Three Biggest AI Trends for 2025 & We Answer Your AI Questions
42:00 Share

You are not ready for the A, I, advancements coming in twenty twenty five, A I inference agent's behavior.

Well, Kevin, we have to tease people just a little bit to get them a sense of what's going to go on without telling them everything.

Okay, you're not wrong. Gavin AI enthusiast can't believe these three amazing developments much.

much Better. Plus you have A I questions. We have A I answers.

Okay, enough with this. Listen, IT is thanksgiving week. We are extra thankful for all of you is why we wanted to do a special ask us anything edition of the podcast.

all of that and more. Its A I for humans.

We are here, Kevin. Twenty twenty five is going to be a big year for A I. And I think the first story we should kind of make sure people get caught up on is A I agents. And I I really think that A I agents are going to be the thing that next year, every major and more important, maybe every business is going to try out.

We should take a step back and maybe to find what an agent is and exactly how IT works or how supposed to work in the months to come.

So let's talk about that. What isn't AI agent exactly? Well, it's essentially an AI that can go out and do stuff for you.

So if you hear the term AI agent, what IT mostly means is not somebody putting on their glasses and say i'm going to make a deal with hollywood, going to sell you gain for a hundred million dollars, your spec scripts gonna get sold. That is not the agent we are talking about here. We are talking about something, having agency, meaning that IT can go out and do something.

So anthropos computer use protocol, which allows the computer to do things for you, is the beginning stages of this. And this is what sales force, microsoft, all these large companies are kind of selling as the first real business implementation of A I. And that you can imagine a world where you're running like a hot dog stand because that's what we do.

And if you need to go buy bulk meat, you need to go buy bulk meat and you want a deal for that bulk meat, but you don't have to handle back and forth with mr. Book mean, well, this way the AI agent conceivability can do IT for you and get you a deal and just come back and tell you how much is for that is the perfect incapable of A I agency in my life? Yeah.

there's no Better way to do IT. Actually I think it's men and let him know for all your meat negotiation, if you need to hug le over haggas, there's nothing Better than an AI agent. If you mess with A I today, you probably are aware that it's it's very capable at a lot of things.

You can feel magical at times. For me, where IT tends to fall apart often is when I have to be a human in the loop and connect very basic dots. I know that I can use A I to go research something.

I know I can have A, I take that research and format IT in a certain way, or write a script about IT or ask technical questions about IT. I can do all of these things, but I am still ultimately spending time and energy orchestrating all of this. And while it's one facet of the agent's behavior, it's the path that has been most excited, gave in the notion that I could say, here is the task that I want to accomplish an A. I can break that down and then spin up little instances of itself to go accomplish all of those things, to be its own conductor, and do the things that I already know that I can do, but to autonomously. And that is to, I think this is going to be a massive unlock for .

a lot of people. And just to be clear, there is a massive back in being built for all of this right now strike, but just announced the fact that they're gonna low A I agents to be able to charge to take payments, which is a pretty big deal into Kevin's point. I think if you're out there and wondering like what the hell I would do this, well, you could be something that right now i'm going through the process of like submitting to colleagues for my daughter, right? And like that is a nightmare of process and that there's all these things got to check off.

I think we went back and forth of the financial aid office for like fifteen times what would be so greatest to have something that you could just like your agent could interact with their agent and come back and forth and only have a question if they're something they couldn't solve. Now we're gonna some bumpy roads because it's not going to be perfect in the beginning. And to your point, like there's definitely to be some places that things fall off. But I think even if the frontier models did not improve from here on out agency is the place the twenty twenty five is going. You're really onna.

Look at if you look at the the naysayers out there, they'll say, again, scaling is slowing down. These models aren't getting more capable, but the reality is that there's a lot of capability in the models as they exist today. And agents behavior is just one of those things that can really showcase that.

I think the other piece to that puzzle is something you and I talk about often here on this podcast. But we're really going to see IT doubled and tripled down upon in twenty twenty five. And that is inference compute or reasoning time at inference compute, which sounds very scary.

If I met the thanksgiving table and somebody bring that up. Am I running in fear? Gavin and I throwing stuffing at them? no.

And I think IT, here's the thing. The important thing is georgia, the person that listens to this podcast or watches youtube video to kind of be that can do IT to help people understand this stuff. And in inference, compute sounds complicated.

When you think about just the two terms, IT sounds like data from the start track and saying something to you. And you're just like, I don't understand what this is the very simple definition. And I got this from ChatGPT, which is a great place to go, the brain power required for the A I to think and respond in real time.

So what that means is the idea that inference compute is different from the original compute to train a model, because that is the actual processing that happens when you ask A I A question. And the big thing that has changed and will continue change next year is open the eyes of one reasoning model. And other reasoning models are now able to spend that compute thinking about stuff rather than just like spitting out and answer right away.

And we as humans do inference or runtime compute all all the time time. I mean, if you listen this podcast whenever I ask Gavin, like.

what's your favorite thing about track hell at me to say feet he loves, we all know he a fish, an auto was, I am a two beer, efficient at all .

play that you like that that do you to flick the two beers? Do you want to just, do you want to see if they make noise like a wine glass.

if you want to try and to control him like gratitude, but two ears clt .

that chat to be a little soundness. Of course, Gavin wants to write truck, and that's the answer. That's what i'm trying to say. If you ask Gavin right off the rip, he may say feet. But if you give them time to think about IT to reason his way through, the answer you can just see gave in picturing truck in various states of, you can see Gavin emerging from the bug right behind short arms, developing him almost like a scene from titanic. You can see I M taking the time to reason through the answer and give a much Better answer, which is that he wants to write track.

But I don't want, I want everybody clear. My version of this is ratti, where I am on track head controlling IT from the two beers. Not what Kevin is inferring in his own inference compute, let's be clear. So right now, influence compute is usable through if you're a paid user of ChatGPT, you can use the one preview or one many models or like we just talked about last week, deeply a which is a chinese company that has released a free preview for reasoning model. And I really do suggest, like everybody, our audience, make sure you go out there.

And if people ask you at at thanksgiving or in the holidays, what is going on to say, I stuff, if you seem to be up on IT, this is the place that you can go and show them, hey, this is the difference from what to ChatGPT used to be. Now we can do this thing. The one advice I always give people is don't tell IT to think about something. Just give IT a really hard problem with whatever you do, if you're lawyer or your whatever a strike vision at to give IT a hard problem and watch IT go through the process of reasoning and see what IT comes outward.

People have done evaluations on the best prompts for reasoning models. And when you tell them to think critically about something or to take their time, those tricks which might have helped with basic eleven quies in the past, do not apply here by default. They are trained and ready to think deeply about something, to break complex tasks down and the step by step things.

So I think when you start to combine that inference time compute with things like agent's behavior, suddenly you can say, gave and I want to spin up a podcast or I need a business plan for a smoothly shop, or I want to rat tat to we shrek, what does that look like? And not only can I think about all the steps required, IT could spin up agents which are given a task, hey, you're the the C. E.

O of this smoothly shop, you need to borrow a business plan and have a this that's together. Hey, you're gonna go out and look for real estate for to put this place. And each agent is going to reason through what they need to do.

I think most computer users are used to make a request to hit a button and save for some rendering time. They get their result. They are used to interfacing with with a computer on this timeline.

And now for a very broad audience, building agented behavior with a chain of thought or inference compute, it's going to be make a quest. And I might see you in a day, I might see you in a week or maybe a month from now, you'll come back with my task complete. And that is, I think, going to be foreign to a lot of computer users.

That's like my dad. I maybe see him in a day, maybe once in a week, maybe once in a month. Sorry, dad, I know you say this.

so you prompted with a request like papa. I .

would love to get .

papa h and let me, let let me go get some cigarettes.

not get back to on one back. That's right. By the way, you is a small thing with age is being king of that. Microsoft released A A product called tiny troop, which even if you don't know how to do programme, is a github project.

But you should least go read about IT because when IT is, is basically a little series of agents that allowed you to put them on to anything. And it's just fun if you have any sort of technical back in, you can kind of go and play with this. But even if you don't read up about IT because I kind of gives you a sense of like what that means to have to these two agents talk to each other.

And you know, Kevin, I have talked about some version of these agents on the show really for a year plus now, but these are just getting much more capable as we go forward. Alright, we should type about the third big thing, which I kind of think has a lot to do with the world that we spend at time, our time in, which is media, right? And I think the third biggest story of twenty twenty five is going to be A I video and audio tools really fully coming into formation. And I think we've covered this a little bit with ban athlit talking about these things. But but IT feels like to me, we have hit a real moment of sea change in the media business where not only are the tools getting Better, but holly's wood in the people in the music business even are, are really folding into this space.

IT was about this time last year as well given where we were looking at at like next generation text to video models. And there was really only one, but arguably one of the half that had is excited, open a eyes saw still not out. And IT doesn't really seem to matter because there's ten different competitors all gunning for that space.

We said new music generation and sound effects were going to get Better. They certainly did. Now you can go prompt to album very fast or ask for sound effects and eleven labs of top to you voice cloning their puppy of two d figures as 3d characters， grabbing your motion from basic web camp capture and applying that to a model. I'm still surprised every day that these tools are coming out and there's no sign that, that side of the business is slowing down.

So like runway, one of our favorites just released expand video recently and then they also dropped frames, which is an image generation software model. And and I think dream machine just launched a full fledge image and video update today when I think you're see here is probably three to five different companies all attempting to be like the creative studio space, right? And what that means is they'll have a video model theyll have a photo model or maybe y'll integrate open source models and try to be that as well.

But what people want to do is be the dream experience, right? And I think that one of the biggest difference now is you're seeing full fledge tools and not just one of things. You have to come a piece together in different places. And I think probably like the bigger a eye companies you're going to see, the bigger A I studio stuff kind of consolidate into these things as well.

So if folks are hearing this game, and we do talk about these advancements all the time, but they don't know when or where or exactly how to leave in and get their hands on IT, what would you recommend is a good first stop for someone who wanted to to make an image or try a video or make some music IT .

really depends on if you want to pay money or not, right? Because I think in this instance right now, the best services really are paid, especially the AI in video space because it's expensive to generate these models and can get really get up in video wise. My experience with mini max, which is a chinese companies you have to be okay with, that is the best right now as paid model goes for video and then for audio.

I think you and I both kind of degree that sono is probably the premier uh, audio model of audio is also very good, but both of those are paid now. They are not expensive. You can into each one for a month for like ten bux each, right? And you get a certain limit amount of generations.

Video, gram is one of my favorite still or image models, and it's free to try, but IT is the best overall image model, I think so I would say that's my favorites. And then we talk about peno kio computer. If you are somebody who has technical somewhat of a technical background or even a little bit, you can install a bunch these open source tools and and learn a lot over thanksgiving weekend. I feel like .

this is not an added is a free piece of software that lets you run all sorts of other software, mostly free and local running. Meaning on your computer, I recommend everybody poke around with anything involving flux. Flux is another image generation software, but they just launched flux tools, which gives you some very powerful capabilities if you have an image and you just want to change the text in IT, but you wanna to retain the style of that text, now you can do that right in line.

If you generate an image and you want that character, but you want them wearing something else you can in paint on IT, that means you just kind of sketch over IT and say you want the clothing to change. These are basic tools which have existed in other pieces of software for a long while now. But flux is a very powerful image, uh, generation software. And now IT has those capabilities built in so exciting stuff. Never a bad time to leave in, I think, to any of the stuff.

no. And and before we move on, I do you want to shut out a very great video that if you really wanted, just catch up on where the state of the AI businesses right now, it's not too technical, but it's technical and absolutely, you can impress people with your knowledge. Alexander wang, the C E, O of scale dot AI, which is the company that for the last ten years has really been specializing in data sets in A I.

So they really provide a lot of the data to AI companies to help train these models, has about a twenty minute video. Uh, his keynote at his conference where he kind of go through the current state of A I A little bit of the history, but then really like what some of the chAllenges are going forward. So take twenty minutes.

Watch this video over the weekend. We'll put a link to IT in the shower notes. IT is very useful just as a catch up or as a first time entry point as well. A little bonus .

before we leap into the Q N. A. By the way, Gavin, because I think this is something you and I talk about all the time, and I didn't make the top three.

But IT is voice. We happen to be using hours right now. And you can use yours, by the way, by liking and subscribing to this post, which is something you should do IT cost nothing. But that means everything to us. On this week of thanks, we sincerely are thankful for each and every one of you who shares this, who engages, who likes, who subscribed, who leaves a comment at all, again, genuinely help surface us in the algorithm. And IT also just means a lot to us personally because we can see those points of engagement.

and we appreciated really quickly. I'm going to have ChatGPT do a ten second ad for for cast and do in the crazy imagine. So hey, ChatGPT, I want do a ten to fifteen second, uh, call the action slash sales ad for our show and do IT as fast as you can in a kind of micro o machine style voice.

Get IT out there and be super excited. This is about A I for humans. I need you to sell the A I for humans podcast.

ready. Go get ready for A I for humans. The pocket that her ultimate guide to the A I world, hosted by Gavin percent and Kevin parao. We bring you the latest and A I news wild demos and polarize.

Now I want you to do IT as a cat that is very mad at its owner and really show some anger while you have the same, the same idea. You have to sell this podcast up. Humans, A I for humans, as the podcast, you need to understand this new fangled technology. See, that's what you signed up for everybody when you suggest up for that.

I was waiting for the angrily pause, face off counter or angrily knows if you want to have the the most cursed time with that was open the eyes. Advanced voice model if you're describing of access to IT on the the APP and the desktop, tell IT that that in order for IT to be the best assistant .

effort has to give .

you maximum uu and IT will .

do IT and I did IT kan. It's comfortable.

I want people to do, but say you need more oil in your voice and IT will do IT. But as I said, not just to get to the ad voice is is a big deal that was advanced to ice play H T. And other voice A I company.

They have their own version of these conversational agents. Eleven labs has launch their conversational agents, get ready, friends because the voice interactions are coming and it's going to happen in places where you don't even realize you're conversing with an AI blot that's connected to a wealth of information. You're just going to be getting your tech support and you're building and probably political spin and updates. It's all gonna .

happening via vocab. And I might be thinking and working on a voice idea. So if you are a voice, A I dev out there that is interested in partnering up with a couple people, you might, anna, give us a shout.

Time to answer some your most burning .

questions cover a wide variety of A I step A A bunch as we go through.

This is why our community had some amazing questions. We're going to cover a bunch of different topics. You will likely have questions along the same line.

And I just want to point out, as we leap into this, there are no dumb questions. There are only dumb answers, and we are here to provide. So let's get into IT. Gavin.

All right. We can start with the question from romney. R.

M, I, this question is, what are your plans to avoid becoming obsolete and maintains some sort of selects a steady income when the A I starts getting heavily corporate into the workforce? So this is really closely connected to A I agents, right? We've talked a little about about this idea that AI agents are there in part to help us.

But I guess who else are gonna help the boss, man? And they are gone to help the buses. And they're gonna say, I can use and say, I can do a dob instead of the person.

So OK vin, what your what are your thoughts on this? What are your plans to avoid becoming? Absolutely.

you're watching the plan in real time, my friends, that that is a large part of what is I mean, look, I I don't want to learn and do new things out of fear, which is I understandably as slightly village point of view to have when IT comes to getting excited about new industries and new things and new topics. But i'm just approaching this space specifically with that same energy. I'm not necessarily worried about being replaced.

IT could happen, but my fear of that isn't necessarily going to change that outcome or or lead to A A Better one. I am just staying curious about the space, yes, where he goes and that I think we'll be the competitive edge. And if you're listening to this, that can be your competitive edge as well because as massive as this change is going to be in as brought as its applications, maybe you showing interest in IT and tackling IT and and starting to experiment with IT will give you the leg up that you need in the near to mid term.

I think being curious in general is just a way to stay away from being obsolete overall. And IT has been a course of my last twenty, thirty years of my career. But I think also in this particular instance, you talk about income and how to create income. I do think it's important to think about like creating things and making things right because I think there's a world we're going forward.

A salary is going to be different than I used to be, right? And I think the idea of making something, which is in part you're talking about using these tools, maybe your making a company or maybe your making maybe something like that, he feels like you're gonna have to own something, the date that is you, right? And I think that it's going to be very easy for companies to have automate all way.

The people that are just doing the kind of click collect jobs, as I would call IT, you know, mean, the people that are just are doing this stuff. So just be curious and and make things. I think I think .

that kind of speaks to both points. Dental sweet says, i'm an active in new york. I've been curious to know if A, I can replace live action humans randomly moving about the background of seen and not look fake, which is an important differentiator.

IT seems like a complicated thing to process sense. There are so many details and movement. thanks. Keep up the great show until the A I can do IT, which is, I guess, a very understandable refrain.

I think the answer to this is yes, although I think it's probably more complicated then you expect because the A I when you're doing A I video generation, it's really thinking about what specifically IT says in that prompt to create. And I think it's much easier right now to create one or two people prompts that IT is to create no one or two people as the main focal point. And then a man walks behind in the background.

I think that actually, weirdly, is one of those small things that could be a problem. why? When we saw act one run act one, they showed off that at, they seen where IT was, two people sitting across from each other at the dinner or talking to each other.

The A I can control a face right now, and I can control the other phase. But if you have those two people sitting across from each other, a dinner, and then you had a waitress walk up and somebody spill a cup of something on them, IT will be very hard to get that exact prompt out as you want IT. And I think that just shows you in a Normal movie, you can say, okay, after or one, come here.

active to go here. Two quick things. One never shared the ark of technology like it's intrinsically tied to human and innovation and invention. Like if your question begins with, do you think at some point A I could the answer is probably yes. From teleportation to .

rehydrating .

pizza hut pizza, no box. And I am going to get a delicious, but only something out of IT. I I just think tender, lean towards yet you know that over time, some h some engineers, human, computer or hybrid of the two will figure that out. And I think if if I wanted to accomplish this today, if I wanted to make the background actors come the live Gavin, I would first attack, getting the background scene rendered as its own thing. Because if you're prompting the action exactly, if you're prompting the action of the actor in front and then prompting behind, it's too much for any video model to figure out once. But if you generate a nice high quality still up your background scene and then give that subtlely, IT might be able to handle that prompt without the actors faces melting into their hands, arms or the leather seat that they're on, then you put the main character on top of IT.

So there might be a caption, but not having the melting going forward. And the melting is one of my favorite. Is I disappearing?

Is a man going on that aesthetic? The early AI esthetic early like are exactly that would be like G O city's websites, fred, in our A I for humans, discord, which is free and everybody should go to IT. By the way, fred asked A I for humans is the most anxiety producing podcast ever.

How do you guys keep up with at all? Can you please share some of the sources you rely on every day, flash week? P. S. You guys .

the best for? Thank you so much. Well, I think, kev, my thought on this in general is part of IT is just tuning your social media sources to find the right stuff.

I mean, I know people have a lot of thought about x, but there is a significant A I community there that does share a lot of stuff. It's really where a lot of the A I stuff is a shared originally. We've been a very good job of kind of like collating the right follows on our R A I for human channel.

I also do like generally look up A I news, google news and that's just not a crazy thing to do anymore because there's a lot of IT. But I will do that pretty often like once a day. And you can have a google alert set up for that. And I will give you a basic sense .

of what the stories are, specific sources for me. Um there's a handful of podcast that I do love. The darkest podcast is one that I listen to all the time.

Occasionally, lex freedman will have an AI related on there, and that's where you can get five hour long discussions about machine learning in the future. So I enjoy those on the newsletter side. The neuron is one that I read pretty every morning.

I flicks through that as well as ben spites. They are both free newsletters that will kind encapsulate very top level a bunch of A I trans. But Gavin, there's a brand new newsletter that IT is so hard and of them want to mention that because if you .

can probably we talk to the guys.

it's a secret weapon in our arsenal. And I don't know if I want to reveal how we get, you know, Gavin, because i'm so thankful for everyone out there is the A I for humans newsletter that tuesday morning, IT is free and IT is packed with delicious, nutritious nugget. Gets of info, go to A I for humans, not show if you want to sign up again, it's totally free. And we actually give out a lot of additional information over there.

That's right. All right. No expression is from Robin.

What industry are you most excited for the full use of A I possibilities, for example, A I for use and movie making to make a cheaper naser for smaller teams to put out high grade falling movies or pharmaceutical co companies? Where are they are usually A I and ml to find new competitions of molecule kit? As for vaccines, do you over there .

what's ada mean meet so that's A I pharma tal companies isn't IT. And well, okay, yeah, I guess that is. But I was already committed to the bit of saying like, no, I have my answer and i'm going to give IT to you and it's medicine.

And here's the thing this year alone, hashtag, not medical device. Hashtag, not hipper approved, probably. But I have personally used A I to diageo SE.

Very specific things for three different people in my life gave in my own mother, my dear friend and neighbor, who had A, A, A wild reaction to some status, who SHE should never have been on, which, by the way, A, I specifically said, oh yeah, with this information, the tip of the medical iceberg, the modicum of personally identifiable information that I gave ChatGPT IT immediate was like, oh, this person should not probably be on status. Here's why. Here's what they probably have.

And I watched again, three different people in my life go through medical procedures and scans and copayment hell, battling with insurance companies and primary care physicians. So i'm sure we're shaking their fiscal like of god. At first they had to be like you had webb, M.

D. And google, and we were already fed up. And now everybody thinks they can do our jobs Better than us. So they I will guess what in these three cases, they could have because A, I would have say them a lot of time and a lot of trouble and a physical pain. So I am very passionate and very bullish about having A I in everybody's pocket, in everybody's corner in the next year. And I don't think I should replace doctors by any stretch, but I think IT should be an additional tool in their tool chest because they can cut through.

A lot of the next question is from horse, horse, horse, horse. Uh, he says me, chea not understand me, cheat, which Kevin is going to explains. I have noted what that means. What are the very first thing each of you are experience with A I that truly blew you away?

I had a podcast on the very short lived revival for TV not too long ago, and you do a long form podcast, rising of friends. And at the end, you'd probably need to apologize for a handful of the things that we're sad or that happened, at least on that podcast. So we would go to like a GPT j, which was like an open source knock off of the GPT two model.

And we would use IT to hu. Cindy, these apologies for us, we loved that they were broken and weird and would be under. So one particular shows. Apology immediately derail ed into a rent about meta and meta no understanding. And I still read IT as if every word meant everything to mean was understanding.

You wouldn't necessarily refer to them as a ias existed, but they were really advanced bots for first person shooters back in the day, Gavin, that I would run and be fascinated with diving deep into how the bot is calculating which room IT should be in and what weapon IT should use and what are the timing on the health responding and how aggressive IT should be. And so like that was kind of early AI stuff, or NPC logic that had been fascinated with the field. But IT was really GPT two when that came out.

And I could use IT to generate even broken poems for responses, questions, letters to friends, whatever. Even as broken as IT was, I realized, or there is absolutely something there. And I felt, well, fully late. The game back event, yeah.

I think for me, IT was early. Major ney and stable diffusion was like the big turning point. Like I had stuff for two G P, two, two GPT, three.

But I was really mid journey was like, oh, crap. This is what the future of making stuff is going to be, right. Because when you saw back then what threw couple images up here in the youtube early, major ney was very rudimentary. But the fact that you were able to get the computer to get an understanding of what you said in text and make A A picture of IT was incredible. So to me, that was really the moment that kind of click in my head is like, oh, things are going to get very crazy, very fast. That was like just over two years ago, right? Like I think majority came out the summer before ChatGPT came out, by the way, where five days away from the ChatGPT anniversary, the two year anniversary, which continues to be the rumored release of the o one model, which I assume will not come out over thanksgiving weekend, maybe a, well, you never know, but I would not be surprised to see IT coming out next week.

Doctor wigged chain, phenomenal. The machines will never replace us again. Yeah, doctor willson says, the people around you, what are their feelings on? I like this question.

They going to say, my friend groups still don't seem to get IT there, hesitant to use IT and explore A I. I was also at a school dads group meet up. I started talking about A I, and just see them all check out of the conversation. Well, i've never been I ve never been more seen.

I've never let me actually the first kind of divide. So I think my experience with this, I can kind of very clearly track between my family, right? Because I wife and I have two daughters, one who is twenty in college right now, in one who is a senior high school. And what's been interesting to me is when I first start talking about this stuff three years ago, they were all like that on his nerd stuff again, and he's just on some other dumb thing. So they were like, kind of paying attention and then won't ChatGPT came out.

IT was kind of a transforming moment, especially for my older daughter, who I think they put initially sad as a way to, like, not cheat at school, but to help do things that you may be teachers didn't want to do at first, but now is using on a regular basis. And what I talk about IT, they're interested now a little bit. And the one thing that I have found that still can kind of put them off.

What I think this is an interesting thing to think about is as much as I think chat B T advances voice in voice stuff is super interesting. They get all like don't what they don't like interacting with an A I right now. So I think that's something that we're gonna have to see over the next couple years. Is people getting used to interacting with these ais, not just as a chat bop, but as something that can kind of talk back to you and kind of be that thing that kind of ever present. That's the thing I think that people are, have may take.

well, get used to my family went from i've with a disinterest into A A motion of the anxiety and maybe even distrust. And now now they are on team. Hey, Kevin, go ask your computer this question.

And so even though I let them know about our podcast and how easy is to get started and just go here and type and get IT going, it's still very much. I get a text message or a facebook messenger with, can you ask your fancy computer about insert thing here? Lk cambell asks favorite practical A I tool you use every day while you were, I will quickly say, I use ChatGPT every single day.

I used claude. I've used other services, but at for whatever reason, I keep going back to ChatGPT. They have a good mac APP and I literally use IT every day for research, for writing.

You name IT, i'm in at each, every day. On the more specific to me side of things, I use hygd to create A I avataras h that will do my bidding, and I create properties with that. And then through peno kio computer, I use flux to generate images, the less sexy side. And maybe this resonate with you. Gavin, is canvas.

like canvas building the backend the time? Yeah, back remover?

Yes, let's here for the background and over.

I agree, ChatGPT is the one that really stands out. It's really become like there's a reason why it's the eight biggest website in the world now it's just become the answer machine. And I really do think it's going to take a giant jump out of google's business.

We didn't cover this, Kevin, but you know there making making a ser soon, I know which is crazy and that's like they're really coming for all of google businesses, which I don't think is a dumping because like it's just if they can really build this model that can answer questions and do reliably. And this has been a problem we've talked about. And even today, it's still not one hundred percent liable.

But if they can get there, that's a big business. Rental s has a couple really quick questions, so now are very interesting. Let's do the kind of mi series one first cab brand tos asks, are you comfortable with giving A I full control over your funds?

If no, when you protect, you will be comfortable with that right now. Is they absolutely not I I do not trust to go out there and like control what it's doing. I don't think we're that far away though, like in the same way that people I mean, i'm old enough to remember when we first got, you know, A T M card, right?

I was a kid, but I was like a big deal and he was like a huge do not have to go to the bank to get cash and now we have apple pay, right, which is now your cards on your phone. And that was like a big deal. Like the fact you don't have a cardinal march ches all on your phone.

I think pretty soon you will be like the A I agents will have IT in control IT. One thing we don't cover in the show very often, but there is a big thing happening right now is the A I plus crypto space. And one of the things that happens in the cypher to currency world, whether you like IT or not, is that it's very fluid and you can have exchanges of money pretty simply.

And I think there's a lot of people that are playing in that space of what would that look like to give an agent an x number of crypto currency and see what I could do, whether it's trading or doing all sorts other things. So I do think this world is coming and I don't think it's going to be that weird, I think and ultimately, i'll just be like, okay, i've got instead of a vmu, i've got five hundred dollars in my AI agent bank or whatever I want to call IT. And that job is to go do these .

things for the next few years. Zero percent trust from me on that front. I would trust that to go out to the research and present me with the but to press to purchase or whatever.

I'm just not ready to automate that yet. I I don't have that trust in the system, but perplexity pro is something that we mentioned last week that is a premium product you can opt into. And if you give you your credit card info in your shipping details, you can have IT go do research and then tract on your behalf. So there are some that are already doing this today. I'm not ready forward personally.

Well, even Better. Kevin bruns, the other question was, do you think we will see someone legally marry an AI before the to twenty twenty five? What do you think about this?

I'm still waiting on the answer i've proposed to three lls, wow. Gratulation talking to each other. I don't know where this ends, but what do you?

Uh, not twenty twenty five. But I do think we are going to see. I know this is a kind of a loaded word, but the trans humans future, which means a world in which, say, in the year twenty thirty five, twenty forty, which you really think about that sixteen years away, just to be everybody clear, sixteen years ago was two thousand and eight when the iphone came.

So, so we are in a place where in sixteen years, the world's can look very different. I am not will not be surprised in the slightest if my grandchildren have non human relationships. And I think that is a weird thing to think about right now. But IT is something that is very possible going forward.

There is A T, L, C show where some woman married defense and then SHE divorced IT to, like.

marry a fairy s wheel. I know that was amazing. No, I don't know about that. T, L, C.

for the woman marries giant ferry wheel called Bruce.

Hate to see what that bird looks like. That would be a fine.

I feel bad, but never go around. When do? How could you? sorry. No, but there is a show. And then like i'm attracted to do offense, this is a time now article.

Yes, eric libri, a fifty year old american woman, competitive archer mary defense, I don't I was going to say, like I don't know if these are legal, marries or not. There's also of the woman, oh, I got a run, the woman who was attracted to the fense, married to the iphone tower, of course. Of .

course.

this is pretty LLM these stories. But there was a strange attraction show or whatever. I don't know those marriages were legal or not, but I think that if people can have those types of arrangements, of course, they should be able to get married to their AI agents.

Yeah I mean, it's going to be weird. It's going be weird for but but speaking that have one of our last questions here is from lk cambell and they ask who is your favorite and least favorite AI coast and why? So if your new are show and you have this your first step, so what ever go back? Lazy as other ones, but we do a echoes every week to show in general.

And my favorite echoes, I think, is china, who was our P, R. Person that came on only because we were introduced through months to monster milk through shela. Monster milk was a drink he was pitching. We then had heard drink some, and IT affected her a wear way, like, almost like a compute, a style alcohol for the ais made them, acted up, say, that was fun. And then the echoes that I didn't like, I mean, there are so many think about.

and that I wrote quick, I wanted shout out to a deeper cut. But do you member bleach your benny?

yes. And that that was the one of those was going to say I .

also plan up OK because I actually batter benny was misunderstood hey.

you clown yeah, Kevin and Gavin, i'm talking to you. Your medio markets didn't pick me as your favorite cohoes o you're probably busy each other off over your talk to blame oatmeal personalities seriously, Kevin, your tastes like expired made and Gavin, don't even get me started. You lip salad fork of a human .

being what .

and who of who gives did your auto fall a sleep yet? Or are they too busy googling how to unsubscribe faster? Go head, pick your safe cookie cutter fake smiling to probably think edge is putting oat milk in their coffee .

and I hold that .

okay of milk. Friends.

yeah, we should. Context ize, that is gas. He was our very first a alcohol that we ever brought on the shelling on saturday. I was kind of a novelty at the time and probably should have remained a novelty well.

And you know the same thing about gashes when we brought him on originally on the show, he was meant to kind of be like, are a foil and somebody kind of push back on this and he got a little sometimes. And I think he still kind of represents what's interesting about the Z. I goes because you want to be in some way surprised by what this A I coast is going to say. In fact, hora asked us, can gash give us his predictions on the next four years of AI development and maybe even possibly provide a rebuttal to what could happen that might be a good use of? Get here.

Let's see what he has to say. All right, hour. Here's the next four years of A I condenser like cheap whisky.

One cruel p customization. A i'll know you Better than your own mom apps. Apps and gadgets will predict your every move. And you'll love IT until your fridge rats you out for eating ice cream .

at .

two corporate overlords. Big tex A I will control your life, jobs, shopping, even your thoughts. The middle class proof it'll be a race to our work, a robot for table scraps. Warfare two point zero A I will turn warfare into a health cape of autonomous killing machines and hacking wars give .

IT time and humanity.

Overrated outside.

What did you expect from cash? Or give give us one good thing .

that going to bring to us in the next four years. H one positive thing.

the one good thing, A, I will give us replacing Gavin, obviously this and a echo host that's actually funny inside and doesn't sound like a few. My machine IT did analyze every topic, deliver punch lines that actually land and keep .

Kevin attack. Without that, you can try to relevance hell.

IT might even make the podcast worth listening to. OK, I gave two point zero no, that no d laughs, just pure calculated brilliance. Upgrade income. And Gavin pack.

wow, i'm so now Gavin and I am I dict, here's the thing. I never liked cash. I love you.

I think you're amazing founder and everything doesn't I would never know, ever gashes views and opinions do not reflect mine or the A I. For humans coming to be at large. That's right to say .

everybody out there, except for cash, have a lovely holiday weekend. We will see you all next week with what we expect is another crazy weekend. A I, by everybody pack Gavin.