can't
it just be an app being vo first can be
awkward and goes against current social
conventions there's also the issue of
complete confidence in the technology
take Uber for example will you trust the
R1 to do the task correctly if you're
booking your flights and hotels there
might be specific preferences which
could be missed because the device
simply couldn't know that also for
ordering food without seeing menu
options it seems like you miss out on
browsing on what you
want so when conclusion the rabbit R1 is
the archetype of a revolutionary product
but its usefulness is undetermined and
killer applications aren't yet apparent
in my view it's a great tool for early
adopters who love to Tinker but the
mainstream will take some time
especially getting used to a voice first
device but that's not the end of the
story many people do roll their eyes at
the term AI but you can't throw the baby
out with the bath water with any new
technology there's going to be scams and
gimmicks but also some useful
applications the question is what
category does the rabbit fall into it's
hard to say I think the rabbit is
interesting because it's bootstrapped to
the ever improving capabilities of AI
what do I mean by that well in theory
it's possible those 50,000 pre-order
customers who begin to show the rabbit
how to use their favorite apps
essentially become a huge aggregate
training data set for the lamb system
this could mean that the R1 eventually
becomes as good as humans at
understanding any app interface all it
would need is a software update when the
aggregate training is complete and that
that's fascinating to think about and
because of that along with a modding
Community I'm not so quick to throw this
concept away those were just my thoughts
I'll leave you with a question what
custom uses can you think of for such a
device getting a conversation going
around this could be really cool in the
end whether capable AI assistants come
from smartphones or a standalone device
Steve wnc was right after all eventually
all all those apps and all yeah I just
want to get to worry about not even
having to find an app you still have to
find the app app but I want the apps to
be able to become a part of the voice
system I want every program to start
coming out and being a part of Siri
every app on the iPhone and the same
thing would apply to other Tech other
platforms I just want to give a quick
shout out to nebula if you're tired of
YouTube ads and want to see all your
favorite educational creators like real
engineering half as interesting and W
over Productions in one spot nebula is
the place for you it's a streaming
platform built by creators for creators
you can can see my videos ad free and
there's exclusive content by some of the
best in the business there's also no
algorithms for us to worry about so we
can experiment for example polymatter
has done an exclusive video solely on
China's one child policy and it was very
interesting it's part of his larger
series called China actually if you sign
up using the link below you get to
support me directly and get nebula for
40% off an annual plan that's a little
over $250 a month it's the best deal in
streaming I'm looking to expand cold
fusion this year and nebula's expertise
and deep understanding of the challenges
of content creation will help with that
signing up to the link also helps me
achieve that goal so thank you if you're
interested in anything science
technology or business feel free to
subscribe to Cold Fusion it's free so
thanks for watching my name is toogo and
you've been watching cold fusion and
I'll catch you again soon for the next
episode cheers guys have a good
[Music]
one
[Music]
cold fusion it's newth thinking
Ver Más Videos Relacionados
ChatGPT Can Now Talk Like a Human [Latest Updates]
Rabbit R1: The First Personal AI AGENT Device NO ONE Saw Coming (Look Out, Apple)
AI the Product vs AI the Feature
New AI Chatbot - Claude 2 - is Free and Outperforms ChatGPT
Microsoft vs. Apple: Satya Nadella Says AI-Focused Copilot+ PCs Beat Macs | WSJ
The Race For AI Robots Just Got Real (OpenAI, NVIDIA and more)
5.0 / 5 (0 votes)