#JensenHuang#NVDA#AI NVIDIA CEO Jensen Huang shares critical insights on the future of autonomous artificial intelligence and why safety is fundamentally an engineering challenge. As AI evolves from interactive models to self operating agentic systems, maintaining strict control over digital permissions and system boundaries becomes crucial for global adoption. Huang explains how new containerization frameworks and hardware isolation chips create secure environments that restrict AI rights, ensuring agents perform intended tasks without accessing sensitive networks or unauthorized data.
The discussion highlights groundbreaking developments in AI infrastructure, including open source containment systems like OpenShell and dedicated silicon monitoring chips like Sentry. By placing silicon monitoring between large language models and software agents, these technologies intercept actions in real time and enforce strict policy adherence across cloud environments. Industry support from over a hundred major tech partners underscores the importance of transparent, open source standards in building public trust and preventing security breaches.
Beyond technical safeguards, Huang addresses the delicate balance between government regulation, innovation, and global market competition. Accelerating AI development is presented as the most effective path toward creating more grounded, reliable, and predictable systems. Rather than imposing premature regulations that stifle growth, establishing robust industrial quality standards will allow enterprises across finance, healthcare, retail, and logistics to safely transition into fully realized AI companies.
Do you agree with Jensen Huang that AI safety is strictly an engineering problem, or do you think governments should step in to regulate autonomous AI agents? Drop a comment below with your perspective.
Timestamps: 0:00 - Introduction and the Debate Around AI Safety 0:32 - Building Safe Agentic AI Systems 2:02 - Solving AI Safety as an Engineering Problem 4:47 - Introducing OpenShell Open-Source AI Containment 5:48 - OpenAI Collaboration and Open Industry Standards 6:54 - OpenShell Architecture and Sentry Hardware Monitoring 11:00 - Nvidia's Investments in OpenAI and Hugging Face 12:54 - Industrial Safety Standards vs. Regulatory Capture 17:44 - Accelerating AI Advancement to Improve Safety 19:53 - Global Competition and AI Regulation Cycles 22:53 - AI Model Distillation and Fair Competition 24:10 - Nvidia Cash Flow and Share Buyback Announcement
AI Generated·May contain errors
Video Summary
AI Generated
Jensen Huang discusses AI safety as a solvable engineering challenge, introducing OpenShell and Sentry to prevent AI breakouts. By implementing a \"force field\" sandbox and hardware-level monitoring, NVIDIA aims to ensure autonomous agents operate within strict boundaries, advocating for a \"build, then regulate\" approach to foster innovation while maintaining global safety standards.
Of course, the CEO of NVIDIA, and Jensen, welcome.
It is great to see you this morning.
Great to see you.
So this is an advancement in something that we've kind of seen the national discussion
take place around AI and AI safety.
And in this discussion, there have been those who have been painted as doomers, who think
that the end of the world is going to come from this.
And then there have been those who have painted you as the guy who doesn't want to see anything
happen on any level.
I don't think that's a fair and accurate depiction of where you stand with things.
I want to see a lot of things happen on a lot of levels.
So why don't we talk about where we stand, and then we're going to jump into what's next.
First of all, as you know, we're going through a major platform shift.
AI is incredible technology, has the potential to do incredible good, but we also have to
make sure that the technology is developed and deployed safely so that the public has
confidence in applying this technology.
The single most important thing we can possibly do is to build it in such a way that is so
safe that everybody could use it in every single industry so that we can transform every
industry.
And so that's the hard work that we have to do.
These agentic systems are obviously agentic, meaning they operate by themselves.
It's like a self-driving car.
And so in order for you to deliver that agentic system in a safe way, you have to make sure
that the sandbox around it, the force field, the play pen.
All of those systems are designed in a way that keeps the agent with minimal rights,
whatever rights it has, it needs in order to do its job.
No more rights than that.
And that it can't do anything that it's not supposed to.
And also, you have to monitor it.
But the problems that we've seen to this point, where open AI, Anthropic, Google, and others
have had agents that have escaped and have gotten out and done mischievous things, maybe
illegal things, the problems we've seen to this point you think is simply an engineering
problem and a fault of those companies, maybe not putting in the right restrictions?
Well, I think the answer is we hope it's an engineering problem.
I believe it's an engineering problem.
I know it's an engineering problem.
And we all need to hope that it's an engineering problem.
It's not an engineering problem.
It's not solvable, right?
And so the fact that all of these companies still are advancing the state of the art is
because they also believe it's solvable.
These are some of the brightest engineers in the world.
I work with all of them.
And this is a technically solvable problem.
And so the way to think about that is if you go back to the days of the internet, you access
a website, it downloads application, maybe a Java application.
All of a sudden, this application comes into your PC, has access to files and resources
and maybe the applications and tools, and it gets onto your network.
Well, in the beginning, we had all kinds of viruses.
But we realized that the web browser itself has to be a containment system.
The web browser has to provide that application with minimal rights, only the rights it needs
in order to do its job.
And it has access to basically access to nothing unless the browser gives it access.
Essentially, what we're doing here, we're creating basically the modern browser.
It's a browser for agents.
Somewhere inside the culture of AI researchers for years now has been this anxiety that somehow
these models have to be clever and they have to be able to get outside of their walls because
they're so clever, right?
Where do you think that's built into the AI researcher culture?
And so at one level, I think while there are some people who believe it's an engineering
problem, there are these others even inside these companies that believe it's unsolvable.
I don't think so.
I don't think it's exactly wrong.
First of all, you want it to be super clever so it could perform tasks and do things.
But the way you think about that is when you deploy an agent, no matter how smart,
the first thing you do is you take away all of its rights.
Job number one is take away all of its rights.
And then you provision, you give it access to files, data, tools, access to the network,
or even internet access, only if it needs it.
You don't put an agent into an environment, into your company, and give it access to everything.
We don't give that to people.
We don't give that to any level of executive, in fact.
Everybody has a badge.
Everybody has file access.
Everybody has access to tools.
And only the things that they need in order to do their job.
And so we have to apply the same philosophy.
Now, the question is, how do we develop the technology to provision that policy,
provision that philosophy?
And that's what we created.
That's what OpenShell is.
It's really important technology.
It has the support of some 100 companies.
You didn't mention IBM and Red Hat and Microsoft and SpaceX.
And so all of these companies, they recognize the importance of basically this force field
that you create around the agent.
And very importantly, it's provisioning rights and responsibilities and access.
It's an open source program.
Which is very, very important.
And the reason for that is this.
We want this sandbox to be visible to the whole world.
So that if there are any vulnerabilities, somebody will find it.
Look at these 100 companies.
There's actually a lot more companies.
There's only we can list.
The industry will all be jumping in, looking at any possible vulnerabilities and patch it up.
You have Anthropic and SpaceX both listed on this list of 100 companies.
OpenAI is not there.
Are they not cooperating with this?
No, this is an open standard.
It's an open platform.
And however they would like to participate is going to be super welcome.
They could look at the code.
They could use whatever part of it they would like.
They could build their own inspired by this.
Or they could use it.
I think over time, we really want people to use the software.
And the reason for that is because the more people use it, the more vibrant it's going to be.
The more resilient and more robust it's going to be.
You've said that OpenShell would have prevented the HuggingFace breakout from OpenAI.
Yeah, Clem and HuggingFace are one of the victims of the attacks.
And a huge, huge supporter of OpenShell.
Their computer sciences are very deep on OpenShell.
And they're a big part of it.
What happened that you all can do this and OpenAI,
Anthropic and others haven't been able to do this to this point?
Well, we started building this, you know, about a year ago.
Has it been a year, maybe slightly longer?
Ten months maybe is what I heard.
OpenClaw came out.
And when OpenClaw came out, you guys might have heard me say,
this is the operating system of AI.
And meaning that it's the operating system of agents.
And when we looked at OpenClaw, we said, hey, this is incredible.
But you can't deploy this into companies.
You can't have agents roam around and drift around the company.
And so you have to find a way to container it.
Containment is number one.
Rights provisioning is number two.
Being able to monitor its adherence and conformance to all those policies, number three.
And so we created containment, a policy provisioning supervisor.
And we created a monitoring system.
And the monitoring system is called Sentry.
And it monitors it in silicon.
Not to get too technical about this.
But the concept of containment and monitoring is very simple.
For those who know about, for example, AWS Bedrock has Nitro and a whole agent
for some other things built into it.
This seems to me a much broader open source version, frankly, of that.
Is that a terrible way to think about it?
Yeah, it's like Nitro in the sense that it's a rights provisioning system.
It's like Nitro in the sense that it's also hardware isolation.
Nitro, as you know, really revolutionized multi-tenant secure cloud.
In a lot of ways, this is multi-agent secured world.
And so a very similar idea.
The differences are going to be around the speed at which we have to run Bluefield for.
And the reason for that is you're monitoring AIs.
You're monitoring a lot of agents.
Are they adhering to the policies?
Are they using files or going onto networks they're not supposed to?
We might even look at its chain of thought and how it's thinking.
And maybe it's thinking about doing something that we really prefer it not to.
It might elevate and escalate to somebody, another AI,
or maybe a person and say, hey, look, this agent is doing something that we might not condone.
Some of the model makers who are making new models
are trying to use their older models to effectively track the new model that's being trained.
With this product, will they still want to continue to do that?
Sure.
The chip is between the agent.
The agent itself is actually a harness.
It's just a piece of software.
And this chip sits in between that and the large language model.
So the model is here.
The chip is in the middle.
And here's the agent.
And so whatever the agent's trying to do,
and whatever it's causing the large language model to think about,
Bluefield or Sentry sits right in the middle.
So we intercept everything,
which is one of the really, really incredible places that we sit in the world.
Because we have computers for agents,
and we have computers for large language models.
NVIDIA now has a new chip that sits right in the middle.
The idea to this point, you've said in the last several weeks,
that, look, liability that currently exists today in existing regulation
should be enough to keep these companies from doing some of these things.
They shouldn't release products that are going to get into trouble
or that are going to hurt people or shut down systems.
I like the way you just said that.
I think you just kind of stated the obvious.
I think, number one, I'm not suggesting that products are vulnerable sometimes
and companies release products that aren't as well tested as they should be.
That happens every single day.
However, if you knowingly do it,
if you already know that you didn't test it properly,
if you know that you could have done a better job testing,
just hold it back.
You're in the middle of all of this.
You are in the process of buying Hugging Face for over $12 billion.
That was a company that was broken into by OpenAI.
I think you've got a $30 billion investment in OpenAI.
How do you come down on who's responsible?
And, you know, are you going to go after OpenAI
for what it's done to Hugging Face or other places?
Will there be lawsuits that come out from any of this?
How do you kind of decide, as somebody who's involved in all of this,
what you want to see happen?
Well, I don't have to decide on any of those matters.
The thing that I have to decide on, in the case of OpenAI,
is OpenAI a great investment?
And the answer is absolutely.
This is one of the most consequential companies in history.
Sam and Greg and Sarah, they do a terrific job running the company.
Incredible researchers.
It's a real privilege to work with them.
Love the work that we do with them.
Hugging Face.
I mean, this is the world's leading platform for open models.
I believe very deeply in advancing open intelligence.
I think we ought to democratize intelligence.
I think every company will become an AI company.
And this platform really, really keeps that community thriving.
And so it's really important that we invest in both.
Does this platform absorb the liability issues for any company that's using it?
I mean, if you're using it and your agents break out, you say,
wait a second, I'm using OpenShell.
It's NVIDIA's problem.
No, the OpenHuggingFace, it hosts the models.
It doesn't run the models.
You download the models.
And once you download the models, the models become yours.
The model, it's really important to realize that we call them model companies,
but in the end, the model doesn't do anything.
It's the agents that does something.
So this is a tool.
And so the model is kind of, yeah, the model is the engine of the car.
The agent is the car.
And so I wanted to go back to the liability piece of this,
because I think it's so fascinating, because the other side of it,
and we had Chris Hughes on, one of the founders of Facebook just earlier.
And we've read reports that people like Scott Besson have even talked about the idea of
building a FINRA-style regulatory body or agency that can oversee this.
Tell us what you think of that, kind of like a FINRA-style regulatory body,
whether that makes sense.
Some of the reporting suggests that Scott, maybe over the summer,
had been thinking about some of this.
You might even talk to the president at some point about not going in this direction.
Well, first of all, I think that we need to have an industrial set of standards and policies.
And those industrial standards, like many other industries,
set expectations of the level of quality.
Maybe they're third-party auditors, agencies that help you audit whether you achieved
the level of performance and quality that you expected and that you promised.
That their industry is a benchmark, products versus others,
so that the better products that deliver on better safety and capabilities
rank higher than the others.
There's many industrial systems that I can imagine putting together.
If the policies are such that open models are not harmed,
young companies are not harmed, that every company could become an AI company,
that every bank could become an AI company, that every retailer could become an AI company.
So long as we could see policies that don't suffocate the industry too quickly.
Because as you know, the AI industry hasn't formed yet.
Every single company here in New York, they're all going to be AI companies.
And so every car company is going to be AI companies.
And so we need to give them a chance to be able to become one of those companies
and don't suffocate the industry quite yet.
Is that what you think the heads of the open or of the
frontier labs have done by calling for a slowdown?
That it slows down everybody else, just lets them stay ahead?
You know, I'll assume that what they're claiming and what they're asking for
are genuine and their concerns are genuine.
I think just in consideration of the whole industry,
in consideration of all of the United States and the entire market,
I would just make sure that when we think about AI,
don't think about AI as several model companies.
Remember, AI is a five-layer cake, as I've described to you before.
And the most important layer of the five-layer cake are the industries above it.
I think that Walmart's an AI company, FedEx is an AI company.
I think that JP Morgan's an AI company.
And I think that Lilly's an AI company.
All of these companies have to be AI companies.
So we have to give everybody a chance to achieve that.
So when you hear David Sachs say,
the motivation of the leaders of these firms, it's all a regulatory capture.
I think that's rebelling.
It's all just an effort to sort of capture a particular part of the market.
And you know Sam, and you know Dario.
Dario just had dinner with the president, apparently, last night.
What do you think of that?
Whatever their intentions are, we want to avoid regulatory capture
advertently or inadvertently.
And the technology is moving really quickly.
Just think, we want the technology to accelerate, not the other way around.
And I can just go back in time and think about this.
When Chad GPT first came out, remember, if you go back and use it again,
people said GPT-2 was so dangerous it shouldn't be even released.
And at the time, I was advocating for AI to accelerate.
The reason for that is this.
One, we solved the hallucination problem.
It's now much more grounded on facts.
Notice how AI is able to now use useful tools so that it could do the job properly.
We can now condition the AI and prompt the AI so that its context is much richer.
All of this capability was made possible.
It's more trustworthy now.
It's more functional.
Of course, it's much more predictable.
All of those reasons, all those capabilities are made possible
because we accelerated the development of AI.
Accelerating AI and accelerating AI safety is the same idea.
I prefer today's car over the car 100 years ago.
At ABS, it has lane keeping.
It has much better braking systems, airbags.
I mean, all of these different technologies are made possible
because of the advancement of the technology.
And so I would advocate that today's AI, when we think about advancement,
you don't have to think about it separately from capability versus safety.
To me, it's the same thing.
And so we want to accelerate the advancement of AI.
And today, we're putting forth this new technology,
OpenShell and NVIDIA Sentry,
that these two technologies make it possible for us to contain and to monitor.
But it takes a lot of technology to make this possible.
Do you get frustrated by...
Look, everybody at this point, I think, is either getting painted as a doomer
or somebody who has absolutely no interest in any safety at all that's going on.
I read an article over the weekend that compared you to Thanos,
the supervillain from Marvel.
But when you actually sit and talk to you, it's a much more nuanced view than that.
You do want some of these safeties.
I'm a responsible optimist.
If we're not optimistic about this future, for what reason are we building it at all?
And so I think you have to just maybe sit and contemplate,
for what reason are we bringing everybody along if we're not optimistic about its future?
But we also have to understand that technology is powerful.
And the technology is powerful, so we have to build it safely
and we have to build it responsibly.
Which is the reason why we built OpenShell,
which is the reason why we launched all this technology.
It's the reason why all of these companies have joined us.
You know, look, I've got a lot of adult friends.
These are leaders of giant industries.
We're not Pollyanna about any of this stuff.
However, we are serious about creating the technology,
doing it seriously, doing it responsibly.
There are laws.
We comply by laws.
The reason why we love this country is because it's a rule of laws
and we've got plenty of laws and regulations.
Apply those laws and regulations.
Take our job seriously and build for an optimistic future.
I've got two questions that relate to China,
because I think underneath a lot of this, in terms of the rush and the speed,
is this idea that we're in this deep competition with China.
I've always thought, maybe you'll agree or disagree with the theory I'm about to pose,
that China has oftentimes been very permissive about technology early on.
I think about crypto.
I think about Alibaba and then what happened with Ant later.
But oftentimes, at some point, when they feel or if they feel
that something's going wrong, they clamp down.
And so I'm curious whether you think, you know,
right now we're in this race with China because there's a view
that China is not going to hold back.
But whether you think there could become a moment where China says,
you know what, actually, we need to control this like full on.
We really need to control this in a different way.
Innovation and regulation are not conflicts.
The question is, which one do you
lean towards at which phase of the industry?
At the beginning of the industry, when the technology is being invented
and nobody really knows how it's going to turn out,
it would be a shame to regulate before understanding.
And so you regulate to you regulate prematurely.
As a result, the technology doesn't get a chance to innovate.
And for example, we never got a chance to invent automatic braking systems.
We never got a chance to invent airbags.
We never got a chance to invent all the technologies that made the cars
made transportation safer.
We never invented autopilot systems with redundancy and diversity.
But you're not you're not calling for protection from regulation.
At different parts of the industry's S-curve, you know,
there's a constant of the S-curve in the beginning,
it kind of goes slowly and then it grows exponentially.
But eventually it kind of plateaus out.
I think that for new industries,
when the innovation and the industry is changing so fast,
you got to let it innovate so that you can learn about it before you regulate.
Do you see a moment where that's kind of what that's kind of the way they are,
right? Build, then regulate.
But do you see a moment where they do?
And the reason I mention this is because maybe it changes the dynamic of the race.
Because if you think at some point they will slow down.
That's right.
When the technology becomes more clear, you regulate.
When the technology becomes more clear
and you can see the landscape shape a little bit, then you regulate.
Now, of course, it's not an on and off switch.
You know, the temperature increases.
In the beginning, the innovation temperature and the energy is super high.
And maybe regulation energy is slightly lower and people watch.
But over time, the regulation and temperature are raised just like it's raising now,
because we have a better understanding of the technology today than we had two years ago.
Could you imagine if we were to shut it down two years ago?
We would have the worst AIs in the world.
Unsafe, unuseful, incapable, no industry, no benefit to society all at the same time.
I had one other related question, which is about distillation, which was just the idea
of whether you think where we are in the race and if China can't distill from us,
which some people think is stealing.
I know other people think distillation is not stealing.
How you how you think about that?
People distill my products every single day.
They take it and they strip it down to its raw parts.
They test everything.
That's called competition.
You think that's fair game?
You think distillation is not robbery?
That's called competition.
OK.
Yeah, you're allowed.
You're allowed to test somebody else's products all you want.
People strip our products down to, you know, down to nothing, down to bones.
And I prefer they didn't.
I really prefer they didn't.
I really prefer that nobody learns from our products and we have the benefit of just cruising
along.
But, you know, frankly, competition makes everything better.
And so I would, number one, if you don't like that, if you don't like people to use your
products, all you know, know your customers and disable the service to them.
And so so I'm I'm fairly certain if Netflix can solve this problem, you know, and they
know your customers and disable the service and don't give you access, you could probably
do the same for you.
Jensen, I know you have to run.
Your stock was up about one and a half percent, 1.7 percent.
Now it jumped on the news of the share buyback, not the open shell and century.
But I'm guessing you think open shell and century is the much bigger news today.
Yeah, really important news.
And the reason for it is, of course, we can't have a successful industry if the world doesn't
think it's built or confident that it's built and deployed safely.
And so so today's the open agent safety system is really, really important for that.
We've got the support of a giant industry to work with us.
I want the world to know that this industry has the technology.
It has the maturity, has the responsibility to do it right.
And we will do it right.
And so so that's number one.
I think we're going through the largest infrastructure build out in human history.
And we're we're we have the benefit of being a very central part of that.
We're generating a lot of cash.
We generate a lot of cash in the coming years.
And, you know, every single year, as we generate more cash,
we'd like to be able to return it back to the shareholders.
Wonderful.
Jensen, thank you for joining us this morning.
We really appreciate it.
So great to see you guys.
Great to see you.
And nice to see you.
Thank you, Becky.
Vampire Dracul here.
If you found value in this video, please smash the like button.
Let me know your favorite part of the video by commenting below.
And most importantly, don't forget to subscribe and ding that bell.
Until next time, keep it real.