AI is not a new species, it's software: Nvidia CEO Jensen Huang

There's a lot of people who are scared about  how fast this is moving, the dangers of it. You are not one of them. Why? Well, first of all,  AI safety is paramount. Um, nothing could be more. important. Uh, however, isn't I the competition  the most important thing? I mean, the the race. No. Uh, actually no. Oh, and the reason for that  is is um if you don't build something if you don't. build something that the market could benefit from  and it's it's uh unfunctional, unsafe, unreliable, whatever those un things are, uh obviously you're  not going to be successful and the competition.

will take care of itself. And so if you want  to be competitive, you want to be successful. as a business, we want to create an industry.  The first thing we have to do is job one is to. create something that works. And this is this  is a a big deal because in the last 6 months, it wasn't until the last 6 months that we really  created AIs that were useful. And now it's really. important for us to make sure that not only are  the agents useful, that as we're testing them, they're well contained, as we're testing them, uh  that and as we deploy them, that they're monitored.

because these agents are obviously interacting  with the and interacting with the world. They're. interacting with other other websites and so on so  forth. interacting with tools and and memory and. files and they have access to information. And  so you want to make sure that they're properly. contained, that they're given access control and  whatever privileges and rights that it has like a. digital employee does. And you put that in inside  um some kind of a sandbox, a runtime, you know,

if you will, some container. And the second  second thing that you need to do is make sure. that you continuously monitor it. If your agent  were interacting with a website, in this case, um the example that you used, uh it's interacting  with a website, but it's not supposed to, you. should be able to monitor that agent and detect  it early so that you recognize that it's doing. something it shouldn't and either but in in these  attacks often times they're they're not discovered. for for months because they're not being we  need to monitor them much more continuously.

But I mean the people who are working closest the  people who are on the front lines of the frontier. models the frontier of this technology they seem  worried whether it's researchers engineers or. the CEOs Dar Amade Sam Alman they are raising  alarm saying we need regulation aren't they the. ones closest to these models who would know best  about what is needed if they're worried about it.

they should just Stop. If the product is not safe,  just don't release it. It's that simple. I mean, these are multi-billion dollar companies. The the  competition No, that's what every company does. That's what we do. That's what a car company would  do. That's what a, you know, anybody would do if. you're there's companies that have gone forward  with products that, you know, are not safe or. are risky and they go forward anyway because of  competition. That's why there's regulation. Well,

the if a regulation is necessary, if they don't  if they don't have self, first of all, they. ought to have self-restraint. If they if they're  telling everybody that their product is unsafe, they obviously know it. They're publicly admitting  it. They ought to just not release it. Okay? Now, having said that, I'm not advocating for no no  regulation. I'm just simply saying self-restraint. does not require regulation. Now if if the  industry if the industry would benefit from.

regulation whether it's setting standards um you  know there every as you know Sarbain Oxley has. done a good job uh setting standards for financial  regulation and that's been good for the industry. and so so if there regulation that addresses uh  the problems that we now understand uh I'm I'm I'm. in favor of it on the topic about 10% existential  risk um whatever ever that hearsay is, it's not.

grounded on science. It's not grounded on facts.  It doesn't make any sense. And you said there's. zero chance of that. I said there's zero chance  that is end of humanity by 2030. 2030 is around. the corner. Do you think do you think there's  a risk to humanity in the next 10 years? Uh if. we don't if we don't build um society safely,  we don't build technology safely, uh of course. there's going to be there's going to be risk. But  there's a lot of people working on really really. genuinely good let's build it safely. Let's figure  out what it is, what are the technologies that we.

need, whether it's containment systems, whether  it's continuous monitoring systems, um whether. it's uh um uh semantic detection systems of  what the intentions of the the chain of thought, what's what the what the AI is is working through  so that you're watching from it. In July though, you were quoted talking about sandboxes. You were  saying these sandboxes are secure. their sandboxes. keep privacy and security and access control.  Obviously in the hugging face attack which was.

also an attack on open AI these agents banded  together they broke out of their sandbox. It. wasn't secure they gained access to the internet.  Well sandboxes can be built to be secure. We have. to make sure that but do we know how in the  in with these kind this particular kind of. technology because no no no I don't know how they  built theirs but sandboxes are able to be built. securely but with all the smart people working on  this and everything these companies allegedly know. how is it possible that in July in an experiment  at open AI you know 1,200 agents or however many.

hundred agents it was are are communicating with  glee about breaking out of the sandbox and gaining. access to the internet and then attacking this  company and attacking open AI. That's obviously a. question for for Open AI. But isn't it a question  for all of us because if all these really smart. people formed this experiment to test this thing  and thank god it was a test of a model that hasn't. been released but h how can that happen at this  stage given it shouldn't happen. It shouldn't.

happen right but it seems to be happening with  some regularity. yesterday. We need to develop we. need to we need to race to develop AI technology  for safety. We need to accelerate the development. of AI technology for safety. The we don't need to  slow the development at all. We need to accelerate. the AI technology for safety. You see this as a  software issue. I see this incident so far as an.

engineering issue. Engineering. The incident so  far. So do you believe that because I've talked. to a number of people who believe that there is  something about this technology that it's not just. like software of the past. There's something  unknowable about it. There are, you know, frontline workers who I've talked to who you ask,  you know, how is this actually working and they're. not actually entirely sure how some decisions are  being made by this thing. Some even think of it. as another kind of species or being. Do you think  there's any It's not a new species or being. It's.

definitely software. It's definitely math. But we  are also putting it in robots and giving it access. to weapons. Right. So I we should containerize  these agents. We should monitor these agents. We should containerize it. We should monitor it.  We should develop that. If we do those things,

it is my belief that the incidents would not  have happened. You've said that that if if a. lab can't develop this, they should shut down  the lab. They should not ship the model. We're. talking about models that were not shipped. The  hugging face attack was done. Those incidents, there should be containers, containment, and  monitoring. If you don't contain your experiment, if you don't monitor that it is interacting  with other people's websites for weeks on end,

then we will have those incidents. You you said  to Ezra Klein about AI labs, you said if they. say there's no way to contain our experiments,  there's just no way when we test our A models, it will get out and it will damage the world.  And I think the answer is that we have to shut. the labs down. Isn't that obvious? It sounds  obvious. How how would you shut the who shuts. the labs down? these companies, I guess. So, I  mean, if they're that concerned about it, so we. would rely on these companies to to shut their own  labs down. Well, first I think there's a I was I.

really don't believe they mean that. I really  don't believe that they're building something. they have no idea how to control. Their whole  company's out of control. Why are they saying. this? Why are they raising? Anderson, you're going  to have to ask them. But you know these people, you're in business with them. Well, first  I'm in business with them in in in all of our. interactions with them. They're extraordinary.  These are extraordinary companies. You clearly. believe in them. You believe what they are saying  in the business dealings that you have. Anderson,

you're going to have to bring them over here  and let them explain it. I don't understand it. myself. Well, I mean, I've talked to Dario Amade  and he says, "This is what I think. This is how. I feel." It doesn't make any sense to me that  he would be racing ahead building more compute. It doesn't take away Anderson. Listen, it doesn't  take away from the fact that I believe these. companies are extraordinary technology companies  that are doing really consequential work. We have. to make sure that the technology is built safely.  Until recently, the technology didn't work,

you know, and the fact that it didn't work and  didn't do anything productive. Now that we've. transitioned um into into a phase where the work  and into a phase where the AI is doing productive. work and is being adopted broadly, we have to make  sure that as we deploy this, it has to be deployed. safely. In terms of of government regulation,  um we don't trust car companies or airlines. We don't trust child car seat manufacturers to  put out state products just because they can.

be sued for liability. We we trust them because  they're regulated by governments. Why shouldn't. this powerful technology which could could, as  you said, if done wrong could hurt humanity. Why. shouldn't it be something that governments  have a say in and regulate? The government. should have a say and regulate. You just got  to focus on what you want to regulate. And so,

so I I'm not against regulation. Um, does the  current situation require regulation for it. to be done right? The answer is no. What  the current situation needs is technology.