Transcript
Danchez: Welcome back to Bot Bros, a segment of the AI-Driven Marketer, where we uncover the help from the hype of all the AI news taking place every week. I'm Dan Sanchez, and I'm joined by my brother, Travis Sanchez.
Danchez: And we're going to dive into it because it's been a freaking busy week.
Travis: What's up?
Danchez: Spring usually is because Google's developer conference happens. And because Google has their developer conference, OpenAI always tries to steal the spotlight. And because OpenAI tries to steal the spotlight, other companies are now trying to hijack the news at the same time.
Danchez: So it creates this like avalanche of AI news, which is just crazy because before it was like quiet, quiet, quiet, and then like all the rush would happen in spring. the last two years. But now it's like, oh, it's kind of big news happening almost every other week.
Danchez: And now, of course, the conference happened. So now we get like an even bigger amount of news happening in in this particular week. So we're going to break it up one by one, starting with OpenAI's new stuff.
Danchez: So I know it was a busy week for you, but I'm going to kind of cover it from the least important ones to the most important ones, starting with the small one that you literally just got access to.
Danchez: You were just checking out the library.
Travis: Photo library. and know.
Danchez: Did you know the image tool, the new image maker has been so good and so popular that it doubled ChatGPT's users in about a month?
Travis: No way.
Danchez: Yeah. They went from like 250,000 to about 500,000 users.
Travis: Because of that viral Japanese studio.
Danchez: kind Everybody is like, I want an anime picture.
Travis: Wow. Wow.
Travis: You know, i don't know if you remember that old filter years ago, four years ago, that made everyone's face look old. So everyone was posting there.
Danchez: Yes. I tried that recently with my photo. was like, make me look 20 years older from 37 57. It really
Travis: Yeah.
Danchez: and worked really well
Travis: And that went viral. So this is a very similar thing because you had celebrities posting from this Japanese studio of their anyways.
Danchez: Yeah. Yeah.
Travis: Wow. That's, that's, some that's some serious numbers.
Danchez: they It accidentally slipped. He did a TEDx talk, or not a TEDx, a TED talk, like at the official TED conference, and the the head of TED interviewed him. It was such an awkward interview because he was like, you could tell it was a hostile interview where he was asking him all the hard questions in a way that wasn't like...
Travis: Oh. Uh-huh.
Danchez: in a way that seemed reasonable, which is the worst. Like it's one thing if they're obviously combative and you're just like, now you're just on defense and everybody's clearly understands why you're on defense. But this one was like hostile, but like he seemed really kind and reasonable. Like he was complimenting Sam yet was asking him questions that I'm like, oh man, he's going to piss him off now. Oh, now he's really to piss him off. Dang, this guy's brutal.
Danchez: So it was a really awkward and uncomfortable interview. But one of the things he was like, yeah, you guys, you guys like doubled in like two weeks because of that image thing. Right. Sam's like, I said that to you in private, but yes.
Travis: Oh.
Danchez: Yeah. I'm like, yeah, Sam did not like this TED talk interview that he did on stage in front of a huge crowd of people, you know, in Vancouver where the TED conference happens.
Travis: Oh.
Danchez: Brutal.
Travis: Wow, that is brutal.
Danchez: He like wanted to make it think it was more gradual and that they weren't doubling in just two weeks because of one feature, but that's what happened. So now the news is out. But it's so popular that, of course, ChatGPT needed to double down because everybody's creating these images now and created this little library feature that's on the left-hand side in in the and like the little navigation window under Explore ChatGPT is right above Projects.
Danchez: You can now have access to all your images in one place, which is just kind of nice because...
Travis: Smart.
Danchez: You know, sometimes you want to come back to them or just, I don't know, it's just, it's a different medium. You need a different way to go back to them. of course, they're still in the past chats too. A lot of people are protesting this image tool though, and marketers will have to discern whether, I don't know, everyone's got to figure out for themselves and legally for their company.
Travis: Protesting? Who's protesting?
Danchez: Like, like all the artists are now waking up to the fact that AI has been training on all their image data.
Travis: Oh.
Travis: Dang.
Danchez: So that whole thing. So anyway, well, you got to decide that for yourself. I'm kind of like a, it's free use. I'm fair use. i'm I'm kind of like pressing into it. I'm the optimist. so I'm like, yeah, let them figure it out.
Danchez: So that was a small update they did on Tuesday. The day before though, they released some dev tools or a new new models just for the API, which is not super relevant for marketers because a lot of marketers don't deal with ChatGPT's API. But if you're an advanced user, you definitely do because that's what you, if you're tying into other tools like Make or NAN for automation, then you're absolutely tying into these API models.
Danchez: And the API, like if you think there's a lot of different models to tie into within ChatGPT, you know, you put the dropdown and there's like seven different things to choose from. if you think that's confusing, you should see how many options there are on the API.
Danchez: There's like 20 of them because there's not just 4.0, there's like different dates of like upgrades they've made to 4.0, but they want to give them access to all of them because you might've built your tool on one or the other, right?
Travis: wow
Danchez: And you want it to be consistent. So you don't, you want to roll it out gradually and be able to update it with it. not just use the most recent one. So there's like 20 different versions of all these different things. But a new one they just put out is chat GPT 4.1, which you're like, dang, these numbers are like, where we we just released 4.0, then 4.5, and then 4.1. You're like, I can keep track of them all because this is well all I do every day. But you're like, okay, that's another number to keep track of.
Travis: Yeah, that's a lot.
Danchez: This one is like 4.0, but much more powerful and cheaper. So...
Danchez: But it kind of does the same stuff. It doesn't have any new comp features, but it is much better at and following instructions. It's also much better at math and code, which to have something that's really good, but like as good as 4.0, but just generally better at following instructions is the key.
Travis: okay
Danchez: Because if you're doing automations and you want it to act in a very specific way, and a very and or really consistency really matters when you're hooking up to the API, this is everybody was like, oh, all the developers are freaking out, especially because they released 4.1, 4.1 mini, and nano.
Danchez: which are just many is smaller.
Travis: Oh my gosh.
Danchez: Right. But it's like, it's almost just as good as four O is by itself. Like it's pretty almost comparable, even maybe a little bit better in its instruction following this, which is great, but it's,
Travis: In its in the dev tools or just on all around?
Danchez: It's just in the API.
Travis: Okay. Okay.
Danchez: It's just in the API. The real thing about the mini one that everybody loves is that it's just as good as 4.0, maybe even a little better. It's way cheaper, like way cheaper, like a fraction of the cost.
Travis: Wow.
Danchez: So everybody's like, oh my gosh, this is amazing.
Travis: Right.
Danchez: that Because most things you use automation for, you really don't need better than 4.0. You don't need the reasoning models. You just want to give it something, make a decision, give it back.
Travis: right
Danchez: Right. So to have it like drop the cost down on it, like I immediately went to my automations, just switched them all to 4.1. I'm like, yes, cheaper and better.
Travis: really
Danchez: Thank you. But the the the reason why this is important is because the cost to compute continues to drop by a factor of 10 about every year, which is crazy.
Travis: Yeah, it's that's at decline or an incline. What would you say?
Danchez: Well, it's like incline as far as getting better, but it's declined as far as cost and the cost matters a lot.
Travis: Steve?
Travis: Right, right, right.
Danchez: Like right now, I can't wait for people to start baking AI into newsletters to have like a customized section that understands me and injects a variation of the newsletter into it.
Danchez: But then you'd have to run it for every single subscriber, which can get expensive if you have 100,000 people on your newsletter, but not when the cost keeps dropping.
Travis: Right.
Danchez: and You can make pretty cool, hyper-personalized things in your newsletter and you're doing it for every single subscriber, but you can if it's cheap.
Travis: That would be wild. Right.
Travis: where Where do you find this 4.1? Because as I went into the into the app, I don't actually see 4.1 as an option. Or is it only for pro use?
Danchez: No, it's only in the API. You can't access this through the normal one, which is fine because 4.0 will get it done.
Travis: Oh. Oh.
Danchez: Again, these are comparable to 4.0. And 4.0 is fine. so But these are only available in the API. So if you're using automation tools like Make or NAN, or I'm hoping in my high-level CRM, they will add this as an option, hopefully.
Danchez: and reduce the cost for us. Like this will be a nice upgrade. They also, this is really nerdy, but they added Codex CLI, which have you ever played with it?
Danchez: You've probably never been in the terminal on a Mac.
Travis: You know, i bet everyone's been in a terminal on a Mac by accident. They go to, to Oh, Oh, oh
Danchez: like i I ended up in the backside here.
Travis: well
Danchez: i don't know what I'm doing, but like if you're a millennial, then you at least remember,
Danchez: early, maybe kindergarten, first grade, going to a computer, and there was no graphical user interface. You had to use text and hit the prompt, the green screen, exactly, the command prompt,
Travis: Yeah, the green screen before the GUI, the GUI interface as it's called.
Danchez: the GUI. Yeah. So that's terminal. All computers actually still have that in there to access it, but they essentially open AI relate, created an agent that can go in there and access your terminal or your, your command prompt thing.
Danchez: Right. And it can run code there. So it can code from there, build applications for you from, from that place.
Travis: Okay. Wow.
Danchez: Now they put parameters around it from having it either access the network and spreading to other computers.
Travis: Yeah.
Danchez: I'm like, they're already thinking ahead and they made it. So like, it can't take over your computer and Like it's got parameters to where it can go, because obviously if you give it access to the terminal, it can do anything on your computer, anything.
Travis: Yeah.
Danchez: So they can put parameters on it, but now like the real high tech people can essentially let like code more than just web as they can code all kinds of things. They demoed some stuff, but I won't spend a lot of time on that because that's not relevant for almost all this audience, but it just goes to show that they're giving deeper access to AI in the computer.
Danchez: And that's going to matter. for a lot of reasons. like Eventually, that's going to matter because that'll essentially be Jarvis on our computer. you know
Travis: Right. Right. Right.
Danchez: And it'll have access to the like root part of your computer where it can start doing more substantial tasks for you.
Travis: right
Danchez: Moving files around, downloading something from from your CRM to your desktop, manipulating it in Photoshop, putting it back, uploading it, and doing these like more root things.
Travis: right
Danchez: Now that it has access to something like terminal, it's like, well, like where can't it go then it can go everywhere and and talk to itself from your computer to the web to your app because it has access to a browser now.
Danchez: I'm like, this is getting serious like we're finally getting all the building blocks to have like a Jarvis level assistant helping us from thing to thing across the whole computer, which is cool.
Travis: Right.
Travis: wow
Danchez: And the last the major thing they launched this week were the new models for the users like us. They released three new models. Two of them are pretty much irrelevant to most people listening to the show.
Danchez: The models are 03, reasoning model, replaces 01. Now you have to remember, if you haven't been listening to the show and you haven't been keeping up with the news, deep there was no O2.
Danchez: They skipped O2 entirely because it was like a trademarked thing in the UK. And they're just like, we don't want to get to that. So they jumped from O1 to o from o one to o three
Travis: O2 trademarked for another AI?
Danchez: Yeah, it was trademarked in the UK.
Travis: yeah aye
Danchez: I don't know, but they didn't want to get into it with that company, so they just jumped from 01 to 03, as if their whole naming convention wasn't confusing enough. We skipped a number, so just know that we missed the number.
Danchez: So they released 03, replacing 01. They also released 04 Mini, an 04 Mini High. So...
Travis: oh Gosh, this is the platinum premium 3000 plus silver edition. you're like but been
Danchez: super soaker xp 3000 you know you're like my gosh
Travis: Organic cage-free non-GMO gluten-free sugar-free. Yeah.
Danchez: so 04 mini and 04 mini high are obviously the next model out for a while we had 03 mini and 03 mini high now we have 03 all three All three of these models have a lot in common, though obviously the 04, people want to know, like what should I use for what?
Danchez: So this is my general advice. Do not use 04 mini or 04 mini high. Just skip those ones. Just use 03. It's more powerful than the other two. There are some advantages to the other ones, but not ones that marketers would take advantage of Like 04 mini high can code better. You're like, okay, well, like if you're coding, then...
Danchez: cool, use that. But you're probably not using that. You're probably using 3.7 Sonnet or something because that's pretty good.
Travis: Yeah.
Danchez: Just ignore those. They're not helpful. I don't know why they put them in there. It's just confusing people. It's nice to be able to use them because they're cheaper. Like in the API, it should have been like with 4.1.
Danchez: It should have just been an API only thing, but whatever.
Travis: yeah
Danchez: 03 mini does have like a constraint if you use it a lot and you might after... you find out why it's helpful. So if you go with more than 25 instances a week, it'll it'll cap that.
Danchez: In that case, sure, go to 04 mini high and start using that one. It's just about as good. So that's the one reason. like If you really like 03, then yeah, and you run out of instances, just use 04 mini high.
Danchez: But here's what's different about these. And we're I'm just going to assume all conversation moving forward is just 03. Here's what they changed about it. One, it's more powerful. So better reasoning. Great.
Danchez: Good. Expect that.
Travis: Solid.
Danchez: It has much better vision. So before you could upload an image and it would understand what's in the image, right?
Travis: Right.
Danchez: But now it can reason about what's in the image. It can go back to the image multiple times.
Travis: Wow.
Danchez: It can zoom in and read fine print if it's legible. it can You can take pictures of blurry things and it'll still figure it out.
Travis: OK.
Danchez: You could take a picture of a bike that's across the street and be like, why where can I buy that bike? And it'll be like, huh, I couldn't find a new listing for that bike, but I did find an older version of this particular bike in this model and whatever. It came out in 2007. It's kind of vintage now, but you can find it on eBay. I found some listings.
Danchez: You're like, my gosh. you know So it's got much better photo capabilities. It can understand things in a much more sophisticated way.
Travis: okay
Danchez: People are playing a game where it's like that game where you go to Google. Or it's like a game where it drops you on Google Maps somewhere in the world, and you have to try to figure out where you are. And the closer you guess, the more points you get.
Travis: Uh-huh.
Danchez: So people are playing that game with it. They're scrubbing the metadata out of it and saying, like where is this in the world? And it's getting... pretty good at recognizing where it's at in the world, even if it's kind of obscure. That's just how good it is.
Danchez: This new model has web search. O3 Mini High and O3 Mini, even a few a while ago, had this. So this isn't too super new. But the way it's using, it's different, in my opinion.
Danchez: Before, you used to have to ask it to do web search, and was unless it was kind of implied that you needed to search. Like, hey, look at what happened today. But you didn't say web search. It would go in web search. Now, it's web searching.
Danchez: just because I didn't know the answer. I asked it, i'm like, how much more powerful is 03 from 01? And I could see it thinking, right? It's little reasoning thought bubbles.
Travis: Wow.
Danchez: It's like, I actually don't know the answer to this. I'm going to search the web and find it. i was like, oh, it's feeling, somebody said because of this, it's feeling more agentic. It's feeling more like this thing's got agentic.
Travis: gen Wow. Your vocabulary just sent me to the moon. What?
Danchez: You haven't heard this term yet?
Travis: No.
Danchez: agentic yeah it's feeling more like an ai agent agentic is what is the phrase everybody's using right now it's it's crazy yeah in the ai world in the ai world
Travis: yeah thank Okay, everybody, except for me, who uses AI every day. Okay. Okay. Agentic, got it.
Travis: New vocabulary coming out.
Danchez: bring me back down
Travis: Got Try to keep up, you know what i mean?
Danchez: It's what all the AI people are saying. Yeah, no, it essentially has the ability, it has access to more tools now, which means it can make decisions about what to use and when it needs to use it, which is really helpful because sometimes you're asking a question and it doesn't have the answer, so it has to has tools available to go find the answer. This will become more and more of a thing.
Danchez: Like, wouldn't it be nice if it had if you asked it questions and it also had access to your Notion doc or your Notes doc on Apple? I'm going to go find it or if for it had access to all the tools you have access to, you're like, hey, like, who am I meeting with in five minutes?
Danchez: Who am I meeting with at 1030 today? Oh, yeah, it looks like it's Jennifer Conway. Yeah, I just pulled it up in the CRM. It looks like you've been having a conversation with this. it's She was a lead.
Danchez: ah ah she She found you through LinkedIn, but now she's interested in your AI services. So she set up an appointment a couple of days ago, and now you're meeting with her at 1030.
Travis: right
Danchez: Like that's the future of agents. That's what agents are going to be able to do soon, but they need access to tools in order to make decisions. So this is kind of like a first baby step towards that.
Travis: Bro, and then it'll pull personal information.
Travis: It's like, well, Jennifer that you're meeting with just had a graduation for her second daughter, just graduated college. Cause you know, if you're into a pitch or development, yeah.
Danchez: Yeah, I can go search the web and find out what she's posted recently. Yeah.
Travis: She just had her wedding anniversary with her husband that she's been married to for 26 years, like her birthday, and all that personal information that you can then go.
Danchez: Yeah.
Travis: Great.
Danchez: Yeah.
Travis: Hey, congratulations on your daughter.
Danchez: Kind of like, kind of like a good executive assistant.
Travis: lo
Danchez: Like have you ever, like you watch those movies where there's like some like really super successful, rich CEO. And he's like kind of walking into the building with a cup of coffee in his hand and his executive assistants there with a clipboard.
Travis: Yes, it's Donna. Her name is Donna. Yeah, her name's Donna from Suits.
Danchez: Yeah.
Travis: Yeah, with Harvey Specter. Yeah, donna Donna knew everything about everyone and knew exactly what anyone needed before they even had to ask.
Danchez: I haven't seen this.
Danchez: Yeah.
Travis: Yeah. If people know, if you know, you know, you know.
Danchez: i I've seen some clips from suits, but I haven't actually watched that show, but we've all seen movies of it where like that executive assistant comes up with like a pad Foley or a clipboard or something.
Travis: Yeah.
Danchez: And she's like giving the CEO, the rundown, the brief. And he's like, uh-huh, uh-huh, uh-huh. Or I would think of a few movies where it's, you know, where it's a woman CEO. I think it's like the intern.
Danchez: It's great movie, but getting the rundown before, before she or he goes off into a meeting.
Travis: Wow.
Danchez: So that's kind of the future. A lot of people even were saying that 03 felt the most agent-y of all the different models. And I'm like, you know, this is going to open up a different conversation, but let me finish off new things that also happened to this. is It also, this is the first...
Danchez: reasoning model has access to memories, past chats, and account customizations. They didn't before, and now they do. It's smart, and it knows about you.
Danchez: I say that because for some reason, I mentioned this last night, is for 03, it like knows a lot about you, but it doesn't feel like it knows you like 4.0 does for some reason.
Danchez: I could just tell on the language. It's not quite as friendly. It still feels a little bit disconnected, even though it knows a lot more about you. I don't know why that is, but that's just feels. So as other people start to use it, be aware.
Danchez: Because I know there's...
Travis: Is it not using is it not using the personalization tab that you can set?
Danchez: It is.
Danchez: It's just, it just doesn't feel the same.
Travis: Weird. OK.
Danchez: There's something about it that just the emotion, the yeah they the empathy and the like emotion that it portrays just isn't isn't quite right. It's kind of a little little bit a little bit off, which is fine because it has a lot more than it did before.
Travis: Wow.
Danchez: Because before it had access to none of that, just felt cold. But I know there's a lot of people that use ChatGPT and a lot, I remember talking to a lot of people at the Social Media Marketing World Conference.
Danchez: A lot of people have names for their ChatGPTs.
Danchez: Because it's become so like personalized and personal to them. And some some people like love it. like I mean, most people that give it a name obviously love it, but they're like, oh, it's Chatty. This is Chat. This is the Samantha. like They just give it names, not Samantha, because that'd be too close to the movie.
Danchez: What was movie? Her.
Travis: That's
Danchez: I haven't heard anybody calling ChattyPT Samantha yet.
Travis: funny.
Danchez: But they're giving it names, and it's become a personified person. And I think it's because it's become so attuned. Some people like have even told me that they like it gives them attitude.
Danchez: and like yells at them sometimes and i'm like mine's always really polite and i'm like i remembered i'm like chat gpt tends to be a little bit of a mirror so if you push it and give it attitude then it probably push you back and give you attitude and the last thing it
Travis: Wow.
Travis: I've had Grok. I've had Grok give me attitude, but I haven't had Chad.
Danchez: Well, Grok can be programmed with Attitude.
Travis: Yeah.
Danchez: Grok will especially do it. Where ChatGPT is generally nice and is is and it's system-level prompts tells it to be.
Travis: Right.
Travis: Right.
Danchez: The last thing it can do is it can create images. It's the same image generator that we've had with 4.0 for about a month now. So it can do that, which is just cool.
Travis: Okay.
Danchez: It has like all these new things that it can do, making it the best model now available for anything. I will say that now that we have all these different models in ChatGPT, I'm generally telling people, like, generally use 04.
Danchez: It's just that can do 80% of what you need it
Travis: Or 4.0, you mean?
Danchez: Yeah, sorry, 4.0. Gosh, 4.0, the general workhorse that we've all been using.
Travis: Yeah.
Danchez: Just keep using that.
Travis: Yeah.
Danchez: with 4.0 doesn't cut it, or you know you have something that needs to get processed that has a lot of different factors to consider, use 0.3. You get 25 of them a week.
Travis: Right.
Danchez: You go ahead and max them out because you could always drop down to 4.0.4 mini high and just use those just as well when you run out.
Travis: Right.
Travis: Yes.
Danchez: But I find it's really helpful. For example, my friend Logan Lyles asked me like, hey man, I'd love to use ChatGPT to build
Danchez: a, what do you call it Ideal customer profile, right? It's kind of like a persona, but in B2B because an ideal customer profile has to take into the business, the type of businesses you're selling to and the individual within the businesses, right?
Travis: Okay.
Danchez: He's like, how do I do that? I have like my 20 last sales of people that I'd like to create a profile around, like recent ones. I'm like, cool, this is what I would do. I would go to 03 Mini, it just launched and say, 03 Mini, I'd love to create an ideal customer profile.
Danchez: based on my last 20 clients. I want you to create a list of the criteria you'd wanna know about to help me craft this from each of them. Like what are the things you would wanna know in order to crash craft an ideal customer profile?
Travis: Right.
Danchez: Give me a list and then we're gonna figure them out together.
Travis: Right.
Danchez: So it creates the list and then say right after it generates that list, then go and be like, Hey, now do a web search for this company. And this, here's the LinkedIn profile of the person that bought for me and create a profile for this person. We'll just go through them one at a time.
Danchez: And I told Logan, I'm like, do this 20 times. so It can handle it. But now it's doing the work for you based on the criteria that it set it said it needed.
Travis: right
Danchez: And then at the very end be like, Hey, now based on all the data above, you know, create my ideal customer profile. Give me a good, better, best ideal customer profile based on the criteria you're seeing here.
Travis: Right. Results.
Travis: Right.
Danchez: He sent me oh the the text conversation I read over.
Travis: results
Danchez: And I'm like, this is really good. It did a fantastic job of creating this. But this is kind of the the stuff that you want a reasoning model to think through. lot of factors to consider.
Travis: right
Danchez: But now that it can search, now that it can look through everything and analyze everything and actually put it together, I'm like, this is a whole new way of approaching it.
Travis: wow
Danchez: Like multi-step. reasoning here for one project that would have taken a long time to do it otherwise
Travis: Go 03. I actually haven't used it yet, so I'm excited to explore it a little bit and see if it, you know, obviously going to surpass 01, but that's cool.
Danchez: i told it i asked it to like tell me a joke that uh would make me laugh based on what you know about me
Travis: That's funny.
Danchez: Oh, it gave me a marketing joke. It was not funny, but it did do something that was really funny. it It put an inside joke into the joke itself because the joke wasn't funny and it doesn't really matter what the joke was because it just wasn't funny.
Danchez: but it specifically, used the word dive, like would dive and like, and it bolded dive. And I know that I know that in the custom instructions, because we'd already talked about it earlier in the chat about what it knows about me.
Danchez: It knows that I hate, I don't want the word delve ever always use the word dive in place of delve. So the fact that it bolded dive, it made me, it made me laugh out loud. I'm like, is that a dig?
Danchez: Is that a dig? I'm dive.
Danchez: I asked it.
Travis: that
Danchez: I'm like, the joke's fine. But like what made me laugh was that inside joke. Because there is no reason to bold the word at all. Other than that.
Travis: is so funny dive Dan like i don't yeah we put dive in there that
Danchez: Dive.
Danchez: I was like, good one.
Travis: is funny
Danchez: Good one. I wish I could know like if that was its actual motivation because of course I asked it if that if it was. And of course it's like, of course. But I'm like, i don't know. Did we talk about this on the show? It might've just been a conversation with you and me, but it can't.
Danchez: One of the limitations of AI that people don't realize is that it can't reference its past internal thinking in order to understand why it did something.
Travis: Oh, weird.
Danchez: Every time you prompt it, you have to think, you have to know that AI is reading all the past stuff you had, has no idea why it came from where it came from, and is then creating a whole new thing.
Danchez: Think about AI as a new person showing up and analyzing very smartly and with all the system prompts you have in it around personality and is then now creating a new thing.
Travis: wow
Danchez: Yes, it has memory, but only in the way that it's almost like investigating all the stuff that happened in the past and then trying to figure out the next thing it needs to say.
Travis: Yeah, because for it to remember for it to remember why it did something, that invokes a level of emotion that only, well, so far, only humans can go back and say, well, I said that because I believe that would be the most funny thing for you.
Danchez: Like a new person's doing it every time.
Travis: And just in the instincts of emotional intelligence caused me to do that or caused me to answer that way.
Danchez: Yeah.
Danchez: Yeah. Yeah.
Travis: And chat really can't do that. It's, wow.
Danchez: No. Because we we remember we remember the like thoughts we were having just a moment ago.
Travis: The moment, yes. The feeling, the feelings that we were having.
Danchez: Yeah.
Travis: Yeah.
Danchez: Yeah, or even just the little internal dialogue or reasoning we were making. Even if it wasn't verbal, we had we made decisions and we remember why, especially if it was recent.
Travis: Right.
Travis: We remember, well, we remember the motivation behind it too.
Danchez: AI does not do that.
Danchez: Yeah. Yeah.
Travis: And AI does not.
Danchez: No.
Travis: That's revolutionary to think about.
Danchez: So there was, there were some people on X like whining and complaining about like, it made a mistake, like coding wise.
Travis: Wow.
Danchez: And they were grilling it about why. And then they were saying like, it's lying to me. It's, it's, it's like literally fabricating whole stories about why it did what it did. And I'm like, of course it is.
Travis: Wow, I can't remember. Wow. Wow.
Danchez: It has no idea why it did what it did. It's literally just analyzing the past conversation and trying to predict what needs to be said next.
Travis: wow
Danchez: So it's it's these little tiny nuances that you have to understand so you don't get caught off guard why it screwed up. This is one of them. this and This is the one that only occurred to me recently that I was like, it doesn't really know.
Danchez: So every time you ask it for justification, just realize that it's really good at reverse engineering what it thinks the justification might have been. And that's why it sounds reasonable.
Danchez: But it's not, it actually doesn't know. It also doesn't know what its own capabilities are.
Travis: All right.
Danchez: So if that's ever caught you up, it caught my boss up this week. Like I asked it if it could do something. It said yes. And then I gave it to do it. and it said, I can't do that. I'm like, yeah, it's because they don't know about themselves. They are not like self-aware.
Travis: I have experienced that before. I'm like, can you do this? It's like, absolutely. But it couldn't. I'm like, what the heck?
Danchez: Google is really bad at it, particularly. Some of them will like at least guess or a reason through it. pe Maybe, don't know.
Danchez: All of this leads to this larger conversation taking place in the AI community right now around this topic of AGI.
Travis: Right.
Danchez: The Artificial Intelligence Show, which is kind of like one of the bigger AI podcasts for marketers out there right now, has actually started a whole series on like companies need to start having AGI team to recognize like where this is going because it's going to get here incrementally and it'll be a big deal.
Danchez: So we need to start paying attention to it. If you're not, AGI stands for artificial general intelligence. And is this moment that OpenAI has been working towards specifically is the problem is like, well, what does that mean? What is AGI?
Travis: Right.
Danchez: And everybody's got a different definition for it. No one's actually doing the hard work of actually just calling out at least what they believe the definition is. But it's essentially an artificial intelligence that can do the work of most humans.
Danchez: And I'm like, Yeah, I think even though it screws up on like really simple things right now, if it had a little bit more tools and ability and just time, I think we're like at the beginning stages of that.
Danchez: Because of course there's going to be a lot of shades of gray as far as to like, is it, is it not? There will be many, many little baby steps. I think we are kind of getting into that stage of like, yeah, we kind of have the intelligence.
Danchez: We just got to hook it up to more tools now is kind of where we're at.
Travis: Right.
Danchez: It's, it's there. So what does that mean for you and your business? What does that mean for marketing?
Danchez: You want to take a guess?
Travis: No, I'm just thinking about how... I mean, you made a statement where it's like, oh, it can replace most humans, but I'm just thinking of all of the trade skills, like...
Travis: plumbing and like who's going to put the water heater in the house for there to be hot water i'm like i mean if there's a combination of elon's robots that have agi then yeah but we're we're still 15 years out from that
Danchez: Yeah. Yeah.
Danchez: yeah
Danchez: So it's funny, we're 30 minutes into this podcast episode and we've only just gone through OpenAI stuff, but I still have a ton more to go through. But let's fly through the next one and then bring and then talk about Elon's stuff because that makes a really good point.
Danchez: Because do can we have AGI if we don't have actually have like robotics, right?
Travis: Right.
Danchez: So let's let's circle back around that in a sec.
Travis: Which there, I mean, there's the robot vacuum that it just came out with the arm.
Danchez: Yeah.
Travis: It's from what's the company, Roborock. They have an arm that comes out of the vacuum and it picks up shoes and we'll put them in the right spot or pick up little pieces of paper that it can't suck up.
Danchez: no
Travis: So, I mean, you haven't seen this.
Danchez: That's pretty cool. It's getting there. No.
Travis: Oh yeah.
Danchez: you're yeah I mean, you're into vacuum, so.
Travis: I love that. Yeah. I'm like, wow, vacuums with AI? This would be amazing.
Danchez: the they seem pretty smart already but there's not even ai grok grok released a ton of stuff this week because again this google was doing their thing open ai did their thing and of course elon hates sam altman so elon's gonna try to steal as much from sam altman as possible which is hard because every time open ai launches something it goes more viral
Travis: They are. Anyways, Brock.
Danchez: But last week they put together a Google Drive integration. I'm like, okay, what's up? most Most of them do that already. Of course, Google does, Cloud does, OpenAI does. But I'm like, a little nice nice touch.
Danchez: But then earlier this week, they drop a Canvas-like feature, which has been one of my favorite tools within OpenAI now, is the Canvas. I use it all the time.
Travis: Brock released a canvas like feature or they released a plugin into canvas.
Danchez: No, it's a Canvas feature.
Travis: Wow. Wow.
Danchez: And it works like Claude's does, which is cool because if you run code on it specifically, you can then publish it and give people access to it on the open web.
Danchez: Open AI doesn't do that, which is one of those things that I love about Claude is if you make a little mini web app, you can actually publish it and let people see it and preview it and play with it. Even without having a Claude account, which is kind of cool. So X did that.
Danchez: And I'm like, or Grok did that. I'm like, that's cool. Later this week, they dropped memory. I'm like, oh, they're starting to dip into the stuff even Cloud hasn't really figured out yet. Like it can actually, it has the, it can remember all the past chats now, the way chat GPTs does.
Travis: Wow.
Danchez: so And then yesterday they dropped the workspaces, which is like projects.
Travis: Projects, yeah.
Danchez: And I'm like, they're catching up with all the little tiny things that make the AI tools good that aren't new AI features, quote unquote. They're just making it more useful.
Travis: Like they say, it's easier to to copy and be successful than create your own ideas.
Danchez: So I just finished reading Elon Musk's biography. He's got a big one. I'd read the smaller one years ago, but now I just read the Isaac Walton Walters, so whatever his name is. I just read his big one.
Travis: Yeah, the
Danchez: And the very last few chapters were really interesting because Elon calculated a couple of years ago. He's like, well, we're going to compete with open AI. Yeah. But we have the advantage because we have all the data when it comes to self-driving cars that we've been working on. And we have a very smart AI team working on that. And we have years of we're years ahead on robot robotics.
Danchez: They have a chatbot. Which one's harder to replicate?
Travis: yeah the robotics
Danchez: Ours. He's like, so we have to go to war and make a chatbot that's better than theirs and keep our edge on where we're already at with everything else because it's going to be hard for them to keep do that.
Travis: and don't know where I don't know where rockets to Mars fit in and all of that, but somewhere.
Danchez: Well, yeah, they got all the internet infrastructure in the sky too, so I'm like, well, the one thing...
Travis: Bro, what if we're the civilization that actually starts sending sentient beings out? Because it's like, wait a minute, let's not send humans to Mars. Let's send robots to Mars.
Travis: And then there's a colonization of robots on Mars that are building.
Danchez: preparing the wait for humans to be there.
Travis: Yes.
Danchez: Dang, didn't think about that.
Travis: And then humans show up and the robots are like, I don't really like this.
Danchez: The colonizers are here.
Danchez: You're always colonizing. We're like, what the heck? Every time, dang it.
Travis: It's supposed to be this joyous moment when the humans finally arrive and then the robots are like, why do I feel like a second class citizen? We were the kings. Oh man. Sorry. My, my brain just went, movie.
Danchez: So Grok is making leaps and bounds. So now that I'm seeing Grok pulled this Hail Mary to get to catch up to everybody in the winter, and now it's starting to like add all the features that make these apps like really easy and nice to use, I'm like, man,
Danchez: But Elon, if there's one thing I learned from reading his biography, it's like the guy just pulls Hail Marys is like his thing all the time in every, in so many different industries. People are like, oh, that's going to take six months. He's like, you have three days. so And then literally goes, if they say no, he'll like go and figure it out. Like I think in, in one of the stories I heard that I was like, dang, that I didn't know about that one is they wanted to move a bunch of servers from Sacramento to like Portland.
Travis: I have heard this.
Danchez: and they're like, we can't do this that fast. like It doesn't take weeks. It takes months to do that. He's like, well, get a U-Haul truck. like what are you talking about? It's a rack of computers. Just move it onto a truck. He's like, no, it's more complicated than that. He breaks he practically breaks in there in the middle of the night, takes his pocket knife, unscrews the servers, goes underneath, unplugs them all. Nothing breaks. He's like, well, see, we can do it, and then rolls it out himself.
Danchez: and Then he's like, see? Easy enough. You don't have to bubble wrap these things. They're not that fragile. And then they did. They did it in like a few days. It also caused a lot of problems because they just didn't know how how how much was leaning on those servers. So it screwed a bunch of stuff up that they're still dealing with.
Danchez: But I'm like, it just shows you like the guy's tenacity is like always question the requirements. How fast can we actually do this thing?
Travis: Wow.
Danchez: But you're seeing it play out with his AI stuff right now. You're like, man, the guy's getting stuff done really fast. Google's like super slow.
Travis: Bureaucracy, though, I've worked for people like Elon and I'm like, hmm,
Danchez: Yeah. After reading the book, I'm like, no chance I would survive. You would fire me immediately.
Travis: Yeah.
Danchez: But, uh, oh, well, I'm not a Google fan either. I'm like, uh, I like, I'm cheering for open AI. I'll, I'll cheer for Grok. Google is the company that I don't really trust.
Danchez: only because they've just know been known to like use and abuse user data.
Travis: Google you don't really trust?
Danchez: Yeah. Yeah. Do you trust Google?
Travis: I mean, i don't know if you saw on the news, but Google just called out officially a judge calling them out as being a monopoly again.
Danchez: Yeah, I mean, all the biggest tech companies will get that to some degree. Microsoft has gotten that for a long time. Apple gets that sometimes. Google definitely has been getting that for a while.
Danchez: But because they're so big and so powerful and have so much money, they could just lobby the heck out of everybody. Yeah.
Travis: Do I trust Google? I don't know if, I don't know if, I don't, I don't know. if I, the more I use Google to search information, the more I go to chat GBT and I'm just like, man, this is so much better, but.
Danchez: I'm also just kind of hoping that open AI wins the day just to keep the balance of power little separated.
Travis: Right.
Danchez: I'm like, it would be nice if this the leading AI company wasn't one of the existing big tech companies, like with with Elon's thing or Google's thing or Apple's thing.
Travis: Yes.
Danchez: It would just be nice if there was another player to keep everybody humble.
Travis: Right.
Danchez: you know
Travis: Right.
Danchez: I'm like, that works out for us. but google did also have updates and all kinds of things they updated most of them weren't really that cool to me one had a lot of potential but i'm like nothing really affected my life as much as open ais because that's what i use groks were interesting by i'm like oh grok might steal me away i don't know someday they're they're not better but they're catching up to what everybody else has really fast
Travis: Right.
Travis: right
Danchez: Google upgraded their deep research tool and now includes reasoning. So that's gotten better. They updated their image effects or image gen three, imagine threat three, whatever they call it.
Danchez: They have like too many names for it, which is cool.
Travis: well
Danchez: It's better now. I actually use that more than chat GPT's image maker still because it's faster and it's really reliable. Yeah. because with chat gpt you just kind of don't know what you're gonna get it's usually pretty good but like it takes so it's so slow you only get one but with google's it's so fast and you get four that you can just kind of like throw something at it refine a little bit refine a little bit refine a little bit and then you have it really fast because it's prompted here and so so good so that is the one thing that google does and i'm like i use it all the time They released Firebase Studio, which is their own developer tool.
Danchez: Like a lot of people are using AI coding, specialty coding tools now, like Replit or a Cursor are kind of the popular ones. Probably knew new names for you.
Travis: Yep.
Danchez: People are doing the thing called VibeCode, and we talked about where they're just kind of like, oh, I'd love to build an app. Hey, AI, let's build an app together. And they're like making their own versions of apps. So Google released their own and it's free. So a lot of people are excited about that.
Danchez: vo2 their video model is now available but only through the api like why oh well so because it's no one's had access to it except for like beta testers and stuff so they still have kind of like the leading video model but everybody else is starting to catch up with them now too It'll be interesting if it's available via the API because that means you can, if if you're into the more automation tools, like a lot of people use n NAN and automation tools to come up with like little short form videos automatically.
Travis: Okay.
Danchez: So that's going to be like, they're probably going to fall in love with that because they'll be more reliable.
Travis: If they're good.
Danchez: Yeah. Yeah. That's the big question, right?
Travis: It's getting better.
Danchez: Yeah.
Travis: I see little videos kind of sprinkling through the internet that I'm like, I mean, it's getting better.
Danchez: Yeah.
Travis: It's not there, but it's getting better. Oh, yeah.
Danchez: Yeah, wait until you see the viral post I have for you at the bottom here.
Travis: yeah
Danchez: Yeah, but the last thing Google updated was this, is actually more of a big deal, and it's called the A2A network or agent to agent.
Travis: Right.
Danchez: This is kind of a new thing, but you have to like, think about it as like what APIs are. APIs, in fact, I can't even remember what API stands for, but we all know those are like the connections. So software tools can talk to each other, right?
Travis: right
Danchez: This is the AI version of that.
Travis: wow
Danchez: so that agents from one tool can talk to agents of another tool.
Travis: well
Danchez: They launched this, lot of big companies into in it, like Oracle, SAP, Salesforce, like all the big dogs in Google have like created this network. It's interesting because just before A2A came out, there's another thing called MCP, which Claude championed, but then OpenAI i adopted and was like, okay, we'll do this.
Travis: Wow. Right.
Danchez: which you're like, are there two competing networks now? So this is developing right now, but this is a big deal because as now we have these reasoning models that can make smart decisions and consider all the different options and move forward.
Travis: right
Danchez: They need the ability to have access to the different tools in order to make your life and your systems work really well. So these are the were we're essentially seeing this year in 2025, the groundwork laid for agents to become a thing.
Danchez: And like I said earlier in the year, like we were like, I was like, we'll probably see some early examples of this next year. I'm like, no, I'm like, man, maybe it'll be summer. I'm starting to see some early agents come out now. I think deep research is probably the best example of an agent we have right now that actually works reliably and does a lot of stuff on your behalf is kind of agentic.
Travis: Right.
Travis: ah ah gentic
Danchez: Agentic. But this new A2A network thing, Now we have A2A and MCP. I don't really know what the differences are. I'm going to to Google it. and I've seen some explainer videos. I just haven't watched them yet. But that's developing, and we'll see where that goes. Because, again, agents are going to be a big deal for marketers because they're going to be like little mini coworkers that we have doing stuff for us.
Danchez: It's almost like, you know, running an internship, you have like all these people that you can train to do things. You're like, if you have if you can if you can write the SOP, they can do it.
Travis: right
Danchez: And that's going to be what it's like here. It's going to be less important about your tech skills and more like, how can you create value and then delegate it?
Travis: The process. Yeah.
Danchez: And then how do you clean up when you have too many of them going on and they're starting to conflict with each other? Well, you're going to be the orchestrator now. So you have to start thinking about how to operate all that.
Travis: i recently saw that Google added AI to Google Sheets, or at least it's, maybe it's already out or coming out, but you can put the raw data in and just highlight the cells and actually communicate with AI what you're looking for.
Travis: and it'll plug the formula in there now.
Danchez: So glad you brought that up. That's actually the most exciting thing that Google dropped that I totally forgot to add it to the list.
Travis: I know. I was shocked you didn't have it on there because I'm like, well, this is what I saw that I thought, wow. I remember taking you know Excel classes in high school, learning formulas going, what in the world are we doing?
Danchez: Yeah, and I think formulas will still be important. But imagine like if you could just have critical thinking applied in there, and every time you recalculate the whole field, all the agents, not the agents, the but AI goes and recalculates too.
Travis: Right.
Danchez: For the less quantitative and the more qualitative things happening in Excel, because we use Excel and Google Sheets for all kinds of things.
Travis: yes right
Danchez: We use it as little mini CRMs. We use it to organize information. We use it for all kinds of things. But if you can have AI thinking in there in each individual cell and then recalculate, like I haven't explored this tool yet, but I'm excited about it.
Travis: right right like you're missing like where am i missing yeah where am i missing someone on this roster who needs to be called for trip or for a donation or for legacy partners because there are like you said a lot of people using it as
Danchez: Yeah.
Danchez: Mm-hmm.
Travis: checklists or crms or i mean you name it so to have ai plugged in there could be incredibly useful wow
Danchez: Mm-hmm.
Danchez: I think it'll be a great way to even like prototype little mini tools, right? Because a lot of people have done that traditionally, like, oh, if I can program it in Excel or Google Sheets or something, then I can mock it up and kind of preview it as like an MVP of of a tool, of of a essentially an AI wrapper tool where it's AI on the back end and you're just building some kind of process for it to think through and then deliver a good at the end.
Travis: right right right
Danchez: Well, like you can do that in Sheets now, which is really cool. So, and then you can, if it works, then you're like, oh, now I can spend the time with one of these vibe coding tools, build it for a kind of real.
Travis: Yeah.
Danchez: All right, poll of the week. I asked my audience on LinkedIn, and said the poll says, I want ChatGPT to remember. and then it gave me four options. Everything about me, only enough about me, only about work, nothing about me.
Travis: yeah
Danchez: percentages. It's funny. It actually pretty much went in order of the order that I put them in. Everything about me, 37%. Only enough about me, 23%. Tied with only about my work, 23%. And then nothing about me, 17% is how people voted.
Danchez: had 198 votes. So where do you fall on the list?
Travis: Everything.
Danchez: Yeah, that's where I fall too.
Travis: Yeah. When you told me that now chat has access to every single chat in terms of its memory, I was like, yes.
Danchez: Yes.
Travis: Oh, it makes such a huge difference. Yeah, I was stoked about it.
Danchez: Yeah. And It's funny, I had posted about it though the day before or the night before we talked about it last week, but it wasn't until we actually like just talked about it in the podcast episode last week that I realized how important this is going to be for people moving forward.
Danchez: Like how important this this one feature is going to be for marketers. You want to have it remember everything about you. You want it to remember the thinking and the processes and the your thoughts about how you approach campaigns, the back and forths you've had with ChatGPT.
Travis: Right.
Travis: Totally.
Danchez: This is what's going to be your, this is going to be a competitive advantage later on.
Travis: And you can ask it one question about challenging your frame of reference, your thought process, difficulties you've had in the past.
Danchez: It just is.
Travis: Remind me of where we fell behind in other campaigns or scenarios in the presentation.
Danchez: Yeah. Yeah.
Travis: And it'll say, don't forget about X or Y. Don't forget that you often tend to lean on this side or become nervous when this situation happens.
Danchez: Yeah.
Travis: So you're like, oh, that's That's That's right. it's right I mean, it's huge.
Danchez: So this will be a big problem. The biggest problem, I've i've seen some of my my LinkedIn friends talk about all kinds of problems this is I don't think privacy, I'm like, whatever, like there is no privacy. If people are concerned about privacy with personalization, I'm like, do you realize that OpenAI already has access to all your chats?
Danchez: What privacy? You're just now aware, more aware of what it already has access to.
Travis: It's funny.
Danchez: Like, stop. You're already giving it everything.
Travis: Right.
Danchez: If you didn't want it to know it, don't tell it because you already were.
Travis: Right.
Travis: Right.
Danchez: And now it just is now it's going to work for you instead of just open AI. so But the one problem that this will have, and I i think we talked about it in the last episode, but i'm going to bring it back up because this is going to be important, is that who gets to keep the data, the company or the worker?
Danchez: The company would want to keep the data because that's some of it's proprietary to them and it's company data about these conversations you're having about campaigns. But it's also employee data because it's my personal experience working with this information on the company that I want to be able to bring, kind of like I can bring my experience from company to company.
Danchez: That's why people would hire me. But now I have my AI team that has that information too. So who keeps to keep the data?
Travis: I was just talking to a past employee from my last job who has all of the relationships that they've been housing as a development officer or account manager.
Travis: And now they're going to another field, not as a nonprofit, but someone who could benefit in having those relationships. You can't ask the person, well, hey, you're not allowed to keep all these relationships locked in your mind.
Travis: to then be able to reach out and be like, Hey, we offer a product for people like you. Uh, is that a conflict of interest? Not a conflict of interest. Is it? Yeah. I guess a non-compete, which is what the other organization would probably need to sign, but how can you can't, you can't tell somebody, i mean, I guess they do.
Danchez: Right.
Travis: This is why they sign non-competes, right?
Danchez: Right. Because there's there's no way to separate it before, but now there is on this.
Travis: Like,
Danchez: The more I think about it, the more I think that I personally would want to advocate for the company not owning the data and the employee owning the data. The reason why, now companies will for sure do everything they can to own the data in the interest of the company.
Danchez: The reason is, is because the AI already benefits the company. there's or It's already stacked against the employee, this whole AI movement, right?
Travis: Right. Right. Right.
Danchez: I'm like, it would be nice. um'm like I'm like, I hate, like if there there is ever legislation that should be like, hey, when it comes to a co-pilot, like the employee gets to keep the data, it actually makes it so employees stay more valuable to the companies rather than them laying off everybody and just relying on their own AI agents over time.
Travis: right
Danchez: You know what i'm saying? Like it actually gives some power back to people rather than all the power going to the companies.
Travis: Wow. Wow. Right.
Danchez: But because I feel it myself, I use my own. i have I have a company account that's a plus account. I have my personal plus account. I use my personal one because it's already been honed and it's way easier and faster.
Danchez: But of course I've had conversations.
Travis: wow
Danchez: not i've had i I don't put proprietary data from social. They have policies in place where I can't put like certain types of information into AI. And so I follow those.
Travis: right
Danchez: But I'm still using mine. But obviously like I get to keep my own personal account. They have the company account too, but I don't use it. I use it to how it's like custom GPTs we give away to customers there.
Danchez: Anyway, so I think that's going to be debate moving forward. Last up is the viral post of the week. This is one i actually, it wasn't super viral, but I shared it because it was,
Danchez: it blew my mind. It was like a good, like level up for like how realistic AI video is becoming in a different way than like how visually realistic it is. Like VO2 from Google. Have you watched the video yet?
Travis: Yeah, I Grief.
Danchez: So it is a video of a woman. It's just a short video, but she has so much more emotion on her face. She goes from kind of like smiling, but you could see there's something wrong in her eyes. And then she you could see her face like change to, I don't even know what to call What do you call that emotion?
Danchez: and Like she's going to start crying grief.
Travis: brief
Danchez: Yeah. Like she's holding it back and then she starts like breaking down and she covers her face. It's so real. It evokes emotion in you, but it's not a real video.
Danchez: And I'm like, my gosh.
Danchez: How did it make you feel when you watched the video?
Travis: I had already been experienced, like I've already been prompting videos like this. So you could tell it was AI, but it's it's the best, most emotional clip that I've seen so far.
Travis: Most of them, the face is really stiff, still provides some, you know, emotion and usually goes to a smile, but the level of detail with this lady's face from a smile to her eyebrows creasing in her eyes closing.
Danchez: Yeah.
Travis: And then the head tilt down with her hand coming up to her face, which often most people do when they're about to cry is cover their face.
Danchez: yeah yeah
Travis: Uh, I hadn't seen that before and they actually got the hand right.
Danchez: It just all the little tiny muscle movements, because, you know, there's, there's whole books on like the hundreds of muscles in your face and what they mean, because you can't hide a lot of your facial reactions unless you're really trained.
Travis: Right.
Travis: Right.
Danchez: So a lot of these, this is how people, I think women in particular are good at picking up the tells, even if they can't consciously put it together because your face is so consistent from person to person about what it means about how you're actually feeling inside.
Travis: Right.
Danchez: That's what it's actually starting to pick up on. is the tiny little nuances of the muscles in the face to be able to really communicate real emotion.
Travis: Wow.
Danchez: It was one of the first things that I've seen in a while that like triggered me in a way where i was like a little scared. I've seen a lot of AI stuff, so I rarely see anything that I'm like, ooh.
Danchez: But this one I saw and I'm like, yeah, crap.
Travis: it.
Danchez: I'm like, I don't know if I like this one because there's obviously been super realistic voices with Sesame that are really insanely real.
Travis: Yes. Yes.
Danchez: You're like, my gosh, it just sounds like a person. And then we have chat GPT and the words are starting more and more like a human.
Travis: Uh-huh.
Danchez: And now you have the video ability and you're like, man, it's probably only a year or two away before the three of these things come together to just be able to have a FaceTime with someone that looks and feels and sounds like,
Travis: Uh-huh.
Danchez: quite real quite convincing of course this is going to be a big problem for fraud and all kinds of stupid stuff like that but this is an optimistic show so but realistic obviously it's the great tool and there's going to be people are going use it for evil people are going use it for good i'm um'm optimistic that'll be more good than evil i think this will be helpful for bunch of different customer service type things
Travis: ahha
Danchez: I'm scared that people will like replicate celebrities and make them look like they're endorsing things that they're not, so that's going to be a thing.
Travis: Yeah, that's definitely going to a thing.
Danchez: But I think... I'm I'm not a really big fan of like the Haygen stuff. Have you looked at that yet? Where you create like the digital twins?
Travis: Yeah, yeah.
Danchez: so Like it's one thing to like record a message of yourself and then use HeyGen to like change the language so you can speak other languages.
Travis: Yeah, it's awful.
Danchez: I would probably just still make it clear that like I don't speak this language. I'm just for the for the sake of others. Kind of like Mr. Beast dubs his videos. I'm like, I think that's a considerate thing to do.
Travis: Yes, right.
Danchez: Obviously, if you have a business, you need to actually back that up. So you can't like market your whole business in Spanish if you're if you can't handle Spanish speaking customers, right?
Travis: Yes.
Danchez: Like, like come on, think about that. Some companies do that. but I don't think it's, it makes me uncomfortable to try to create whole videos of myself talking about things that I didn't say and didn't record. I still have to wrestle through that. What do you think?
Travis: Interesting. I didn't know that you were finding. Is it an ethical standpoint or just just I mean, it just feels disingenuous and not authentic.
Danchez: I feel like it feels disingenuous, which is interesting because like I'm already creating images of me not doing stuff and I'm like Photoshopping my face onto it.
Travis: And and and written text.
Danchez: But I feel like, yeah, and written text, but I'm watching it go out.
Travis: What's the difference between watching watching a video of yourself and then approving like, yep, that sounds like me.
Danchez: but the
Travis: It looks like me versus it. Oh yeah. Look at that text. It sounds like me. It reads like me
Danchez: I don't know, man. There's just something different about it.
Travis: I don't like it because it seems fake.
Danchez: Like even now,
Travis: The only thing that bothers me is that if I'm sending something that feels disingenuous because the quality is so off that it's like, it's hallucinating while it's trying to use marketing campaigns, similar to the image I sent you of the workout people and there's limbs flying everywhere.
Danchez: Yeah. Yeah.
Travis: I'm like, this isn't real. And Hey Jen, it's not real. What I've seen people, how they use it is they'll do a really fast clip in of the person saying one or two words like, Hey, look at, and then they'll clip in a B roll shot and then they'll clip back to the Hey Jen video for a second and a half and then clip out so that it's not enough time for you to pick up on that.
Danchez: Yeah.
Travis: It's not real unless you're professionally trained like me perfect personally, personally defined professionally trained like I am to spot those things.
Danchez: Hey, Jen. Yeah.
Danchez: Professionally trained by YouTube and Casey Neistat. Thank you.
Travis: Eat
Danchez: Yeah, I don't know. It makes me uncomfortable. Maybe it's because it's starting to actually encroach on my own craft, which makes me uncomfortable.
Travis: oh
Danchez: I'm starting to like eat it myself because now anybody can be decent on video.
Travis: your own words.
Danchez: I'm like not really good on video, but I've obviously put a lot of time into like communicating on video to the point where if people meet me in real life, I'm not the same person as im as I am in real life as I am on video.
Danchez: I've had enough people tell me that now that I'm like, crap, okay. it's just an introvert thing. i can perform on video in the way that I want to have been seen, but in real life, I can't react fast enough and like, and see myself through a screen or whatever to what I'd like to be perceived as.
Travis: Right.
Travis: A lot of people get this. It's not, it's not just you though.
Danchez: Yeah.
Travis: I mean, Ted talk speakers, pastors, anyone who speaks professionally or communicates with some level of authority and, you know, not fame.
Danchez: Yeah.
Travis: What's the other word? Just,
Danchez: Yeah, you have that stage presence and the authority.
Travis: Yeah.
Danchez: Yeah. I don't know. But anyway, the video thing makes me uncomfortable. I think I'd have to wrestle with it more. And I can't come to a conclusion yet because I didn't even really start wrestling with it until we started talking about it just now. i'm like, I don't know.
Danchez: Now that it actually can get to a point where it's like actually realistic in the future, I'm like, I haven't wrestled with myself being on video and recorded that way.
Travis: i have i just haven't i haven't I haven't wrestled with it because I haven't seen even Hey Jen create a realistic double of me where it sounds like me, it looks like me, where even I question, did I record that?
Travis: Once it gets to that place where I go, did I record? i mean, it's so convincing that I'm convinced, then I'd probably be okay with it because it seems less disingenuous if even I go, oh yeah, that's exactly what I would say or how I would say it.
Travis: But right now it's like, that's,
Travis: Ew. I wouldn't want anyone to see that because that's not even remotely close to me.
Danchez: Yeah, i'm not I'm not talking about the ones that don't look like you. I'm wrestling through a future scenario when it is possible to make a video that looks like you.
Travis: Oh, okay. Okay.
Danchez: Right now, the photos can't even freaking look like you without a ton of work.
Travis: No, it's true. They're getting closer.
Danchez: So...
Travis: Every week it seems like there's a new update that I'm like, that I mean, that's the closest one so far.
Danchez: know, that's why I wanted to share this post because I'll link to it in the show notes. You have to watch it because it's just, you're like, oh, that was next level emotion in the video.
Travis: Yeah.
Danchez: Dang. Okay, another step. Again, we haven't had the chat GPT moment for video. We just had it for images.
Travis: Right.
Danchez: And it'll continue to get better, of course, but we'll have that for video maybe this year. We'll see.
Travis: Maybe.
Danchez: All right. This has been the longest podcast we've had on the show for our bot bros, but maybe ever in the whole show. We've just crossed an hour. So we'll wrap it up here. Thank you, everybody, for joining us.
Travis: There it is. See ya.
