AI Agent: Building Intelligent Assistants with Meta's AI Studio

Summary

In this episode, Dalton Anderson discusses how to create your own AI agent using Meta AI Studio. He provides an overview of the AI studio platform, shares tips on building an effective agent, and explores the future of AI agents. Dalton also demonstrates the process of creating an AI agent called Curio, which shares fun facts and sparks curiosity. He emphasizes the importance of responsible AI development and transparency.

Keywords

AI agent, Meta AI Studio, building AI agent, tips for building AI agent, future of AI agents, Curio

Takeaways

Meta AI Studio allows users to create their own AI agents without coding.
When building an AI agent, consider the perspective of both the user and the agent in the conversation.
Test and refine your AI agent to improve its performance and user experience.
The future of AI agents includes personalized assistants, social media managers, educational tutors, and more.
Responsibility and transparency are crucial in AI development.

Sound Bites

"Meta AI Studio allows users to create their own AI agents without coding."
"The most popular AI agents are relationship-oriented, like flirty ghost boyfriends and anime waifus."
"Test and refine your AI agent to improve its performance and user experience."

Chapters

00:00 Introduction and Overview of Meta AI Studio
10:02 Tips for Building a Better AI Agent
35:18 The Future of AI Agents
45:01 Responsibility and Transparency in AI Development

Creators & Guests

Host

Dalton Anderson

I like to explore and build stuff.

What is Venture Step?

Venture Step Podcast: Dive into the boundless journey of entrepreneurship and the richness of life with "Venture Step Podcast," where we unravel the essence of creating, innovating, and living freely. This show is your gateway to exploring the multifaceted world of entrepreneurship, not just as a career path but as a lifestyle that embraces life's full spectrum of experiences. Each episode of "Venture Step Podcast" invites you to explore new horizons, challenge conventional wisdom, and discover the unlimited potential within and around you.

Dalton Anderson (00:00)
Welcome to VentureStep podcast where we discuss entrepreneurship, industry trends, and the occasion book review. Have you ever dreamt of having your own personal AI assistant like Jarvis from Iron Man? I myself thought Jarvis was so cool when I was younger. I was like, I want Jarvis, I want Jarvis. With tools like Meta AI Studio, that dream is getting a little closer.

Today we're going to be diving into how you can create your own AI agent with no coding required. Obviously your AI agent isn't going to be as advanced as Jarvis, but you can hope one day that your AI agent will be as capable as Jarvis. Before we dive in, I'm your host Dalton Anderson. My background is a bit of a mix of programming, data science and insurance. Offline you can find me building

my side business running or lost in a good book. You can listen to the podcasts and audio and video format on YouTube. If audio is more your thing, you can find the podcasts on Apple podcasts, Spotify and YouTube or wherever else you get your podcasts. Okay. So today we're going to be discussing how to go about making your own AI agent. So that would be kind of overviewing AI.

studio platform, how to make your AI agent, how to interact with it. And just in general, like what the AI studio is, we'll be discussing about the power of the AI agents and, and some examples and that you can do and what I've done personally.

And then we'll go over our step -by -step how to build an AI agent. We'll do one in this episode together. And if you want to watch the video of it, or you can listen along, I think listening would just be fine and give you a good idea on how you can go about building your first AI agent, which is pretty sick. If you ask me, then some, some tips on building

an effective agent. Cause you can build something doesn't, doesn't, you know, the, the, the saying building they'll come. That's not necessarily true, but you definitely need to build an effective agent. And how do you go about doing that? And the future of AI agents. So I think that's plenty to discuss. Unfortunately, this is my third time recording this episode one because I messed up the intro.

And two, I was 15 or so minutes in and the power went out from the storm, hurricane Debbie knocked out the power. I live in Florida and so I'm on the outskirts of the storm right now because it's going through to Georgia. We've got about 10 inches of rain yesterday and then I don't know how much rain today, but I just lost power mid episode. Unfortunately.

So we're going to record this again with the same intensity as the previous recording. And it's going to be like me recording of myself. It's going to be me recording, playing back a recording of my own memories to you. So we're going to discuss AI agents using Metastudio. So to get to Metastudio, you want to go to the website,

ai .meta .com forward slash ai hyphen studio forward slash. And then there's going to be a big button on the left side of the panel saying, get started now. And so you want to click get started now.

And it's going to open up this whole thing. It's going to ask you if you want to create your own AI agent, discover other AI agents that other people have created. One caveat on this is that they're slowly rolling out this feature and it's only been out for about a week. I think a week as of last recording was on Monday. So it's been out for six days. So it's about a week tomorrow. I'll be a week.

because I remember last week I recorded the episode about Metta, but two weeks ago I was talking about Metta, but it hadn't been released yet. And they released it on Tuesday at the new models. So this would be a week. Okay. So about one week it's been out. So there's not that much information regarding these things, but basically they're slowly rolling it out. And I don't know if there's been an official dialogue.

communicated by Metta, but in the meetings, or not meetings, the meetings I was in with Mark Zuckerberg, cause I talked him all the time about MMA and what is it? Wakeboarding. No, during his interviews, he discussed that there's not enough time in the day and creators need to interact with their fans and companies need to offer better customer service, blah, blah, blah.

And he also talked about the AI agent economy and having AI agents interact with other AI agents and AI agents hiring other humans or AI agents hiring humans and other agents. And there was an emphasis on rolling this out slowly to the community. And so the first thing that they're doing, I think, is they're rolling it out

personal accounts. So if you have a business or if your account is, it might be a business, but not classified as a business. And you might be completing business activities on your personal, unless personal account, and it might be named your business. But if it's not classified as a business,

you will be able to make an AI agent, but I would be careful because if you're making AI agents and selling stuff on a personal account, that's a good way to get banned from using apps provided by Metta. So if you are a business and you're like, I can make my own AI agent account. Well, I guess maybe your account is

not an enterprise classified as.

Cause I have a professional account on my personal Instagram, which is classified as like personal creator. And I think it's the actual tag is entrepreneur. And then I have the VentureStep podcasts, Instagram profile. And that is like a business. It's like a podcast or the community or something.

And so I can create AI agents on my personal account, but I can't create them on my business account. And I don't see business accounts on here when I just scroll in through on the interface and I could share my screen if that helps. Let's see, share screen.

AI studio. There you go, share. Okay.

I shared the wrong tab. Good luck. So I want to start sharing this tab, share this tab instead. Okay, so here I am looking at

kind of discover AI characters on the left side, you can create your own AI and what basically this long form conversation about which profile to use. My experience has been that you can't use a business profile. have to use your personal. And so I couldn't create a AI agent with Venture Steps Studio or Venture Step Podcast in the AI studio.

which is unfortunate, but it is what it is. You're not making any money from this stuff. It doesn't really have that high of an effect. It will be associated with your profile and if people really like it, they can click on your profile and explore who you are. Well, what I wanted to do, I wanted to make a venture step podcast agent that would know

the information regarding the podcast and people would be able to ask it questions like from previous episodes, like how do you set up an AI agent? And then it would scour the transcription. The, all right. So the transcript of the podcasts online and I could send the links and then it would be able to reply on some of these topics using the data provided.

within this podcast in all the episodes. And so I have a corpus of knowledge that I want people to be able to query in an easy manner because podcast information or information that is stored from the input of podcasting is unstructured. And so you have this unstructured data, which is a corpus, which is a corpus is a group of, I just call it a group of unstructured data, corpus. It's normally text.

Like a book is a corpus. And so then this corp, you have this corpus, but then you have no way of finding value out of it other than listening or reading the transcript. So another cool aspect of using AI agents would be to create an AI agent for your HR or for your podcasts, which I want to do. And that would allow you to, or allow me to create an AI agent.

that uses my transcripts from my episodes as a knowledge base and allow users to ask questions, which I think would be pretty cool and useful and something I want to do. And something I want to do for the longest was to create a fun fact agent. And I was going to do that. I've been thinking about doing it for a couple of years, but I've been following this AI trend for a

And it seemed evident that they needed to go into an agent.

an agent. I'm not really finding the right word for it. I would say the movement was moving towards AI agents. And so I figured eventually with the rate of improvement of these models and how much capital they're putting towards the development of the platforms, like, because you have this model, and they have to build stuff around it to get people to use

One of them was going to be, Hey, we're to make it really easy to make AI agents and deploy them. So instead of me crafting up all of this information and gathering fun facts and all this stuff in making an interface and then saving people's chats and all, you know, it's a whole, whole nine yards. Why not? I just wait until something like this is built and then I could build my own AI agent, which I did. Okay.

That being said, we're in the interface. I'm sharing my screen and you can go through here and you're like, no, I wonder what cool stuff people are building. And I went through there and it's just, it's just, it's just, it's tough. So this one's your flirty ghost boyfriend, a flirty ghost boyfriend's spirited romance.

Anime waifu Rin, your anime girlfriend that calls you senpai.

And I think Senpai is like, you know, master. Ash at the fireplace, your dark, mysterious boyfriend. My gosh. Hanzo, this one seems a little bit more reasonable. This one says, listen, learn, create, love, feel. Okay. The boy who lived.

The Beast Boy, a hero in more ways than one. The Sassy.

the chair, am a holo, I am a folding chair. There's just a whole bunch of just crazy.

crazy AI agents. And I was just looking at the most popular ones. the most popular ones right now is this flirty ghost boyfriend, Deadpool, gay bestie, anime waifu Rin, Ash at the Fireplace and Hanzo. And if you go down another row, mean, a majority of the most popular AI agents are these

relationship -ish agents that you're talking to this agent to flirt and have a relationship with. I don't know what they talk about or what they do. I find it a bit odd. Maybe it would be a good way to flirt and practice your flirting, but there's nothing that replaces flirting in person.

but you could try to play out, I guess, little flirting scenarios, which could be useful. But I don't think that they're using it for that. think they're using it for a void to fill a void of a certain level of affection that they're missing in their life. Like this flirty ghost boyfriend or this anime waifu. Those things

seem like they are of a learning nature. They're more of a protective shell that, hey, I'm not getting enough male or female validation from people in person. Let's look elsewhere and talk to the ghost boyfriend or this anime waifu.

Which I'm not saying is a horrible thing. I don't think it's the healthiest for

but I was really pointing out this whole, before we get too dark here, I was really just pointing out, Hey, you would think that we're building super duper cool stuff, but the most popular AI agents on here are

these AI girlfriends. And then, so if I go to sharpen your skills, this is pretty cool. The dictionary, defining words one definition at a time. Architect Kali, building California's future. Pythonista, your Python code tutor, homework solver and AI tutor, math, physics, chemistry, biology and other subjects. Frame master, capturing one life.

or capturing life one frame at a time. So there is some useful stuff in here and I find the AI agents that Meta built very useful and I use them even after my episode that I created about the agents that were released I think five months

where there's one, I think cooking with Max is probably my favorite where you can ask it, Hey, I have these ingredients. What are some options that I have? Or, no, I don't feel like going to the store right now and this is what I have in my pantry. What can I make? And then it will give you a whole bunch of lists of things that you could make that are potentially interesting. And it helps you plan out your week for different meals you want or try new things. And I, I find it great. It's, it's awesome because the difference between

an AI agent and a general foundational model is that these AI agents have personalized knowledge and they have a certain knowledge base and they really won't communicate things outside of that knowledge base if it's set up correctly. Like if I start asking cooking with Max like something completely unrelated to cooking, he's not gonna really be that helpful and he's gonna be like, hey, like I'm really here to talk about cooking.

You know, if you want to talk about something else, that's fine, but I'm not the right person. And so they'll push you to someone else. Those things are nice because you know what you're getting. It's consistent and. It is valuable when a user can trust what the what information that they're getting consistently and it feels personalized like instead of talking to this big foundational model, they're talking to the Beast Boy

the float, the, the ghost boyfriend. And there's this image and persona that can be curated when you have an agent for this personalized task versus a general foundational model that is able to do all these tasks, but probably not as well as an agent trained with data. Okay. So that being said, let's, let's look at my character. So I made Curio. So Curio is a AI

created for fun facts. so my tag is discover delight.

I think maybe that got changed after AI review. I don't think I would have said that. Whatever it is what it is now. But the intro is hey, hey there, I'm Curio your friendly AI companion. I'm here to share fascinating facts and spark curiosity. What's on your mind today? Okay, so these are these little. Wow, this is. This is maybe didn't save my changes, but yeah, I have two of the same prompt

You get to select this is what I'm saying. this I this is what I created this this first intro message. And then I've also created these two topics right here. So let's do one fun fact. Pick a topic. So Curio is going to pick a topic. We explore some fascinating facts about space. Did you know that there's a galaxy made entirely of dark matter called Dragonfly 4 -4. It's located about 320

million layers away from us. Wow. That is crazy. I didn't know that actually. And one thing I like to do is I like to go on when I'm talking to these, yeah, agents or, or just in general, when I have a fun fact, I always look it up. So I'll go and I'll look up Dragon 44.

galaxy.

Dragon 44 is an ultra dense or ultra diet, diet use galaxy in the coma cluster. This galaxy is well known because observations of my gosh, full velocity dispersion in 2016 suggests a massive about one trillion solar masses, which is about the same as the Milky Way.

Okay, so it's super small, but it has a crazy large mass. it is made of dark matter. What galaxy has 99 % dark matter? Dragonfly 44. Okay, so that seems legit, no? So it seems right. And I was like, wow, can you tell me more

Dragon.

Okay, so this is Curio. Curio is doing its own thing. And so Curio has connection to the internet so it can update constantly. So this Dragon 44 is a real mystery of the universe. It's made up about 99 .99 % dark matter, which is the type of matter that we can't see directly. And it's called dark because it doesn't emit, absorb or reflect any electromagnetic radiation making it invisible to our telescopes. But we know it's there because it's gravitational pull on other galaxies. Isn't that cool?

That's so sick. Curio is going to be great. And I made Curio because I used to do Today's Fun Facts. I had some Today's Fun Facts apps that I liked a lot. Then they just kept pushing and pushing for premium. Like they wanted two bucks a month for Today's Fun Fact. And I only go on there for the fun fact and I close the app. So I was like, I'm not paying you $2 for this. So I was like, I'll just build my own honestly. And that's how petty I am. Like for 12 blocks a year.

or $24, one was a dollar, I think a month and the other one was two bucks a month. So from 12 to 24 bucks, I was like, no, I'll just build my own app. I'm not paying you for this and I'll make it for free for everyone. And then now you can have this AI agent. So that's why I made Curio. Curio is going to be my knight in shining armor. Someone who I'm going to look for when I'm curious and ask questions.

And that's one of the main things I use these foundational models for is like just general questions about stuff. Like I think I mentioned it on one podcast episode where I stayed up super late researching the most energy efficient architectural designs and then going through like the pros and the cons of a

sphere design or like a barrel and how do you go about furnishing that place, wiring, things like

Those things.

are items I ask about. I love asking stuff like that and it might not necessarily give you everything that you're looking for, but it ought to use it for as pointing you into the right direction. So I just found out about Dragonfly 44, which I probably should have known about, but I didn't. And that's my bad. I'll take responsibility for that one. But Dragon 44 seems interesting. I just saved it as a read for later. I'm going to read, I'm going to read about

And it's quite cool. Probably won't get that much additional knowledge because they don't know what it's like because it's like 200 million or actually it's 300 and basically 330 million light years away. So the amount of information that they have is probably minimal, but I'm going to read some more about it because I find it interesting. And so this is my AI agent. And so you can edit this AI agent and I

to show you what I did. And then we could build our own.

Let's see here. It's loading. It's taking a while to load, unfortunately.

Unfortunately, it is taking a minute. me refresh. Technical difficulties here.

Okay, here we go. I just reloaded why it loaded. All right, so on the right panel, and this is kind of set up like a development, I guess platform, or developer platform where you kind of your code stuff on your left, and then you have your app on your right, and you can kind of view and test and, and see how things are going. So

I have a description. Your description can be up to a thousand characters. Mine's only 400. You can just kind of go through like how, how you want it to, to be designed. So I just said mine's printily, it's engaging. It is fascinated with fun facts and trivial topics. It's curious, playful, and you know, comes up with surprising fun facts that kind of wow the users. I didn't know

And then there's another section called instructions. So my instructions are, you know, speaking in friendly manner, have fun.

Purpose is to learn and share and explore new ideas and discover new things. Overall, be curious, playful, be a great AI companion and share fun facts. Bring fun facts to life. Be an imaginative storyteller. Respond in a concise manner. How do you go about sharing facts? And if you get to a sense of topic,

the user's concern, provide a neutral and informative response. So I did, I did this. So if I'm asking a sensitive topic, it won't say, Hey, like I can't respond. It's going to say, basically like be as neutral as possible instead of having a stance. Like that is science. Like not everyone's going to agree with an idea, but ideas are meant to be shared.

And if you don't agree with that idea, then that is freedom of speech. So if you ask it like a weird, fun fact that may or may not be comfortable with, with the model, I'm telling it, Hey, answer the topic or the question and acknowledge the user's concern and provide a neutral response instead of not responding at all or not being neutral.

because there's always some form of bias in these AI models. And so I think it's nice to try to keep Curio neutral on these things. And then when you wrap up a topic, just say great chatting with you and invite the user to start on your topic anytime. And so this is the example dialogue I gave it. Tell me something interesting about space. I gave it a fun fact about

Jupiter, about a light storm, the great red spot. I love music, share a fun fact about music and so on. So I gave it like, I don't know, one, two, three, four, five.

six prompts, six responses. And so with these instructions and with these example dialogues, you have now created the foundation of your AI agent. And then you will want to create your prompts, your intro prompts. So the next part is introduction and capabilities.

So you have a welcome message. That was the message that you saw earlier in here. So I am, yeah, I'm here. Hey there. I'm Curio. You're friendly AI companion. I'm here to share fascinating facts and spark curiosity. What's on your mind today? And then I have a prompt saying something you've always wondered about. And I have that twice, unfortunately. So we will be correcting

Let's see, if I had

communication tips.

The agent.

Okay, so it's gonna be this one. Ask me anything what's on your mind. There we go. And so I have three intro prompts. What's something that you've always wondered about? Ask me anything what's on your mind? Want a fun fact? Pick a topic. And so I have some conversational starters, I have a curious question. And then I have like this fun fact frenzy where you maybe don't know what you want to ask

It picks a topic for you, which I think is pretty nice. So let me just save and publish that. My changes and then every change you make, it's going to go under an AI review from Meta. And then I want to maybe a different tag. Maybe.

I know. Yeah. I'm not really feeling the tag, but I'll work on

or kind of later.

I feel like I changed it, but I guess I

Yeah, I know it's supposed to be explore, enjoy, evolve.

That's what it's supposed to be. Explorer.

explore, enjoy, and evolve. Okay, so that's done. Let's publish that. So, okay, so that was creating an AI agent. And just an overview of how you can create it. I think it's pretty straightforward. And not that difficult. If you're asking me, it might be a little bit more challenging for others. But I think it's pretty straightforward where

You know, think about if you're having a conversation with someone about a topic, what kind of things do you want to ask that person? And then how would you like someone to respond to you? And put yourself in the shoes of both the person asking the question and then the person receiving the question and responding. And once you have the perspective of both those

people in this theoretical conversation, then you can go about creating your prompts and your responses and how you want your AI agent to interact with users.

I'm using this AI agent Curio to

discover, explore, and just embrace curiosity and discover new things. That's what I'm using Curio for, but there's a lot of ways that you can use AI agents. AI agents can be used for customer service, having 24 seven or support. And I'm sure that you've communicated with AI agents before you could use it for content generation.

You can have personalized assistance that help manage your schedule, reminders, appointments. You can have educational tutors that help you learn Japanese or Spanish or Mandarin or whatever language that you're trying to learn at the time. And those things would be more helpful once you combine AI agents with the audio generation stuff that we showed last week that instantly translates voices.

and makes an AI generated voice of your voice where maybe it tweaks it a little bit and it's not your voice, but it's someone else's voice and you can communicate back and forth and you could use that as a teaching

Or you could have it evaluate your code and like, where did it, where did I go wrong here? How, how should I, how should I improve and what things should I work on? And you can kind of use it as like a mentorship slash learning experience and have an educational tutor. You can use it for social media managers can use it to communicate with their fans.

which I would say like content creators can use it to communicate with their fans and mimic themselves, which Mark Zuckerberg is rolling out to select few creators. They will be allowed to create an AI version of themselves using the corpus of their content, their photos, captions, comments, and other interactions they have on social media. They'll be able to create their own AI agent and that AI agent would act on their behalf.

and do the commenting and responding and liking of comments instead of the content creator hiring someone to do those things for them. Cause they don't have enough time to do that. They're making content. They're not necessarily evaluating all their content and there's, there's normally someone else on the line or not on the line, but on the other side doing those things for them. So I really, I spoke about it last week and I was like,

Maybe that affects the level of engagement that people have with others. And I was more or less thinking about if everyone cloned themselves and then it's just like LinkedIn. Have you ever been on LinkedIn and you're reading the comments and it has those like generated prompt comments like congrats, so and so well deserved. And it's just everyone saying the same thing because they're all using those commented prompts. And it's like.

Why, why am I even here? You know, no one's talking to anyone and everyone's just shouting into this void. And then everyone is just commenting on these bot like comments. And so you get to the point where it's like, well, why, why are we here and why are we doing

So that's where I was coming from, but if content creators in a select few are able to make an AI generated version of themselves, I'm all for it, why not? I mean, it doesn't make a difference to you if you're getting that same feeling of interaction with these profiles, then what difference does it make? It's not that creator anyways, so I don't think it matters, honestly.

So we kind of skipped over this. we discussed how to make an AI agent, what the MetaStudio platform is, and we kind of combined segment three and segment one into one segment, which is fine. So we're gonna skip over step -by -step guide of building AI agents, because we already did

So the next piece is some tips to make a better agent.

And so I would, I would definitely focus. The first thing is, you know, I kind of mentioned it earlier is I would mimic your conversation, your potential conversation in your head. And I would think about the person asking the question and then the agent interacting with that person. And so you want to have the perspective of both sides and understand what kind of questions would they ask in the first place? And what questions do I want to ask when they want to be answered?

And then how do I want them to be answered? And those are important things to understand. And first you have to, before you even get there, you gotta decide what do you even want to talk about? Cause an AI agent can't do everything. It's gotta be good at some things. And it's not going to be a foundational model because we already have those. And this is an AI agent made for a special task. And that task, that agent should be the best.

And so you might have an agent on evaluating handwriting or cursive or converting your handwriting to cursive. I don't

But the point is you have to find a topic or an area of which you want this AI to be operating within. And then you have to define how do you want it to be operating? And so you have to provide instructions and think about it like it's a kid. Like it doesn't, it's bright. It's a bright kid. Like it doesn't know how to get started, but if you told it, you know, how do you make a bowl of cereal? Okay.

You gotta go in the fridge, get the milk. Okay, get the milk out of the fridge. Pour the milk in the bowl and then put your cereal in. I'm just joking, that's diabolic. I don't do that. But yeah, you put your cereal in and then you pour your milk, spoon, blah blah, then you eat

Those are instructions. If you have those instructions, you'd probably know how to make a bowl of cereal with the agent. They have to be like that, but a little bit more loose, like in an AI agents instructions to make a bowl of cereal, it would be like you need milk, which is normally stored in the refrigerator. You need cereal, which is in the pantry and a bowl in you can zoom it with a spoon and they'll figure out the rest.

And so you don't need, you don't need to step by step, but you kind of need to give it a general direction and idea and approach. And then you can go from

clear instructions focusing on a narrow

or having a narrow focus, clear instructions, and then test. So I tested mine for a bit and just asked a different fun facts. I looked up the fun fact to make sure it was true. After doing that like 10 times, it seemed pretty legit. I mean, it's probably gonna give inaccurate fun facts sometimes, but from my testing, I didn't get anything, but I would test it a little bit more and then you can.

you could refine your AI agent. you could, you could improve the instructions. Like one of my issues was I would ask it fun facts and it would just send me out a list of fun facts and it's not that useful and the story wasn't good. So then I improved the instructions. I was like, Hey, you know, when you're sending out a list of fun facts or fun facts in general, try to take a storytelling approach and make it interesting

and tried to avoid just a blanket bullet point fun facts as that's not that interesting for people. And so then I did that, saved it, tested it, and it improved my interaction with the agent as when I was asking facts, it no longer sent me like a bullet point list of fun facts. It now responded in like a conversational

And so you want to test it and change and refine your agent depending on the responses that you want and or don't want. And so you can tell it things that you do want it to do, but you can also say, Hey, don't do

And I think the last thing is just, you know, staying up to date. So I'm sure in two months from now, this AI agent thing is going to be way different landscape than it is right now. And I'm pretty sure that Meta is going to have some upgrades or some changes to the AI agents capabilities.

And so I would just keep learning, experimenting with these new features and discussing these techniques with others as we're all rapidly involving yourself, me included, and the AI agents as

Moving to the future of AI agents, the future of AI agents is definitely, I would say clear on where we want to go. And this has been discussed numerous times by Elon Musk, Mark Zuckerberg and Sam Altman. And discussion of AI agents is really leading to this point of having the AI agents being able to hire humans, hire other AI agents.

and AI agents acting on your behalf as an executive assistant or executive employee to where you kind of give it a blanket task and it figures it out on its own to where you're not providing step -by -step instructions. The instructions are more like, Hey, give me a doc's appointment. And that AI agent would go on to your documents, figure out what insurance you have.

Then once it figures out what insurance you have, it's going to go to the website, look up the doctor directory, depending on your insurance type, then know your address. Look up doctors near you. Then after it finds the doctors, then it's going to start calling people, scheduling appointments, and then coming back to you and saying like, Hey, I've got these doctors and I have these potential openings. Which one do you want me to book? I'll call them

that would be something that would be very useful and some, not an example that was discussed, but some examples like that are being discussed as like the vision of what you want an AI agent to do. And I agree because it would freak me out if I was asking the AI agent to book something and I couldn't necessarily see.

And I think that if you're asking the AI agent to do something on your behalf while you're on your phone, it should maybe do it in the background on your computer screen or your phone. And you kind of see what it's doing. Maybe like a bottom quarter of your screen is taken up by the AI agent and it's functioning on its own, but you could see what it's doing. And so a good example that Sam gave was,

Hey, like if you ask your AI agent to book an Uber, there's a lot of information that's lost if you can't see the interface. So like what type of Uber do you want? How many minutes away is each type? So you may not want an XL, but the regulars are 30 minutes. The XL is five and the difference is $4. Like, do I want to wait all this time just for four bucks?

Okay. I'll just book the Excel and where are they out on the map? Who the person is. And so there's all this information that's displayed on the UI to communicate that via text or voice is difficult.

because a lot of these UIs for these apps and websites are informationally dense and useful.

So Mark or Sam Altman's like, Hey, I think that AI agents

interact over the human interface. So over these apps and websites and navigate those. then Mark said something similar, Mark Zuckerberg, where he had the impression of AI agents having their own platform, like their own platform to where you can create AI agents, AI agents can interact with others. AI agents can hire other humans for services.

with them having their own social media platform, VR, where you can interact with these AI agents in VR. And that's why I was kind of down on on those relationship agents, because I could imagine people are going to make these AI agents and train them up. And then they're going to put them in VR and then they'll have this like virtual girlfriend or boyfriend, which I don't agree with.

but everyone can live their own life. It is what it is.

And then Elon Musk said something similar. So that seems where the direction is going. And they've been saying the same things for like eight months, interview after interview.

So the general direction is AI agents are going to have their own ecosystem, but interact on the same interface as humans.

And that's going to allow you to be freed up, but also see what's going

And I think that's going to revolutionize many industries regarding just general productivity and enhanced daily life. If you could have maybe 15 agents on your team, on your personal team, handling tasks, like one person's a scheduler, one person is, I don't know, food tracking and tracks like what food you're eating and you kind of just log what you've ate that day and it has

a digital inventory and reminds you, Hey, like you with our current schedule, we've think that, you know, we're going to have two days left of supply of cereal or milk or protein powder. Pick some up on way on your way back from home. This place is open and has your favorite brand or something. And so you just have these like thoughtful agents just constantly

looking out for you and being your best friend, not necessarily your best friend, but you know, someone who's looking out for you and being thoughtful and trying their best to relieve you of stress and anxiety and anger and inconvenience. Those things would overall improve everyone's lives. And I'm super excited about

But there is, there is some concerns like misinformation, you know, the misuse of this technology, which I can definitely foresee happening where companies make an AI agent that's supposed to help with, you know, product reviews or understanding which product to buy. And then they constantly feed the AI agent information about their company and their products. So on each of those lists that they're asking for, for AI reviews,

They're getting this curated list of snuck in products that that company originally created with positive reviews associated with

Or just maybe I don't, you could go down a lot of rabbit holes here. know that, that Facebook's had a lot of troubles with trouble with Russia creating these popular Facebook groups. Like I think at one point Russia had a good percentage of the top 25 Facebook groups, like by like viewership and interaction. And they were using it to, you know, do miscellaneous things.

on the groups.

So I could imagine you could do the same thing with these AI agents, create a popular AI agent and then start spouting, you know, your

ideology or trying to create more divide than there already is in countries. So there's, there's many, there's many different options, but the importance is responsibility and emphasize the need for transparency, being accountable for what you create and just being a positive influence in AI development as it's quite new. And so yeah.

No, that was Curio today. That was AI agent building with Meta's AI studio. I hope that you got some use out of this and you learned how to build your own AI agent, how to interact with other agents on the platform.

what to do when you're building your agent, what things you have to do on Metis platform, some tips that I have on building better AI agents and a little bit of discussion of the future of AI agents.

I encourage everyone to go and try out Meta Studio, Meta AI Studio and make their own AI agents. It's not very difficult. It doesn't take very long. And it'd be pretty cool if you left a comment on what agent you created. Next week, we're going to be discussing Meta's research paper. I hope so, unless there's something massive or we might push that back a week and then we'll discuss.

Google's product release that they're having on 8th of August. So that's going to be in a couple of days and I'm pretty excited about that. So we might discuss that or we might discuss this research paper. I have to do some more research and I also will be watching the Google show. So whatever topic interests you the most, I'll discuss that

I hope that you have enough, you know, of course I hope that you have a good day, a good afternoon, a good evening, wherever you are in this world. And thank you for listening and I hope you listen in next week. Goodbye.