different from the the fancy UI that we had in the video, so you can see the Smart Zoe toggle button down there.

So right now Smart Zoe is turned off. Let's just, let me just quickly introduce what the game looks like. So the life simulation

game, I'm simulating a family of three, David, Grace, Katie. Um, they are family of three. Grace is the mom, let's move on to

the David is the father, um, Katie is the cute daughter over here. Okay, cool. Um, it's a coincidence they are famili

of K. All right, so, um, now let's play the simulation for a little bit before we turn on this Mar and you can see the

action queue down there. That's the action queue. So this is running the traditional AI simulation logic. So it's going to choose

what to do based on some rule-based AI. So now Gracie is talking about the choice of walk and so forth and you can see all

the other characters are doing the same thing and you can just sit, relax and see what these characters do, but that's still traditional AI.

What's the one biggest limitation of this traditional AI? You don't really understand what they are doing. They

are just making some random actions, something after something, but you cannot tell why they are making those actions. So let's turn on Smart Zoe and see what

the biggest difference is. So now we turn to Smart Zoid and what happens is on your single GPU now it's going to run LLM

inferences with small language models. Can you pause it here? So now the big difference is you can see the, and now

she is trying to study professional textbook in SE app. You can see the inner thought inside the Zoe. Now she's saying, I

wanted to, um, make the most of my time by studying while organizing my book. Okay, that's cool. Let's have one more action.

Um, so now, um, cancel the action, just stop the action. So AI is running behind the scene to choose an action, next action.

What she is doing, she is trying to share to great places to eat in town, sounds like a good idea. Why? Um, cool. So she wants to

treat David and Katie to some new food experiences. So every action is going to be accompanied by this kind of thought

happening inside the zoo and there's nothing scripted. Everything is going to be generated on the fly with AI, so

right, so as such. So let's start using the uh infinite characterization customization I just mentioned. So click

the pencil button over there. So that's the button where you can literally type in whatever you want to

specify as an extra characterization customization. So let's see, Grace won, she is the mom. So let's say, let's say you

love your daughter so much. Um, Katie, you love your daughter, Katie, so much. So then you can see that the mother, the mom

becomes a bit obsessed with her daughter. Let's see. All right, so um now we specified it, let's run the

simulation. Okay, so now, um, once this extra specification is entered, now the SRM is running the next inference. It

takes like a few seconds. Um, so now she decides to ask them what they appear to be and if you look at the this

the rationale behind it, curious about Katie's aspirations, what she wants to achieve in life and that's basically very consistent with

what just typed and you can specify this for every character to your plane and that way you can see a completely different way of

them living their lives. Well, I want to show you just one more before your laptop. Let's um fast forward time to

sometime close to midnight. I want to show you to um show show them writing their Diaries. All right, so okay, close to

midnight, can you make Anne sleepy? All right, so they're pretty sleepy. So um so when they are sleepy um

and near midnight, they'll probably go to bed. When they are going to bed, just before they go to bed, they will think

about how their day went. So now let's pause it over here. So Grace's inner thought, the very last inner thought at

the end of the day is literally the diary and language water wrote. So let's look at it. This morning, um blah blah

blah, she was talking about caffeine addiction and you can see that in the middle of the um inner thoughts, I

started organizing my bookshelf, had some fun conversation about local age. That's what she did in the beginning, right? So

you can see that the diary is summarizing what they did and also at the end, they also feel like I I didn't

finish this, I want to do next time and all this diary is going to be fed to the

language model so that the agent, this PC agent, can plan the next day better than they used to. So well, um with that, thanks

Joe, um thanks for running the demo. Um let me go back to the last slide.

250108 CES 2025 크래프톤 엔비디아 간담회 인조이 스마트 조이 기술 시연

Scripts