How NVIDIA constructs Payments Foundation Model for PayPal

In theAgency CompanyIn the fifth issue, Simon Taylor (Head of Tempo Market Development) and Bam Azizi (Mesh CEO and founder) invited Pahal Patangia (Head of Global Industry Development and Payments) to discuss the following topics: open-source model in financial services, Agentized workflow as intellectual property in business。

Time axis:

00:00 Introduction
05:03 Based onTransformer Structurepayment base model
10:44 Adoption of open source models for financial services
17:53 COST AND DELAY BALANCE IN AI REASONING
20:24 ON AIToken Economyand efficiency
23:21 Agentization Workstream as Intellectual Property in Business
25:45 Trends in protocol integration in Agency Corporation
OpenSHIELD on open source running for Agent's safety
33:33 Stable currency advantage in Agent-to-Agent micropayments
35:36 Search for actual landing in Agent faster than payment

Takeaways:

1. The essence of Agentic Commerce is “outsourcing in context”: the context of past controlled consumption decisions is being transferred to Agent, and capacity to pay is no longer an executive but part of the decision-making chain。
2. The Payments Foundation Model is the core variable: input traditional tabular financial data into Transformer, generating user behaviour embeding, which is the key infrastructure that Agent can “consumer like human beings”。
3. The search has matured and payments are still in their early stages: the real landing of the current Agency Company is concentrated on the search and referral links, which are still in the Sandbox and experimental stages。
4. The underlying reason for the explosion in the financial sector is not technology, but regulation and control: interpretability, controlability, fine-tuning, and more important than performance。
5. The performance gap between open-source and closed-source models has narrowed to “negligible areas”, making cost, compliance, and deployment flexibility the dominant factors in enterprise decision-making。
6. Token economics became the new generation of “paying economics”: the core constraint of AI is no longer just the cost of handling, but the combined optimization of token consumption, reasoning costs, delays, energy consumption。
7. Many Agent systems are the main battlefields of the future: assuer, acquirer, merchant, enterprise internal systems will all evolve into Agent, and commercial processes will be completed through machine-to-machine interaction。
8. Agent workflow is becoming a new core enterprise asset: formerly API and SaaS, now Agent's decision path, implementation logic and feedback cycle constitute a new “business IP”。
9. The stabilization currency has a structural advantage in the Agent-to-Agent scenario: micropayments, real-time settlements, global availability, which the traditional card network cannot support。
10. Agent's trade growth is exponential: The traditional payment system TPS model does not carry this paradigm shift, with two transactions per day for humans, possibly 2,000 transactions。
11. The payment track will not be replaced, but will coexist in layers: Carnets apply to human interactions, and stabilization coins apply more to machine interactions, and the two will run in parallel in different scenarios。
12. THE LLM LAYER IS CURRENTLY IN THE “EARLY STAGES” OF LLM: MULTI-NEGOTIATION PROMOTES INNOVATION AND IS BOUND TO CONTRACT TO A FEW STANDARDS IN THE LONG RUN。
13. Security has become the infrastructure issue of the Age of Agent: a runtime such as OpenSHILD is needed to keep Agent in isolation from the Sandbox and prevent the spread of systemic risk。
14. The core use of AI in the area of payments has not changed: anti-fraud, authentication, individualization remains the most central value, but only in a way that has evolved from the rule-based model of Agent。
15. The real breakthrough of Agenic Commerce is not to pay, but to “automate decision-making”: when search + recommendation + implementation is fully automated, payment is only the last step in mobilizing capacity。

Simon Taylor:
Welcome to Tokenized, a program that focuses on stabilizing coins and real-world asset monetization agencies. I'm Simon Taylor, today's host, and I'm the author of Fintech Brain Food and the director of Tempo's market development。

Today we're going on with the Agenic Commons series, along with Mesh's CEO, Bam Azizi. How are you, Bam

Bam Azizi:
I'm fine, thank you Simon for inviting us again。

Simon Taylor:
This series is really taking off now. I think that Agenic Commerce has become one of the most popular topics in the world, and it really attracts everyone's attention。

Today we also have a guest from a company that is equally very interesting — in fact, one of the largest companies in the world — who have done something that most people do not realize in their support for Actic Company。

So today we invite NVIDIA's Global Business Development and Payments Manager Pahal Patangia. Pahal, how are you

Pahal Patangia:
I'm fine, Simon, thanks for the invitation. It's good to be here, and we look forward to this dialogue with all three of us。

Simon Taylor:
INDEED, ALL THIS IS GATHERED TOGETHER. IT'S ALL I LIKE: PAYMENTS, NVIDIA ACCUMULATIONS IN VIDEO GAMES, BUSINESS, STABLE COINS... ALL THESE GOOD THINGS。

Before we begin, however, I would like to remind the audience and the audience that the views of the guests of this programme represent only individuals and not necessarily those of their companies. At the same time, nothing we say constitutes tax, legal or financial advice, and you are invited to do your own research。

Okay, from a macro perspective, what exactly does it mean for a company like NVIDIA? Why is a GPU company, an accelerator company, an AI company, a hardware company involved in payments and business

Pahal Patangia:
Of course, Simon, that's a good question. And I'm glad you asked from the perspective of GPU, Hardware, Accelerating Calculator, because it's really the perception of NVIDIA over the last few decades。

But what I'm trying to say is that this perception has actually been evolving over the past 20 years。

OVER THE PAST FEW DECADES, NVIDIA HAS BEEN TRANSFORMED INTO A FULL-SCALE ACCELERATOR PLATFORM, AND THE COMPANY IS PROVIDING CAPACITY FOR AI APPLICATIONS THROUGHOUT THE ECOSYSTEM。

Before we go into Agentic Commerce or AI, it's important to understand the location of NVIDIA at the platform level and the capabilities that we provide -- which are actually driving the AI explosion you see every day。

We usually use one."Five-storey cakeTHE CONCEPT DESCRIBES THE ABILITY OF NVIDIA TO PROVIDE ECOSYSTEM-BUILDING AI APPLICATIONS。

THIS "FIVE LAYERS OF CAKE" IS MADE UP OF DIFFERENT "CONTENTS" THAT MAKE IT POSSIBLE TODAY TO BUILD AN AI APPLICATION AND BUILD AN AI PLANT IN A SCALABLE MANNER。

THE BOTTOM IS LAND, ELECTRICITY AND ENERGY -- THIS IS THE BASIS FOR DOING ANYTHING ABOUT AI。

THIS IS ABOVE THE CHIP LAYER, INCLUDING HARDWARE, GPU, CPU AND ASSOCIATED NETWORK SYSTEMS。

And up there is the system layer, the data centre layer, how these chips are organized, and we see them as different units that eventually combine into a “big computer”。

Because in the past we understood computers as personal equipment, but now the data centres are a computer in themselves, and that's the system layer。

And above that is the base model layer. These basic models include knowledge, industry understanding and capabilities. Many partners in ecosystems, such as OpenAI, Meta, Mistral, are building these basic models。

However, these underlying models need to be further broken down into specific industries, specific scenarios, specific issues, and this is the fifth — application level。

THE NVIDIA PLATFORM RUNS THROUGH THE FIVE FLOORS AND COMBINES THIS SET OF CAPABILITIES. DEVELOPERS CAN USE THESE FIVE FLOORS TO BUILD THEIR OWN APPLICATIONS FOR EXAMPLE。

In the area of payments, a key application is Agentic Commerce。

Our goal is to embed our hardware, software and modelling capabilities into these eco-playing houses so that they can build these applications on a large scale. That is where we are and the way we are driving the whole of ecological development。

Simon Taylor:
And what's interesting to me is that when we talk to a lot of people about Agenic Commerce, we all agree that there's a lot of software behind it, that there's a lot of hardware running it, but you've been in this business for a long time, and you really understand how these bottoms work. What do you think

Bam Azizi:
Yeah, it's funny, actually. I had a post on LinkedIn before. It was pretty hot。

It's like Pahal just said. I was talking about the basics, distributions, layouts and connections. My point at the time was that the connecting layer was the most important -- a little personal, of course, because Mesh was on that level。

BUT I'M ACTUALLY CURIOUS, FROM THE POINT OF VIEW OF NVIDIA, WHICH FLOORS DO YOU THINK ARE THE MOST IMPORTANT? WHICH FLOOR DO YOU INVEST THE MAXIMUM AMOUNT OF TIME AND MONEY

Pahal Patangia:
Yeah, that's good. From our point of view, I think two very critical phenomena are taking place in the payment sector。

WE'RE INTRODUCING AI INTO THE PAYMENT INDUSTRY ON A MASSIVE SCALE, AND USUALLY ONE PHENOMENON LEADS TO ANOTHER。

The first phenomenon is the emergence of the “payment base model”。

If you look at the whole process of Agenic Commerce, you'll find that it's actually been compressed. For example, the billing process has actually been compressed。

In the past, as a human being, you have mastered the context. You know what you're going to buy, you know how you're going to finish the check, and the context is in your head。

But now the question is: Where does Agent get these contextes

Agent must access these contexts by learning about the behavior of the user, the image of the user, the preferences of the user, and the restrictions you set for the transaction (e.g. all rules from SKU to the final transaction)。

So, Agent, how do you get these powers

This leads to a new trend, which I would say is somewhat “underground”, but is rapidly gaining attention — the “payment base model”。

Because in the financial services sector, especially in the payment and banking sectors, most of the data historically existed in the form of structured tables。

What you used to do is feed these data to machines to learn algorithms, and then build trend models, such as predicting what users might buy and what transactions they might make。

But with the emergence of a new generation of algorithms, especially Transformer structures -- It is the basis for generating AI - there is a new trend to expose these structured data to Transformer models。

This is the concept of the “payment base model”。

and these models create something called "embeding."。

in short, embedding is a semantic expression of user behaviour. for example:

• Pahal, what does this man do

• What's his latest dynamic preference

• What's his long-term behavior pattern

Transformer models can integrate this information into this embedding。

And then these embeddings will be entered into Agent, which will then execute the action on the basis of this information, such as completing the transaction。

THAT'S WHERE THE TWO WORLDS STARTED TO MERGE -- AI AND PAY。

These embeddings become the “texture layer” of Agent, enabling it to be better implemented, more iterative, and to ensure that all actions are within the established rules while learning and optimizing。

This is an important trend that is currently driving the development of the Agency。

In addition, I would like to emphasize in particular that another trend we have seen in Agentic Corporation is:

If you break the whole process into "search" and "pay,"

Now the fastest-growing and most mature part of the story is the search component。

The search has been under study for many years, and now there are better algorithms to solve it, so the wave is very effective in “search”。

That is why the user experience is now more personal and sticky。

We're also doing a lot with PayPal. PayPal wants to bring the capacity of Agenic Commerce to their commercial ecology, about 19 million。

MOST OF THESE BUSINESSES ARE SMALL AND MEDIUM-SIZED, AND THEY'RE ACTUALLY MORE "BLACK BOX"-- THEY DON'T UNDERSTAND WHAT'S GOING ON。

PayPal ' s approach is to provide these capabilities to these businesses through the platform。

Their way is:
The open source models are fine-tuned to fit the PayPal environment and specific examples。

It is natural for businesses to use these capabilities without having to understand the bottom techniques themselves。

Simon Taylor:
I have just heard a lot about you, and I would like to try to repeat it to see if I understand correctly, but also to make it easier for the audience to understand。

A lot of people would ignore the fact that apart from the models of Anthropic, ChatGPT, Gemini, there are many open-source models, and NVIDIA is an important participant in this。

Like yoursNeMoAnd Netron, these models, they've been leading in performance。

And business customers like PayPal will bring these skills to the business。

In the payment industry, creating value for the business is everything. Businesses are the core of the world's operations. If you can't serve the merchants, you're nothing。

They sell goods, they're your clients, they pay you. So you have to create value for them。

Stripe had also previously issued a payment base model, which had had a good impact on anti-fraud。

But I'm curious, beyond anti-fraud, what else can the payment base model do

if i have a very rich, multi-dimensional embedding that can understand the preferences of users, how can those abilities help businesses to sell more and serve their customers

AND IT'S LIKELY THAT THE MERCHANTS ARE RELUCTANT TO GIVE THE DATA TO THE LARGE AI LAB。

So they tend to use open source models。

And now the gap between open-source models and front-line models is about six months, and it's a performance gap。

For most daily uses, differences are virtually invisible。

For many small and medium-sized businesses, these models are far better than the free version of ChatGPT they now use。

So PayPal can give them a very good experience, and the bottom is actually NVIDIA's ability。

I don't think many people actually realize that。

AND I'VE SEEN A STUDY THAT SHOWS THAT 65% OF FINANCIAL INSTITUTIONS ARE ALREADY USING AI, AND 84% OF THEM SAY OPEN SOURCE MODELS ARE IMPORTANT FOR THEIR AI STRATEGY。

So I'd like to ask you: Why is open source model so important in finance

Pahal Patangia:
Yeah, that's a good question。

The financial sector has been “slow” in adopting new technologies。

The reason for this "slow shot" is:
Regulation
Explanatory requirements
And no trust in the Black Box Model

Financial institutions would like to be able to understand what is happening inside the model, so that it can be used in the production environment。

So they prefer models that can be controlled and fine-tuned。

At the same time, as you just said, the performance of open source models is now very close to the large closed source models。

This “performance approach” shift the focus of the discussion from “model performance” to other dimensions, such as:

• Cost

• Control

• Compliance

• System resilience

Businesses wish to have more choice in constructing these applications than relying on a single supplier。

Of course, we also see the providers of basic models as important clients and partners。

At the same time, however, open source models would be more appropriate when enterprises needed more flexibility。

NVIDIA's Netron model, for example, and the NeMo tool chain can help enterprises to fine-tune models more easily。

And it's going to get more and more important in Agenic Corporation。

Simon Taylor:
This is an interesting trade-off。

Bam, I'd like to ask you, how do you see open and closed sources from the perspective of building a company in the area of stable currency and payments? Does your client care about this

Bam Azizi:
I don't think they really care about open or closed sources from the client's point of view。

This is a matter of concern to the technology community and is important for scientific and technological development。

But the client only cares about one thing:
There's no best way to help them run the business。

However, open sources are very important for the industry, and we need to push them as far as possible。

And the other thing that struck me was that Pahal was just talking about NVIDIA's position。

In the past, NVIDIA was more like a layer of hardware, and then there was a layer in the middle, like ChatGPT, Cloudmaker, and so on。

But now that you're working directly with companies like PayPal, does that mean you're skipping the middle

Does that mean faster, cheaper, more efficient

Is that a threat to companies like OpenAI

Pahal Patangia:
Not at all。

Our idea is to “support developers where they are”。

If developers want to use our large partners, such as the providers of basic models, we fully support them and will help them achieve the best results。

If they want to use open source models, we also provide tools and platform support。

This depends more on business needs and decision-making within the enterprise。

We provide a complete platform where they can freely choose。

Simon Taylor:
I think this is really interesting。

Pahal, how do you direct payment companies like PayPal to make these decisions? For example, if they were to provide those capabilities to businesses, how would you help them weigh the different examples? What's your feedback from these paying companies

Pahal Patangia:
That's a good question。

And in this field, when you start running increasingly sophisticated models, from today's models to the future of Agent, to more Agent systems, there are many factors to consider。

First, of course, is accuracy. But when you optimize accuracy to a certain degree, it is the other factors that really determine the outcome。

The first is cost。

For example, you're going to serve 19 million businesses, and that's going to generate a lot of reasoning every day. You have to think about how to optimize the cost of these reasoning calls in your case。

The second is delay。

Nobody wants to wait there, just like the little snake game in the browser after the network is broken。

What you need is a millisecond response。

Models need to be thought, reason, access to information from different data sources, context and decision-making within established rules in milliseconds。

all of this requires a great deal of token, a great deal of decision-making, complex processes, and all of this must be dynamic and intelligent。

This can be done if Agent is properly fine-tuned and operates under the right constraints。

You do it once, and there's a feedback loop。

This feedback cycle creates a Data Flying Wheel:
You will continue to obtain new data, compare the “real results” with the “ideal results”, and then optimize the models。

Simon Taylor:
Yeah, and then when you extend this logic from individual Agent to more Agent systems, things get more complicated。

For example:

• Internet side of Agent

• Agent on the side of the card

• Agent on the side of the line

These Agents will communicate with each other。

Or within the enterprise:

• A procurement agent in SAP

• It needs to talk to the inventory system

• And the financial system

How does the whole system work? How can we be more efficient

and that leads to the question: token will explode。

that's why "token economics" becomes very important。

not just token use is reduced, but how to achieve optimum efficiency between costs, calculations and delays。

It's even understandable:
"how many high quality token output per kilowatt hour can be produced."。

It's actually an economic model。

If you don't control it, it's easy to burn a lot of money。

Anyone who's played OpenClaw knows that it's easy to spend $1,000 a month, just a few APIs, and then fall into all kinds of rabit hole。

The problem is even more serious for enterprises。

You used to just run some machine learning models, like that on Snowflake, CNN, and so on, but now these AI models have completely different cost structures。

This difference in cost is significant for an enterprise that is committed to user loyalty or anti-fraud。

And of the different roles of card organization, merchants, issuers, each has different requirements for Agent and different needs for token。

So the complexity of the system is very high。

It's not just about controlling costs, it's about getting the system better over time, learning as people do:
“You just made a mistake and don't do it again”。

But if you did use OpenClaw, you'd know that it's really hard to keep the system stable。

SO IT'S VERY VALUABLE THAT NVIDIA CONTINUES TO SOLVE THIS PROBLEM AT THE ENTERPRISE LEVEL。

Simon Taylor:
We pulled the subject back to the electrician。

What is the impact of Agenic Commerce on commerce

Can users really feel these changes when they close the books? Where do these values come from

Pahal Patangia:
Our goal is to support players that really create value for end-users, like PayPal。

At the same time, they will work with large retailers to deploy a consumer-oriented Agent on top of them。

From an industry-wide perspective, some of the trends observed include:

For example, MasterCard has already made fully Agent-driven transactions in some countries。

These are early signs of success。

This leads us to believe that these technologies will eventually become mainstream。

Of course, there are many issues that need to be addressed, such as:
Do these Agents really improve the billing rate
Is it stable enough

At present, more fine-tuning and restraint mechanisms are needed to enable Agent to carry out its mandate with real autonomy。

Simon Taylor:
I would like to mention in particular Sardine, who have done a lot in the area of anti-fraud。

They have a 7 billion-dollar data network, built their own models and documented the implementation effects of Agent。

These historical data and Agent workflows are in themselves an intellectual property right。

These capabilities were provided through SaaS or API in the past and are now Agent workflows。

In the electrician, your Agent workflow is your core IP。

I think it's a very critical point。

Simon Taylor:
Okay, thank Mesh and all the sponsors for making this show happen。

Bam, I don't know if you're like me, but now I hear a lot of different protocol names, and I can't remember。

HOW DO YOU DISCUSS THESE AGREEMENTS WITH CLIENTS NOW? YOU'LL ASK NVIDIA WHAT'S THE PROBLEM

Bam Azizi:
I think the central question now is: will the future move towards integration or will it continue to be fragmented

This is a “billion dollar level” issue. If someone could answer that question, a huge company could be established in that area。

If you ask me, I'd prefer integration, like the development of the Internet。

THERE WERE A LOT OF DIFFERENT AGREEMENTS IN THE PAST, BUT EVENTUALLY WE ALL UNIFIED TO HTTP。

There are also many protocols for communication between equipment, but eventually it is basically harmonized to Wi-Fi and Bluetooth。

Even on charging interfaces, one or two are eventually integrated from different interfaces。

So I think something similar would happen here。

In particular, recent developments such as their promotion into the Linux Foundation, hosted by a neutral organization and supported by companies such as Stripe, Coinbase, etc。

I'm self-identified and secure, and we've seen similar integration processes in authentication agreements。

So my judgment is that it will be integrated。

But I'm also curious about Pahal's opinion。

Another question is:
Will there be a different deal in the future

For example:

• The human-Agent interface

• Interaction between Agent and Agent

THE TWO SCENARIOS UI/UX, THE AGREEMENT MAY BE COMPLETELY DIFFERENT。

What do you think is going on in the market

Simon Taylor:
I THINK OF A CLASSIC XKCD COMIC BOOK:

"There are now 14 certification standards, and we need a uniform standard."
And then it went, "Now there are 15 standards."

What do you think of this

Pahal Patangia:
Yeah, if I had a crystal ball, I'd like to know the answer。

But from our point of view, I agree with Bam:

In the end, these agreements will draw on a few mainstream programmes。

But in this process, diversity is now a good thing。

Because the protocols are activating more developers and more people are starting to build。

THE CURRENT STAGE IS THE “DEMOCRATIZATION PHASE”, AS HAS BEEN THE DEVELOPMENT OF LLM OVER THE PAST THREE YEARS。

The emergence of different models has given impetus to industry-wide adoption。

The same thing will happen to these agreements。

These agreements will attract a growing number of participants — developers, businesses, users — and build on them。

This will promote interoperability and eventually move towards integration。

In addition, security issues have become increasingly important as more and more Agents have been constructed。

Everyone is building their own Agent systems, but it is important to ensure that they operate in a secure environment。

So we posted on GTC something called OpenSheld。

OpenSHIELD is securely reinforced when it is an open source running between Agent and infrastructure。

It can provide Agent with a sandbox environment where they can operate in a controlled environment。

This would limit the scope of impact even if problems arose。

Simon Taylor:
Yes, that is crucial。

Many do not realize:

When you're building Agent, and you have a production environment, do you want to put Agent in a production environment

If there is no isolation, if there is a problem, the impact will be great。

So a sandbox mechanism like OpenSheld is very important。

Simon Taylor:
AND I THOUGHT OF ONE EXAMPLE: THERE WAS A WAP IN THE EARLY DAYS OF MOVING THE INTERNET, SOMETHING LIKE THAT, BEFORE THE SMARTPHONE CAME OUT, PEOPLE TRIED TO PAY FOR IT。

At present, it may be at a very early stage。

So I wonder:

How do you allocate your energy now

You're focusing on stabilization coins
Or is there an interaction between people and Agent
Or the interaction between Agent and Agent

Do you do all of it or do you focus

Pahal Patangia:
That's a good question。

From my point of view, we are primarily concerned with the most important current trends:

• payment base model

• Agency Company

However, new sub-trends will continue to emerge。

Like a stable coin。

We see the stabilization currency as a complement to the existing system, which brings with it new users and new ecology。

A new generation of users may be more accustomed to using stable currencies than credit cards。

But at the same time, the two will be integrated。

IN ESSENCE, HOWEVER, AI ' S CORE USAGE IN THE AREA OF PAYMENTS HAS NOT CHANGED:

• Anti-fraud

• Organisation

• Personalization

These remain the most important。

Simon Taylor:
Yes, in essence, it is the added value of payment。

These problems exist whether you use a stable currency or a card network。

Simon Taylor:
Bam, I'm curious what you think. What do you think of the relationship between Agenic Commerce and stabilization currency

Bam Azizi:
I think Agenic Company can use different payment tracks。

Now, for example, users search for goods like a pair of shoes or a T-shirt on ChatGPT, Anthropic, or Portexity, and then Agent can help the users pay。

This payment can be made with a credit card or a stable currency。

In this scenario, the two are parallel。

However, in cross-border payments and international transactions, a stable currency would be more advantageous。

And in the Agent-to-Agent scene, I think the stability currency is an absolute advantage。

The reason is:

These transactions are usually micropayments。
For example, $ 0.0005。

This amount cannot be processed in Visa or in the traditional banking system。

At the same time, these transactions require:
Real time
Global
Online

The currency of stability meets these conditions。

The other thing is the transaction frequency。

A person may make two deals a day on average, but Agent may do 2,000 a day。

THIS TPS IS SUPPORTED ONLY BY BLOCK CHAINS。

Traditional payment systems are not designed for Agent; they fail。

So I really appreciate the use of stabilization coins in the Agenic Company。

Simon Taylor:
It's really an outburst, isn't it

I remember there's about 4 million e-mails on the Internet every second, and it's just mail, not even a video。

In a world like this, it is clear that the capacity of traditional payment systems to deal with tens of thousands of transactions per second is insufficient。

But let's get back to reality, Pahal. From your point of view, where is the real user demand? Where's the real deal

I've often joked that there are more agreements now than payment agreements。

You may be the closest person to the bottom infrastructure — even “infrastructure infrastructure”。

So where's the real need? Where are the real examples

Pahal Patangia:
I think there are two ways to answer that。

The first is from an ecosystem perspective。

As I mentioned earlier, we can divide the whole process into two parts:

• Search

• Payments

At present, this part of the search has become more mature and may even be said to have been resolved。

This component of the payment is still at a substantial experimental stage。

many sandbox tests are in progress。

And that's why I'm looking at tools like OpenSHIELD, because it helps ecosystems build these Agents in a safe environment and gives them the ability to trade。

The second is in the long run。

I watched a lot of Agent developments。

In the world ahead, different Agents interact and collaborate。

And our role is to help these systems become better:

• Loop through feedback

• Through a safe operating environment

• through various binding mechanisms (guardrails)

Of course, a lot of fine-tuning is needed to ensure that these Agents are implemented as expected, without deviation。

These are the future directions of our focus。

Simon Taylor:
i think one of the important themes of today's discussion is “token economics”。

In fact, when we were talking about token, Bam and I laughed, because in the area of stable currency, we understood token economics as another logic。

But now you'll find:

everything has become "token."。

there's token in the id
it's in cyber security
Visa, Mastercard, network token
there's a token in the open bank
stable currency is token
It's in AI, too

the word “tokeen” is actually confusing in english, because it would have meant “some alternative”, but now almost everything can be called token。

But anyway, you have to understand the economic model behind it。

ULTIMATELY, EITHER AI OR THE PAYMENT NETWORK, THE USER EXPERIENCE OR:

• Speed

• Cost

These two factors will keep bringing us back to reality。

Simon Taylor:
Pahal, thank you very much for sharing today. As a long-time focus on NVIDIA and as a member of the paying industry, this conversation was really interesting. If you want to know more about you or NVIDIA in the field of payment, where can you go

Pahal Patangia:
Everyone can contact me on LinkedIn or through my mailbox。

IF YOU WANT TO KNOW WHAT NVIDIA IS DOING IN THE AREA OF FINANCIAL SERVICES, YOU CAN VISIT THE NVIDIA OFFICIAL NETWORK, AND WE HAVE A DEDICATED INDUSTRY PAGE ON OUR WORK IN THE PAYMENT, BANKING AND CAPITAL MARKETS。

WE WANT TO BRING THE CAPACITY OF AI TO THE ENTIRE ECOSYSTEM AND WE ARE HAPPY TO BE PARTNERS。

Simon Taylor:
Good, thank you. Bam, what do you want to do if you want to access Mesh or contact you

Bam Azizi:
You can visit meshpay.com or search Mesh Pay on Twitter, LinkedIn. If you want to find me, you can search Bam Azizi on Telegram or Twitter。

Simon Taylor:
You can also find me on every platform, or you can visit Finlandfood.com. I recently wrote an article on "Invisible Commerce" to discuss some of the issues that might exist in Argentina Commerce. If you like the show, remember to subscribe, compliment and share it with friends so that more people can see it. I'll see you next time。