Focus On Your Biz Focus Group with Mackenzie Bowe, Founder of GPTBoss

Date recorded: January 26, 2023
Length: 48:46
Transcript: See below
Topics covered:

• Tour of GPTBoss
• Bertha.ai new tools (Chat, and image search)
• What is prompt engineering?
• Behind the scenes of creating a custom prompt
• What is a token?
• What does this AI revolution mean for us?
• How can we use AI in our businesses?

Resources mentioned:

• GPTBoss
• BerthaAI
• The Sizzle Newsletter

Join us live next Thursday at 5pm ET!

Correction: I mistakenly said the Bertha AI image search was using Stable Diffusion, but it is actually Lexica. To be fair, I thought I was wrong and definitely said “Don’t quote me on that!” 

Note: I inadvertently cut the recording a bit earlier than intended. (Whoops!)

@0:02 – Stephanie Hudson

Where were we? AI. Welcome everybody to this week’s focus group, where we are gonna talk about some geeky stuff as usual.

Today, we’re gonna talk about artificial intelligence. We have a lot of the usual misfits and bandits here, but we also have a special guest, Mackenzie Bose, the creator of, oh, Ken’s back.

What happened? We lost Ken. There he is. Okay, he’s the creator of GPT Boss, which is super cool. Vicky introduced us to this last week.

She mentioned it and boom, look, we got the creator right here. So Mack, tell us a little bit about GPT Boss and how, like, just give us a quick little overview of what the app does.

@0:48 – Mack

The sound bite that I perfected on TikTok is GPT Boss. It is an app that gives you access to a suite of AI employees and generators that automate small business tasks so you can run a company by yourself.

@0:59 – Stephanie Hudson

Okay, so that really means… It’s nothing to anybody though, when you travel it off like that. So can you do a screen share and show us?

Because the cool thing is that, I mean, there’s so many AI tools out there right now, right? The things that can help you run your business.

So what Mac has done is he’s built, it’s maybe the reason I like it so much is because it reminds me a little bit of FocusWP.

So basically he’s giving you your whole team right there, except for each different team member is its own little AI persona and character that has presumably been trained or led to a certain style type of knowledge, whatever.

I’m talking like an idiot right now. Cause I haven’t, I just learned about this a week ago. I haven’t even had time to like fully try it out.

So is that accurate? Should I share my screen? Yeah, please do.

SCREEN SHARING: Mack started screen sharing

@1:54 – Mack

All right, how are we looking?

@1:57 – Stephanie Hudson

Coming up, there we go.

@2:01 – Mack

Okay. So yeah, this is like the main dashboard. This is what it looks like after your account. There’s oh, hello.

Are we still on? Yeah. Yeah.

@2:23 – Stephanie Hudson

Guys, this doesn’t bode well.

@2:34 – Jason Rutel

Just now he’s back.

@2:42 – Stephanie Hudson

So basically things are going really well tonight.

@2:45 – Michael Lofton

Yeah. Well, the stars are aligned this week, so we’ll see.

SCREEN SHARING: Mack started screen sharing

@2:48 – Mack

Maybe I was making like an infinite zoom here. So okay, wish me luck. Anyways, so this is the employee recently changed this.

The idea was that this would just be a list. of every employee. I do have to do work on like filtering the content.

But when you log in like this is all empty. This is all empty. And then you have Luke Fraser as your kind of default chatter.

And I need like a UX audit horribly badly. But like this is kind of the main feature of it.

@3:19 – Stephanie Hudson

Right.

@3:20 – Mack

So when you when you log in, the idea is that this is like your first time in business. You’re like, OK, I am going to foray into something that I haven’t done before.

And that is like entrepreneurship. Like how like where do we even start with this? And the reason for that is because I think a lot of people are going to get laid off this year.

Like we’ve already seen 200000 engineers get laid off from Fang and everyone was like, oh, yeah, engineers are recession proof.

Like nobody would ever fire a software developer. Two hundred thousand in two months laid off. Probably not true either.

There’s going to be a lot of people that get laid off and they’re like, well, what do I do now?

Right. And maybe I’ll try being an entrepreneur. So this is to like help them, right. It’s to like move through all of the grit that we all had to go through, like figuring this out and failing and falling on her face.

The idea is that it’s supposed to just help you get it done. So you talk to Luke. Hey, Luke, I’m thinking about starting a business selling crocheted sweaters.

@4:30 – Stephanie Hudson

Pick the hardest word.

@4:34 – Mack

I’m working on getting this streaming, so it will stream back like chat GPT soon, but it’s just like processing this idea, generating the whole thing, and then it’s going to pop back in the chat.

It’s a lot more impressive on TikTok where I have like editing tools to control the flow of time, but.

No, that’s okay. So. He is here to help you. Like you have an idea, like you’ve seen like something on TV or there’s something that you could do.

And Luke is like to bring it to reality, right? Like where do you get started? Well, crocheted sweaters. Does anyone even want to buy that?

Right. So he lets you know, you got to research your customer demand. How much are you going to sell it for?

Like, is the market able to take the price that you need for your crocheted sweaters? Here’s how you plan for growth.

Here’s how you focus on marketing. And the idea is that the chat like goes back and forth. Right. So, well, how do I actually do customer research, customer demand?

Then you kind of go back and forth with him on on this chat where like you’re asking questions, he’s providing answers, and then you kind of learn entrepreneurship.

And so Luke is the one that’s there by default. Let’s see what he says back. before moving back to the employees.

I could say that a common problem, a lot of people when they get started with this, they get themselves into a situation where the worker is saying like, I’ll send you an email about that.

Luke would say, let’s book a Zoom call to talk about this. Luke can’t do a Zoom call.

@6:22 – Stephanie Hudson

He’s like an AI.

@6:24 – Mack

But yeah, but there’s like the picture here and there’s like this description and they’re like, is this like a real person?

Like, can he do a Zoom call? And so they email me and they’re like, he’s not getting on the Zoom call.

I’ve been waiting for 20 minutes.

@6:36 – Stephanie Hudson

What’s wrong with your app?

@6:40 – Mack

And yeah, so I included these kind of responses to say like, okay, like they’re definitely not real. They will never email you.

They talk about emailing you. Not all the time, but like often enough. I hear about it like once a day, at least.

@6:53 – Stephanie Hudson

Anyway. It’s kind of fascinating that people, I mean, like that’s kind of cool that people are actually like, yeah.

It’s a sort of a two-edged thing, right? Because you’re working well enough that people are thinking they’re actually talking to a human being.

@7:07 – Mack

So that’s kind of cool. Yeah. And the advice, some criticism that I get on this is that it’s basic, but the person who’s saying this is basic has already been through the ringer, right?

Like how do you know that it’s basic? Because it’s true, right? Like you’ve done it and this is where you start.

So if you like, there’s nothing wrong with it being basic especially because it’s for like a first time person.

@7:28 – Stephanie Hudson

Where if you’re- I thought you meant it was basic, like wearing UGG boots and having Starbucks.

@7:34 – Mack

Yeah, like the business advice, like, well, I think some people really like magic bullets, magic bullets for business, like researching customer demand, like, oh, there’s just a website that you buy a report from and it’ll tell you how to make a million dollars.

A lot of people feel this way that there’s like some magic bullets they have to find. And in reality, like when they buy the course, what they find out is like, oh yeah, you gotta do the work.

You gotta like do cold calls. And here’s how you do cold calls. Here’s how you feel less bad doing cold calls because it’s horrible to get rejected all the time.

But the reality is you just gotta like do the cold calls. And so I think part of that like comment of like, this is basic info is somebody saying like, well, it’s not giving me magic bullets, right?

@8:12 – Stephanie Hudson

Right. Okay, so then you can add on to your team.

@8:18 – Mack

Yeah, so we’ll go back to the team. So I’ve added a few in here. The, whoops. So the way that this works is you start with just Luke and then there’s nine slots.

@8:30 – Scott Tambling – DesertDigital.Design

Nice.

@8:31 – Mack

In the future, like, so if you fill it up, it’ll just add kind of more slots so you could edit these.

But in the future, I’m going to limit it to like 10. And then if you need an 11th guy, whatever, five bucks.

But you can just kind of throw them in for right now. And so the way that this works is that you just have like easy access to them, right?

Like it’s free. It’s like during beta, we’re still really early. So it’s free. Any of these employees you could just take and fire at your will.

And this is kind of the gamut of. Jobs. There’s an accountant, legal advisor, SEO guy, a react developer, because I’m like snooty about development.

I might put a WordPress guy in and he might be better than the react developer because there’s a lot more WordPress content, but I just haven’t.

Haven’t done it yet. Um, marketing manager, the business coach, we saw the copywriter and then some of them, like I, I have to duplicate all of them, there has to be like a guy and a girl kind of for each role so that you can build your own team, a big complaint that I had for a long time was like, why are they, why is all the employees men?

@9:28 – Vikki

Like, yeah. Can we say this is full of testosterone when Mackenzie started?

@9:35 – Stephanie Hudson

Why are there no redheads in this?

@9:38 – Vikki

Yeah.

@9:40 – Mack

That’s kind of like what I was thinking is like, like, if you want to ask weird questions about like, why isn’t it women?

And then I don’t know, like it’s an AI. They’re not real people.

@9:51 – Scott Tambling – DesertDigital.Design

So it’s not, I think it’s brilliant because you really, like my biggest issue is, um, it’s you’ve taken, it made us so user friendly and I maybe, like you were saying, maybe too user friendly for some

people if they actually believe that they’re real people. Quick question is, I know you’re using the TBT API, but do you have to like, like use funnels or filters with each one of these employees that, that will get different responses or why are there different?

Why are there so many different employees when basically it’s pooling out of the same place? I’m just curious.

@10:27 – Mack

So prompt engineering. They like, that’s another part of the user interface that I wanted to do, because you could theoretically go to the GPT-3 playground and like engineer your prompts and do these generations and save all of your prompts into like a text file on your PC and then like load in your prompts that work really well.

And play with them. And, but that’s like such a CF, like who’s doing that. Um, so the idea is that these employees are like tested prompts that I’ve engineered that I’ve had really good results with for these employees that you just get to like start talking to.

Right. You don’t have to do any of that, like testing or kind of. Like, um, open AI calls them identities.

You don’t have to do identity management for your prompts.

@11:06 – Vikki

Mackenzie, I have a question about that. So I know like you’re doing the prompt engineering, but with the number of people that are using GPT boss, are they also learning from the people that are using the tool?

@11:20 – Mack

I don’t have any pipelines for that set up yet, but it’s something that I can do. So we’ll talk about the terms, right?

Because if you’re putting your business idea in here, like what does that mean for your copyright? Right. Right. What does that mean for your intellectual property?

What if you’re like an author, right? What if you’re using the copyright or, well, we haven’t even gotten to the generators yet, but there’s a lot of like prompt engineered like text generators that do a lot of different things.

Like the content writer, I think is the most powerful one. There’s a third grade and fifth grade content spinner.

So you could like simplify your words. So you’re not like getting like zozled by, you know, fancy $5 words.

And this is, well, the content spinner is really good for websites too.

@12:00 – Stephanie Hudson

Words, sozzled.

@12:02 – Mack

Sozzled, yeah. You know what I mean by that.

@12:05 – Michael Lofton

And is there a plagiarism at the same time tester?

@12:09 – Mack

So there’s no plagiarism test on here, but I’ve had people use their results, like people are using the content writer to write blogs, and they’re putting it into the available like AI detectors, and they’re passing it like 94% human.

Okay. It’s because of the prompts that I’m using. So chat GPT is the, I think, what most people are using.

And these generators are a lot like the prompts. So the chat GPT prompt is act as an AI assistant.

Like that exact string goes with every message that you send to chat GPT. So it kind of acts a little bit differently, but these generators and the employees act as like people, or the generators are like, so for the content writer, the prompt is act as an AI content writer, but it still passes with like 94% human because the perplex-

complexity of the output is different than what ChatGPT does.

@13:03 – Michael Lofton

So could you dive in like a content writer prompt?

@13:06 – Stephanie Hudson

It is- I would just like to actually pause for one second, Michael. I want to make sure that we haven’t lost anybody here.

Raise your hand, ask a question in the chat, speak up. If you need more clarification on how these terms are being used, do you need to understand more of what he’s talking about, prompts, prompt generation- Well, go ahead and explain prompts.

Okay. Could you then explain that, Max, in sort of semi-layman’s terms? I mean, we’re all geeks here, but this is new tech.

@13:44 – Mack

Sure. So the best place to do it is actually not on my website, but it’s on openai.com slash playground?

Nope. Yeah. Oh my goodness.

@14:01 – Stephanie Hudson

What’s wrong?

@14:02 – Mack

Oh, just so many buttons I have to click to get to this thing.

@14:09 – Stephanie Hudson

I mean, I feel your pain.

@14:18 – Mack

Oh my god, are we stuck here?

@14:21 – Stephanie Hudson

Okay, great.

@14:22 – Mack

So, finally. And anybody can go to this playground? Yeah, anybody can go to this playground. So, the idea, like basically the way that my app works is that there’s like the preamble and then the user entered input on the other side.

So, for the content writer, the preamble is act as an AI content writer, respond to any requests for content and return at least 500 words in markdown.

Something along these lines. And that’s like on my end. And then somebody writes into the chat box and they’re like, please write an article about AI prompt engineering.

So these go to open AI together.

@15:19 – Vikki

And I love that you say please because I always say please too and thank you.

@15:23 – Mack

Me too. Yeah. So in this case, these are other parts of the prompts on the sidebar here. So this ended early because I had a maximum length of 60.

It goes to 4,000. So let’s keep it going. I’m adding this feature soon, like this like streaming kind of one token at a time.

So TBT boss would look like that. But there’s differences in this. So that is like one prompt. That’s like the content writer prompt.

And then another generator that I have is the fifth grade content. So rewrite this content to be appropriate for a fifth grade reading level.

And then I like to pass quotes around like the thing that I want rewritten.

@16:20 – Michael Lofton

So if you want to adjust your prompts as far as the variability of them, can you do that?

@16:26 – Mack

Not right now in the app. So like in my app, there’s no, I’m thinking about putting in a toolbar like this, but I have like a thousand other things to do first, like fix this email issue.

@16:37 – Michael Lofton

Yeah. And the reason I asked that is for an example, if you want the response and content writing to be first, second or third person, for example.

@16:47 – Mack

So you could do that, right? In the content writer.

@16:55 – Michael Lofton

So you can add a variable on top. Okay. All right. As well as maybe. Email or mail voice for.

Yeah. Yeah. Okay.

@17:06 – Mack

No, my site just crashed. Okay. We’ll try it in here. So because it’s pretty much the same thing, it’s the day for it.

Yeah. AI content writer. Um, so like, imagine that this is like my prompt and then your prompt could be, um, please write a 500 word article, uh, explaining why Harvard is a great school.

And the first person is a female voice.

@17:42 – Michael Lofton

Excellent. Okay.

@17:43 – Mack

Right. So you can kind of do this on your own, but the idea with GPT boss is that you shouldn’t have to know that you have to do this right.

It’s like, you should just be able to say like 500 word article Harvard. And that does it for you.

@18:03 – Michael Lofton

And you mentioned something about your token. What do you get for a token?

@18:06 – Mack

So tokens are, tokens are parts of words. The AI doesn’t consider like a word to be a word or a letter to be a letter.

It just, it doesn’t think like that. We get really like not good results when we force the AI to count things in this way.

So we use something called a semantic token, which is a small part of the word. It’s about 0.7 words, right?

So some examples of like what a token is, is the number here like 150 might be a token by itself.

A space, capital N-O-B might be a token by itself. The whole word Nobel might be a token, very small chunks.

So on gptboss.com, because I’m using OpenAI, I have to pay OpenAI for their tokens. They like, they bill me, right?

All of these generations I had to pay for. And so on GPT-BOSS, you also have these tokens and it’s just like feeds forward on that cost.

And that cost is like, the only thing that it really pays for is just the electricity of running the AI.

@19:13 – Stephanie Hudson

Right, and those, what you just generated that had the green text, that probably cost you a couple cents, right?

Like that’s, it’s not super expensive. Not at all. Small little things like that. What, but once you start talking about like having multiple clients saying like, write me a thousand word blog post and you’re gonna pick and choose all and rewrite a bunch of stuff.

And you’ve got like hundreds or thousands of people doing it. Like that’s when you start to, at scale, get into some dollars.

And so that’s where you get these apps that are charging money for things, but also there’s added value because of like, you’re not just going to the playground or to chat GBT or whatever and train.

meaning it while you’re asking the questions. So apps like Boss or like Bertha, for example, is another one where they’ve set up a structure for you and like served it up because we have sort of an idea of what you need already.

So we can kind of get you in the ballpark and then you just feed in what you need. Did anybody see my Max Headroom joke in the chat?

Because that’s really what I feel like is happening to David. Is there anybody old enough to know what Max Headroom was?

Yeah, I saw it.

@20:32 – Mark-Andrew

It was just smiling, laughing and emotional.

@20:34 – Stephanie Hudson

Because what is happening, David? It’s bonkers what your video is doing. They need to plug your thing in tighter.

That’s actually what it looks like to me. Like, does your camera have a loose connection?

@20:46 – David Browne

No, I have no idea. There was still no internet on as well.

@20:50 – Stephanie Hudson

I might have too many things roaming or something. No, I just suggested you maybe dipped it in Guinness, David.

I think that probably happened. Okay, so I wanted to show you a couple of things too. I wanted to pop something up here with Bertha just to show another sort of thing.

Hang on one second.

@21:19 – Michael Lofton

Does anyone currently using AI in their business? I mean, except for the one that owns AI or has a business that deals with AI.

Okay, Vicki, you do, right, in your work.

SCREEN SHARING: Stephanie started screen sharing

@21:34 – Vikki

I use it to find like synonyms because I deal with search. So I need to find like as many alternate words for, you would know with SEO, for as many alternate words for concepts as I can.

So I’m often using it, you know, what are other words I can use for. you know, insurance policy or whatever, words or phrases.

So I use it a lot that way. But I’m very, you know, working for a company like State Farm, they’re very wary of things like OpenAI right now.

So, you know, making sure that we don’t put any State Farm code in there or anything like that. So I kind of having to use it on the DL, but I do use it.

@22:31 – Stephanie Hudson

So this is just a simple example, a more simplified example of what Mac was just showing. So here we’ve got this like city-based pages.

So if this is something that you want to add to your site for the SEO value and whatever, or if you’ve got somebody that one of your clients is doing local, so you can come in here and you can put their company name or yours.

You could put what city it is and then a description of service, what’s the call to action. keyword and it’s take, so it’s basically giving you like mad libs.

For those things, you could also go into like the chat or the ask me anything or chat GPT or whatever, you know, like any of these things you can go in there and say, like, create a city page that follows the, you know, that does X, Y, and Z for the brand of, for the brand, blah, blah, blah, in the city with this service description and make sure you include this call to action in these keywords, right?

You can write all of those things out and give it those instructions and get similar or, you know, results, but these are, this is.

A prompt that has been created and trained so that it will spit it out in a certain format. We also have one coming, for example, for, I don’t think it’s hit yet.

Um, for generating, uh, like a description for your, uh, course, like a course description, cause we’re doing a little partnership with lifter LMS.

So it’ll ask you. like what’s the title of your course? What’s the certification? What’s this? This asks a couple of questions again, Madlib style.

And then it will spit out like an introduction, a bulleted list, and a certificate, like whatever the other, it’ll format it in a certain way.

So it’s basically teaching the AI how to format its answer to you based on the context of where you’re inputting these things.

@24:29 – Vikki

So those questions like front loading the prompt? Yeah.

@24:35 – Stephanie Hudson

Yeah. Yeah, so these are all created with that stuff built in the playground like Mac was just showing. So we go in there and we come up with something and we structure it in a certain way.

So like we give instructions, and then we give an answer, and we format in a certain way. And then it’s just like with a little kid, like when you teach about it.

tie their shoes or something like that. You say like, you do this, then this, then this. Okay. No, you do it and they do it and it’s a mess.

And you go, okay, not quite let’s tweak it. And let’s try this. You know, and you just keep doing that, doing that, doing that until the kid can tie his shoes 10 times correctly on his own.

And then he can go to school and he can be okay. You know, or whatever it is. So that’s the, that’s basically what’s happening in the background.

Now in Bertha, we don’t control the tone and we don’t control the personality or like the, the skill set, basically.

That’s what Mac is doing. So Mac is, is telling these people, you know, cause you can, you can say like, uh, you to the AI, you just basically tell it what it is or say act like, or, you know, so you say like act like a developer and answer this question versus act like a designer and answer this question.

You guys can see how that’s like, like what, what’s involved in creating a website to a designer is a different question.

than to a developer. Theoretically you would get different answers from different people on his team. Whereas with ChatGPT, there’s no personality or like skill set or job description included in that innately, again, you can still type all of these things in.

So all of these apps and tools are layers really. They’re layers on top of this technology that is building in shortcuts to get you to where you need to go.

And so if you are looking for a team like that, like where you can just ask different questions to different people without having to try and break down like all this stuff.

So anyway, does that help clarify it to you guys?

@26:49 – Scott Tambling – DesertDigital.Design

I mean, I understand all this theoretically and it really is bringing stuff home about what all these apps are doing.

It’s just really making like Bertha simplifying the prompt. So we don’t have to learn them all and making it easier.

When are we going to get a full tour of Bertha? What’s going on with all this? Sorry. I’ll take it away from Mac.

Sorry about that. But I didn’t realize you get pages and, and doing it’s all very web specific. Obviously I haven’t dug in yet.

@27:21 – Stephanie Hudson

So, yeah, so because, because Bertha is a WordPress plugin, um, that’s why it’s very web specific now. I don’t, I mean, I mean, you guys are coming up with the Chrome edition though.

Right. We are adding that on too. Yeah. So I didn’t really clarify with, with Mac too much. Like this isn’t really a sales pitch either for GPT boss or for Bertha or for any tool, what this is to me that I wanted to come together and talk about tonight is like education, like let’s start demystifying this stuff a bit and helping people to understand it, we are on the cutting edge, you guys like, this is a time like th this is one, this is the pivotal point.

in technological history. Not to be overdramatic or anything, but it is like, this is like, I mean, when the internet was created and how it changed everybody’s lives or when the iPod or iPhone were created and it made such monumental changes to so many people.

Like this is one of those pivotal times in history. And like, we are geeky enough to kind of understand it and ride this wave.

Like that’s huge. You know, like that’s so cool to me. And so if we can sort of like break through the barrier of like figuring about understanding it, because we can much quicker, but let me tell you it’s coming for your mom.

That’s not like a, your momma joke. Like it’s coming mainstream. It will be mainstream. Like what ChatGPT did more than anything else is like, forget the functionality of what it did.

It broke it out. It broke it out and put it viral. It still is. inaccessible and mysterious and not really like, that the use case is not clear to most people.

A month ago I put on my Facebook, like does anybody like non-geeks, do you know what chat GPT is?

And all you bozos, my whole thing was filled with people explaining to me what it is, like I didn’t know.

I wanted to see like how many people knew about it. The only people who weren’t geeks doing that, geeks explaining to me were people who were like, never heard of it.

So it like, as much as we see it in our feeds and things, like it really isn’t full blown, it’s like mainstream to geeks, but it hasn’t really broken through to like mom and dad yet.

@29:46 – Michael Lofton

You know, like your grandma. Well, I asked the mainstream geeks in here who’s using it and it’s very low percentage.

Yeah, that’s what I mean. So it’s kind of curious, what’s holding you back? I can’t believe how much I’m using it.

I’m using a daily research.

@30:00 – Stephanie Hudson

It’s great for research. Like I, I’m so glad I’m sitting down to hear that Michael Lofton is using artificial intelligence.

I know. Well, I’m always 10 years ahead of shit.

@30:11 – Michael Lofton

So.

@30:14 – Scott Tambling – DesertDigital.Design

You’re almost crazy not to, you know, I I’ve been playing and tinkering with Jasper for a year, but not until the last, I kicked myself.

It’s like, why am I spending time writing or doing any of this when, and again, it’s not a replacement, but it’s an absolute enhancement.

And I think we need to, we have to embrace it and not be scared of it because as technical people and creatives, it can be a great tool and, and yeah, you’re right.

It’s coming for your mama.

@30:39 – Michael Lofton

So we better stay ahead of the game. And I think when 4.0 hits this year, it’s going to just jack things way up.

What do you think about that? Matt?

@30:49 – Mack

Sam Altman is saying that it’s overhyped. He’s trying to like manage expectations.

@30:54 – Stephanie Hudson

He’s saying a lot of people are way too.

@30:56 – Mack

Way too excited about what you can do for us. He also, so. But like there’s that he’s like GPT-4 won’t be nearly as good as you think it is.

And then at the other time, like the impact of what GPT-3 is has not made it through society yet.

So he also says that like as GPT-3 starts getting integrated into more and more businesses, our like social fabric will substantially change.

@31:18 – Michael Lofton

I remember sitting in a room with some engineers and he holds up this and it’s five years before it came out.

So this is stuff that we’re aware of right now is really it’s not so much cutting edge. It’s been around a while.

It’s just that certain people are starting to be introduced to it and they introduce, you know, the bell curve thing.

So it’s going to raise up and everybody’s going to catch on sooner or later. But this is I mean, think about the technology they really know.

They’re 50 years ahead of us basic human thinkers. But that aside, I’m just really curious why more of us are not utilizing it on a.

daily basis. I from the SEO standpoint, which is kind of where I’m coming from, digital marketing, customer conversions, SEO, I can’t use it at a certain level because it doesn’t have current event referencing.

So it can’t tell me when it gives me metadata how many people exactly are searching. I can’t go deep into keyword research with it, but it can give me some simple feedback as far as meta tags, descriptions, H tags, that sort of thing.

But it cannot duplicate what I’m able to do from an SEO standpoint to actually dig deep and to verify that those tags will actually work.

It’ll give me some examples, but not the deep research because it doesn’t give me current event.

@32:49 – Mack

I think, yeah, I think this is something that’s worth working on. There’s a startup that I found today called Splite, SPLE, that has a process for embedding.

local, like your company’s like knowledge base. So if you have like certain processes for like sales or customer support, now you have a chat GPT like interface where you can just conversationally ask questions.

And the embeddings include the source. This page in our notion is where we wrote down that process. This is exactly how you take like customer funds if they live in Saudi Arabia and get charged to get tax, right?

So that kind of thing, this like embedding, I think that’s the startup that I need to be working on really is this embedding pipeline tool so that you can like execute a search, store the embedding, and then that’s all like an automated pipeline.

So that the model automatically updates and then you get this like SEO tool that knows new things. It’s not that hard either.

I just got to finish up my day job and then like go completely entrepreneur mode.

@33:48 – Stephanie Hudson

So simple. So simple you guys, if you’re Mac. So the other, there’s a couple things about that that are fascinating to me and I’m glad I’m recording.

this because I’m going to tell you something right now. The thing that is going to blow your minds the most, this is the prediction that I’m going to make.

And you watch. It’s going to happen. Bing is going to overtake Google. I know, David, I know. It’s going to happen because Microsoft, like, Google’s already using AI.

In tons of things that it does, tons, like that’s not new. This is not news to Google, but it’s going to be a little bit more conservative in how it’s doing it.

But the CEO of Microsoft is going all right. Write it down, Michael, put the date.

@34:38 – Michael Lofton

What’s the date?

@34:40 – Stephanie Hudson

January 26th, 2023. Because… Oh no, what’s the date that they’re going to take over? Oh, I don’t know that.

But it’s going to happen. Watch it. Like, let’s put it on the calendar for next year this time and let’s see how we’re progressing.

Because it’s going to start incorporating things… things like ChatGPT into search. Because a lot of people have been saying, first of all, I hate how they just refer to it as ChatGPT only, because it’s the underlying AI technology, and it’s so unwieldy to say.

It’s just annoying. It’s like, is that the best name they could come up with, really? But so anyway, the AI, so some people are saying things like, ChatGPT is gonna replace search.

Well, no, it’s not, because it’s different. It’s not gonna replace search. It will fill certain voids, and it does work for you, and it can research things, sort of.

But if I wanna go and research, so say I wanna do a recipe. If I want a recipe for Tuscan chicken, or something like that.

Okay, so I’m gonna go, and I’m gonna see, actually, you know what’s a better one is margarita, because I actually just looked this up recently.

So I wanna go see a margarita. Well, then I’m gonna see all of the tequila brands have it. So.

A, I could go to my favorite tequila brand and see how they recommend. Or I could go to somebody else who I already know the reputation of, who I can, who I know I trust, like if I’ve seen other recipes of theirs before.

So, you know, I can do something like that, or I can go read three or four or five and say like, okay, I get the gist of this it’s tequila and lime and triple sec.

And I’m going to do a splash of OJ cause I saw somebody mentioned that. You know what I mean?

Like you can kind of make your own thing. Whereas if you go and say to chat GPT, how what’s a gimme a margarita recipe.

You get one, boom, one recipe. You don’t know what it’s like. You don’t have anything to compare it to any reference for it, anything.

So when it comes to stuff like that, I think there’s just no, it’s, it’s apples and oranges.

@36:46 – Michael Lofton

Of course you can always prompt GPT to give you more and it’ll give you more.

@36:50 – Stephanie Hudson

Sure. But there’s no, there’s no reputation behind it. There’s no like comparison. There’s no like really the nitty-gritty is there’s no citation.

And that’s another whole big thing with all of this, but that’s a, that’s a bit pretty deep well. Are you laughing at that?

@37:09 – Mack

No, Splite fixes that. Like this just got released like a couple of days ago on product time. Like they’ve already, so that’s, this is the thing that’s like so crazy about this is that the schedule of these tools getting better is like blazingly mind-blowingly fast.

So we had ChatGPT that had this problem that you’re talking about on November 30th, that was released. Yesterday, Splite was released and it has citations and it can give you, like, you can ingest web pages into something like Splite.

And so when you’re searching chicken recipes, it’ll say like, well, here’s one that I found. This is the source.

And you could, you could conversationally edit that prompt. So I need a chicken recipe that’s under 400 calories a serving.

I need a chicken recipe that’s under 200 calories a serving. I need a chicken recipe because I’m like very hungry, a thousand calories, like whatever it is, I need a chicken recipe that uses triple sec.

Splite can have like this, like progressive ingesting of embedded data can solve those problems. And. it’s only been two months.

Like what’s gonna happen two months from now?

@38:05 – Michael Lofton

So- Was slice spelled like it sounds? S-L-I-C-E?

@38:09 – Mack

I’ll type it.

@38:10 – Stephanie Hudson

Okay, thank you. Oh, Mark put the definition of it. Extending the legs at right angles to the trunks. Got it.

That’s pretty good. That’s a pretty good name for it, I guess. So the other thing that’s gonna happen is sort of, and I don’t know, so there’s giving the AI certain data to search, which is super valuable if you wanna go on to your own, like especially if you’ve got like a big blog or a very robust e-commerce store or like a lot of content on your site, if you can get an AI, you can feed it your site basically, have it digest it, and then have a chat on the site that could answer questions and talk to people.

and do things like that’s powerful, compared to any stupid chatbot that you go to now that is completely useless.

So you can do things like that where you’re injecting stuff, you’re giving it the information it needs to teach it.

The other thing is if you could actually have the AI search, and I don’t know if that’s the right way to say it or not.

I don’t know if it’s gonna be continuously consuming in real time, just continuously consuming, or if it’s gonna go, you know what I mean?

It’s probably not really search. It’s probably just still continuing to consume and then give you multiple answers. Mac, is that what you would say?

@39:40 – Mack

So there was a contest last week for finding inverse scaling problems for large language models, and this was one of them.

The larger that a model gets, the less it asks for new information. Because just like us, when it knows a lot of things, it’s like, well, I know everything.

I’m not gonna ask, I’m not gonna check. So that’ll be interesting to see too. Like if we set up this embedding pipeline.

And it knows a lot of things. More likely over time, it’ll stop asking for new information. So search is unlikely.

It’s going to be unlikely to ask to search for things.

@40:11 – Stephanie Hudson

It’s also fascinating. But you watch, guys, because Bing is going to start incorporating this stuff into… Is somebody leaving?

@40:19 – Suzanne Black

Yeah, me.

@40:20 – Stephanie Hudson

Oh, bye, Suze. Glad you came.

@40:24 – Michael Lofton

Have fun seeing…

@40:24 – Stephanie Hudson

Get some sleep. It’s going to be really interesting to see what happens with Bing, though, because I love to rag on Bing about how pointless it is.

And now it’s going to like… I’m not saying it’s going to eclipse Google completely because Google has so many other things like with maps and Google My Business and all those things that are going to remain relevant.

But like Bing as search is going to start to be something that you might actually choose to go to instead of Google.

@40:52 – Michael Lofton

Well, I think Google is also, like you said, it is AI generated at a certain level and they’re using more of it, hence another 12,000 laid off.

@41:01 – Stephanie Hudson

the other day.

@41:02 – Michael Lofton

They’re finding ways to trim and that’s one of them.

@41:07 – Stephanie Hudson

Yeah. So ways for us to use AI in our business, like Michael asked. So there’s a few different ways.

When we’re talking now, we’re talking future, I’m talking like just sort of big picture. You can utilize these generative AI tools to generate content, obviously.

Like you can do that today. You can go generate content. You can use them to learn. Of course, all of it needs fact-checking.

Like that’s just side note, like asterisk on everything. Like you have to fact-check all of this stuff. Cause it’s, what is it?

What’s the word that they say? It’s like fantasizes or something like that. That’s not the right word. What are they saying that it does Mac?

@41:53 – Mack

That it- Hallucinates.

@41:56 – Stephanie Hudson

Hallucinates. Yeah. It’s like, it just invents things and states. as facts, like it does not question itself or, um, or anything, you know, so it’s like, you have to fact check and everything, but basically like you can, you can use these tools to generate content, to streamline your workflow, all of that kind of stuff.

But what if you decided to do like Mac did or like Andrew and Vito did with Bertha and like create something that could then be a tool for others to use.

So that’s another opportunity. Can you guys think of any specifics or other general ways that, that you could use AI to, um, facilitate growth in your business or even to start a whole new business potentially?

@42:39 – Michael Lofton

I used it to write three press releases this week. Boom.

@42:44 – Vikki

Awesome. I use it to write for a four page.

@42:50 – Stephanie Hudson

Um, funny jokes. That’s cool.

@42:52 – Scott Tambling – DesertDigital.Design

Anybody else? I’ve been struggling with SOPs for the last two years.

@42:56 – Stephanie Hudson

I’m like, duh.

@42:57 – Scott Tambling – DesertDigital.Design

Let’s have Jasper help with my SOPs.

@43:02 – Stephanie Hudson

How are you doing that? SOPs are very personal, like your specific methods. So how are you?

@43:09 – Scott Tambling – DesertDigital.Design

Again, not perfect, but it helps.

@43:12 – Stephanie Hudson

It’s a start.

@43:13 – Scott Tambling – DesertDigital.Design

It helps, right? It cleans things up. You can just write fragmented bad sentences and it makes it pretty.

@43:25 – Stephanie Hudson

The use cases if you are involved in imagery, like Jean for photographers or other people that are involved with imagery.

Photoshop is already adding. I know Scott put something in here.

@43:39 – Scott Tambling – DesertDigital.Design

Alcade art. Alcade art is free now. It’s crazy, crazy powerful. Because you can mix it. You can, if you’re a Photoshop user like myself, you can manipulate things a lot quicker than just using chat alone.

@43:57 – Michael Lofton

And then speaking of prompts. chat. And Andrew brought this up after hours last week, and we played around with it quite a bit.

You can create the prompt and chat and then move it over to mid journey and drop it in there.

And then that will help with your image.

@44:17 – Stephanie Hudson

Yeah, there’s a little thing like that on the image generation in Bertha, where you can put in a thing and then like basically get Bertha to help with it.

And it sort of ups it. Oh, there’s one other thing that I do want to show you guys that is, hold on, let me find my screen here.

@44:34 – Jeane Bope

I mostly enjoy the thing that you were describing with the meal plans and that kind of thing, as opposed to search, because it is so easy to say, give me a diet for seven days that’s gluten-free and under a thousand calories a day and doesn’t have this item, this item or this item.

And as soon as it spits it out, you say, give me a shopping list to make these recipes. And it gives you like, that would take so much longer.

on Google. That kind of usage for the tool is really cool to me.

@45:06 – Stephanie Hudson

Or like, here’s what’s in my fridge. What can I make?

@45:10 – Vikki

And I love that you can just say, and put it in a table. And it just puts it in a table.

@45:15 – Stephanie Hudson

I’ve never done that one.

@45:18 – Vikki

Oh my God. Put it in a table. It’s so much easier to read.

SCREEN SHARING: Stephanie started screen sharing – WATCH

@45:24 – Stephanie Hudson

Okay, did I share the right thing? I have so many browsers open. Okay, so, hang on a second. What was I gonna show you?

Oh, so BerthaArt, that reminded me. Oh, come on. Do I not have my plugin updated? Oh, the worst. It’s been like just a really winning tech day.

No, this is the current one.

@45:47 – Michael Lofton

Why didn’t that work?

@45:49 – Jason Rutel

Let’s see.

@45:50 – Stephanie Hudson

Okay, here it is, here it is, here it is. Okay, so there’s two choices now when you go into BerthaArt.

You can just do the regular thing, create it. But they also added this thing where you can search like.

What do we want to search for?

@46:04 – Michael Lofton

I don’t know. You were saying margaritas.

@46:08 – Stephanie Hudson

Breweries.

@46:09 – Vikki

Not the salmon.

@46:10 – Michael Lofton

She doesn’t want to hear that word again.

@46:12 – Stephanie Hudson

What did he say?

@46:15 – Vikki

I said not the salmon again.

@46:18 – Stephanie Hudson

No, I did think of it. So you can do like…

@46:24 – David Browne

Making this more like…

@46:26 – Stephanie Hudson

Frosty margaritas with lime on the rocks. And then you can search.

@46:34 – Vikki

Oh, you can get rocks now. You’re going to have like a salmon thing going on there.

@46:39 – Stephanie Hudson

Good. So, I mean, we’re going to get some terrifying stuff like yowza, but this is it right here. Cat drinking it.

So what this is, guys, is AI-generated art from Stable Defusion. I think this is the Stable Defusion database. Don’t quote me on that.

@46:59 – Scott Tambling – DesertDigital.Design

So because AI-generated…

@47:00 – Stephanie Hudson

Free AI generated art cannot be copywritten copyrighted because in order to get a copyright, you have to have personhood and a machine can’t have personhood because it’s not sentient.

Sentient.

@47:17 – Michael Lofton

So, um, so they are copyrighted free.

@47:19 – Vikki

How does that, Kevin, I mean, there’s such a benefit of all this too.

@47:24 – Stephanie Hudson

So the answer is, there’s no copyright on it, which means that all, like if somebody else is a lot better at doing the prompts and getting the styles correct, then like, what, what is the AI trying to say that there’s given me all these cats with margaritas?

Did I even say that? They must record your data that you have a cat. Single lady Thursday night searching for a margarita recipe.

You do the math. She’s got a cat. Yeah. So, uh, anyway, there’s another one anyway. Uh, so you can go in and

search for AI generated art, and then you can just download it, save it to your media library right here, whatever.

And you can even grab the prompt from it if you wanted to modify it or create other things. And these don’t take your credits because they’re not generating, it’s not generative on here.

How rad is that? I think that is a super cool thing. Anyway. All right. I am gonna hit stop on the recording now because I like to keep a little something back for those who show up live.

So we’re gonna keep chatting. Michael loves the after hours chat. So I’m gonna pull the plug–