Contextual Guidance in Chat GPT for Effective Marketing

[00:00:01] Jess Bahr: And we are live.
[00:00:03] Jess Bahr: Welcome back to another exciting episode of Marketers Talking AI, where Pavel and I gather I being Jess to talk.
[00:00:11] Pavel Konoplenko: Hey, world.
[00:00:11] Jess Bahr: AI.
[00:00:12] Jess Bahr: Hello.
[00:00:13] Jess Bahr: Hello, world is the first thing everyone, I think codes.
[00:00:18] Pavel Konoplenko: Yes.
[00:00:19] Pavel Konoplenko: That's like a classic phrase now.
[00:00:22] Pavel Konoplenko: It's going to go down in history.
[00:00:23] Jess Bahr: Yeah, except for kids these days, they don't know what it is.
[00:00:26] Pavel Konoplenko: Kids these days, they don't know what Xeroxes are.
[00:00:30] Pavel Konoplenko: And I was listening to a YouTube video where it was just like the modem sound, like the dial up sound.
[00:00:36] Pavel Konoplenko: And it was like the welcome to AOL message.
[00:00:39] Pavel Konoplenko: I felt so nostalgic.
[00:00:41] Pavel Konoplenko: It was kind of unsettling of how I was comforted through time.
[00:00:47] Jess Bahr: This noise that used to be jarring and just, like, great at your ears.
[00:00:50] Jess Bahr: I used to back in my little early hack, not even a hacker back in the day.
[00:00:58] Jess Bahr: I used to drive around when I first got my license, I would drive around my neighbor.
[00:01:03] Jess Bahr: This sounds so bad as I say it.
[00:01:04] Jess Bahr: I would drive around my town and I would see whose WiFi was open and I would just park out front and I would like, hijack their WiFi and try and get into their system.
[00:01:14] Jess Bahr: I never did anything, but I was like, I'm going to learn how to be a hacker in my 7000 person town in the Midwest.
[00:01:22] Jess Bahr: Sorry, Oregon neighbors.
[00:01:23] Jess Bahr: It might have been in your Internet.
[00:01:27] Exploring the New Chat GPT Plugin Feature

[00:01:27] Jess Bahr: Speaking of kids these days, I saw that Chat GBT rolled out a new feature that you've been playing around with.
[00:01:36] Jess Bahr: Tell us about it.
[00:01:37] Jess Bahr: Let's open tell.
[00:01:37] Jess Bahr: Tell the world about your experience.
[00:01:40] Pavel Konoplenko: So I finally got access to the chat GPT plugins.
[00:01:45] Pavel Konoplenko: So I signed up for the waitlist maybe two months ago.
[00:01:48] Pavel Konoplenko: At this point, I forgot that I signed up and I saw some Twitter video of someone showing some of their latest plugins.
[00:01:57] Pavel Konoplenko: And I was like, oh, I wonder if I go there, what happens?
[00:02:01] Pavel Konoplenko: Because I didn't think I had the plugins.
[00:02:02] Pavel Konoplenko: So I went there, opened up the store, and I had all of the plugins available.
[00:02:07] Jess Bahr: Nice.
[00:02:08] Pavel Konoplenko: It's like 49 pages that you could scroll through eight plugins per page.
[00:02:13] Pavel Konoplenko: So it's like, oh, my God.
[00:02:16] Pavel Konoplenko: Math.
[00:02:17] Pavel Konoplenko: Almost 400 plugins, just different plugins.
[00:02:21] Pavel Konoplenko: Some that are able to pull in research papers and go through the information there.
[00:02:27] Pavel Konoplenko: Some are more geared into product management.
[00:02:31] Pavel Konoplenko: Some are for financial predictions and advice.
[00:02:34] Pavel Konoplenko: So a lot of the plugins felt like they were preloaded prompts, basically, in which it set the plugin to act as a specific agent.
[00:02:43] Pavel Konoplenko: Plus it gave them additional functions that the plugin can do, whether it's like searching the web or pulling in one of my favorite plugins.
[00:02:51] Pavel Konoplenko: It's able to create diagrams, like a visual diagram of whatever process you input it for.
[00:02:58] Pavel Konoplenko: So it gives the Chat GPT more functionality.
[00:03:02] Pavel Konoplenko: It reminds me a lot of the plugins that we have on our browsers in general that aim to supercharge performance is.
[00:03:10] Jess Bahr: It like the third parties that are building the plugin for.
[00:03:16] Pavel Konoplenko: It'S.
[00:03:17] Pavel Konoplenko: Random independent developers and I don't know what the submission process is, but they're constantly introducing new ones.
[00:03:25] Pavel Konoplenko: They also have now the ability to browse the Internet through Bing, so it's able to pull in more recent data directly through Chat GPT, which it also reminds me of the App Store.
[00:03:42] Pavel Konoplenko: Yeah.
[00:03:42] Pavel Konoplenko: So Microsoft is really pushing to become like a leader in this know, they missed the entire mobile world, basically.
[00:03:51] Jess Bahr: I called it, I talked about it the other week.
[00:03:53] Jess Bahr: I think Microsoft is the winner of all the AI stuff.
[00:03:57] Jess Bahr: Microsoft is the winner of Apple's announcements when it comes to augmented reality and all that.
[00:04:02] Jess Bahr: I like how Microsoft will see an opportunity and they will capitalize on another company, building something or another network to get into versus saying like, we want to own it.
[00:04:12] Jess Bahr: I like it for Microsoft.
[00:04:13] Pavel Konoplenko: Good for them.
[00:04:14] Pavel Konoplenko: Well, the partnerships in general, I mean, Google and Apple had a partnership too, and that helped Google become the default search engine for iPhones.
[00:04:22] Pavel Konoplenko: That was through partnerships.
[00:04:23] Pavel Konoplenko: I think people forget the value of partnerships in marketing and especially with being able to build out these new AI ecosystems.
[00:04:32] Pavel Konoplenko: You need to go in with a partnership because especially with AI, there is a lot of money that needs to be invested to develop your own base models and to create your own fine tuning of the model.
[00:04:43] Pavel Konoplenko: It is getting cheaper, it's getting more accessible, but the monopolies have an advantage here, so it makes sense that they're leveraging that.
[00:04:52] Exploring the Impact of AI Tools and Functionality on SEO and Advertising

[00:04:52] Jess Bahr: So what do you have plugins that you really like so far?
[00:04:55] Jess Bahr: What's your initial reaction to them?
[00:04:58] Pavel Konoplenko: Yeah, so there's the one that I mentioned that allows you to create diagrams.
[00:05:02] Pavel Konoplenko: Big fan of that.
[00:05:04] Pavel Konoplenko: I like to illustrate workflows if I'm creating something.
[00:05:08] Pavel Konoplenko: So it's a good place to start to have a pre made diagram of a particular workflow.
[00:05:14] Pavel Konoplenko: There's another one I used for SEO.
[00:05:17] Pavel Konoplenko: You're able to just give it any website and it starts to pull in the ads that they're running.
[00:05:22] Pavel Konoplenko: The ad copy the keywords, so you're able to do really precursory like SEO search just directly through Chat GBT.
[00:05:31] Pavel Konoplenko: This doesn't replace needing to use tools like SEMrush for example, or RFS because those do give you a much more in depth access.
[00:05:39] Pavel Konoplenko: And there's things you can do with integrations where you can automatically integrate a tool like SEMrush with Chat GPT.
[00:05:47] Pavel Konoplenko: But that's kind of an off the shelf solution.
[00:05:50] Pavel Konoplenko: So it's good to see that the plugins give you basic ability to do SEO research right there when you're creating your prompts or doing any sort of research for idea starters, for campaigns.
[00:06:00] Pavel Konoplenko: So I was a fan of that.
[00:06:03] Pavel Konoplenko: There's another one that I liked with the research papers.
[00:06:05] Pavel Konoplenko: So it searches specifically for different research papers and pulls out information that you're asking about.
[00:06:12] Pavel Konoplenko: And research papers aren't things that normally get a lot of publicity.
[00:06:17] Pavel Konoplenko: You don't really see what the current research on marketing and digital interactions with data, you don't really come across that in your day to day because a lot of these academic papers get written almost kind of forgotten or inaccessible.
[00:06:37] Jess Bahr: Statistica has a ton of great graphs, but a lot of them are behind like a $300 paywall to get into.
[00:06:44] Jess Bahr: So it's just like they're not accessible to the common person.
[00:06:49] Pavel Konoplenko: And I think the best thing for people is to sign up for that waitlist.
[00:06:54] Pavel Konoplenko: I don't know what the exact website is because they are unrolling.
[00:06:58] Pavel Konoplenko: It the waitlist.
[00:07:00] Pavel Konoplenko: They're letting people in.
[00:07:01] Pavel Konoplenko: So if you get picked and you can have a chance to go through all of these plugins that help make the most of the functionality from Chat JPC.
[00:07:11] Pavel Konoplenko: And I think part of it too is it makes the input a little bit easier because you as the user are now priming yourself and priming the agent to talk to solve a particular problem because otherwise prompt writing is too open ended, I think, for a lot of people to make the most use of it.
[00:07:29] Jess Bahr: Yeah.
[00:07:30] Jess Bahr: This is also like an actual waitlist that has an actual product.
[00:07:36] Pavel Konoplenko: As opposed.
[00:07:36] Jess Bahr: To a lot of the waitlists we're seeing where you're just a waitlist for nothing.
[00:07:40] Jess Bahr: But it makes sense they have a waitlist.
[00:07:41] Jess Bahr: They're rolling out something that's going to take resourcing to run.
[00:07:45] Jess Bahr: There could be bugs.
[00:07:46] Jess Bahr: It makes sense for that and it's an actual product.
[00:07:49] Jess Bahr: Yeah.
[00:07:50] Pavel Konoplenko: The waitlist.
[00:07:50] Pavel Konoplenko: Have you seen any other fake, fake waitlists coming out?
[00:07:55] Jess Bahr: This might be controversial, but I feel like Meta's entire announcement of them now focusing on AI and this suite of AI generative tools for advertising is just like them announcing their intent to do something.
[00:08:10] Jess Bahr: They're not announcing anything.
[00:08:12] Jess Bahr: There's nothing that's coming with it.
[00:08:13] Jess Bahr: They're just announcing that.
[00:08:17] Jess Bahr: Yeah.
[00:08:17] Jess Bahr: Which we would expect we've seen through the course of our careers at least we've seen third party tools build on top of Facebook and then Facebook, buy them, integrate them.
[00:08:28] Jess Bahr: So I would 100% anticipate that Meta is going to go buy another startup that's doing the thing they want to do and fold it in.
[00:08:35] Jess Bahr: I don't think they're really going to generate something new and novel that they're originating and I think it takes them a long time to develop shit.
[00:08:43] Jess Bahr: And they want to capture the attention of AI right now they want to get out.
[00:08:48] Jess Bahr: And so they're almost doing worse than a waitlist.
[00:08:51] Jess Bahr: They're not waitlisting you.
[00:08:52] Jess Bahr: They're just saying that they're going to do something eventually because they'll do a waitlist for the actual thing.
[00:08:59] Pavel Konoplenko: Is this the curse of they're like the late comers to everything and they're just they are.
[00:09:06] Jess Bahr: I think they used to be better at embracing better ways to use their tools when they would buy these third parties.
[00:09:14] Jess Bahr: And now it feels like they're just trying to chase a hype I mean, I'll say the same thing for metaverse.
[00:09:20] Jess Bahr: Their second life came out in, what, the early 2000s?
[00:09:26] Jess Bahr: They're chasing this thing to try and be innovative, but they're a legard.
[00:09:30] Jess Bahr: They're rarely ever first to market.
[00:09:33] Jess Bahr: And when they do, it's not always like the best experience.
[00:09:35] Jess Bahr: I don't think oculus, I mean, I'm saying this off the cuff, I don't know what the stats are, but I feel like oculus doesn't own a larger portion of the VR world.
[00:09:44] Jess Bahr: We anticipated that Oculus would do well because you're already in like there's large portions of the world that use facebook as their entry point to the internet.
[00:09:53] Jess Bahr: They have a large user base, but I think they've really failed in getting that user base activated to newer technology.
[00:10:00] Jess Bahr: And so, yeah, they're kind of behind on a lot of things and I think they're going to be behind on AI.
[00:10:07] Jess Bahr: Third party tools will come out to leverage AI to better your Facebook advertising experience.
[00:10:12] Jess Bahr: And I think if Facebook is wise, it'll just buy those and gobble them.
[00:10:16] Pavel Konoplenko: Yeah, yeah.
[00:10:18] Pavel Konoplenko: I'm curious to see how AI impacts the big revenue generators for the know facebook and Google.
[00:10:25] Pavel Konoplenko: Google has been embracing some of the AI generative content.
[00:10:29] Pavel Konoplenko: I think they started rolling out their, I think it's called like performance max or something, allows you to use generative AI to create the ads so it's know companies like Facebook and Google, they built out their dominance on advertising and AI aims to disrupt a lot of that.
[00:10:45] Pavel Konoplenko: So this is going to be, I think, a make it or break it technology for Facebook or for Google to see how they respond to this, even for SEO for know, how is that going to shape out?
[00:10:56] Jess Bahr: I was just saying I think SEO is a perfect example where Google, which I will say, I think google innovates quicker than Facebook does.
[00:11:04] Jess Bahr: Can we all remember Facebook's phone, like that horrible launch?
[00:11:09] Jess Bahr: But I think google innovates fast on Facebook.
[00:11:11] Jess Bahr: But Google's in a spot where if they don't figure it out, they are going to start to become at the whim of AI.
[00:11:18] Jess Bahr: And so all of a sudden, we're going to see AI really dictating what's ranking in SEO because AI is able to understand it's, able to identify the algorithm changes.
[00:11:27] Jess Bahr: Again, SEO is just writing for an algorithm.
[00:11:30] Jess Bahr: It's writing for a computer.
[00:11:32] Jess Bahr: AI.
[00:11:33] Jess Bahr: Is able to detect those patterns and those changes and figure out the formula to rank well a lot quicker.
[00:11:38] Jess Bahr: And so if google doesn't get ahead of it, it's going to just be at the whim of these AI programs that are generating things to optimize for SEO.
[00:11:47] Jess Bahr: I think google has a different risk than Facebook does when it comes to the AI game.
[00:11:52] Pavel Konoplenko: Well, Google has been playing this cat and mouse game with SEO consultants and black hackers and wikipedia for over a decade.
[00:12:02] Pavel Konoplenko: And I feel like it ebbs and flows sometimes.
[00:12:06] Pavel Konoplenko: It's clearly that Google is like, owning that conversation.
[00:12:09] Pavel Konoplenko: I think right now, the SEO consultants who are using these programs were able to build out these tools, could potentially have a leg up on Google until Google does their own version of, you know, when they completely change everything around.
[00:12:24] AI Impact on Future Technology and Misinformation

[00:12:24] Jess Bahr: Plus they have cyborg panda.
[00:12:28] Pavel Konoplenko: Cyborg panda.
[00:12:31] Jess Bahr: Internet terminator panda.
[00:12:34] Jess Bahr: I was watching panda videos the other day.
[00:12:36] Jess Bahr: Actually, this is not related at all.
[00:12:39] Pavel Konoplenko: AI will never replace panda videos.
[00:12:42] Jess Bahr: Yeah, AI will never replace the joy you get when you see a little panda.
[00:12:45] Jess Bahr: There's a little panda bear.
[00:12:46] Jess Bahr: And it was like playing with this.
[00:12:49] Jess Bahr: I think it probably had water or food in it and then it accidentally got under it and it couldn't get out.
[00:12:53] Jess Bahr: So you just see this big bowl running around because the panda stuck.
[00:12:57] Jess Bahr: AI can't give you that joy.
[00:12:59] Jess Bahr: They can't give you that serotonin until.
[00:13:02] Pavel Konoplenko: It generates that video.
[00:13:03] Jess Bahr: In a year, right.
[00:13:07] Jess Bahr: The impact will have on know, Google interesting, definitely is going to, I think, be a massive wild card for them if they do crack AI, I think it's still going to be like revenue impact portion relatively small.
[00:13:21] Jess Bahr: But I think we're going to see these other players, like nvidia nvidia outperformed the analyst projections for them because their data center branch outperformed by like I'm going to say it, and it's going to be wrong, but I think it was like two or three x their projection like insane.
[00:13:38] Jess Bahr: Growth in data center.
[00:13:39] Jess Bahr: Because all these data centers are being updated to be able to handle the computing needs of AI.
[00:13:44] Jess Bahr: And so their gaming division, which gaming typically is recession proof, it's growing throughout the regardless of economic turmoil.
[00:13:54] Jess Bahr: Gaming was down, data center was up because of AI.
[00:13:57] Jess Bahr: We've talked about it before.
[00:13:59] Jess Bahr: The people who make the pitchforks are the ones who really profit during a gold rush.
[00:14:03] Jess Bahr: And so I'm kind of thinking too, horror, revolution, depending on where you're at, I want to see someone come out with the new Faraday cage to protect yourself from AI.
[00:14:17] Jess Bahr: Someone's going to be selling that well.
[00:14:21] Pavel Konoplenko: I mean, we're going to be seeing more demand for fact checking.
[00:14:25] Pavel Konoplenko: And yes, I think the election, this coming election, I feel the election is always like this pivotal moment where the zeitgeist of technology and culture and society come to the forefront.
[00:14:38] Pavel Konoplenko: Every election year, there's some new technology that we must wrestle with.
[00:14:42] Pavel Konoplenko: And I think this year specifically fact checking AI videos.
[00:14:46] Pavel Konoplenko: AI photos deep.
[00:14:48] Pavel Konoplenko: Yeah, there's going to be a big need for security from misinformation and falsely created news from I think that's where we're going to start to see the deeper second order, third order AI businesses that are now preventing us from the people with those.
[00:15:10] Jess Bahr: Excuse me.
[00:15:12] Jess Bahr: I would say, Pavel, that you and I are pretty tech savvy.
[00:15:16] Jess Bahr: And there's stuff because I've been watching a lot of these comparisons where it's like, is it AI.
[00:15:21] Jess Bahr: Generated or actual.
[00:15:23] Jess Bahr: And it's usually celebrities in really awkward situations.
[00:15:26] Jess Bahr: And you try and guess.
[00:15:27] Jess Bahr: And there's times where I'm like, oh, clearly they have like 50 teeth.
[00:15:31] Jess Bahr: Clearly, that's AI generated.
[00:15:33] Jess Bahr: Or clearly there's something a little bit off on this that I think that it's not quite right.
[00:15:38] Jess Bahr: But it's not easy to discern between AI generated content and human generated, I think to my grandparents who fell for a lot of telephone scams, and there's going to be a lot of people who don't.
[00:15:51] Jess Bahr: Even the gentleman, the lawyer who submitted a briefing that he used Chat GPT to write that had fake cases referenced, there are going to be a lot of people who don't know how to discern AI from actual and even pre AI.
[00:16:06] Jess Bahr: We had a huge fake news issue.
[00:16:09] Pavel Konoplenko: I mean, we have that with print.
[00:16:11] Pavel Konoplenko: There's news magazines that are stretching the truth.
[00:16:15] Pavel Konoplenko: And my take on this, this is like a human problem.
[00:16:20] Pavel Konoplenko: It's us as a society adapting to the new technologies that we're able to create and entertain ourselves with.
[00:16:28] Pavel Konoplenko: And I think this is going to be just that new wave of just understanding.
[00:16:34] Pavel Konoplenko: There was a really funny Tumblr post that I saw.
[00:16:38] Pavel Konoplenko: Our boomer thing is going to be we won't be able to know holographs from reality.
[00:16:45] Pavel Konoplenko: And then we're going to be talking to our kids and be like, I can't believe that person wasn't talked to me.
[00:16:49] Pavel Konoplenko: It's like, oh, they blink way too regular.
[00:16:52] Pavel Konoplenko: That's clearly a hologram grandma.
[00:16:55] Pavel Konoplenko: We're not going to know.
[00:16:56] Pavel Konoplenko: And I feel like we as a collective are very good at picking up these nuanced signs into what content is generated.
[00:17:08] Pavel Konoplenko: So the example I like to think about is we as a group of consumers are able to pick up if content is created by corporations.
[00:17:19] Pavel Konoplenko: We know what corporate speak sounds like.
[00:17:22] Pavel Konoplenko: We can't put our finger on it.
[00:17:23] Pavel Konoplenko: We know that this came from some bureaucracy.
[00:17:27] Jess Bahr: You know, legal reviewed it.
[00:17:29] Pavel Konoplenko: Yes, and you don't care about it.
[00:17:32] Pavel Konoplenko: That's like the corporate side of the world.
[00:17:34] Pavel Konoplenko: But when an artist, some individual person creates it, you're like, I feel it.
[00:17:39] Pavel Konoplenko: This came from a real person.
[00:17:40] Pavel Konoplenko: No one teaches us to pick up those cues and symbols.
[00:17:44] Pavel Konoplenko: But I think as we start to get exposed to it, people will have to adapt and learn.
[00:17:49] Pavel Konoplenko: Like, this is what AI generated content feels like and sounds like and moves like, and how often they blink in their videos.
[00:17:56] Pavel Konoplenko: And some people will be better at catching it, and the kids tend to be better at it because they're the youth are observant.
[00:18:07] Pavel Konoplenko: But yeah, it's going to cause some upheaval.
[00:18:10] Jess Bahr: Yeah, well, then the AI is going to pick it up and realize that you have to blink, like, intermittently, but no longer than you can't go longer than X amount of time without blinking.
[00:18:20] Jess Bahr: They'll learn.
[00:18:22] Jess Bahr: They'll pick it up.
[00:18:22] Pavel Konoplenko: They learn.
[00:18:24] Pavel Konoplenko: And then some new thing will come out.
[00:18:25] Pavel Konoplenko: Now they're blinking a little bit too irregular.
[00:18:27] Jess Bahr: Yeah.
[00:18:28] Jess Bahr: And then people start changing their behavior to feel less AI.
[00:18:33] Pavel Konoplenko: Yeah.
[00:18:35] Jess Bahr: Do you ever fear failing a captcha?
[00:18:40] Pavel Konoplenko: Do electric Sheep dream of captcha prompts.
[00:18:45] Pavel Konoplenko: I've seen some really convoluted prompts now I think I was doing a prompt with Discord.
[00:18:53] Pavel Konoplenko: I don't know.
[00:18:54] Pavel Konoplenko: I don't know.
[00:18:54] Pavel Konoplenko: Maybe this was even Twitter.
[00:18:55] Pavel Konoplenko: I was setting up a Twitter account and I think I had to do some puzzle.
[00:19:00] Pavel Konoplenko: It was like a lot.
[00:19:01] Jess Bahr: It was like, do a slide.
[00:19:04] Pavel Konoplenko: Yes.
[00:19:04] Pavel Konoplenko: There's rotations, rotations and sliders.
[00:19:07] Jess Bahr: It was just like, also, why can't an AI figure it out?
[00:19:10] Jess Bahr: I feel like that's not that know.
[00:19:13] Jess Bahr: See, what I liked about Google's Captcha was that you knew there's a secret intent behind it.
[00:19:18] Jess Bahr: Like, yeah, they're trying to get security, but they're also doing it because now you have this massive labor force, people that are really cataloging and helping train their AI for free.
[00:19:27] Jess Bahr: Sorry, their machine, their ML databases, llam, the acronym, Llamas training, their Llamas.
[00:19:38] Jess Bahr: But it's interesting.
[00:19:39] In-depth Discussion on AI in Marketing: Its Reach, Potential, Limitations, and Bias

[00:19:39] Jess Bahr: Now, you talked a little bit about second and third order, how AI can come into play, and I feel like you and I have talked about it a bit, but I think it could be good to chat about more widely for people to hear about as they think about how they're using AI.
[00:19:54] Jess Bahr: There are a lot of tools that are pretty surface level with what they actually offer that people feel like, I'm going to sound like an asshole as I say this.
[00:20:05] Jess Bahr: People feel like I'm using AI, look at this cool thing I'm doing.
[00:20:07] Jess Bahr: And when you look at it's like, well, it's really pretty surface level.
[00:20:10] Jess Bahr: You can get deeper with it.
[00:20:11] Jess Bahr: You can use it to be more impactful and it's not as hard as you may think it is.
[00:20:17] Pavel Konoplenko: Yeah.
[00:20:18] Pavel Konoplenko: The way I think about it is right now with AI, we have access to tremendous new capability in the same way that word processing and the Internet opened up our capability to create content.
[00:20:32] Pavel Konoplenko: It would be like if someone gave you a computer and then you're like, wow, it has built in spell check.
[00:20:37] Pavel Konoplenko: And all you're doing under this fancy new computer is just spell checking everything that works.
[00:20:45] Pavel Konoplenko: It helps and improves the thing.
[00:20:46] Pavel Konoplenko: But you're underutilizing the entirety of the model that you can use to whatever you're creating as a marketer, as a coder, whatever it is.
[00:20:58] Pavel Konoplenko: And I think a lot of the surface level tools right now are necessary in our space.
[00:21:04] Pavel Konoplenko: We see a lot of these content creation tools where you can create blog posts, social posts, and a lot of them feel the same, like they have the same value, add the same features, and a lot of them are very individual based.
[00:21:20] Pavel Konoplenko: You give it your prompt and it does the blog post or you're able to write and then it fills in your sentences.
[00:21:27] Pavel Konoplenko: But the issue is you still have to now sit and sit down and write and then for there to be a sentence to complete, a paragraph to complete, it calls upon a lot of initial work and it doesn't have this meta layer over it.
[00:21:40] Pavel Konoplenko: I think what is interesting to start seeing over time is rather than the AI working on an individual level of here's your perfectly crafted article.
[00:21:49] Pavel Konoplenko: And by the way, there's no perfectly crafted article that we'll ever get to, but right now the AI is giving you these perfectly crafted bits and pieces, bits and pieces, and there is very little ability outside of you creating like embeddings and vectorizing.
[00:22:05] Pavel Konoplenko: The databases is for the AI to look at the entirety of all of the content that you published and start to make recommendations out of that, looking at the content you're producing and looking at the analytics for it, looking at the user behavior, and then starting to make larger scale decisions.
[00:22:23] Pavel Konoplenko: Not into how you craft a specific article, but what direction of content you should go into.
[00:22:28] Pavel Konoplenko: What are the themes and topics that are being covered too much or are there gaps in your coverage in terms of what you're trying to rank in?
[00:22:37] Pavel Konoplenko: Keywords right now, the AI has no visibility into it and a lot of these tools don't give you that.
[00:22:43] Pavel Konoplenko: They're very centered on.
[00:22:46] Pavel Konoplenko: You can change the tone of this and I have such a gripe against this as a shallow feature of being able to change the tone.
[00:22:54] Pavel Konoplenko: It's like you click more casual, more professional, and it's such a again, I'm going to sound like an asshole too, but it's such a dumbed down version of what you're trying to create because if you're talking to a marketer, the differences in tone are not like, oh, we're more casual, more professional.
[00:23:13] Pavel Konoplenko: There's a lot more depth and nuance to what your brand is, your tone and your style is, and a lot of these tools where you just have this simple pre prom to make it more professional, more casual.
[00:23:24] Pavel Konoplenko: It winds up, I think, reducing the effectiveness of the tool because you're now giving it this very vague prompt that you're telling the machine to be more professional without any additional guidance, without any additional what that means.
[00:23:40] Pavel Konoplenko: Examples of previous content.
[00:23:42] Pavel Konoplenko: And I think that's where a lot of marketers can fall prey to some of these easy features and then be like, well, this doesn't really change anything and I still have to sit and proofread it and rewrite the whole thing anyway.
[00:23:54] Jess Bahr: Well, I feel like it also introduces a lot of bias because my gripe with professionalism as like a topic itself and executive presence, especially where I feel like executive presence is often a bucket to say, my idea of what professional is doesn't align with you.
[00:24:14] Jess Bahr: And we have seen that be a massive barrier for a lot of people of color to be able to be seen as professional because it doesn't match the white view of it essentially.
[00:24:27] Pavel Konoplenko: The language.
[00:24:28] Pavel Konoplenko: Yeah, the bias is built in, and that's something that we used to, I feel like, talk about more as a society.
[00:24:36] Pavel Konoplenko: When we were testing out the AI models, maybe like a year ago, a year and a half ago, there was news of that Twitter bot that was let loose on the Internet and became like, a Nazi.
[00:24:48] Pavel Konoplenko: So there was a lot of these conversations of bias, but I feel like haven't bias hasn't been talked about.
[00:24:54] Pavel Konoplenko: We see these chat GPT, we put it on a pedestal.
[00:24:58] Pavel Konoplenko: We're like, wow, we're so close to general intelligence.
[00:25:01] Pavel Konoplenko: This is the greatest thing we've ever created.
[00:25:03] Pavel Konoplenko: This is it.
[00:25:04] Pavel Konoplenko: And because we put it on the pedestal, we forgot that, hey, this is built on top of the Internet, which we know is super biased, super discriminatory, and has all of those elements, and the machine doesn't know.
[00:25:16] Pavel Konoplenko: It doesn't have a metal layer of morality, if you will, on it.
[00:25:22] Pavel Konoplenko: And it's not able to check its own bias because it's an endless feedback loop system within it.
[00:25:33] Jess Bahr: What would be checking against too?
[00:25:35] Jess Bahr: The model in the data set is biased, and so if it does check against that and says, if this is unbiased, I'm unbiased.
[00:25:42] Jess Bahr: But it's like, well, no, just the creation of it.
[00:25:44] Jess Bahr: All of your language is gender.
[00:25:46] Jess Bahr: To assume that the person you're talking to is a man.
[00:25:48] Pavel Konoplenko: Well, English language languages themselves are embedded with preconceived meanings and context that you can't get away from.
[00:25:59] Pavel Konoplenko: And we haven't solved that problem.
[00:26:02] Pavel Konoplenko: And that's like, we're going down a very philosophical role of like, what is the role of language in giving us meaning and the role of culture?
[00:26:11] Pavel Konoplenko: Because it's trained on American data, it's trained on American Internet activity, trained on English books, and they're slowly expanding it.
[00:26:24] Pavel Konoplenko: But how do you get away from that bias?
[00:26:27] Jess Bahr: Yeah, we've kind of gotten a little bit of rabbit hole.
[00:26:30] Pavel Konoplenko: This is a very philosophical take, but I think it goes back to when we think about prompts, when we think about everyone's searching for the perfect prompts, like, oh, if I can get the perfect prompt to write out a marketing plan.
[00:26:44] Jess Bahr: Yeah, the number of little clickbaity or like, lead magnets I've seen that are 101 prompts you need to know as a marketer.
[00:26:51] Jess Bahr: And when you look at them, they're all single prompts.
[00:26:57] Jess Bahr: There's never like three or four steps to get a better result from that initial prompt.
[00:27:01] Pavel Konoplenko: Yeah.
[00:27:02] Pavel Konoplenko: And I think a question we often get is, how do you write better prompts?
[00:27:05] Pavel Konoplenko: How do you find this?
[00:27:06] Pavel Konoplenko: And my advice is to take a step back.
[00:27:11] Pavel Konoplenko: I think when we come to AI, when we come to Chat JBT, we have our in mind.
[00:27:16] Pavel Konoplenko: We're like, hey, give me this thing, give me this social post.
[00:27:21] Pavel Konoplenko: You would be able to get a much better result if you're taking a few steps back and you're creating the story for the AI, like, this is our company, this is the campaign that we're running.
[00:27:32] Pavel Konoplenko: This is our targeting, this is our ICP, this is our values.
[00:27:38] Pavel Konoplenko: You've copy and pasted your mission statement.
[00:27:40] Pavel Konoplenko: Give it more context into who you are as a company, individually begin the conversation of like you're building out a marketing plan.
[00:27:49] Pavel Konoplenko: If the goal is to get to the social plan, you're building out a marketing plan.
[00:27:55] Pavel Konoplenko: Give it a broader directive from which to go on once it puts out the marketing plan.
[00:28:00] Pavel Konoplenko: Overall, say, let's work on the social plan in the context of what we're building out of home, where we're building with content and with email.
[00:28:09] Fine-tuning AI for Effective Marketing Plans

[00:28:09] Pavel Konoplenko: Let's now work on the social plan.
[00:28:11] Pavel Konoplenko: You don't have to fill in anything for the email, but you want to prime the machine to know that this is in the right direction.
[00:28:17] Pavel Konoplenko: Because LLMs function a fancy next word system, it just picks the next best word for its sentence.
[00:28:26] Pavel Konoplenko: So you want to put it into the right river of stream of consciousness, basically for a lack of a better word, where it goes down the correct path, where it chooses the next best word or the next best idea based on something that you help determine rather than it just starting from scratch.
[00:28:46] Pavel Konoplenko: And it picks one route of where it's trying to go, constantly picking the next word, next word.
[00:28:50] Pavel Konoplenko: And it could go down some hallucination because you didn't really prompt it into where you wanted to go.
[00:28:56] Pavel Konoplenko: So I feel like being able to.
[00:28:59] Pavel Konoplenko: And this, I think, highlights the necessity of the marketing fundamentals.
[00:29:03] Pavel Konoplenko: Marketers are always told, okay, we're going to start at creating the ICP, but how many marketers have the ICP on hand, especially with early stage startups?
[00:29:14] Pavel Konoplenko: It exists somewhere in the minds of the founders.
[00:29:16] Pavel Konoplenko: It exists in the minds of the salespeople, but it's not on paper, it's not refined.
[00:29:21] Pavel Konoplenko: But once you have that, you can constantly refute it to the AI.
[00:29:24] Jess Bahr: It might be super big too.
[00:29:25] Jess Bahr: When you're starting out your ICP, you don't really know what it is until you have product market fit.
[00:29:30] Jess Bahr: And so it might be pretty broad, but it sounds like the crux of it, if I can boil it down, is giving your prompt a lot more information up front, being specifically and guiding.
[00:29:44] Pavel Konoplenko: It towards the right thing.
[00:29:45] Pavel Konoplenko: Because it doesn't know what's the correct answer in the sense that there's no AI, doesn't know what's a good marketing plan it knows what's the next best idea or the next best word is.
[00:29:56] Pavel Konoplenko: But it needs to know that it needs to spit out an effective marketing plan as opposed to a marketing plan.
[00:30:04] Pavel Konoplenko: Because within the data pool, it has world class information, but it has some random essay that some college student wrote about what makes marketing positioning.
[00:30:14] Pavel Konoplenko: And it doesn't know, unless you tell it that this is more valuable for us to reference.
[00:30:20] Pavel Konoplenko: Because it literally doesn't know.
[00:30:22] Pavel Konoplenko: It just looks at any mention of what marketing positioning is or brand positioning is, and it tries its best to predict what you want.
[00:30:30] Pavel Konoplenko: So you need to guide it as much as you can, take a few steps back away from your output, and build out the logical flow of how you get to the best.
[00:30:42] Pavel Konoplenko: And you may have to even describe what makes an effective marketing plan, what our objectives as a company are.
[00:30:47] Pavel Konoplenko: Is it to drive more sales?
[00:30:50] Pavel Konoplenko: Is it to penetrate a new market?
[00:30:52] Pavel Konoplenko: Are you rolling out a new product line?
[00:30:54] Pavel Konoplenko: The more context you give it, the better that it does.
[00:30:58] Jess Bahr: Yeah.
[00:30:58] Discussion on Prompt Writing

[00:30:58] Pavel Konoplenko: So that was my rant of prompt writing.
[00:31:01] Jess Bahr: Well, that rant is perfect timing.
[00:31:04] Jess Bahr: We're submitted over.
[00:31:05] Jess Bahr: But thank you, everyone, for joining us.
[00:31:07] Jess Bahr: And make sure in the comments below, we have Linked to the next episode.
[00:31:10] Jess Bahr: If you're watching a replay on YouTube.
[00:31:12] Jess Bahr: Yeah.
[00:31:12] Jess Bahr: Any other platform subscribe, like, hit the bell for notifications.
[00:31:16] Pavel Konoplenko: I don't know if there's bells.
[00:31:17] Jess Bahr: I don't know either.
[00:31:19] Jess Bahr: Yeah, stalk us.
[00:31:23] Jess Bahr: We will see you next week live on LinkedIn.
[00:31:26] Jess Bahr: Yeah.
[00:31:26] Jess Bahr: Or we'll see you the next Thursday on your calendar, live on LinkedIn at 115 Eastern Time.
[00:31:33] Pavel Konoplenko: Bye.
[00:31:34] Pavel Konoplenko: Bye, world.
[00:31:35] Jess Bahr: Bye, everyone.
[00:31:36] Jess Bahr: Bye.