What Are These Companies Hiding?

Summary

视频讨论了两款即将上市的AI助手设备：Rabbit R1和Humane AI Pin。Rabbit R1是一款200美元的设备，已经预售超过10万套，而Humane AI Pin则是一款700美元的设备，需要额外的月费。这两款设备都声称能够通过简单的语音指令完成复杂任务，如预订全家欧洲旅行或计算食物的营养成分。然而，视频中提出质疑，认为这些设备的营销可能过于乐观，实际上它们目前的功能与智能手机相似，但不如智能手机强大。同时，视频还指出了这些设备依赖语音交互的问题，以及Humane AI Pin的激光投影技术可能过于噱头。

Takeaways

• 🤖 两款AI助手设备即将上市：Rabbit R1和Humane AI Pin，它们都提出了令人印象深刻但值得怀疑的功能承诺。
• 🌍 这些设备展示了通过简单的语音指令执行复杂任务的能力，如预订全家欧洲之旅或计算食物的营养成分。
• 💰 Rabbit R1定价200美元，已预售超过10万单位，而Humane AI Pin定价700美元，外加每月25美元的订阅费。
• 📱 这些AI助手设备被认为可以替代手机，但目前它们能做的基本上就是手机能做的事情，只是不如手机做得好。
• 🎯 这些设备基于AI构建，可以响应语音命令执行任务，但与手机相比，它们的互动性和功能性有限。
• 🚫 这些公司没有直接回答为什么这些产品是硬件而不是应用程序的问题，可能是因为需要操作系统级别的访问权限，而这是苹果和谷歌不会提供的。
• 📱 语音交互是这些设备的主要交互方式，但与在屏幕上的精细调整相比，仅依赖语音命令可能会导致用户体验下降。
• 📹 一些演示视频被剪辑以隐藏响应时间，这可能会误导用户对设备性能的期望。
• 🔄 Humane AI Pin的激光投影技术在实际使用中可能存在问题，如设备过重和穿戴不稳定。
• 🐑 Rabbit R1的“大动作模型”（Lamb）是一个有前景的特性，但目前可能还需要大量用户数据来训练才能变得实用。
• 💡 尽管对这些产品的市场推广持批评态度，但视频作者认为这些尝试是技术进步的一部分，并希望这些产品能够成功。

Q & A

兔R1和Humane AI pin是什么类型的产品？

兔R1和Humane AI pin是两款即将上市的人工智能助手设备。兔R1是一款200美元的设备，拥有可爱的外观设计、小屏幕、摄像头、模拟滚轮等硬件配置。Humane AI pin则是一款700美元的设备，需要额外支付25美元的月费订阅，它使用投影仪将用户界面投影到用户的手上。

兔R1和Humane AI pin的主要功能是什么？

这两款产品的主要功能是通过人工智能助手来简化用户的生活，例如通过语音命令预订全家欧洲之旅、计算食物的卡路里或蛋白质含量等。它们旨在通过自然语言处理来执行任务，减少用户在手机上的操作步骤。

兔R1和Humane AI pin的销售情况如何？

兔R1在预售阶段已经售出了超过100,000台，而Humane AI pin由于其700美元的价格和额外的月费订阅，销售情况似乎没有兔R1成功。

为什么这些设备不能仅仅是应用程序？

这些设备不能仅仅是应用程序，因为它们需要访问操作系统的高级权限，例如密码、信用卡信息、麦克风、相机和GPS等，这些权限通常不会开放给开发者。另外，苹果和谷歌正在开发自己的类似产品，一旦推出，第三方应用程序将难以竞争。

兔R1的large action model（大动作模型）是什么？

兔R1的large action model是一种AI模型，它可以根据用户的提示执行动作，如鼠标点击和滚动。这种模型可以通过训练学习新的动作，例如通过观察用户如何在Photoshop中编辑照片，然后在未来自动执行类似的编辑任务。

Humane AI pin的投影技术存在哪些问题？

Humane AI pin的投影技术存在的问题包括设备体积较大且重，容易在使用时晃动，以及在演示中没有允许人们实际体验，存在过热的问题。这些问题表明产品可能还没有准备好上市。

为什么说这些设备的语音交互可能存在问题？

这些设备的语音交互可能存在问题，因为它们依赖于语音命令来执行任务，而人类在使用应用程序时通常会通过屏幕进行微调，如查看评论、照片或价格等。去除屏幕交互，仅依赖语音命令可能会导致用户体验下降。

这些设备的市场营销策略存在哪些问题？

这些设备的市场营销策略可能过于乐观，展示了设备的理想状态，但实际产品可能无法立即达到这些展示的功能。此外，一些演示视频经过剪辑，隐藏了设备的响应延迟，这可能会误导消费者。

兔R1和Humane AI pin的潜在风险有哪些？

兔R1和Humane AI pin的潜在风险包括依赖于未经充分训练的AI模型，可能导致执行任务时出现错误，如错误订购大量披萨或预订远超预算的假期。此外，这些设备的实用性和用户体验可能不如宣传的那样好。

这些设备的未来前景如何？

尽管这些设备目前可能存在一些问题和局限性，但它们是技术发展过程中的一部分。通过不断的尝试和失败，最终可能会为消费者带来更好的产品。

消费者在购买这些设备时应有哪些预期？

消费者在购买这些设备时应预期到它们可能无法立即执行所有宣传的功能，可能需要时间来训练AI模型，并且可能需要携带手机作为辅助。此外，应有心理准备，这些设备可能在实际使用中存在延迟和用户体验上的问题。

Outlines

00:00

🤖 两款AI助手设备的介绍与质疑

介绍了两款即将上市的AI助手设备：Rabbit R1和Humane AI Pin。这两款产品都提出了令人印象深刻但值得怀疑的功能，例如通过简单的语音指令就能预订全家欧洲之旅，或者通过展示一把杏仁就能计算出其卡路里和蛋白质含量。尽管这些设备声称能够简化生活，但视频中提出了对这些公司和产品的质疑，包括它们的销售数据、实际功能以及为何不以应用程序形式存在。

05:00

🗣️ 语音交互的局限性与市场营销的批评

讨论了这两款设备主要依赖语音命令的交互方式，以及这种方式与我们在手机上通过屏幕进行的精细调整相比的局限性。指出了演示视频中可能的误导性编辑，如剪辑等待时间，以及对Humane AI Pin的投影技术的实际应用和潜在问题提出了质疑。同时，批评了市场营销策略，认为这些设备的实际功能与宣传中的描述不符。

10:02

🧠 Rabbit R1的'大动作模型'与未来展望

详细讨论了Rabbit R1设备的一个特色功能——'大动作模型'（Lamb），这是一种能够根据用户提示执行动作的AI模型。虽然这个概念很有前景，但视频中对其实际应用和用户预期提出了担忧，认为目前这些模型还没有足够的训练数据来实现其宣传的功能。同时，对可能的AI错误和这些错误可能导致的严重后果表示了担忧。最后，虽然批评了产品的市场营销策略，但也强调了尝试新事物对于技术进步的重要性。

Keywords

💡AI助理设备

AI助理设备是指使用人工智能技术，通过语音交互等方式帮助用户执行任务的硬件产品。视频中提到的Rabbit R1和Humane AI pin就是此类设备的例子，它们被设计用来简化用户的生活，如预订旅行、计算食物的营养成分等。

💡语音命令

语音命令是指用户通过发出口头指令来控制设备执行特定任务的一种交互方式。在视频中，Rabbit R1和Humane AI pin主要依赖语音命令进行操作，如订餐、查询信息等，这种方式旨在提供更自然的交互体验。

💡硬件

硬件是指构成计算机系统或其他电子设备的物理组件。在视频中，Rabbit R1和Humane AI pin作为AI助理设备，它们的硬件设计包括屏幕、摄像头、扬声器等，这些硬件是实现设备功能的基础。

💡市场宣传

市场宣传是指企业为了推广其产品或服务而进行的广告和公关活动。视频中提到，尽管Rabbit R1和Humane AI pin的实际功能可能还未完全成熟，但它们的市场宣传却过于乐观，可能会误导消费者。

💡用户体验

用户体验是指用户在使用产品或服务过程中的感受和满意度。视频中讨论了AI助理设备在提供便利的同时，可能存在的用户体验问题，如设备响应时间的延迟、交互方式的局限性等。

💡价格

价格是购买商品或服务所需支付的金额。视频中提到Rabbit R1的价格为200美元，而Humane AI pin的价格为700美元加上每月25美元的订阅费，这影响了消费者的购买决策和产品的市场接受度。

💡隐私

隐私是指个人或组织的信息保护状态，防止未经授权的访问和使用。在视频中，提到AI助理设备依赖于语音交互，可能需要访问敏感信息如密码、信用卡信息等，这引发了对隐私保护的担忧。

💡技术发展

技术发展是指科技领域内新理论、新产品或新应用的创新和进步。视频中讨论了AI助理设备作为新兴技术，尽管目前可能存在局限性，但它们代表了技术发展的一个方向，有可能引领未来的产品创新。

💡操作系统集成

操作系统集成是指将硬件和软件组件结合在一起，形成一个协调一致的系统。视频中提到，苹果和谷歌正在开发自己的AI助理设备，这些设备可能会与他们的操作系统和合作伙伴紧密集成，从而提供更好的用户体验。

💡大动作模型

大动作模型是一种AI模型，它不仅能够理解和回应语言，还能执行鼠标点击、滚动等操作。这种模型可以通过训练学习新的任务，从而在未来能够自主做出决策。视频中提到Rabbit R1具有这种模型，它能够根据用户的指令执行复杂的任务。

💡AI幻觉

AI幻觉是指人工智能在处理数据或执行任务时出现的错误或不准确的表现。在视频中，作者提到如果AI在执行复杂任务如预订旅行时出现幻觉，可能会导致严重的后果，如订单错误或预算超支。

Highlights

即将上市的两款AI助手设备：Rabbit R1和Humane AI Pin，它们都提出了令人印象深刻但值得质疑的功能声明。

这些设备能够通过简单的语音指令执行复杂任务，如预订全家欧洲之旅。

设备能够识别物体并计算其营养成分，无需手机或其他应用程序。

Rabbit R1的设计简约，具有小型屏幕、摄像头、滚轮等硬件配置，预购量超过100,000台。

Humane AI Pin定价700美元，需要额外的25美元月费，使用投影技术将用户界面投影到手上。

这些AI助手设备的核心功能与手机相似，但执行效果不如手机。

AI助手设备基于AI操作系统，可以通过语音命令执行任务，但目前这些产品的功能可能被过度宣传。

这些设备为何不以应用程序形式存在，而要作为硬件购买，这是由于它们需要更高级别的操作系统访问权限。

苹果和谷歌正在开发自己的类似产品，未来可能会在硬件和合作伙伴集成方面超越第三方应用。

AI助手设备的设计旨在吸引注意力，因为如果只是应用程序，可能不会引起如此大的关注。

这些设备主要依赖语音命令，但人类的在线决策交互通常需要屏幕的实时反馈和微调。

演示视频中的等待时间被编辑，没有展示设备实际响应时间，这可能误导用户对产品性能的预期。

Humane AI Pin的激光投影技术在实际使用中可能存在问题，如设备过重和不稳定。

Humane AI Pin的激光投影可能是一种噱头，因为市场上尚未出现将手机屏幕无线投影到手上的技术。

Rabbit R1的Lamb功能是一种AI模型，可以根据用户的提示执行动作，如鼠标点击和滚动。

Lamb功能需要大量训练数据才能变得有用，目前可能还无法满足预购用户的期望。

营销材料可能过于乐观，没有充分揭示产品当前的实际功能和潜在风险。

尽管存在问题，这些产品的尝试是技术进步的一部分，应该鼓励而不是批评。

Transcripts

there are two of these AI assistant

devices that are shipping really soon

there's the rabbit R1 and then there's

the Humane AI pin and both of these

companies have made some very impressive

but questionable claims as to what these

products are capable of doing so they've

shown examples of you know these devices

you can just talk to them have two or

three voice commands and it's booking an

entire trip for your family to Europe

it's wild and they also have examples

where they're like you know showing the

device uh a handful of almonds and it's

able to calculate calories or the amount

of protein that's in it and you don't

need your phone there's no launching of

apps the idea is that these devices will

make your life easier like you can use

AI to assist you with those different

repairs uh you can do like live

translation they allude to the idea that

you could replace your phone with one of

these devices down the line now as we

get closer to these ship dates I

couldn't help but just look more into

these companies like why is it that

these things are weeks away from

shipping and no one has handled one of

these things yet right it's kind of

weird so I went in and having done a

little bit of research I feel like we

might be getting fleeced okay I'm going

to start off with the rabbit R1 so this

is the more popular of the two this

thing is a $200 device cute looking

Hardware it's got some teenage

engineering design language going on has

a small screen a camera a analog scroll

wheel a speaker a button it's got very

simple hardware and at the time of

shooting this video they've supposedly

sold over 100,000 units on pre-order

this thing has popped off and then the

other device the Humane AI pin this is

by contrast a

$700 device and it also needs a $25

monthly subscription but this is also an

AI assistant but instead of a screen

this uses a projector that projects the

image of the UI onto your hand it's a

neat party trick maybe a little gimmicky

but this device is not handheld it's a

pin that can attach to your clothing now

they haven't revealed sales figures or

anything but it seems just by looking

online that this product hasn't had as

successful as a launch as a rabbit R1

probably because of it $700 price tag

with its monthly sub description now the

the first question you might have having

seen these two things is what do they

even do and that's a really important

question right but every single time

I've seen these companies answer that

question they they skirt around it they

they talk about these like AI buzzwords

they talk about you know contextual

Computing I I think the reason why they

don't like answering this question is

because it's an uncomfortable answer the

answer I think of what do these things

even do is that right now these products

they basically do what your phone does

but they're just not as good at doing it

your phone's better at it that means you

can use it however you want like what

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 gotten stuff that you do just all

the time how would I know without seeing

a screen that somebody was texting me

and I'm your phone is more versatile

it's more powerful it's more private

because these AI assistant devices lean

on voice interaction to be able to do

what they do you can't really watch

videos on these things or play games on

them and because you still have to carry

your phone around with you you now have

two devices you have to manage but these

AI assistant devices are rooted in AI

like they're built from the ground up

with operating systems that are entirely

based on AI and because of that you have

the ability to be able to talk to these

devices with voice commands and they can

carry out a task when is the next

eclipse and where is the best place to

see it the next total solar eclipse will

occur on April 8th 2020 4 one of the

best places to see it is Naz Durango

Mexico it can give you answers it can

perform a sequence of tasks that would

normally take like five to maybe even 10

clicks on a phone so there is value to

that right if you can just talk to

something with your natural language and

this device will just do

stuff however the follow-up question is

okay if that's what this thing can do

why is it not an app like why is this a

piece of Hardware that you have to buy

why isn't it just something you can

download from the app store or the Play

store okay this is a very important

question and every time they've been

asked this question these companies

again they Dodge it I feel like they're

just they won't just admit the the

simple reason as to why it's a piece of

Hardware well there's a few reasons

number one in order for an app to exist

that has the type of capabilities that

we're talking about here this app would

need elevated access to the operating

system like you would need access to

like passwords credit card info it would

need access to like the microphone the

camera the GPS like all of that stuff

off of a single tap of a button and

those things are locked out from

developers rightfully so right Apple and

Google there's no way they're giving

developers that type of access on one

click certainly not right now but the

other reason is that even if in the

future if apple and or Google decide hey

you know we're going to allow app

developers to have access to this off a

single click Apple and Google are

actively working on their own versions

of these things and when they do come

out like it'll have awesome Hardware

integration and integration with

partners and stuff they're in app would

absolutely destroy any kind of third

party app so right you don't want to

touch apps but the third reason and the

main reason is for attention because if

this was an app no one would care I

wouldn't been having this conversation

with you if they had an app that did

exactly what the rabbit or the AI pin

did but it was just an app that was on

your phone you had to pay 25 bucks a

month truly no one would care but

because it's a beautiful piece of

Hardware particularly the rabbit we're

talking about it like if you think about

it just my hot take half the reason why

the one got the attention that it did

cuz it looks like that it's got teenage

engineering Aesthetics it's not an

official collab or anything it just has

te hot sauce all over the design

language and so now you know what it is

and why it's not an app let's talk about

problems so uh the first problem I have

is that both of these primarily run off

voice commands where should Ken and I

grab dinner

tonight here are some recommendations

for you Sushi Ron shisen and elephant

sushi they lean on voice as the main way

to be able to interact with these two

devices but right now when we use our

phones and our regular apps there is a

lot of fine-tuning that we do like on

the screen by poking things and reacting

to stuff that happens on our phone in

real time like it doesn't matter what

you're doing if you're like uh you know

ordering food or trying to book a hotel

or a flight you are making decisions and

adjusting your thoughts looking at the

screen like you're seeing reviews or you

know an appealing photo or pricing like

there's stuff that's actively affecting

what you're going to click and tap next

and same with an Uber like I'll often

adjust my pin be like you know pick me

up across the street because that's

going to save me 5 minutes for like the

the turnaround right so there's stuff

that we actively do in our apps because

apps are built and optimized for human

interaction right now but to remove all

that stuff and to just have voice

commands with like a little bit of

adjustment on like a tiny screen that's

not how we use things the moment there's

some kind of like option assessment that

we have to do I think this whole voice

interaction thing falls apart it's

unfortunate but I think that's just the

way that humans interact with online

decisions right now and to expect an AI

to like my wife will spend two or three

days like full ass days focusing on how

to book a vacation you know what's the

exact thing that we're going to do you

think an AI can do it in two voice

commands bro come on man that's not

realistic and the fact that they're

showing it like that is I think

misleading the other thing they have

shown in some of their demonstrations

and also in released videos like on

their social media people interacting

with these devices like kind of the

pre-release devices and they're editing

the videos on the wait time like they'll

ask it a prompt who designed the

Williamsburg

Bridge the Williams and instead of

letting the video run so that the

viewers can get a sense of how long

these things actually take to respond to

things they'll cut the video and just

edit it to when the response happens it

is so weird it's like that weight time

that lag is a big factor in determining

how good or useful these things really

are there's the argument to be made like

okay this is pre-production engineering

stuff so the wait time that's live right

now isn't actually representative of

what the product will be in the future

sure throw some text up why are we

hiding the latency of this stuff behind

edits I hate that okay another thing I

want to talk about is in regards to the

Humane AI pin specifically so I remember

when they first showed off this laser

projector Tech at a Ted Talk and it was

really cool right it looked like it

could be really small but now that it's

out it turns out it's actually a fairly

chonky device but it's quite heavy so

unless you're wearing like a thick

material or you're wearing like really

tight fitting clothing this thing tends

to just flop around when you're using it

now it is strange to me that we're I

don't know a few weeks out from their

ship date and they still have not

allowed people to demo this thing live

like if you went to mwc you couldn't

handle these things you had to watch The

Humane employees demonstrate the product

for you to look at it was it's so weird

it's like people should at this point

have a good idea of what these things

feel like when they're being worn and

there were reports of the devices

overheating because the laser projectors

were being used too much like it's just

so strange that this is the state that

the product is in just weeks out from

the ship date I honestly think that the

the whole idea of the laser projection

is like it's a cool idea but I'm worried

it's a gimmick and my simple reasoning

is that if that feature was actually

good if a laser projected screen onto

your hand was something that people

actually wanted we would be able to go

to the store and buy that like you can

create that Tech today like you could

have a wireless like a device that just

your screen from your phone is

transmitted wirelessly and it's

projected to your hand you could make

that but it doesn't exist because I just

don't think people want that but Humane

is telling us that well that product's

not good what you just described Dave

but if you add AI to it and just make it

voice operated now it's good I don't buy

it I really don't okay the last thing I

want to talk about pertains specifically

to the rabbit R1 so this device has

something kind of special it's their

large action model the lamb and it's the

feature that gives me the most hope for

what they're trying to do with the R1 so

the easiest way I can think of to

describe what a large action model is is

that it's an AI model that performs

actions based on your prompting so

instead of just reacting to language and

text and stuff like that this can

actually do like Mouse clicks and

scrolling and it can be trained just

like a language model so if you just

showed this lamb how you edit a photo in

Photoshop let's say and then you got

5,000 people to show this lamb how they

do it it would learn and then in the

future it would be able to make its own

decisions or the lamb would be like hey

if you want to make a warm looking photo

this is what you do because I saw what

all these other people did in Photoshop

I click this I click that I scroll here

and that is a lamb a very primitive

description of it now I think the idea

is very cool but when they showed it

during the presentation they were first

of all they were

showing the use of their lamb on

services that have apis so I don't even

know if they're actually built Lambs

that function the way they did in the

presentation but also I imagine that

people that are pre-ordering this device

think that these Lambs are going to be

available right off the bat and they I

don't think they will be how could they

be I think the 100,000 people that

pre-ordered this thing they will be the

ones responsible for training these

things to be able to do whatever they

end up doing in the future so if you

bought one you're expecting to be able

to do anything other than like simple

GPT instructions or perplexity

instructions I don't think it can how

could it that has no training data as of

right now at least not enough of it to

make it actually functional or useful

and that I felt was a little bit

misleading now now they're they're a

business right they're they have to show

the best case scenario of what this

thing can do and I respect that but it's

just when I see the presentations when I

see the demos I'm like come on man

there's no way this thing can reliably

book a vacation in three voice prompts

like there's no way and also errors like

AI hallucinations right now when you

have an AI error like a hallucination in

mid Journey you just get a dude with 12

fingers we laugh at it it's hilarious

but if you have and hallucination with

large action models you're talking about

ordering 50 pizzas or like you know

ordering a booking a vacation that's

$20,000 over budget like this is real

stuff here because these are action

models I feel like it's irresponsible

for them not just to market it like they

it can do it but also to even have that

ability right now like it should be

super Super Beta and people should be

aware that this is not something you can

do right now the thing I have to keep in

mind and I think you guys also after

hearing my rant about this the thing we

have to keep in mind is like when it

comes to the technology that we have

access to today the things that we use

every day all of the stuff we use came

from lots of companies trying lots of

weird stuff and most of them ended up in

Failure but the result in the end is

awesome products for consumers so I

think these two products that's part of

it right that's part of the process to

get there I respect that and I want

these guys to succeed I don't want to

see them fail I want to see this stuff

pan out the way that it could but the

marketing right now is just so strangely

optimistic and like both of

them it doesn't make sense to me how can

you how can you be responsible and put

marketing material out like that that's

all okay

um that's it hope you guys enjoyed this

video