#25: Stripe Sessions, AI-Based Payments, and a GPT-2 Comeback?

Welcome back to the junction, a podcast by Venn

Technology about AI and automation without the jargon.

Hosted by me, Mel Bell, and my co host Chase Friedman.

We're thought leaders, industry experts, and visionaries,

here to unpack the latest trends in

AI and how we're using it.

Listen for practical advice and a little bit of

banter on how to improve your business and career

by being at the junction of it all.

We're so excited you're here.

Let's jump in.

Woo.

We made it back from San Francisco, flew across

the country to go see a conference about payments. Yeah.

So for those of you who don't know, Chase

and I are based out of Dallas, Texas.

Dallas Fort Worth, Texas.

Or just DFW like everybody else calls it.

Something like that.

The d up front, behind the FW.

Behind Fort Worth.

It's not FWD.

That's true.

Anyways, we flew from DFW to San Francisco to

attend the stripe sessions conference, which was pretty epic.

It was an all week.

Really was like a couple days.

Yeah, there was energy in the room.

So much energy.

I didn't go last year. It was really cool.

You went last year, right?

Yeah, it was very different.

Somebody told me that the lighting was

like 1.2 million, you know, that I

mean, that club lighting was on point.

It just gave you like a sense of

like, ooh, like, what's going on here? What do you mean?

It wasn't great for my photos,

but otherwise, no, it was.

It really did hats off to

the events team at that point. Oh, for sure.

I don't know if she watches this.

There's this one individual that just,

this was her first event.

Like, she ran the whole thing and she's gonna

be like, why didn't you say my name?

And it's cause I can't remember it.

But the folks that we were talking to,

everybody was nervous because this was her first

event and I think she kicked it out

of the freaking part, killed it, the vibe.

I think they doubled registration attendance as well.

There was a ton of people there, lots of really

cool sponsors like OpenAI, anthropic bunch of AI companies.

But then also like a lot of the people

we were talking to, startups using tons of AI.

Well, if you're in the process of taking money

in the door, which is basically every organization that

I've ever heard of, and you want to know

how to do it better, faster, stronger, all those

things, you probably have a reason to be there.

And I think we saw a ton of people

there because of the capabilities that stripe has just

brought on and delivered year after year.

It's kind of like nobody asks when you go to

the grocery store like, hey, who's your payment processor?

Just give me my groceries and move along.

But do you like me kind of geek out

when you see a stripe browser or something?

You're like, oh, I know that one.

Yeah, well, their UI is super

nice, their APIs are the best.

It's as a techie, right.

It's like real easy to get excited about

it because you know, whatever you build is

going to work for the most part.

I thought it was interesting hearing their, the co

founders speak on what they saw as the gap

in the market a long time ago when they

were a different company, because I didn't realize that.

So they kicked off in 2010.

So they're really not that old of a company.

I mean, 14 years.

I mean, they're pretty young.

They've got like 7000 staff globally, a lot

of really smart people that work there.

But product innovation has just been monumental, I

think over the time that they've been around.

Well, the whole industry prior to stripe was

built around this idea that you can only

use certain providers, right, and there's only certain

ways you can do this, to be honest.

There's a ton of regulation around all of

that, obviously, because there's a ton of fraud.

So it makes a lot of sense.

And then when you want to combine technology with that,

I mean, historically, the industry's moved so slow that it's

like trying to get the dinosaur to move on so

you can bring in something else, but. Right.

I think it was one of the guest speakers

had said something about how just AI and payments

and how the cloud enabled SaaS companies to really

like open up, like their market, their potential.

Potential.

It just like opened up the borders, if you will.

Now you can just quickly, like, no matter where the user

on the other end is that has the credit card, it's

like in their foreign currency and in their language.

And can you imagine how difficult that would

have been for, um, a business 30, 40

years ago, even 1015 years ago, I guess.

Well, I mean, if you're trying to scale and you've

got 10,000 customers, you need to build $20 a month.

Like you need an arm 30, 40 years ago, you

need an army of accountants that could send that out.

Right.

Um, and then prior to that, prior to

stripe coming on, you could build it out

internally, but that takes an army of programmers.

Right.

And so stripe came along, built out this platform where

you can do that and I think that's why we've

seen, and it's not just stripe, because there's some others,

there's some other competitors that are maybe similar.

Um, but they've enabled this ability to

scale your business and not worry necessarily

about payments coming in the door.

Now you can focus on your product.

That's so true.

I also didn't realize how big

their ecosystem of ISP's is.

So we, you know, we are a partner of Stripe.

We actually did have a booth there, a little kiosk guy.

That was fun.

It was hoppin'it, was, I mean, really, the

whole show was, but they've got over 150

apps in the stripe app marketplace.

Um, and just, that's really obviously been a very smart

part of their strategy is you mentioned the APIs, you

having worked with APIs, saying APIs are amazing to me.

I don't get in and work with APIs on

a daily basis over, but like, you know, I

trust when you say that you love working with

them and they've made it so friendly for developers.

And even then, when you think of the

companies that have essentially embedded the stripe payment

platform into their offering, that is Uber, Instacart,

those American Express, I mean, there's so many

like big name companies.

But then I talked to so many small

startup type business owners that are like, yeah,

they just made it really easy for me

to collect payment for my handful of customers.

So it's not this like

prohibitive enterprise only solution.

Well, what is, what is interesting about this,

if you're building out your company, is that

it doesn't cost any money to start, right?

Which is maybe different than something like Salesforce, because you

can't, I think you can start a free trial and

if you're a nonprofit, you can get in for free,

but if you're for profit, you have to pay money,

right, on Salesforce to get on a Salesforce license.

And maybe you can get like a

super cheap license, but there's not like

a free forever HubSpot actually does this.

And maybe it works, maybe it doesn't.

But the difference between a CRM and a payment

processor is I'm going to be solely focused as

a one person startup on building my product and

getting customers and then making sure I get paid.

Because if I don't get paid,

then it doesn't really matter.

And if I don't have a product,

then nobody's gonna pay me anything.

Like then I'll focus on the CRM.

So we saw a ton of founders and

startups and CEO's and c suites of companies

that are like three employees big.

But that's the ideal thing for them because they don't have

to pay any money to get in and they can just

start integrating stripe and start collecting payment, solve all of their

PCI compliance right out of the gate and not have to

worry about, oh, like, I've got a credit card number, I've

got Mel's credit card number on my server, in my, in

my closet, in my college dorm room.

They remove that friction. Yeah.

Also a lot of the features that we were, you

know, that were being announced during some of the various

sessions coming out, of course, AI enabled in many ways,

a lot of automations, but one of the things that

I did not know that has stuck with me is

that 70% of carts or abandoned.

Oh, yeah, high number. Yeah, it's huge.

And then they talked about how they're basically

trying to solve for, by taking all these

data points that they know about buyers and

specific personalization around a buyer.

Like chase in the evenings is more of an idle shopper.

Yeah, but Chase is more, you know,

apt to buy at like 07:30 a.m. Type thing.

And so those types of insights and being able

to like, toggle up or toggle down or like,

basically introduce offers or different ways to pay.

Like, they have so many cool things that are

coming out around that what's like, it's automation, right.

That is sort of doing this a b testing.

I know we're going to talk about that in

just a second, but it's giving you more tools

at your disposal to collect money faster.

And I mean, raise your hand if

you're not interested in that yet.

And everybody's going to be interested in that.

Yeah, this AI based payment method, they said, and I

mean, it's fractional, but they increased conversion from 3% to

7% of the businesses that are using it.

That's four additional, like, transaction

points, percentage points in value.

And that's just by turning it on, I think.

Oh, that's right.

Like, just let it do its thing.

Who doesn't want 4%?

And part of that was the A B testing.

So, like, the payment method.

A b testing, the payment method with

100 plus different payment types or methods.

I didn't realize how much they partnered

and how many different payment types were

actually available across the world.

I learned a lot about payments over the two or three

days that we were there that I, you know, frankly, just

didn't know what a vast and complex landscape it is.

Oh, totally.

They just make it so simple.

Like, they make it seem so simple. Simple.

And that's actually the definition

of, like, that's pretty.

If you can take a really complex

concept and make it seem simple. Yeah.

You know, that's kind of what we

are striving to do here at Ben.

And that's something that Jensen,

the Nvidia guy, the CEO.

The guy, the CEO said is that

simple things are not easy nor simple.

They're extremely complex. But, yeah.

In that demo, they even did like,

a b testing on the fly.

Based on your historical data. Yes. Right.

To tell you that, hey, if you do this,

you might get x percent increase in sales.

I was thinking about the number of abandoned carts.

I basically use the cart as a way to get

the final price, and then I'll circle back around.

So I don't know if they'll solve 70%.

Well, are you saying you don't trust the initial

promo or something that comes across your head?

Well, that, or, like, I might go find

some discount codes or I might be like,

this other thing might be better, you know?

So I'm primarily thinking around, like, computer parts

where, like, I don't know, you could.

There's not like a definitive, like, oh,

yeah, yeti is the cooler brand.

I need this yeti in the computer parts world.

It's like, well, you got Samsung.

You got some new players.

I don't know, things could be better or worse.

But, I mean, I'm not sitting here judging you.

I abandoned my.

Any cart frequently.

A lot of times it is just kind of.

They do kind of make you

click through to see sometimes.

Yeah, to see things.

Other times I'm thrown into my cart and then

I'm like, I'll get back to that later.

And frankly, I don't mind the, like, hey,

you left something in your cart email, but

that gets buried in 300 other promotional emails.

So I think it's a really important step for businesses

to be able to, like, if you're on the stripe

platform, that's one additional thing that you can entrust the

stripe platform to do is to help. Yeah.

Well, if you're in the.

If you're in the business of selling something, which

I think everybody is, at least if you're maybe

selling or giving away or asking for donations, right.

Everybody's trying to get that

checkout rate to go higher.

And if they're using data to tell me that I can, if

I just turn this on, I can get four more percent.

Like, nobody's gonna say no anyway.

I think that's why there's a ton of

energy in that room from last week.

Like a ton of energy.

And all these discussions and what

people are getting excited about.

Well, and the other thing is that maybe some people

don't always think about picking a provider like stripe.

How do I go collect money is the goal.

You kind of mentioned this with PCI,

like keeping somebody's credit card information.

They're doing a lot of

investing in global fraud protection.

That's another area where they're doing like AI powered

assistance to write fraud rules, because there's been an,

they quoted 11% increase in global fraud.

So with more of these tools,

more bad actors, if you will.

So that's another area that we also saw at

the Sage transform conference a couple months ago.

Aaron Harris, the CTO of SAGE Intact, had spent

a lot of time on basically anomalies and error

detection, or like fraudulent anomalies in the books.

Right?

So you managed finance department accountants, and you're

processing, let's say, hundreds of thousands of journal

entries and all sorts of different transactions.

And that, like Sage, AI, is it copilot. Copilot.

They're all called Copilot. All of them.

You just say copilot and, you know,

you're probably talking about the red assistant. Yeah.

So, um, you know, that was another area.

And it's like, this is probably not news to

everyone, but it's just a common theme like these

in the fund flows and the movement of money. If.

What is driving the, which tool do I pick?

I think, um, this bolsters well in the

favor of app publishers like stripe and Sage

intact that are keeping those, like, trends.

And, yeah, we are, we have a lot of

really cool automation that can do all this stuff.

But by the way, there's also been an increase

in global fraud, and here's how we're responding to

that and keeping your business and your customers safe.

Well, there's.

I talked to a decent amount of people

last week, and in my mind, like, we

don't thankfully, have a ton of chargebacks.

And, um.

But a lot of people struggle with this, and you

don't think about it until you get one, right.

And when you get one, you're like, oh, what do I do?

And the real what, most folks, you know,

being in the consulting role, we have lots

of insights into what's going on.

Well, you might get one chargeback a year, whereas

people, some of these companies are getting like 20

chargebacks a day, and some of them are fraudulent.

So you're, you're trying to

prevent that from the beginning.

And that's what stripe focuses on with, with radar, is

that they can prevent the charge from happening before it

goes through because every chargeback you get kind of dings

you on this little score that you have.

And if you have a ton of chargebacks, you're

viewed as a higher risk, which means not only

stripe wants to take your money and so does

PayPal, but Visa basically can blacklist you.

Well, it doesn't matter if you talk to stripe or

patrise or any of these other addien, like, that doesn't

matter because visa isn't going to take your payments.

Well, now you've got a big problem

because now you can't take money, right?

You don't want to be on that list.

You don't want to be on that list.

Like the no fly list, like those crazy

people that end up in viral videos, right?

Yeah, but lots of, from the radar perspective, right.

Lots of AI that'll help you write the rules because

I mean, basically in this product that they have, right.

You have to think like a fraudster, right?

Like this person looks like this and they probably

are using this and da da da da, right.

I think for the most part, a majority of

us don't do a whole lot of fraud.

Defrauding of folks. Right.

And having this AI, basically, it's

somewhat of a chatbot, right.

About it, will write the rules for you.

And then you're like, oh, yeah,

I didn't think about that.

It's just very helpful in writing things

that you are not versed in doing,

which is defrauding people of money. So. Yeah.

Another thing that they announced at the

conference was that they're bringing crypto back.

Oh, yeah.

It's a form of payment.

Do you do crypto?

Yeah, I do some.

I really like the technology.

The NFT hype was way too much, but

I do see value in democratizing currency.

So stripe talks about democratizing payments.

I think there's value in democratizing currency because you

look at the people, like in Venezuela and some

other countries where their dollar is now like 10,000

or more, it's probably, I'm probably way off.

Like your dollar today, tomorrow is now 10,000

or is, is the other way around. Right?

And if you had a, what people call a stable

coin, like one USDC, there's other stable coins out there.

But today it's a dollar, tomorrow it's a

dollar, next year it's a dollar, right.

It's always a dollar.

You don't have the government.

These potentially soon to be defunct governments

like inflating the value of your money

that you worked for in an exchange. Right.

You have something that exists in a

decentralized state where the value can't fluctuate

unless there's really big changes.

Anyway, long story short, some of those have happened,

which is why crypto kind of fell off.

And I think it's given time for some

of these to kind of perfect and get

better just in terms of how they're built.

They did spend a lot of time

on educating the audience on how political

payments have become, citing several examples.

And within that vein of crypto, there was that theme

of this is a way for markets that don't have

as strong of a currency as you're mentioning.

But also, again, I had never considered the impact

that some very, very significant, you know, political events

would have on currency and the ways to pay.

Also, they did mention something.

Have you ever been to Germany? No.

So did you catch when he said, you know, it's

more customary that they want to actually have the thing

in their hand before they trust providing the payment?

Uh uh.

I did not know that he said something.

I hope I'm not misquoting that country.

But there he also cited like, there's also just

like customary norms and like things that you just

might not know or might not consider as a

business trying to transact in other countries.

He cited a few, but I do think it's

a really interesting perspective and one that they can't,

like, not talk about if, you know, that's part

of their strategy around it's increasing the global GDP.

Yeah, totally.

Well, as it relates to crypto, I think it will

take much longer than the crypto fans had hoped for,

but I think eventually were all going to wind up

with wallets that have basically either currencies or cards and

youre going to track your balance in some kind of

decentralized way, assuming were talking about a bank account or

a debit card, not necessarily a credit where you rack

up baby debt. Right.

But yeah, I think thats inevitable.

And so its promising for the crypto world

to see striped accepting crypto payments again.

I mean, I think they make up a

decent portion of the total, like global e

commerce payments or just like electronic payments.

Right.

So if stripe is doing it, you know they're

doing it well, they're going to do it right.

They have too much liability to do it wrong.

So if they're doing it, then

it's inevitable that more folks follow.

I just hope it sticks this

time because they did have it. Yeah.

And then it went away, and now they're bringing it back.

So.

So do you think that there's a world.

You just said we'll have these digital wallets.

Do you think there's a world

in which cash goes away forever?

Oh, man, that's a.

Such a challenging discussion to have.

I I like the idea of cash.

I don't carry cash often because I have so

many different forms of electronic payment or electronic funds.

Right.

But I think until we have a

fully decentralized sort of crypto world where

everybody accepts cryptocurrencies, cash must stick around.

I know.

Not that a lot of people watch our show, but if

there's some people are like, cash has to go away, and

then some people are like, cash is never going to die.

I think it's probably going to be somewhere in

the middle, especially until we have some form of

the true decentralized global currency that has a significant,

like, trillions of dollars in it, and we're not

doing that in the next 50 years.

Yeah, it is interesting.

So I went to an event a couple weeks ago.

It was like a music festival that there's no cash.

Right?

It's all on your wristband.

You load the wristband with a credit card, and you

just swipe it the same way you'd swipe back.

I don't care. I don't know.

Punch in my pin, we're good.

But then in preparing for a trip here in a

few months that we're going to internationally, my husband was

like, oh, we got to make sure we get cash.

I'm like, really? Are you sure?

And we are going to a.

Where there's going to be, like, a lot of

smaller towns, and then we're also going to another

part of the world where I'm like, no, that's

going to be fully, like, credit card everywhere.

But it is interesting because there's all

these, like, different currencies, and you're like,

like, how much do I get?

Do I really need cash? Like, we.

We took out, like, far too much cash, went on

a different trip last year, and it was just like

we were trying to burn through to the local currency

just to, like, get it off our hands.

And so, um, it.

It does surprise me when I do go to places

like that where they are, like, no credit card first,

and you're like, okay, are you going to. I don't know.

It's.

I do see where there is that mentality of,

like, maybe going to a smaller local market.

You might think they don't accept credit cards.

Yeah, but well, like, in terms of

global payment acceptance, Stripe has over 100.

They're coming out with more now.

They're doing crypto.

It's inevitable that you'll find

stripe all over the globe. You already do. Right.

But just more, like, you'll see

the stripe logo more and more.

I used to say this about Salesforce.

It's the biggest billion dollar

company you've never heard of.

Stripe is well on that way. Right.

And I would look forward to them doing an IPo

one day, but actually kind of enjoy that they're private.

Somebody Jensen mentioned on stage, like, it's a good

thing that you're private because shareholders will try to

get you to do anything, you know, to get

the share price to go up, which, you know,

is good for the shareholder but bad for the

people that, you know are customers or employees or.

I don't know. Yeah.

Let's camp on that talk that Jensen

did for stripe sessions for a moment.

So the founder and CEO of. Nvidia. Right? Nvidia.

Nvidia.

It's a.

I've heard it both ways.

No, it's definitely Nvidia.

I said Nvidia.

No, you said it differently. Nvidia.

No, it's Nvidia.

Like you say, the n first.

I really enjoyed his talk. I did too.

He, he was very different than

I thought he was gonna be.

Unassuming, approachable. Oh, totally.

And he talked about how he was a janitor, right.

And I don't know.

You just don't think about that guy at nine years old.

At nine years old being a janitor, that he's going to

run one of the biggest companies in the world one day.

I don't know.

To me, that's like, as he started telling the story, I'm

like, yeah, that's a guy who works really hard, you know,

like, you don't start cleaning bathrooms at nine years old, um,

and then work your way into a busboy.

And then a, you know, he had a great story. I know.

He's probably also.

He's done other things, right? Yeah.

But, man, he was so much more, like, approachable.

Yeah, yeah, I totally got that. Relatable.

For someone who's relatable, literally running

this company for the last.

They've been in business 30 years

that I did not know either.

I remember when I was first getting the

computer game world, building out my computer.

You had to pick, there were some other

graphics cards companies, but Nvidia was coming up

and AMD was kind of around.

Intel had some graphics cards and there were some

other ones that have been bought up or whatnot.

But any of Nvidia was always at like

the forefront of like, had the best technology.

It was like, well, yeah, I want the best because I

want to play my game at, you know, 120 frame rates.

Why not?

Obviously.

And then they got yet done, don't you?

But then they got into the crypto deal, right?

And they figured out that everybody would

need cards for the crypto game.

And now they're in another hot run with the AI stuff.

Okay.

So just total random. They.

That same day that he was on

stage, it was in the morning, right?

In the afternoon, he went and saw Sam

Altman and delivered the first h 200 card.

So all of these large language

models are built on h 100s. Okay.

The h 200 is supposed to be the next big thing.

It's infinitely better.

Da da da da.

And he was hand delivered the first.

I don't even think he teed that up.

No, he didn't say it was

in an Instagram picture or somewhere.

Oh, yeah.

Even that one close to the vest on stage. Yeah.

That's cool.

That was really cool.

And what's hilarious is Sam is just in

a blue shirt, like super, like tech bro.

And Jensen's just hanging out right there.

There's some other gentleman that I didn't know.

And then there's this giant box

that looks like a steel brick.

You're like, what is that thing?

And it reminded me of, I don't think I'm

dating myself here, but it reminded me of like

the big server rooms that you would walk in.

And it's not a server.

Like, that's a single computer.

Anyway, it was pretty cool to like, just

think through, well, he's on stage here and

then he's talking to Sam Hallman.

Oh, and by the way, he just delivered the first

thing that's going to propel OpenAI into the next, like

into the next, I don't know, generation of whatever.

So what was the most, your

biggest takeaway from Jensen's talk?

Yeah.

One, it's really two things.

I'll do two quick ones.

The first one is he talked about torturing people

into greatness and how many direct reports he has.

He said he had 60 direct

reports and my mind just exploded.

He said he doesn't do one on ones.

And he gives feedback direct to the

employee, like his C suite person.

Like in front of everybody else. Right.

And then you think about the word torturing. Right.

Obviously it doesn't do anything, you know, weird.

But like, if I'm gonna be berating you

in front of all of your peers, like,

that's gotta be super high pressure anyway.

But I really appreciated that, just having

insight into how he runs things.

And then the other one was he basically outlined

what he thinks the AGI world is going to

look like, and it's effectively a culmination of these

different agents or these different repositories of knowledge that

are combined by some kind of conductor or some

kind of traffic cop.

And that's what we've been reading in, and that's what

I've been playing with and seems to work well.

So if we're reading about it and we're testing it out and

it seems to work, and then now he's also saying it.

It was a really good kind of confirmation because

I think I've said this on earlier shows, like,

nobody knows, nobody's like, the world ten year veteran

on, like, where we're at right now, because it's

all just coming out on the fly.

So everybody's learning at the same pace, or maybe

not, but everybody has the same opportunity, right?

So those two things stuck out for me.

What about you on your first one?

I will say, as he was talking through that,

and you mentioned giving, you actually said berating.

So berating in front of a bunch of people

definitely reminds me of my days playing soccer.

Um, for those who don't know, I played soccer

in college, and the feedback was certainly not delivered

one on one, you knew if generally you didn't

really need feedback, you knew that you screwed up,

um, right on the spot.

But, you know, we got feedback in

front of our teammates all the time.

And it taught you one lesson.

I don't want to have to learn that lesson that way.

So it taught.

His thing was, why tell someone if I have

60 direct reports and I tell one person, those

59 people missed out on an opportunity to learn.

So he sees that as knowledge is power.

But he also specifically said that around the

one on ones, that he doesn't have one

on ones is that he doesn't view any

information for his 60 direct reports as privileged.

So similar to the giving feedback to one person in

front of the whole team, he believes that by delivering

the information he has to all 60 direct reports, that

is empowering them in a way that wouldn't otherwise empower

them if he was just delivering it kind of one

off or in his one on one.

So I thought that that was it.

All of that resonated with me.

I definitely think that, you know, people

like that push us, push people to

their potential and that's needed at times.

So really enjoyed that part.

I also enjoyed in more in relation to their

product and how, how they've gone about building things

and launching all of the things that they do.

He used the word love, and he says

he uses that frequently, like, love and beauty

and like, there can be beauty.

You can, you can care for and love like a product.

A person.

Like, he was really, he kind of

camped on that for a minute.

And the co founder that was interviewing him of

stripe, they clearly were very aligned in that.

And he, he, too, I think that was the most that.

So this was like Q and a style for anyone that, you

know, we didn't really tee up how the, how the talk went.

It wasn't Jensen just kind of delivering a keynote.

It was a stripe co founder asking him questions.

And the stripe co founder actually just fully launched into

his own kind of perspective on what we do, the

products that we build, they, we take pride and in

the beauty and the craft of what we're doing.

And I think that that's maybe a bit taboo.

Jensen sort of implied that it was a bit taboo

to use those words in, you know, their environment.

So I thought that was kind of cool. Yeah.

Such a, what's it called? What's like a.

You've got on a spectrum of one end

and then, like, the total far extreme.

He's talking about torturing and loving,

like, in the same hour.

I thought that his talk was really good, combining

that with like, he's the CEO of a major

global company and he's in here just chilling, you

know, like, I was a janitor, you know, torture

people into greatness and love your products.

I don't know, maybe I just had different expectations.

It's that whole thing about there's like, you

don't appreciate the view at the top of

the mountain if you did not have pushed

through some pain while you were climbing. Right.

You're like, it truly is, like.

And then there really is. I don't.

He didn't actually say this, I'm putting

words in his mouth, but he implied

that there really is no top, right?

Like, but there's always going to be the next thing.

The next thing.

And so you appreciate it more when

you've gone through challenge to get there.

Well, you have that experience. You get the lessons.

And I think that's why

he talked about torturing people.

And one of the things he mentioned that he hung on and

camped out on for just a second was I, and this was

one of the questions was um, what did you mean by, uh,

I hope that you have a lot of pain, something really close

to that, like, and have a lot of painful experiences.

I've realized this and all the time that we've

got gotten you and I to work together.

And just in general, like, if

everything's hunky dory, like, you don't,

aren't really learning anything, right?

Like, if you're, if you go twelve and o, like

you did everything perfectly, there's nothing to learn from.

And then when you meet that challenge next,

you're not gonna know what to do.

You know, what the lead in was.

I wish you success, but what that really means

is I wish you suffering, pain and suffering, because

that's what gets you to the success.

He had a lot of little nuggets of

wisdom that I think anyone could hear.

And if you were a startup or a founder or business

in that audience, you were probably like, oh, my dad.

My dad used to say, if you do things right

the first time, you don't have to do them again.

And I always thought, well, I'm just going

to do everything right the first time.

What I've realized is that he should have added on

another phrase to that, that if you do things right

the first time, you don't have to do it again.

But if you don't do it right the first time,

you're going to get it right the next time.

Like the next time, the next job, the next

business right, you're going to get it right the

first time then, because you have the experience of

doing it wrong the last time. Yeah.

Um, and every time I, every time I make

a mistake, I'm like, not gonna do that again.

Cause nobody wants to do things

four or five times, right? So.

Okay, so switching gears a little bit off of stripe

sessions, you wanted to talk about chat GBT two. Two.

Why?

Hot take, um, or something. I don't know.

This isn't, I feel like this just came out.

I'm sure it's been available.

Um, anyway, there's a leader board out there

where they measure, like, all of these different

large language models against each other.

Claude's on there.

You got Opus, you've got chat GPT 3.5, you've got four.

There's a lot of mega,

like, Facebook's open source models.

And basically you can set it up where you can type in

the same prompt and have pick model a and pick model b,

and it'll send the same request and you can see it type

out the responses and you can just compare them.

But the leaderboard has become, like, the de facto.

Like, here's.

Are you voting on your preferred

response to create the leaderboard component?

In a sense, I forget exactly how it works.

But that's the idea is like,

this is just the overall general.

It's like up votes and downvotes.

Right kind of idea.

And anybody can vote.

It doesn't really matter.

So it is susceptible to like, well, I'm going to hire

a bot army to come in and click the up vote.

But it's become the de facto.

Like, everybody looks at that to

see which one is the best.

Right now, there's some other, like,

statistical analysis that people are doing.

But if you want a quick look just to see

what people think, like, this is where you go.

Anyway, this was today.

So today's Monday, the 2020 3rd.

29th.

Wow, 29th. Yeah. Catch you up.

Yeah, I'm just way behind.

Yeah, the 29th.

So four chan had a post, and honestly,

this, I didn't check to see the post

timing, but this showed up on Twitter.

This guy posted a picture of the deal.

Anyway, there's a, there's a model in

the arena now called chat chatbot.

Chat GPT-2 um, something about chatbot.

Anyway, by the time you hear this, it's

going to be a week old, right, maybe.

Um, but what the rumor is is

that this is actually chat GPT 4.5.

Um, and they're testing it with a different name.

So people don't just go like, interface.

No, the like or the chatbot chat GPT-2 thing.

The chat two, like incognito.

Yeah, is incognito 4.5.

And I mean, you can quickly just kind of figure

out that it has, is or isn't by asking it,

like, kind of questions around who are you?

What model are you based on?

And it consistently references OpenAI

and your large language model.

Unless you have the weights behind OpenAI.

It's not gonna say that it's OpenAI, but

it doesn't necessarily say what model it is.

Anyway.

So there's a lot of, like, on the fly testing

that our people are doing to try to figure out

and suss out what so interesting, what is happening.

And if you test it yourself, you'll see

that it is a little bit smarter.

It answers questions and riddles in ways that none

of the large language models do right now.

Like a lot of, like, if you go search Google for

the hardest riddle to solve that we have an answer for,

and you type that in into opus or into chat GPT

four, like it might get there, it might not.

It might give you just straight up the wrong answer.

But if you go type that into this

chat, GPT-2 it gives you the right answer,

which is like, well, that's interesting.

You know what this reminds me of?

Have you ever heard of, like, crime sleuths?

Like true crime?

They go on, like, tv show?

No, like people who get really obsessed

with these unsolved cases and then they're,

oh, yeah, there's a subreddit on that.

I'm sure there is. Aren't you a part of it?

I'm not one of.

I'm not one of them.

I somewhat envy it a little bit.

It does appeal to, like, my interests.

I have not carved out any time in

my day, just dedicate anything to it.

Maybe when I'm retired, totally.

But by then, AI bots are

gonna be solving them all anyway.

So it's like, whatever.

But it kind of reminds me of that,

where they're, like, all piecing together all these

clues to try to suss out the perpetrator.

Well, the rumor is that this is, they're trying to,

like, test it against other models and do, like, a

soft drop to get it rated, but not have the

masses come in and just blow it up. Because it's not.

It's just supposed to be like,

let's get, like, a quick take. Low key.

Yeah, like, chill.

Obviously it's publicly enabled. Right.

But this is a way to, like,

get true sentiment versus, like, the market

blowing, like, the news stories going crazy. Da da da.

Right? Interesting.

Okay, so you don't know when this is gonna

be, like, unveiled, but this is, like, the theory.

A theory.

Assuming it's true.

Let's take this assumption here that it's true.

Saw a couple of ideas that they're gonna drop it

during the Google I O conference, which is coming up.

I also saw some thoughts that this is actually

Facebook's next model that they haven't dropped yet.

I think that because it says that it's OpenAI,

it wouldn't be really challenging to do that if

you wanted to hack it, make it up.

But generally, we haven't seen it.

Ton of that, at least. I don't know.

It's certainly possible, but so what's the

so what in all of this?

The so what is. Because it's.

It's answering riddles, like, with the answer

rather than saying, I don't know, or

coming up with a hallucination.

Oh, what did they like?

If you say, like, how many?

Which would you, who would win? In a battle, right?

1000 duck size horses or one duck sized horse or.

Yeah.

Anyway, those, like, weird, like pick this.

It's totally subjective and most of the

models are like, ah, it depends.

And here's why.

This one's more definitive and this one's like,

yeah, it'd be the thousand duck sized horses.

Is it also teaching us or a case

study and like how they might be bringing

products to market differently in AI or what?

Well, that, I don't know.

You're the marketing queen.

Maybe it's just total nonsense and somebody has just found

a way to make a better large language model.

I don't know.

I like the sleuth idea.

There's totally people hammering it right now

trying to figure out all this stuff.

But if it was worse, it wouldn't be a thing, right?

But in somehow somebody has a model

that's better than what's currently publicly available.

So the sleuths are like freaking out.

Do you identify as an AI sleuth?

No, I don't have enough time for that.

Like, I don't have enough time to go pick parts.

I would've thought you would've already built like a whole

slew of bots to be sussing it out for you.

That's a great idea.

Except a lot of reasons why people don't do

that is because they generally, it costs money, right?

To ping the API, you have to

pay money, and usually you can.

Well, if you did that, like if you build out

a thousand, anyway, it costs a lot of money.

The other thing that was interesting is

that OpenAI dropped a product update that

they haven't really talked about.

It's with assistance, and now you can upload, I

think it's like 10,000 files into a single assistant,

whereas two weeks ago it was 20.

And they also enabled something they're

called, it's like JSOn format.

Like people are trying to shove this data in the

large language model and say, send me back this structured

data, and it's kind of okay at that.

You can give it even more pre

prompting to try to get it close. Maybe.

We talked about this on a show.

I've been talking about this amongst my colleagues, but

the best that it's ever been is 85%.

85% of the time it would send you

the right structured data in this particular format.

Well, that doesn't go into production

like 15% of the time.

It errors out or it's wrong.

This is never going to work.

But now they've dropped this JSON format where it

sends back the correct format every single time.

And that's potentially a game changer.

So if I'm going to be building 1000 bots

to do anything, I'll be playing with that.

But I thought in terms of marketing, I thought

it was interesting that they haven't really mentioned that.

Um, so I don't know if that's just like they don't

think it's that big of a deal or if they're trying

to not like get it to be like overloaded again when

it was like three messages per 4 hours or whatever.

Slow playing it to make sure

that they can handle the volume.

It's honestly, it's probably more like that.

I do remember that.

It seems like so long ago now though.

It's like takes forever to be able

to put in your next chat.

But think about how, I mean, it's

changed so fast in such little time.

Yeah, that's why that h 200 thing is a really big deal.

Interesting, but yeah, I've babbled a lot.

Anyway, I'm curious if anybody out there is listening to

this is one of those sleuths come onto the show.

Let's talk.

We actually did get, had several conversations with

folks at stripe sessions last week that I

am going to get on the show. Oh, nice.

Really interested in the similar

things that we're talking about.

And was it Jensen?

No, I'll work on that next time. It works.

I did start following him on LinkedIn. Oh my gosh.

Didn't, didn't go full, full bore.

Asked for ask for the connection

request or anything, but did she?

This total tan took 10 seconds.

Only ten second tangent.

There's a girl out there that just like started her

podcast and then got like p Diddy or somebody.

And her thing was like, I'll give you $250 if you can.

If you can get p Diddy on my show. Yeah.

And then I'm sure I don't remember who the rapper

was, but p Diddy's like, great, give me $250.

And then he was like, are you trying

to connect this to like getting Jensen on

the show by offering up like somebody?

Yeah, let's do it. Let's do it. Jensen.

If you want on the show, I'll give you.

Or if anybody can get Jensen on

the show, I'll give you dollar 250. All right.

Heard it here first. There we go.

Well, that was really fun.

Next week we will not be talking about stripe sessions.

We're going to be unpacking a whole bunch

of other AI and automation related topics.

But more importantly, if you have a question

or a topic that you would like us

to cover on the junction, please email us@thejunctiontechnology.com.

Follow us on any one of our social media channels.

We're on LinkedIn, Instagram, x, although I don't

think that's as much of a thing.

I dropped the X thing.

That's Twitter.

Well, you know where to find us.

We love to hear from you.

Good, bad, ugly.

We'll take all the hot takes and

we'll cycle them through the show.

And also, if you would like to be

a guest on the show, reach out.

We are always looking for new and exciting viewpoints.

So same song and dance, the junction@ventechnology.com.

Sponsors, too, right?

Sponsors.

Keep it automated.