Episode #12: Rabbit r1, GPT Store, UX / UI

Based on the show notes, here’s a brief introduction for your podcast episode:

In this episode, we dive into the innovative Rabbit R1 Device, explore the potential of OpenAI’s GPT Store, and unravel the complexities of the New York Times lawsuit against OpenAI. Join us for a journey through these cutting-edge developments in the world of tech and AI.

Show Notes

BigCheese Podcast Show Notes – Episode: 12

Topics Covered:

1. Rabbit R1 Device

2. GPT Store by OpenAI

3. New York Times Lawsuit Against OpenAI

Rabbit R1 Device:

- The hosts discussed the newly released Rabbit R1, a device designed to provide large language model interactions without the need for a phone.

- It supports SIM cards and Wi-Fi, trained to perform specific tasks such as shopping on Amazon or playing games like Diablo.

- Privacy was highlighted due to the camera design that points away when not in use.

- Comparison to existing tech and speculation on the future of user interfaces.

- Speculation on the potential for Rabbit to watch and learn from users' behavior.

Relevant Quotes:

- "It's about the size of a Post-it note... specifically designed just for a large language model in this thing." – Brandon Corbin on Rabbit R1's design.

- "That hypothetically Apple could just release their own thing that does a lot of the same functionality... it's only 199 bucks." – Andre Harakas on competition for the Rabbit R1.

- "Speaking is such a powerful medium for productivity... from a web perspective, speaking has never been able to be leveraged for productivity." – Sean Hise on voice command devices.

- "They want to get rid of the UI, right? But their entire model is based on a fucking UI." – Jacob Wise on the potential conflict in Rabbit's design approach.

GPT Store by OpenAI:

- OpenAI has opened the GPT store, a place to find and use different GPTs.

- Discussion on how these GPTs can use actions to call external APIs for integrating services.

- Monetization plans for creators who produce popular GPTs.

- Skepticism about the utility of some GPTs and how they may end up providing narrowed or sales-driven outputs.

Relevant Quotes:

- "I'm basically just creating a really good prompt for chat GPT." – Brandon Corbin on creating GPTs.

- "Likey, I'm a plus." – Brandon Corbin expressing approval for the GPT store feature to use actions.

- "What seems to be... there's a conflict though. And you identified it, that they wanna get rid of the UI." – Jacob Wise on the GPT store's approach.

New York Times Lawsuit Against OpenAI:

- The New York Times sued OpenAI for using their content to train models without compensation.

- The potential consequences for content creators and future of publishing valuable content for free.

- Comparison to Google's practices and concerns about scraping the web for training models.

Relevant Quotes:

- "Every single person's coming for their lunch that that data was trained on." – Jacob Wise on the implications of the lawsuit.

- "This is going to the Supreme Court, 100%." – Sean Hise on the potential legal trajectory of New York Times vs. OpenAI.

Additional Notes:

- The hosts mentioned the Big Cheese AI website updates, including podcast episodes, latest news, and AI product features.

Remember to follow Big Cheese AI on Instagram, Twitter, LinkedIn, and subscribe on YouTube for more AI insights and discussions.

- So when I was creating GPTs, I'm sitting there going,

okay, I'm basically just creating a really good prompt

for chat GPT, but with GPTs, you can actually use actions

to call external APIs to integrate your services

into your GPT, thus making it more unique.

Likey, I'm a plus.

Hey, hold on real quick,

I think we have a technical problem.

- She did it.

(upbeat music)

- Let's get this thing rolling.

Everybody, welcome back to the Big Cheese AI Podcast.

I'm one of the world's top moderators,

Andrej Herakles, joined by Sean Hyes,

Jacob Wise, and Brandon Corbin.

This is the Big Cheese AI Podcast.

and smash the subscribe button on YouTube.

Today, we're getting into the cheese.

So let's talk about the biggest news right now.

What's going on with this Rabbit thing?

- Out comes Rabbit.

Sorry, Rabbit R1 is the name of it.

And it's about, I guess it's about the size

of a Post-it note, maybe about an inch thick.

And it's this red orange device

that is specifically designed just for,

it's not a phone, it's a new type of device

that just basically gives you a large language model

in this thing.

And it's got a camera on it,

so it does support a SIM card,

and it can connect to Wi-Fi.

So it's not connected to your phone,

but it can connect to the internet, it can do its thing.

But what it really is, is it's meant to be a,

what do they call it, a large action model,

I believe is what they're calling it.

And they've trained their own model

to basically do specific tasks.

So go buy me something on Amazon, right?

Go buy me, you know, go repeat my order of this.

And they've trained their own model

on being able to navigate Amazon,

be able to click on different things,

and to be able to do it.

So that's actually kind of cool, right?

It's their own foundational model that can do these things.

And you're gonna be able to train it yourself.

So you'll be able to point your Rabbit at your screen,

and be like, and one of the things I saw

is somebody did it with Diablo.

And it trained to be able to survive in the game Diablo

just by watching the screen.

And that's gonna be coming out later.

It's not available in the current model.

- Is there a world where it just,

the evolution of this thing is it just starts

to train on you?

- Oh, sure.

- It just watches you, and does you.

- But what's the time for you to go brush your teeth, right?

'Cause traditionally you do it.

Apple's already doing that.

Every Thursday at about 11 o'clock,

or 12 o'clock, or one o'clock,

I get into my car to drive here, right?

My phone, my watch tells me how long it's gonna take me

to get here, and I don't have it set up

as a calendar event, right?

So it's kind of already doing it.

And that's where I kind of run into a problem.

That hypothetically Apple could just release their own thing

that does a lot of the same functionality.

That again, it's only 199 bucks.

And it doesn't seem like there's any subscription model.

It's a one-time purchase.

So that's the only thing that kind of caught me.

I'm like, "Eh."

For 199 bucks, if I could have that thing

just sitting on my desk,

doing a couple things here and there,

then maybe it does become something that's interesting,

that I don't need to worry about breaking on my phone.

I don't need to worry about it.

So yeah, it's definitely an interesting thing.

There's a lot of comparisons to it in the Humane Pen,

which is that one that you're gonna pin to your chest,

and then use your hand, and it like laser,

shows some shit on your hand.

- Which I thought was super cool when it came out.

But the more I thought about it, I was like, "Eh."

- I know, 'cause I'm looking at my watch,

and I'm like, "Well, that looks a lot better

"than a green laser, War Games-looking thing on my hand."

- Yeah, this one reminds me,

you've seen the movie "Her," right?

- Oh yeah. - This is giving me,

and I wonder if Scarlett Johansson

is gonna do the voiceover for this.

- That would be awesome.

That would be very cool.

- So here's my take on the rabbit thing.

So you watch the keynote, and you go,

"Okay, this makes sense."

There's a couple things that I think

that really make a lot of sense.

First of all, the one thing that they go through is,

the first step when you're really onboarding yourself

to this platform or this device,

is you go through and add all of your services.

So you go through and add your Spotify,

you go add your Instacart.

- But you're going through their web.

- And you're going, I know.

So there's a setup, which is fine.

So let's just get, you can get past, I can get past that.

But you're going through and you're setting up,

here's my life, and here are the,

and you're seeing these massive, you know,

these massive apps that all use OAuth,

that all authenticate through your,

these apps that are obviously apt to do that stuff.

And then you're saying, "This is the stuff that I do."

And so now when I talk to you, Mr. Rabbit, or Mrs. Rabbit,

you're gonna know what I do,

and you're gonna know that it's me.

And when you talk to these services,

they're gonna be able to do it on my behalf.

So when I tell it to play a song, you know,

on my playlist, it's gonna know me

because I'm already logged into Spotify.

When I order some bananas and some milk

and tell them to deliver it in two hours,

it's gonna know me, it's gonna know how to charge my card.

So all of these things are going to be able to function.

And that, obviously, I don't think,

I don't know if that's been done anywhere besides the,

really, I guess the browser or your phone

via different apps.

But no one's really built an app

that unifies all those experiences.

I know that in China, you have these super apps,

WeChat, things like that, right?

Where you're having one experience

and you're accomplishing all these different tasks.

I think that is interesting.

And go ahead, Jacob.

- Yeah, I was just gonna say, you reminded me of,

so the demo was, oh, I'm gonna plan this trip

to London or something.

And it was like, finding flights and itinerary and this.

And it just printed out itinerary with flights.

I'm like, that is not how anyone has ever shopped

for a trip.

I would go and I wanna know,

I have a lot of things, costs I'm weighing.

Like, I might wanna wake up early, but I might not

if the ticket's not that much more expensive.

I need to make that decision.

That, I think, is probably not ready yet.

But add something to my calendar and knowing,

having context of like, hey, dude,

do you actually have something at one?

Are you sure you wanna add something at 1230

that might conflict with that?

That adds a ton of value.

And Google can't do that yet, but they probably will soon.

But authenticating to those types of applications

and having context around all of your data, that's cool.

- Eliminating, so the premise of what he was talking about

is he introduces the product,

and then he goes to this slide where

there's a million iPhones with all these different apps

right behind him.

And he basically makes this premise that says,

these technologies that exist today

aren't saving people time.

They're using it to kill time.

Because he talks about how most time spent in apps

is with entertainment.

Well, that's an arguable thing because obviously,

when you're playing a video game,

you take 20 minutes.

When you go check a Slack message,

you spend three seconds, whatever.

But he's sitting there and he's talking about

how we have app overload, right?

And he's fundamentally making a hypothesis

that says user interfaces are going to die.

That's what they're really doing here.

And I don't think that this product

is going to succeed at all.

I think it's just gonna be a new,

it's just gonna be,

it's gonna be a drop in the hat at CES.

CES is one of those things where they sell out

and then a bunch of people churn.

But the implication of the concept

that they're going after is real.

And I think, I see it all the time.

And it's already happening,

but it's an evolution of the user interface.

Like when I pull up my, like you said,

when you start driving here,

it tells you where you're going.

When I wake up in the morning,

I have the quick order button on my start on Starbucks

that all with one tap.

I don't know, I mean,

even though it's real easy to order on Starbucks,

really easy.

It knows where you order it from,

it knows what you order,

and you have to do a couple of clicks.

Now with the quick actions on iOS,

you literally just click one button, right?

And it's using a third party payment service

to do that with PayPal.

I mean, those technologies are already baked

into those operating systems at a basic level.

And I think what Apple is missing out on

is actually first party integrating

these third party platforms.

Do you know what I mean though?

Like, why wouldn't they have done that a long time ago?

They do that with their apps.

But if you're thinking about,

I'm trying to make it so people can do things

very efficiently and effectively,

the iPhone's the device.

I don't need, honestly, I was offended

by the user interface thing with Rabbit, right?

Because if we know, as a family from a web developer,

coming from someone who understands

design systems and standards,

like the iPhone is an amazing way to experience computing.

The web is an amazing way to experience computing.

And to say that you're gonna,

just because you can input your voice

and it's really fast,

that you're gonna just be able to put some dog shit

user interface out there,

and that's gonna be okay with people.

Like Jacob said, like,

he's gonna wanna look at those flights.

He's gonna wanna be able to interact with that,

swipe away, do all these actions

that have taken years and years and years to perfect.

And so I think the concept is really interesting and great.

I think that absolutely this large action model,

which basically creates these agents,

and that's the question I have for you.

I think that they're onto something huge.

But here's the question I had.

These agents, what the hell are these agents?

Are they literally, he calls them rabbits.

Okay, so you go and you say, I wanna book a flight,

and he's like, we got these rabbits.

And okay, the rabbits are agents.

And the way I'm old school,

so I'm thinking that they got like

one million puppeteer instances running Chrome headless,

typing on a computer, but they're doing vision,

or Jacob mentioned they might be doing,

like, what is it doing?

- I have no idea.

I have no idea how they're actually executing on it.

But what seems to be, there's a conflict though.

And you identified it,

that they wanna get rid of the UI, right?

But their entire model is based on a fucking UI.

And if you don't have a UI,

then you don't have anything to train it on.

And so it is, it's a very interesting,

I think there's gonna be a lot of

extreme opinions on this device.

I think the hardware looks awesome, right?

So this is coming from Teenage Engineering,

is the design firm that did the work.

It brings in some stuff where they've got like a real,

it's a push to talk, so it's like a walkie talkie.

It's not always listening.

And it looks like it's very tactile, like, boom,

like you can just feel it, right?

And then they've got this nice little scroll thing going on

where it's like the real 3D.

And the more I saw it, I'm kind of like,

ah, it feels very much like toys

that we grew up with as kids, right?

But I almost feel, and I look at my phone

and I'm like, there's no real buttons on my phone anymore.

It's all just like this fake bullshit

with haptic feedback and all that.

- You know, Honda, I'm a big Honda fan boy,

but they switched their volume to a digital volume.

And that was like two years ago.

And they switched back to an actual button

because of that exact point,

where it's like people like physical buttons and knobs.

- Again, Apple, out of everybody,

Apple, I think does one of the best jobs

of making a fake button feel like a real button, right?

I actually had one of my phone,

back when we used to have the home button, right?

Where you'd have that little haptic feedback

and pressing it.

I had, something went wrong with my phone

and that thing all of a sudden amped up about 10 times.

And it went, boop, boop.

Every time you hit my home button,

the vibrate was like so loud, everybody could hear it.

And it was just like, oh my God,

it just really shows you that, yeah,

there's something about the real button nature.

- So go to Google CES iPhone keyboard, 2024,

and then scene switch over to it.

- Well, give me one second.

I wanna comment on this a little bit.

- Yeah, sure.

- I think it's one thing to say,

like user interfaces from a phone's perspective,

like you started off with a computer.

When you had a computer, you could use all 10 fingers

to input information and get stuff back out.

Then we moved to the phone.

So you're just using your thumbs to get information in

and get information out.

And that's limiting to an extent,

which then that's why we made incredible interfaces

for the mobile apps and buttons

that can do multiple different things

because you're only using two fingers.

And I do think that there is something really fascinating

about this Rabbit product

where you're not using any fingers.

You're not using two fingers or 10.

You're just using your voice

and the speed at which you can articulate

and get a thought through your voice,

as if it's a walkie talkie to get an answer back

is so much faster and easier

than if you were actually using a user interface.

So I could seriously be more adopted to this.

- It's not that many years until most computing,

especially casual computing and getting things done

and interacting with services like put this on,

order this, do that,

that there will be the input,

the user input will be something that doesn't involve.

It's a voice, it's an action, it's something.

- I mean, what's like the number one thing

in homes in America right now?

And it's been that way for years now.

It's been growing like crazy.

Echoes, Amazon Voices.

I mean, those things make your life convenient.

Whenever someone drops off something

at my grandparents' place,

who are like 65 and 70, by the way,

it, ding, Amazon, what happened?

Oh, your package has been delivered.

They could have realized that by opening the door,

but it's just easier to talk to.

I feel like speaking is such a powerful medium

for productivity.

We just have never been able to really be productive

outside of the spoken word via podcasting,

via talking to each other.

But from a web perspective,

speaking has never been able to be leveraged

for productivity.

And in my opinion, communication, the verbal word,

is the most productive way to get things done.

- You hear a lot, and you had a post the other day

that kind of touched on this as well,

which is like service overload,

or we just have too many things going on, right?

Like we have too many apps,

we have too many things that do the tasks in our lives.

And the problem is we don't have a universal language.

So I have to remember, okay, I downloaded this one app

that did this one thing that was useful,

but I kind of forget because it's been two months

since I used it.

And I do think Rabbit and these types of companies

are onto something.

We have an absolute overload of information,

and we need to find a more natural way to consume it.

Now, do I think that I'm ever gonna get on my phone

and say, order some very specific thing?

And like, unless it's something I order every month

that's the exact same, I still price check that stuff.

It scares the hell out of me.

Like all of a sudden I'm gonna get an Amazon bill

for $5,000 'cause it ordered whatever.

I have a cat, I get litter every month.

I don't just blindly reorder it unless it's through Chewy

and there's a guaranteed price or something.

It's like, I'm still checking that thing

because that's just how it is.

So do I think it's a good idea,

but maybe the wrong point in the wrong direction right now?

Yeah, I think that's probably where it's at.

I'm buying one.

I'm definitely buying one.

Yeah, do it, man.

Let's play with it.

Well, hey, let's check it out.

So everybody at the Big Cheese Pie,

we have an upgrade.

I'm gonna show you right now.

Look at that.

Can we roast them right away?

'Cause they're CDNed.

What the hell is going on?

At this moment, the Rabbit website

has its hero images completely broken

and Jacob has a technical diagnosis

he'd like to share with the world.

Yeah, they have a custom CDN

that's not working right now.

It's a cute name too.

It's like lotsofcarrots.com.

But the only other news and noteworthy thing

coming out of CES was this.

So for those of you that grew up

in the Blackberry generation,

there is now a device that you can buy,

what is it, for $149.

Oh yeah.

And it will provide you with a keyboard for your iPhone.

So I kind of wish they would've spent

a little bit more time to make it

so even if it was just a little bit fatter,

that the keyboard could slide down optionally.

'Cause now all of a sudden,

that's like a nine and a half inch banana.

Oh, you can't slide it up?

No, it's just always,

you've always got this like nine inch sausage in your pocket.

Well, you go.

Aren't Blackberry--

Nine inch banana.

Aren't Blackberry fans--

You mean not everybody's like that?

So that's my only problem with it.

But conceptually, I think it is blowing up

and a lot of people are excited about it.

I don't, it's just, these phones are already big enough

that man, that just makes it look obnoxious.

I mean, again, that kind of goes back

to my original argument on how like,

the people want more ability to get things done faster

and just your fingers is limited.

Well, there's been,

just like we talked about last podcast

with the skeuomorphic stuff,

there's a resurgence in the Taptic

and the old school devices.

And so for Rabbit, you get a lot of points

for being forward thinking.

Can we bring up, sorry, can we bring up a Rabbit

like that actually shows us what it looks like?

Oh, yeah.

Because on their website,

there doesn't seem to be a picture of it here.

Yeah, the whole CDN is not working for Rabbit.

What's it called, R1?

Yeah, the R1.

Okay, let's pull up a--

Just get images.

Yeah, so, let's do this.

So from a design standpoint, there it is, right?

And that's gonna be, that's pretty sexy.

So up on top there, next to the screen is your camera.

This camera is 360 degrees, so it can go forward,

it can go backwards. What does that do for you?

I don't get it.

Well, what's nice about it is that by default,

it's either pointing down or up, so it's always private.

Right, so that's another thing

that they're kind of leaning on here

is the privacy nature of it.

So this way, it can either be taking a selfie

or it could be filming your monitor

while you're doing something.

And then the thing below it is your actual thumb wheel,

like is the scrolling, you know, is your scroller.

Okay, what do you mean it's pointing at your computer

and learning something?

Yeah, so right now, the top is,

like if that was on our desk,

that camera would be pointing to us, right?

Right, but if it's pointing at your computer,

it can learn stuff?

Yes, and that's the piece

that makes it very interesting, right?

And that was the thing I was talking about,

but with Diablo.

So you can point it to your computer

and you can be like, all right, so this is now,

I'm gonna go send some spam to Crafted.

And I go to crafted.up, I fill up,

or crafted.up.com, and then I'm gonna fill out the form

and I'm gonna, you know, send some obscene thing to Jacob.

And it can remember that and then it can replay it,

is basically what their training model is.

And so this train, which is gonna be coming down

later on down the path, that's where it gets interesting.

And so then you could see hypothetically

that they would then allow those trainings

to be maybe shared across the community.

Well, that changes it a little bit

because you're looking at something

where you can crowdsource the agents, you know?

And that changes things

because I was looking at it from the keynote

and I'm looking, okay, so 10 really good apps

like Spotify and Instacart are gonna do great with this,

but everybody else is like, what's the point?

You know, they changed their UI.

And so it's been trained on the buttons over here.

- Yeah. - And if that's how it works,

I don't know. - That's what I'm kind of curious

about with their model is, because again,

with large language models, you don't have to be exact.

- Yeah. - But you can be

a little fluffy, you know, things can shift.

So if all of a sudden we have this big orange buy now button

that becomes a purple, you know, pay now, does it work?

- I'm also very curious how they're gonna do this

for a fixed cost.

- Yeah, the 199 seems like it doesn't scale.

- If you do not think there's a subscription

coming with this, there's gotta be.

- There's gotta be. - Of course there is.

- That was the first question on the Reddit

this morning, by the way.

Reddit, dude, so Jacob's this big Reddit guy

and I'm like, not, but like,

you, if you wanna get just get your ass kicked,

head on over to Reddit. - Honest feedback.

- It is, it's the most brutal, honest feedback.

We were just talking about it.

And I've been there and one of my apps

that I released for Android, I go to the Android subreddit

and I'm like, hey, look at my app.

And then 100 replies that I'm the biggest schlub

in the world because I didn't make it material design.

Right, and I was just, I mean, just brutal, savage.

But at the same time we were talking about it,

it forced me to go and delve into material design spec,

really understand it, really learn it.

I went and I was like, all right,

I'll be back in a couple months.

And I went and I redesigned the whole damn thing

to be absolutely material.

And I came back, I'm like, all right.

And they're like, all right, you're better.

(laughing)

- So what are our thoughts on this Rabbit R1?

Good cheese, bad cheese, what do we think?

- It's good.

- I think it's good.

I'll be curious to see how it actually pans out.

- It's good, but don't act like you're gonna

replace the user interface.

- Yeah, it's good cheese, but it does give me

Kickstarter vibes where like, it's probably going

to be a catalyst for a bunch of really great ideas.

Is this gonna be it?

Probably not.

- Yeah, I'll say it's good cheese.

I think it would be cool.

I like that click walkie talkie with my life.

- Oh my gosh, next sale, bring it back.

- If you can implement any product that actually works

when you press the button and ask it to talk,

'cause Siri, you go, hey Siri.

Wait, oh, I was talking already three seconds ago.

You missed that one, honey.

- All right, so aside from Rabbit R1,

we also have a big update from OpenAI.

We have the GPT store.

Any news on that, what's going on with that?

- Well, they announced on their blog yesterday

that the GPT store was open.

Guess what, it's not a store at all.

And these have been around, so essentially

what happened was, when we've covered it

in a previous podcast, if you haven't seen it,

GPT's killed our startup, that whole conversation.

So you can basically go to OpenAI and build GPT's

and it's essentially, you're gonna give it

a bunch of prompts and you're gonna create

this really specialized GPT.

It might do nothing, it might do something,

it might do something really specific.

- It mostly does nothing, in my experience.

- The GPT's?

- It mostly does almost something, but nothing.

- Well, I've had good experience so far,

but basically what they did is they just created

a page in the app where you can actually search them

and find them and start using them.

- That was kind of my disappointment,

where I'm like, oh hey, they launched it,

and I'm like, oh, I'll go share it to LinkedIn.

Then I actually go through it, I'm like,

well, what the fuck?

(laughing)

- The one thing, and go back to the site.

The one thing that is interesting here is,

they, in their blog post, they said that they're

introducing monetization in Q1.

And so what that means, and it makes no sense to me yet,

is that if you go and build a GPT and people use it,

they're gonna share some revenue with you.

Is that, I think that they're going more

towards the YouTube, Instagram style?

- Twitter.

- Or Twitter, you know, where--

- I won't charge for my GPT, but the adoption of the GPT

would hypothetically give me a revenue share--

- Yes, that's right.

- That's what I think is gonna happen.

- I think that's the best business model

of all time, by the way.

- Well, I mean--

- Because you can't switch it.

That's the one thing that's nice is

it's not guaranteed income.

When I go and I put an app up on the App Store for $3.99,

and somebody buys it, I'm gonna get my $3.99 minus my 30%,

and that's gonna go to me, no matter what.

- Yes.

- Yeah, these creator funds and these other ones,

they're variable.

They can be like, "Oh, hey, we're making money."

They're not, and all of a sudden,

you don't have a guaranteed income.

- Well, OpenAI crossed $1.2 billion,

$1 billion of annualized revenue.

- Jesus.

- So, but we know that their costs are high.

So from a revenue sharing perspective,

you know, they don't have advertisers.

You know, they're this model.

That model, the model's gonna be interesting.

So while they put out a GPT store,

it is not a store, it is a indexed,

categorized search service to discover GPTs,

because the only way you could get on a GPT

is if you had the link, and somebody shared it,

and some people started creating websites

and all these different things to just do

what this just did.

But I did, I mean, I think it's,

I think that we're just sitting here

waiting for the real thing to happen,

and it could be the next FardApp opportunity.

- Can you use these GPTs if you don't have,

if you're not a Plus member?

- Good question. - I don't know.

Well, I did notice that they just added Teams now.

So for $25 a month, I can now have a Teams account,

which gets me guaranteed that it doesn't train on my data,

and some other bullshit.

But real quick, that consensus, by the way,

consensus, and they've just rebranded,

they were like, I don't know, StudyGPT

or something like that, but that consensus

is actually a really cool platform

if you wanna go do any kind of studies on scientific papers.

They have 200 million papers that they've processed.

- I think it's really cool.

- Yeah. - So I do not understand that.

So that's my biggest question with GPTs.

So if I go create a GPT, okay, so go back to consensus.

You go to consensus and you go,

okay, consensus is your AI research assistant,

and it's got 200 million academic papers.

So when they created this GPT,

did they tell OpenAI to go train itself on their data?

- No, so they, I think in their case,

they actually had their own data source.

I used to, Emily came down, she's my wife,

she's like, you should totally do a demo on consensus

'cause she uses it all the time.

- Really? - And so I went

and checked it out just on the website.

This was before the GPT announcement.

- And by the way, this is one of the most used GPTs

out there today. - Yeah.

And so you can just go to consensus,

I think it's .ai or something like that.

- I think it's 'cause students probably.

- Right, but you just type in your thing

and it goes, finds the papers,

and then it gives you a breakdown of it.

So I think what they did then is they just did

the integration with ChatGPT to make that actually happen.

But it's still going to their data source with all of this.

So all of those 200 million papers

I don't think are part of ChatGPT's core.

- So with the GPT you can actually,

you can have it call APIs?

- Yeah, well, that was the cool thing

that Brandon mentioned a few episodes ago

where that sticker company, you create a GPT

that leverages something on your side of the company.

So it was like, oh, I can create using DALI

these images of stickers.

Now I wanna buy 10 of those stickers.

That will send the business directly to that company.

- I'm still so intrigued though

because when I created GPTs, there was no prompt.

It's their actions.

And it's very subtle at the bottom.

There's this actions thing that you can click.

And then you can go and put in additional endpoints

and things like that to configure it.

- So previously when I was creating GPTs,

I'm sitting there going, okay, I'm basically

just creating a really good prompt for ChatGPT.

But with GPTs you can actually use actions

to call external APIs to integrate your services

into your GPT, thus making it more unique.

Likey, I'm a plus.

- Oh yeah, I mean, think about this.

Like think about like an example for Crafted.

How to choose a good development partner GPT, right?

And then it gives you a suggestion

and then it sends you to Crafted, right?

I mean, that's a little example.

That's the only answer.

(laughing)

- Here's a risk question with GPTs.

Are we creating a situation where people

are going to ChatGPT, which they think summarizes

all relevant human knowledge, and they start getting results

that are just salesy or contrived or narrow

because they're actually calling third party data services

that are not part of the model?

- Well, I think that that's that separation

of calling it a GPT.

You can be like, well, that's a GPT.

That's not ChatGPT because we wouldn't answer it that way.

- That's a good point.

Google's at least held accountable in the sense

that it's the most linked or they have metrics

or reasons, cues why that is the first article.

It may not always be right, but at least it's, you know.

- My advice to people using GPTs would be

to understand where their data is coming from

because it doesn't necessarily come from OpenAI.

- We got some New York Times drama, gents.

- Yeah, it's a little old news,

but New York Times sued OpenAI.

And...

- Why did they sue them?

- Because for obvious reasons, I mean,

and I've been saying this about Twitter for 10, 15 years.

I mean, how many journalists go tweet all their news

on Twitter for free and then link back, you know,

oh, here's a link to my article.

They just literally wrote the article on Twitter

and live streamed the game too with all your comments, right?

But same thing with New York Times, right?

They, OpenAI uses their content to train their models.

And so, I mean, OpenAI didn't learn all this stuff

from thin air.

And so you have all this content that's owned.

I mean, these are journalistic properties, right?

And so OpenAI is using the internet to train their models.

And then, you know, New York Times, those are smart people.

You know, they've been in this game a long time.

They've survived.

Don't subscribe for the $250 a year.

Wait for the special, get it for 69.99

and then cancel the next year, get the special again.

But seriously, they have a legitimate case against OpenAI,

in my opinion, in terms of they are not being compensated

whatsoever for the content that is being used

to train their models.

Thoughts?

- So the most damning evidence there is,

and again, I'm not sure exactly what prompt

that the lawyers of OpenAI or of the New York Times used,

but they were able to output one of their articles,

like 90% of it, and they did a side-by-side.

They're like, "Here's our article.

"And here's what ChatGBT output.

"And all of the things in red are the same."

And it was literally 95% of the article

was just output word for word.

That's very damaging.

I mean, that's not good.

Now, again, what you're saying is that ChatGBT is saying,

"Well, you broke the terms of service to execute

"and to pull that off, so it's not valid."

I think they might be screwed here.

And it's going to be a problem,

because at the moment, if they do win,

every other party's coming for their lunch.

Every single person's coming for their lunch

that that data was trained on.

- So here's where this goes, which is,

well, I guess I'll ask this question.

How does Google get away with what they've done

with indexing and crawling websites?

- Because they don't output the results.

- Yeah, they're starting to. - Kind of.

- That's what I was talking about.

- So their carrot is, Jacob, that--

- I mean, at the very least,

Google's always been able to say,

"Oh, no, we're actually doing a service for you.

"We're promoting you.

"As long as you do good content,

"we'll put you at the top of the list

"and we'll give an opportunity."

But I've been saying this for years,

like the cards and the frequently asked

or the Google questions and all the things

that they're showing directly on Google now,

well, I don't have to click on your site anymore.

- That's not taking them to the site.

- Yeah, yeah, and there's advantages to that.

Like if you're selling products,

you can have your product one-click buy on Google now,

which is great, but if you don't have a product

and your product is content,

if you have that publicly available,

Google's already doing that,

and now OpenAI is taking it a step further.

And what incentive is there?

Well, that model is broken then.

Why would you ever promote

or actually publish valuable content for free anymore?

- Right, right.

- So that's the question here that this will answer is--

- So OpenAI's only doing a billion dollars a year,

and they're training their model on the internet.

So is this an outcome where it's the internet?

I mean, is the Supreme Court ever ruled?

I mean, at the end of the day, it's the internet.

You're putting your stuff out there.

- Well, you can legally scrape LinkedIn

as long as it's in the public internet,

and you can do things with that data.

That has been voted on and has been approved,

so you can do it.

I guess the question's gonna come down to,

specifically with these large language models,

maybe not as much as Google,

because they're kind of taking your information,

reformatting it, you know, and like putting it into a card

so they don't actually have to click on your website now,

but with OpenAI, it's literally just like

the same exact copy,

and copyright is very specific to the copy, right?

And so that's the one difference here

where they might just have to say that we can never

make our model so it never outputs the same

sequential guarantee of terms,

and it always has to be shooken up or something like that.

No matter what, it's going to force

these large language models to kind of rethink

their positioning on how they allow

the same content to be output.

That I could see 'em go,

okay, we're just gonna go spend all the time

on making sure it just can't output the same exact stuff,

and instead just rewrite it to be so it's not like

a pure 95% copy of the content.

- Yeah, that reminds me back whenever I was in college,

and you just put it in the paraphrase bot,

and the paraphrase bot gave you a little bit

something different than it was originally put into it.

That's basically what they're gonna do

at a billion dollar level.

- Well, so I think to summarize it,

the internet used to exist, links would be promoted,

and then Google came around, and they indexed all your stuff

but they linked to your articles.

Now, ChatGPT, OpenAI, just took all that content

and created a model and are serving it up in a chat

and saying, look what we did, mom!

And that's just not going to work.

They're gonna have to pay up.

- Yeah, I think they might have to have 'em,

but it's wild, isn't it?

I mean, it's like Google, the whole sphere is shaken up.

No one really knows what to think.

- This is going to the Supreme Court, 100%.

It's going to the Supreme Court.

They're gonna have to figure it out

because AI's so important and they want the business to,

I mean, you saw what the executive order,

part of the executive order was, oh my God,

people are gonna lose their jobs!

But they were also like, hey, let's use this.

We should institute a commission

that's looking into implementing this in all departments,

safely, privately, for the benefit of Americans

and the markets.

But the one thing that the markets will always do

in this country, in the United States of America,

is they don't like it when people steal stuff.

If there's a class action law, I mean,

the law is the law and these judges don't change the rules.

It's been tested in the last few years.

But I mean, they will look at it to the letter of the law

and one of those judges, if it goes all the way

to the Supreme Court, they'll get smart on this

and they will figure out if someone is taking advantage

of someone else and essentially breaking laws.

- Everybody, we are Big Cheese AI

and we help companies launch and market AI products.

If you guys have not checked out bigcheese.ai,

you'll get a full overview on what exactly that looks like.

We have brand new product drops and news updates

dropping on the site daily.

Brandon Corbin is the big cheese, made this website.

It looks absolutely phenomenal.

- We've got all our podcast episodes.

We've got latest news and we've got

all the latest AI products.

- Everybody, this is the Big Cheese AI podcast.

I'm Andre Herakles, joined by Sean Hyes,

Jacob Wise and Brandon Corbin.

Thank you so much for tuning in with us.

Can't wait to talk to you guys next week

about the biggest news in AI.

Thanks.

[BLANK_AUDIO]

Social

Show Notes

Subscribe

Let's Connect