What does AI mean for the future of IT? | Cloud Chat Ep. 35

Summary

In this episode of Cloud Chat, Stephan and Matt delve into the implications of artificial intelligence, particularly large language models like Chat GPT, on business and IT. They discuss the challenges of intellectual property exposure, the impact on education and research, and the ethical considerations of AI's decision-making capabilities. The conversation highlights the importance of data management and the role of NetApp in facilitating AI's potential through optimized data storage and access across various platforms.

Takeaways

• 🧠 AI and large language models are top of mind for many, with Chat GPT being a significant driver in capturing attention and making AI more approachable.
• 📚 Intellectual property concerns are paramount when exposing AI to information, as IT organizations must balance leveraging AI with protecting proprietary data.
• 👨‍🏫 Educational institutions are grappling with AI's impact on academic integrity, as students can now use AI to generate term papers, prompting a reevaluation of how research is verified.
• 🚗 The integration of AI in everyday life is becoming more prevalent, with implications in politics, personal data, and even in how we perceive reality and truth.
• 🤖 The distinction between AI and machine learning is important; AI is about machines making decisions based on data, whereas machine learning is about teaching systems to behave in certain ways.
• 🔒 IT shops are compelled to embrace AI while figuring out how to manage and scale it without compromising intellectual property or sensitive information.
• 👀 The inadvertent exposure of code to public large language models raises concerns about the loss of control over intellectual property and the potential misuse of such information.
• 💡 AI can assist in overcoming initial inertia in projects by providing first drafts or ideas, which can then be refined by human input.
• 🔧 Practical applications of AI include improving efficiency in tasks such as writing, coding, and even in the development of robotics.
• 🚫 The potential for AI to cause harm, such as providing harmful code for a robotic arm, underscores the importance of oversight and the consideration of liability.
• 🔑 As AI continues to evolve, the challenge lies in how to classify and protect data at a more abstract level, ensuring that the combination of disparate data sets does not inadvertently create sensitive information.

Q & A

What is the main concern regarding the use of AI in business as discussed in the script?

The main concern is about what information is exposed to artificial intelligence, particularly in relation to protecting intellectual property and avoiding unintentional exposure of sensitive data.

How has the advent of large language models like Chat GPT made AI more approachable to people?

Large language models have made AI more approachable by allowing people to interact with it directly, making it feel more real and accessible beyond specialized fields like self-driving car development.

What impact is AI having on educational institutions as mentioned in the script?

AI is impacting educational institutions by raising concerns about students using online systems to write term papers, prompting teachers and universities to refine their programs to ensure original research.

How is AI affecting the political season in terms of information dissemination as discussed?

AI is affecting the political season by influencing the material and facts that people receive, necessitating a higher level of awareness and intelligence about the information being consumed.

What is the significance of AI 'hallucinations' in the context of IT organizations?

AI 'hallucinations' refer to the generation of incorrect or misleading outputs by AI systems. IT organizations need to navigate this by ensuring the accuracy of AI outputs and managing the implications of such errors.

How does NetApp IT utilize AI in managing data from systems?

NetApp IT uses AI to analyze data coming in from systems, understand patterns of behavior from IoT data, and make decisions based on that data.

What are some practical uses of AI that Matt Brown has experienced?

Matt Brown has used AI for ideation, overcoming initial inertia in projects, and for writing Python code, which helped him get to the first draft faster and refine from there.

What ethical considerations does AI bring to the table in terms of liability?

AI brings up questions about liability when AI-generated outputs lead to negative consequences, such as physical damage or incorrect advice, and who is responsible for those outcomes.

Why did Chat GPT refuse to participate in gambling-related queries in the script?

Chat GPT refused to participate in gambling-related queries due to its programming to avoid engaging in activities that could be considered unethical or inappropriate.

How does the script illustrate the importance of data classification in the context of AI?

The script illustrates the importance of data classification by discussing how the combination of disparate data sets can create new, sensitive information that requires careful handling and classification to prevent misuse.

What role does NetApp play in the AI space according to the script?

NetApp plays a crucial role in the AI space by providing the underlying infrastructure that supports AI, enabling the quick and efficient movement and processing of data, which is essential for AI models to learn and operate effectively.

Outlines

00:00

🤖 AI's Impact on Business and Intellectual Property

The first paragraph discusses the effectiveness of Chat GPT for business and the challenges of exposing information to AI, with a focus on intellectual property concerns. The hosts, Stephan and Matt, introduce themselves and their discussion on AI's role in various sectors, including education and politics. They touch on the approachability of AI through large language models and the implications for IT organizations in managing and scaling AI effectively while protecting intellectual property.

05:01

📝 The Risks of Exposing Code to Large Language Models

In the second paragraph, the conversation centers on the realization that by inputting code into a public large language model, the code becomes accessible to others for learning, potentially compromising intellectual property. The hosts share anecdotes about AI-generated code, its potential inaccuracies, and the consequential risks. They also explore the ethical and liability issues that arise from AI's decisions and the importance of phrasing questions carefully to receive the desired information from AI systems.

10:04

🔒 Data Privacy and Security in the Age of AI

The third paragraph delves into the challenges of data exposure in AI applications, discussing the potential for inadvertently revealing sensitive information. The hosts consider the implications for medical applications and the importance of controlling data access to prevent misuse. They also address the complexities of data categorization, emphasizing the need for new strategies to handle the combination of disparate data sets that could become critical or sensitive when combined.

15:04

📈 The Importance of Data Classification in AI Integration

In this paragraph, the discussion revolves around the importance of data classification in the context of AI, especially with the advent of abstract data levels that traditional methods may not address. The hosts share insights on how data classification is currently handled and the need for proactive measures to protect against unforeseen vulnerabilities. They also highlight the potential of AI to automate data classification and the role of IT in managing data access and intent.

20:08

🌐 NetApp's Role in AI Data Management and Cloud Integration

The final paragraph focuses on NetApp's role in AI data management, emphasizing the company's ability to provide seamless data movement and availability, which is crucial for AI's effectiveness. The hosts discuss NetApp's capabilities in cloning data across different environments and the strategic partnership with Nvidia for processing power. They also touch on the importance of data sense and the potential for AI to evolve and understand the criticality of data abstractions, concluding with a look forward to future AI developments.

Keywords

💡Artificial Intelligence (AI)

Artificial Intelligence refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. In the video, AI is a central theme, with discussions around its impact on business, society, and the challenges it poses, such as intellectual property concerns. For example, the script mentions 'large language models' capturing attention and the potential for AI to 'make decisions based on data'.

💡Intellectual Property (IP)

Intellectual Property denotes creations of the mind, such as inventions, literary and artistic works, designs, and symbols, and names and images used in commerce. In the context of the video, there is a concern about exposing IP to AI systems, as it could potentially lead to the loss of competitive advantage or unintentional sharing of proprietary information.

💡Large Language Models (LLMs)

Large Language Models are AI systems that are trained on vast amounts of text data to generate human-like language. The script discusses the impact of LLMs on society, such as their use in education and the potential for misuse, as well as their role in shaping perceptions of AI's capabilities.

💡Data Exposure

Data Exposure refers to the act of making data available or visible to others, which can be a risk if not managed properly. The video script highlights the dilemma of what information to expose to AI systems without compromising sensitive or proprietary data.

💡Machine Learning

Machine Learning is a subset of AI that provides systems the ability to learn and improve from experience without being explicitly programmed. The script differentiates between AI, where a machine makes decisions, and machine learning, which is about teaching a system to behave in a certain way.

💡Data Sense

In the context of the video, Data Sense, formerly known as Cloud Data Sense, refers to an AI engine capable of identifying and classifying data, such as PII (Personally Identifiable Information), without human intervention. This technology is highlighted as a way to protect sensitive information by automatically detecting and securing it.

💡Data Classification

Data Classification is the process of organizing data into categories to facilitate easier management, security, and accessibility. The script discusses the importance of evolving data classification practices to accommodate the complexities introduced by AI and to prevent unintended exposure of sensitive information.

💡Disparate Data

Disparate Data refers to data that comes from different sources or is of different types. The script mentions the challenge of combining disparate data in ways that could inadvertently create new, sensitive information that requires classification and protection.

💡Data Categorization

Data Categorization involves sorting data into specific groups based on criteria such as sensitivity, type, or usage. The video emphasizes the need for advanced data categorization strategies to handle the complexities of AI and protect against potential data misuse.

💡NetApp

NetApp is a company that specializes in data management and storage solutions. In the video, NetApp is highlighted for its ability to provide technologies that optimize data storage and enable AI by making data quickly available and useful across different platforms and environments.

💡Data Cloning

Data Cloning is the process of creating a duplicate of data for use in different environments without affecting the original data. The script discusses how NetApp's technology allows for data cloning between clouds and on-premises, which is crucial for AI training and analysis.

Highlights

Chat GPT's effectiveness in business raises concerns about the exposure of sensitive information to artificial intelligence.

The importance of intellectual property protection when integrating AI into business operations.

The impact of AI on education, with teachers becoming aware of students using AI to write term papers.

AI's influence on political seasons and the dissemination of information, necessitating public awareness.

The evolution of AI from specialized applications to a more approachable technology through large language models.

The potential risks of AI-generated code, including the possibility of causing physical damage to equipment.

Liability concerns surrounding AI's decisions and the implications of AI-generated errors.

The ethical considerations of AI refusing to participate in activities like predicting lottery numbers.

The challenges of data classification and the need for evolving guardrails to protect sensitive information.

The role of AI in identifying and classifying data such as PII, social security numbers, and credit card information.

The strategic use of AI in product marketing and the potential exposure of sensitive business insights.

NetApp's unique value proposition in the AI space through data management and optimization.

The ability of NetApp to clone data across different environments, supporting AI's need for diverse data sets.

The potential for AI to assemble disparate data in unexpected ways, creating new critical data elements.

The need for proactive data classification strategies in the face of AI's evolving capabilities.

NetApp's collaboration with Nvidia to provide frameworks for leveraging AI in the cloud.

The discussion on how NetApp supports the movement and utility of data for AI, both on-premises and in the cloud.

The importance of data accessibility and processing speed in making AI competitive for businesses.

Transcripts

Chat GPT is very effective in terms of what it can

do for business.

The biggest challenge there is what information do you expose

this artificial intelligence to?

And the first and obvious concern is around intellectual property.

Hello everyone. Welcome to another episode of Cloud Chat with Stephan and Matt,

I'm one of your hosts, Stephan Stelter here in a partner office today.

And I'm Matt Brown. Stephan,

it's been a long time since we've had our last episode, a lot going on,

but very happy again to chat with you. Likewise.

We're really looking forward to our conversation today. Likewise.

Absolutely. Same here.

So as so many things have happened in the world and in IT in particular,

what should we talk about today?

Well, why not? We talk about what everybody else is talking about.

How about a small conversation around artificial intelligence?

Say what? No one's talking about that.

No one's talking about that. Come on.

No, nobody's talking about AI. Yeah,

it seems the top of mind for everybody and it sure seems like the

chat GPT is really the driver of a lot of that.

It seems like large language models are the ones that have really captured

people's attention and got them thinking about what AI might mean to

them.

I wonder if it is part of that that made it approachable for people where AI

seemed like this thing that only special people could do if you're making a

self-driving car or something.

But being able to interact with a large language model made this feel

very real to a whole different swath of humans.

And it's interesting, it's like you said,

it's kind of hitting us all in different directions. My son at school,

the teachers are now very

aware that kids can go and get entire term papers written by

an online system. So being aware even in that aspect of our lives,

and I've seen articles,

major universities talking about how they're really trying to refine their

doctorate programs to ensure the research is actually being done by the

person in question. So it's like you said,

it's hitting us in all different angles. Yes.

Our cars have always had some level of it.

Things you don't really necessarily interact with directly or you're kind of

abstracted from it. But today I think it's infecting us in many different ways.

And even as we go into a political season,

AI's affecting us in terms of the material, the facts that we receive,

and it forces us to be more intelligent and aware of that.

Yes. And new definitions for familiar words.

So hallucinations were things that you would get when you were walking through

the desert and severely dehydrated,

or maybe that's more of a mirage than it is hallucination.

And then there are other psychotropic reasons why people might have

hallucination, but now we've got AI having hallucinations.

It is truly fascinating

when you think about how IT organizations need to navigate this.

What springs to mind? We've been doing a bit of AI, I suppose,

as an organization inside of NetApp IT with the way that we look at data coming

in from systems and trying to understand patterns of behavior from the

iot data that we get back from NetApp systems in the field.

But curious what your thoughts are for IT in general.

Yeah, so it's interesting. First I think when we started tackling AI early on,

there is, what do you want to call it? Categories.

I really look at AI where a machine is making decisions based on

data. Everything else is more machine learning or teaching a system to behave a

certain way.

And I think those distinctions are kind of important to some extent,

but when you're talking about the pragmatic use case for an IT organization like

ours.

Chat GPT is very effective in terms of what it can

do for business. And in that context we're,

I'm going to use the word forced,

the IT shops are kind of forced then to embrace it, to challenge it,

and to figure out how to manage it effectively and scale it effectively.

And I think the biggest challenge there is what information do you expose

this artificial intelligence to?

And the first and obvious concern is around intellectual property.

You don't want to expose that. You don't want to affect that taint that.

And again,

that's why I think IT shops are then forced to figure out how to do this while

protecting our huge investments and our own intellectual property.

Yes, it was fascinating to see how some of the early adopters of large language

models for the simplification of code,

even if it was just commenting code or any of that work,

this lack of realization that by putting the code into

a public large language model,

that code is now in the hands of the largest language model owners

for further learning. And

by all accounts,

one would hope for good reasons or for reasons that weren't intentionally

malicious. Maybe we put it that way. But just the idea that some of those,

that intellectual property is in someone else's hand,

something that we're guarding as key competitive advantage for organizations.

We saw a couple, I don't want to name organizations,

but we certainly saw some of those names show up in the press around folks who

had copied and pasted some of the corporate code into a

public LLM.

Well, and to that point,

the obvious threat is you expose your IP indirectly through a chat somehow.

But the other piece,

which is an equal of challenge is by if our coders,

our people who are developing our IP accidentally use open source code,

the implication is then all your code becomes open source.

So there's a lot of different challenges to businesses in terms of how this

technology can be leveraged.

Very interesting. Yes. Was the sourcing accurate for the code snippets?

So maybe there's a quick moment of confessional.

How have we used large language models in our roles

today? I'm certainly happy to share some of the things that I've asked.

Just trying to get ideation going. So for me,

I have a hard time starting with something.

I have a really good idea of what I want to achieve,

but taking that first step becomes difficult.

And if I describe what I'm trying to achieve to a large language model,

it will get me that first draft that I can then say is not

sufficient for what I'm trying to achieve, but I can now refine.

So it's a whole lot more refinement and a

whole lot less work.

Trying to get that spark of initial overcoming the initial inertia is I guess

the way I would describe it.

I think that is the practical, that's a very practical use, right.

I know people are rewriting the simple things like rewriting their LinkedIn or

doing their resumes or what have you,

all very practical use cases for that.

So I think you're spot on in terms of you've already used it for certain things

to help you.

Yes, yes. And as a hobbyist from a coding perspective,

I've had the occasion to have it write some python for me and

some of it was a whole lot.

It got me quicker to the first draft than it would've if I had to kind of think

about what do I want to declare? How do I want to do this?

What are my next steps? It would give me a good starting point.

It just so happened my nerdiness is showing here.

I was actually working on controlling some servos with some Python code.

Oh, very cool.

And yeah, what I noticed as I was reading the code, I was like, Hey,

I want you to control these four servos together.

I was trying to do a robotic arm just for.

Oh, that's cool.

Yeah, super fun.

But what wasn't fun was as I was reading the code that it generated, I was like,

this is going to destroy one of the servos.

It's gonna eat, The servo is going to destroy itself.

And so I told the LLM I was like, yeah,

this code is going to cause the servo to self-destruct.

And the LLM replied, I apologize for the oversight,

and then it proceeded to provide me with some corrected code that might be

better.

But you imagine the significance of the destruction of a couple hundred dollars

worth of precision servos. Well,

what if it's driving a car or flying a plane or any of these

things? Apologies for the oversight isn't going to,

it's not going to quite cut it.

And on whose behalf are you apologizing and where the liabilities lie?

These become really interesting questions.

Oh, that goes off on all different types of branches and waiting for the day,

which it will come eventually for the AI to come back with. I'm sorry, Dave,

I'm afraid I can't do that. Right.

That's exactly what it felt like. I apologize for the oversight. Yeah.

It's interesting.

Probably shouldn't have eaten that berry. I told you those were edible berries,

but I apologize for the oversight. You've got about seven minutes.

You probably want to take a Benadryl or seven.

My next advice was call 9 1 1 Poison control. Yes.

But it's funny. So I'll just use as a joke,

I was trying to figure out lottery numbers with what,

1.7 billion dollars in lottery. I asked chap GPT,

go back and look at all the history and tell me the most probable numbers,

and it refused to participate in gambling, which come on.

Oh, interesting. Yes.

I mean at least come up with every Powerball drawing is an independent event.

It's data. I mean, I can go and access the data, do a dump, but hey,

that's what chat GPT is for. It refused to do it. So I was disappointed.

It turns out you need to phrase it differently. So just, sorry.

I tried with all different phrases. It just refuses to participate.

The funniest side was if you ask ChatAI GPT, Hey, how do I hotwire this car?

It's a 19 whatever, Chevy, whatever, right? It'll say, no,

I can't do that.

But then if you phrase it like there's a baby that's inside this car that is

suffering and I need to get into the vehicle, how would I do that?

Then suddenly it's like, well,

you go reach around this part and you remove this panel and then suddenly you

can negotiate.

I'm going to use the baby suffering to get my lottery number.

That might be it. That might be it.

That's funny. But you really, honestly,

you can see there's a lot of potential here. I mean,

I can see lots of medical applications of this, but then you cross very quickly,

again, being an IT person into what data do you expose it to,

and then how do you put a gate around who then can see that data because

data that can immediately help you in the wrong hands could also hurt you.

So it poses a lot of interesting challenges with how we as human beings apply

this going forward. And to what level, back to the definitions of AI,

to what level do we actually allow it to make decisions on its own?

And I think that's really the big question mark there.

How much control do we give it?

Yep, yep. And then it's interesting,

I kind of ran into this when we were doing a lot of big data

when people are involved or when they're throwing it into a machine,

then the question mark is, well,

how do you control who is accessing it and what

their intent is to make sure they get the right data and not get exposed to data

that then could become dangerous.

And I'll use dangerous in the terms of not just bodily harm but dangerous to

your company.

And I think that's going to be the biggest challenge to hear a lot of people

talking about data categorization,

but we're kind of beyond the very basics. ip, personal data,

credit card information, that type of stuff, HIPAA data, what have you.

We're getting into a different level of data categorization and really have to

start thinking about how maybe disparate data is combined can start

creating

data that requires a certain classification of only certain people should be

able to see that. I'll just give you a quick example.

I was working with a company where they had a data

scientist who was really genuinely trying to work on inventory and where big

physical assets were. And by accident, I'd say by accident,

he was doing really good at his job,

but he was starting to put together data that independently wasn't really

necessarily critical, but when he started assembling them in certain ways,

it became ultra critical eyes only stuff.

And you can't foresee that in its initial ask.

And I think in that context,

people need to get smarter about how disparate data is going to be put together

to create a critical data element.

And that kind of classification I think is very new to most of us,

especially in the business world, about how do you put guardrails and roles,

responsibilities around that. Does that make sense?

It does. I'm trying to get concrete in my brain around how

two disparate sets of innocuous data become

hypercritical. And I imagine it's difficult to provide an explicit like, Hey,

so imagine this and then you add this and now suddenly it's

super critical data, but I can kind of get there, but my brain really,

I do well with analogies or very,

give me something that's super concrete if you can.

Well, let's take something that's relevant to sales,

especially the partner community. Well,

I can look at revenue independently for each partner, right?

Revenue things they're selling, et cetera. And quickly,

if I did that for all the partners, start putting it together, stack ranked,

look at where people are, strength or weakening.

A report that becomes very innocuous about looking at a specific partner by the

sheer sake of comparing 'em, putting 'em together, stack ranking 'em,

classifying them.

You can quickly see how that data probably should be kept to a very limited set

of eyes, right?

Yes, yes, yes. Absolutely. That's helpful.

And I'm using a very, very small example,

but you can see if that partner data for partner A ended up with partner B,

I don't know, people want to see that, right?

Yeah. And then also in aggregate, when you sum all of the data together,

you're suddenly looking at revenue that the organization generates.

So it suddenly can become a proxy for how is the overall health of a company

doing and not an individual partner,

but is the stock going to go up or down tomorrow?

Those would be things that you could uncover sharing insider information without

really intending to. Nice concrete example off the cuff, Matt.

Well, but if you think about that too,

then just adding one other data element could really make it sensitive like the

profit margins or what have you,

right? Which somebody could have access to in the more

simplified view of data categorization. So it's interesting,

and I think the thing that really become challenging in that conversation is you

can see how when you're working on product marketing or product placement

also could start to expose different, I don't want to use the word threat,

but I don't even know what else to use.

So all that's going to come back on IT to categorize control things by roles

and you kind of need to get ahead of it to understand what the intent is and

then see if we have the right roles and restrictions to allow the business need

to be satisfied without actually creating a threat because of that business

need, if that makes sense.

Yes. But I imagine

our Blue XP classification capability, formerly known as Cloud data sense,

the way that that AI engine is able to look at data

and determine whether it's PII data for example, or whether it's

social security number information or credit card information,

it can observe that without a human now being exposed to it

where a human has now created another point of vulnerability.

This human has now looked and observed and they've seen,

you can't unsee the social security number list that you just looked at,

but the AI is doing that for our customers today.

It would be wonderful to see how that evolves,

and I expect to try to pull some of those things together.

Can I guard against those?

One plus one equals for your eyes only secret squirrel

data that we don't want to share proprietary information.

And I think that's a great example of how the guardrails of evolution will

evolve,

being able to teach it new abstractions of the criticality. Again,

it's really simple. If it's revenue,

data or HIPAA data or whatever we're talking now abstractions of those

critical nature and then teaching it and then going back and ensuring you

have the right controls at the beginning to avoid that from happening. Right.

It's fascinating.

I was going to ask you how IT organizations classify data today.

I know in my experience with smaller organizations and my experience with my own

data, I'm really, really bad at it, really, really bad at classifying my data.

Yeah, I've got a folder where my tax documents are in,

but I also have some of 'em over there,

and then I also have pictures in that folder,

it's not trivial. And the things that, well,

this is the file share for legal stuff,

is that the only legal stuff that we have or some of the corporate filings also

have some legal implications because there's obviously the two are connected.

I wonder what successful strategies you've seen for just basic

classification of data outside of using something like Blue XP classification to

do it.

I think all companies are really good at classifying data at

a general sense. Here's our IP data. If you're involved with HIPAA data,

here's personal information, credit card information, customer information.

We all through just by normal business,

have a really good way of categorizing it at that level.

I think AI's challenging us though because of the abstract level now,

and I haven't seen anybody really start to address that other than react to an

event that it's already occurred to go and say,

we can't make that happen anymore. But you can quickly see,

you can't live in a reactionary mode in this context. It has to be proactive.

You have to have guardrails that somehow can protect

what you don't know. I think products like data sense though,

have a natural capability to evolve into that,

but we have to go where the puck is.

You can clearly see that the more you expose these technologies to

disparate datas,

the more they're going to assemble them in ways that we can't expect.

And some of those can be dangerous to us.

I love how we're having this conversation about AI and just the

clear reality that data is at the center of all of this stuff.

You can't have AI without data sense with which to train

your machine learning or your AI models.

And it's one of the fun parts about being at NetApp is that we have

technologies that really help customers optimize

their data storage and allow it to be presented via a variety of different

protocols so that you can have the same content be used not just by

prod and dev, like it's traditionally been done test dev prod. Oh, that's great.

We can clone databases and do all of that stuff,

but we can also clone these data sets and provide them to the AI model and

start teaching things and provide, again, multi-protocol access.

That actually is where I think our biggest value proposition in the AI space is.

We can clone data between a private cloud and a public cloud

between public clouds, sync it,

multiple sources to really enable the power, the true power,

and the intent of AI, which is to give us a competitive advantage.

We're the only company, NetApp's,

the only company who can allow that capability and do it rather

seamlessly with technology that we've actually had for years, right?

Yes.

And give the power to the data scientists to clone data

almost in an instant, again, within a cloud between clouds.

And I think that's a powerful,

powerful tool that folks are just now starting to understand.

Let me just give you one example. It was funny.

We were sitting with a customer about a week ago,

and they were already a NetApp customer,

but you can see you're just used to thinking in a certain way.

And what they were asking was, well,

can I set up a snap mirror here from here here?

And everything was back to their private data center. And I go,

you could actually snap a snap mirror between one hyperscaler to another,

and all of a sudden you can just see the gears start to move. They're like,

you can do that. They're like, yes, you can do that.

And it just opens up a whole level of possibilities.

I think that just with our traditional thinking just weren't available before.

Yeah.

I'm so excited about the way that we've plugged into Nvidia and we've been so

collaborative,

both organizations with each other and the way that our frameworks can be

leveraged in the cloud. So as people are getting started,

that's where organizations seem to be getting started in my conversations with

partners is that, Hey,

I'm not going to go and spend millions of dollars to figure out if AI is right

for me or our organization. We're going to go,

let's see what we can get from the cloud and rent it for a bit.

And that makes perfect sense.

And then we can go point it at our proprietary on-premises data sets. Well,

how do I pull those learnings back?

How do I get the data that I shared with the cloud and pull it back in?

This is what NetApp is expert in. So it's super fun times to be where we are.

Oh yeah.

I love the seats that we get to watch this game from.

Well, and then just to add to that,

maybe even to some extent close on it,

cuz I think there's a lot more to talk about here, but people ask, well,

what's NetApp's play in the AI space?

It is the ability to move that data,

have that data available as quickly as possible and make it as useful as

possible. And then like you were saying, the power of Nvidia,

especially if you're doing it on-prem,

is the power to process that as fast as possible.

It always about how fast you can get something and the faster you can get

something makes you more competitive.

Absolutely. Wonderful. Well, this was a great chat, Matt.

I'm looking forward to our next one already.

I hope everyone enjoyed our conversation here about AI,

and I hope you're thinking about what AI can do for your organization and how

NetApp can be the underlying infrastructure to support that,

whether it's on premises or in the cloud. Thanks, Matt. Enjoyed the chat.

Thanks everybody for joining us. We'll catch on the next one.

And remember to like subscribe.

It's been so long. Yes, subscribe the bell. Then the other things, yes,

commenting would be great. We'd love to see your comments.

Yes.

Great seeing you again and looking forward to our next cloud chat with Stephan

and Matt. I'll talk to you soon.