How Frank Mendoza Breaks Down AI's Evolution and Impact

[00:00:00] Hector: Hello and welcome to another episode of the how to scale a business podcast. My name is Hector Santhia Esteban and I'm your host for today. our special guest, the man of the hour, is Mr. Frank Mendoza, and he's the founder and managing partner at Catalytics, and they're helping people with something that I think is becoming really important these days, and that is AI, robots, the Terminator Arnold Schwarzenegger himself anyways, we're gonna get into all of it.

[00:00:23] Hector: Frank thanks for hanging out with us today.

[00:00:25] Frank: Absolutely. Hector. It's an absolute pleasure to be here today and really look forward to today's conversation.

[00:00:30] Hector: I wanna get into it, and I say all that in jest because I'd love to structure the conversation in the way that most people, Are well, they're probably not as into it as maybe someone like you or even me. Chat GPT is a window that's very frequently open on my browser, all moments even on my phone now.

[00:00:47] Hector: And so it's become something that's, I've tried to integrate lot in that. With that being said, I realized that a lot of people It's still new. it's still something that they're unsure about or they haven't fully utilized. we're going to get into that.

[00:01:00] Hector: I think it's important though for you to catch us up on how and why you got here. whatever is important for you to, share. There's some guy who's sitting in their car and thinking well, who the hell is this Frank guy. He could tell me about this whole A. I. Stuff.

[00:01:12] Hector: So catch us up on there and then a little bit about what you guys do a catalytics.

[00:01:16] Frank: Catalytics is company I started about eight years ago. And what we do is we help consumers, C. P. G. Brands really execute actionable A. I. And data analytics strategies to help grow their revenue by Sorry. Tens of millions of dollars, right? And we've helped big brands and small companies as well really start to take advantage of gaining new insights and leveraging these emerging technologies.

[00:01:39] Frank: Myself, I'm actually a certified data scientist. I've also become a deep learning specialist. And all those are just fancy terms for saying I've played with this stuff. I've implemented these type of solutions. And at the end of the day have enabled customers and companies to really get closer to a data driven culture that I think is important in order to gain new insights into their business and their consumers at the end of the day.

[00:02:05] Hector: really good point that you bring up is that A. I. Is not necessarily new per se, in the sense that it's new and it's trendy, but we've always had data and we've had this ability to analyze data and that's been able to grow. Now, what it seems like is That regular human can now interact and interface with that data in ways that was previously inaccessible.

[00:02:29] Hector: And so I'd love for you to just share how field might not be as new as people may be thinking it is.

[00:02:36] Frank: No, it's a great point, Hector. I think. Certainly A. I. Has been around a while, the term was actually coined in the 1950s, 1960s. we've entered in what's called the A. I. Winters where it'll progress and then things will not progress. The technology or the algorithms won't progress.

[00:02:53] Frank: And so we enter these winters where not much is done in the last, 8 10 Plus years, there's been this culmination of a few things that have happened that allowed us to really start to take advantage of it. One is compute power, amount of compute power due to, chips and even cloud computing, has allowed us to do some things that in the past has not been impossible without, huge mainframes and servers and things like that. The second is around storage of actual data. You alluded to this, Hector. Data is imperative to be able to take advantage of AI. You need vast and vast amounts of it,

[00:03:31] Frank: And the cost associated with storing data has just exponentially dropped over the last 10 years, as you can imagine, right? Everybody has a cloud storage now, and it's just, For lack of a better word, hoarding a bunch of stuff whether it be photos or videos or audio or whatnot in the cloud.

[00:03:47] Frank: And storage has really been very cheap. The third has probably been the access and the progression of algorithms. So the ability for algorithms to actually do some amazing things. This includes deep learning algorithms neural networks, things like that. Those things have brought about this AI hype, right?

[00:04:07] Frank: What happened in November of 2022, though, was the introduction of chat GPT, and you mentioned this, Hector, what it's done is it's lowered the barrier to entry for most people in terms of interacting with these technologies, whereas previously with more traditional AI. You had to go make a significant investment to collect data to actually compute* *to identify patterns and leverage the algorithms.

[00:04:32] Frank: Chat GT image generators like Dolly or Mid Journey have really enabled people to,with a few clicks really be able to take advantage of some amazing generative AI solutions that are out there now.

[00:04:45] Hector: I want to get into businesses might use it because I think one big barrier to entry is just use cases and how you know someone who runs a laundromat or someone who runs a contracting business or what we might call these regular brick and mortar or more traditional businesses, how they might integrate it before we get into that, though, I'd love for you to talk about that other person who's sitting there saying yeah, but this is going to end the world.

[00:05:12] Hector: But what about the ethics? What about, there's so many. Reasons why someone might have an issue with it. So I'd love for you to just tackle some of those notions that people might have with regards to ethics or, the Terminator or whatever it is that's holding people back from utilizing the tech.

[00:05:31] Frank: No, absolutely. I think it's a valid concern. The technology is right now moving much quicker than the regulations and the legislation are at this stage, What does that mean? I like to propose that AI is a tool, just like any other tool. It can be used for good, it can be used for evil.

[00:05:49] Frank: I like to equate it to a hammer, right? A hammer, I can create a house with a hammer, right? I can build something new something creative, something that wasn't there before. But I can also act with malice and cruelty and the hammer can actually fill that function as well. So it's really incumbent upon us to recognize that this is just a tool.

[00:06:09] Frank: But that it comes along with that tool comes a necessity to recognize the risk, the inherent risk associated with it. I'm not going to sit here and say there's not disruption along with this technology. There's absolutely going to be disruption with this technology when it comes to people's careers, jobs.

[00:06:28] Frank: Like I said, malicious acts, things like that. for us to recognize that There have been other technologies in the past that also had that same challenge, right? So I'm old enough and I have enough gray hairs to recognize. I remember when the internet first started, right?

[00:06:42] Frank: There was a lot of concern with chats and, children's access and things like that. There will always be new technologies that are being introduced that can be used maliciously. As people or organizations are looking at trying to figure out what use cases, as we just discussed, are possible with this stuff, they have to recognize the challenges and limitations of these solutions as well.

[00:07:06] Frank: So things like bias things like hallucinations, those are things that are going to be part of our post generative AI world going forward and we just need to be cognizant of those and mitigate those risks as much as possible.

[00:07:21] Hector: Let's come back to that because I think that both of those are great terms that I'm like, Oh, I love that. And, but people are, might be thinking, Frank, what the heck are you talking about before we get there though let's talk about, how someone might. it. And I'd love for you to think about it, from that more traditional brick and mortar sense or maybe there are use cases that you've seen, outside the box ways to, maybe inside the box ways to utilize this stuff.

[00:07:46] Hector: But I'd love for you to speak to that angle just so that people can start to think about how might I, leverage this or how other people have leveraged it.

[00:07:54] Frank: Yeah. At Catalytics, we've developed a framework to be thinking about this because as you can imagine, there's a lot of confusion around these technologies,There's a lot of New applications that are being released constantly, right? Daily, even that are really giving people pause in terms of trying to move forward with this.

[00:08:10] Frank: And at times, people have told me in no uncertain terms, I'm a bit paralyzed, right? I don't know where to go apply this. don't know where to even. Think about using this in my organization.we've developed as a framework to think about where it makes the most sense. And how do we think about your unique use cases?

[00:08:29] Frank: So one consideration that you should think about when you're Looking for tasks for A. I. To actually help with is the human creativity perspective. And what I mean by that is when you look at a task that you want A. I. To perform how much actual human creativity is necessary to perform that task,

[00:08:46] Frank: Is how much innovation is necessary? How much creative problem solving is necessary? Think about it from that aspect. Now, the other second thing to consider is your competitive advantage. Do you have a unique process or proprietary process? Do you have a set of unique products that are unique to you or even a strong brand presence, right?

[00:09:07] Frank: Reputation that you look at as unique and a competitive advantage? So if you take those two considerations and let's put it in a a two by two matrix, if you will, with human creativity being on the Y axis and a competitive advantage being on the X axis. If you look at the bottom left hand corner, those are going to be things that are low creativity.

[00:09:27] Frank: Low competitive advantage. Those are amazing opportunities for you to look at to say this is an opportunity for AI to really augment or help my organization and take on those responsibilities, right? Potentially delegate even to some. A good example of that is say you have a website or you offer FAQs to your consumers all the time.

[00:09:48] Frank: That's an awesome opportunity for A. I. To take on that responsibility to answer those low level questions that your consumers might be asking of your organization. That can do it 24 by 7 3 65, including holidays, That's the lower left hand corner. If you move up the access to a high creativity but low competitive advantage, those are situations where these aren't your competitive edge tasks, but they do require more human creativity.

[00:10:15] Frank: This is an opportunity for you really to leverage AI to augment. Your creative staff. So could you use a I to help you brainstorm or work through some of these solutions? To build and on top of the human creativity element. These are also opportunities for you to maybe look for partners or vendors who that does lie in their competitive advantage. And then Hector going on. To the bottom right hand corner. Those are going to be where it is your competitive advantage and it's a lower human creativity. Another opportunity to start with AI augmentation. And then finally, you have the crown jewels. These are the upper right hand corner. These are high competitive advantage, high human creativity tasks.

[00:10:55] Frank: This is actually where I don't encourage people to have their initial steps or,dip your toes in the AI water on these tasks that are really your competitive advantage. These are unique to you. These could be strategic ad campaigns that you're building out. These could be developing new ads for image generation, things like that.

[00:11:15] Frank: These are opportunities for you to augment, but don't delegate necessarily to an AI solution. Does that help?

[00:11:21] Hector: Yeah, I think it gives a framework for how people might be able to utilize it. I'd love to know if you have even more buckets in the sense of I've seen AI be great in terms of. Research. I try and ask it. What are the things that you can see that I might not see? Because we have such limited data set and perspective, they are able to see the matrix, if you will, and see behind it and connect a lot of those things.

[00:11:44] Hector: in that vein, you said automation or delegation. There's this research part. You mentioned customer responding to FAQs and stuff like that. So are there any other kind of buckets, if you will, that we're missing, that someone might be able to leverage it?

[00:11:59] Frank: Yeah. Let's talk some use cases, right? I think you're alluding to one that I think people aren't taking advantage of enough, right? Which is gaining a different perspective than what they might've already done. I call it playing devil's advocate or contrarian, Leveraging these large language models to say, here's my perspective.

[00:12:15] Frank: Here's something that I'm, This is my view of the world, if you will, around a particular topic or nothing or something like that. And then asking these large language models to say, play the role of a contrarian or a devil's advocate. And help me think of, four or five or brainstorm some ideas that I might not take it into consideration that I should, especially when you're talking about a small business owner. This is a challenge, right? Because your world is encompassed around your business. You tend to have a bit of a blinder on and you understand the challenges that you're facing. But these AIs can open up a whole line of things that you might be blindsided by as a result of it.

[00:12:51] Frank: Those are great use cases to leverage solutions like that. Also think about. Leveraging it to brainstorm ideas around your marketing efforts, right? Give me some ideas on content generation that I can do, make sure you include educational, promotional, and call to actions in those ideas, right?

[00:13:10] Frank: And so it can start to generate ideas that you hadn't otherwise considered before. Or even giving you a 30 day calendar of content that you'd want to create. The challenge I would have Hector though, is I always tell people it's your responsibility though, to ensure that what it provides you. Is a line to your marketing message and to your brand, never encourage people to say, Oh, just take what it gives you and publish, It's always important to keep maintain that human in a loop because. You've gained your expertise as a function of experience, lessons learned. And you've gained scars as a result of things that have gone and things that have gone not so well.

[00:13:48] Frank: by applying your perspective and your expertise, I think using it* *as a thinking partner can go a long way. And then we can talk about specific use cases by organizations, if that's interesting.

[00:13:59] Hector: I'd love to know if there are any out of the box or I think people resonate with stories. So if there's any, of your own clients or what, people that you've seen in, popular media or anything like that of just creative ways or insightful ways or successful ways that people have actually been able to put it into use.

[00:14:16] Frank: Yeah let's take one example, Shopify. I'm positive you're, of your audience is probably leveraging Shopify. What they've done is just recently introduced a generative AI solution to help you with product descriptions. So you have the capability of entering some information around a product that you're selling via your Shopify store.

[00:14:33] Frank: and then aligning That description to, for example the writing style or the tone that you're shooting for to really align to your specific consumer. And it generates those pretty quickly and so that you have, a little bit, less heavy lifting when it comes to some of that generation. There's some creative solutions around image generation right now, Nutella, for example has been able to generate these hero packs for their containers. They were able to leverage generative AI to generate just millions and millions of designs. What ended up happening in that regard is they actually ended up generating thousands upon thousands of years, user generated content.

[00:15:13] Frank: As people were purchasing these, as you can imagine, they're unique. Their designs one of a kind. Their social media reach just exploded as a result of this. And they sold out of those hero packs within a month. other examples of people really taking advantage of iteration.

[00:15:29] Frank: If you're iterating on a slogan the ability to brainstorm multiple slogans, the ability to iterate on ad campaigns, right? Images of things you want to maybe shoot for The reality is the cost associated with creating this content has just dropped exponentially,And so your ability to try and iterate on four or five or 10 or 20 or 100 concepts is very low.

[00:15:53] Frank: But then I would also again caution youit probably has to have a human element though of a maintain a human in the loop where you still have to send that to a graphic designer or a copywriter to make sure it's aligned to what your brand, your logo, et cetera. is still what you want to resonate to the audience.

[00:16:08] Hector: like you said, it's a tool and with any tool comes a skill, right? And there's a variance in skill. And so when someone comes in and they, swing the hammer and they hit their thumb or they break a window or that you can misuse it.

[00:16:22] Hector: One of the things that I found to be very helpful is, the more that you can feed. something like a chat GPT the better your output's going to be. the more that you can tell it about your target audience or your target, your ideal clients or your customers or your business, or like you said, the tone or the direction, all of that stuff, the more that you can feed it, the better that the output is.

[00:16:41] Hector: If you just give it, one sentence and say, write me a, a web copy for a website about. Selling boats, it'll write it for you, but it's going to be nowhere near as good as if you can go through and say, here's the history of my company. Here's what we do. here's everything you need to know.

[00:16:55] Hector: Now, can you write me some copy for a website? That's going to be so much better.

[00:17:00] Frank: Yeah, Hector. I one of the challenges I think people are faced with as well right now is these messages that here's one prompt, here's a script of a prompt that will solve your problem or one prompt to end of all prompts, right? Type thing. And the reality is my experience, that's not necessarily the case, right?

[00:17:16] Frank: I like to give. More of a framework for people to think of when they're doing these prompt generations. And that framework consists of providing some context, A good rule of thumb or best practices to give you the chat bot a bit of a role to play. So I want you to act as a social media manager.

[00:17:34] Frank: I want you to act as a PR expert. I want you to act as slogan generator, for example, you want to give it some context. So it understands. What role you want it to play and then give it more context in terms of like you described here's the audience i'm trying to shoot for right single females in the fitness industry or i'm going to be looking for engaging or entertaining content you give it some context and then what you alluded to as well, which is provided input data.

[00:18:01] Frank: There's a concept called. Zero shot prompting, and then there's one or a few shot prompting. So zero shot prompting is exactly what you described, which is, I want you to draft me a blog about X, Without any examples or anything it can constitute as good, then you have one shot or a few shot prompting is when you actually.

[00:18:21] Frank: Inject in that prompt. Here's what good looks like. Here's an example of a post that I've done before that I liked, or here's five slogans that I really want you to get close to in terms of what I'm trying to go after. So it has some ideas of what good looks like, and it's going to do its best to try and get close to that good.

[00:18:41] Frank: Description that you have. So I always encourage people to take those two things into consideration, have some type of context in that prompt and then provided input data like tonality, writing style. Here's what good looks like so that you'll probably be. Happier with the output and the desired results that it's giving.

[00:19:01] Hector: Great tips. mentioned before some hallucinations, which I think is just funny. It's just a great word to say, how is this chat? But, and then there was one other one that, that you mentioned a bias. So I'd love for you to talk about and weave in really quickly over this last little bit, any bits of caution or, things to avoid pitfalls that people might be making.

[00:19:20] Hector: And I'm sure that they might not even know how, Okay. And then if you can expand on this hallucination

[00:19:24] Hector: bias kind of thing,

[00:19:25] Frank: let's start with hallucinations first of all. So hallucinations is a relatively new term now that generative AI has presented itself in our vernacular, right? Hallucinations are where a large language model like ChatGPT or Google's BARD or Cloud2 or any of those doesn't know the information.

[00:19:43] Frank: And so it will make it up. And it will do it very confidently to the point where you don't recognize that this is a made up company, or this is a made up information, If you probably ask it who is, Frank Mendoza, it might make up some information. It's not doing it maliciously.

[00:20:00] Frank: I think it's doing it because it's attempting to predict the next Word in the sentence, right? And so it's been trained on this vast amounts of web information, right? Whether that be web pages, articles, blogs, books Wikipedia, et cetera. And it's just doing its best to give you an answer and try fulfill its, duty,

[00:20:19] Frank: whatever response you asked it to. So it's making up word information accurately and unfortunately confidently. And the challenge I would, again task everyone that's leveraging this is to recognize that it can do that. it's important that you apply your due diligence and you do as much source.

[00:20:39] Frank: Credibility on that, right? Why did it come into this conclusion or why did it have this response? There are becoming more and more solutions. If you leverage Google barred online if you leverage Microsoft Bing which has an underlying chat GPT interface these are trying to do better and better in terms of responses with actual resources identified at the end of the response that you can do some due diligence.

[00:21:07] Frank: So that just recognize the limitations of these solutions and make sure Microsoft Like I've mentioned before, you maintain a human in the loop and that you're applying due diligence around any of these responses to ensure that they're accurate and that you're comfortable before you publish the information.

[00:21:26] Hector: about biases? How might that creep in?

[00:21:28] Frank: Yeah. So again, I mentioned, for example, that these large language models are just trained on vast amounts of internet information. Image generations are trained on vast amounts of images publicly available. as You can imagine There is a bias associated with that information, whether it be stereo typing people, whether it be misogynistic information, all these things are out there and they exist right now in the public domain.

[00:21:56] Frank: And so it's not unheard of and unrealistic to expect that these large language models have also been trained on the content as well. And so recognize, for example, that. When you have a response that it's given, there could be inherent bias associated with that response. I'll give you a prime example here, Hector.

[00:22:16] Frank: There was a recently, a young lady who was of Asian descent, and she was looking at using one of these headshot image generator. you apply a couple of images of yourself and it will generate AI images of headshots, right? After she submitted her images, the images that came back showed a Caucasian woman, right?

[00:22:39] Frank: So as you can imagine, not very happy with that response, but you recognize the bias associated with it. Chances are that image generator here was trained on a lot of Caucasian people to in order to get to that that's just a very kind of one example, but recognize that all these large language models have been trained on some corpus of knowledge.

[00:23:01] Frank: And it's important for you to recognize what biases they can entail that will then the models will Replicate as well.

[00:23:08] Hector: YeAh. It's not necessarily making up new stuff necessarily. it's being trained on this stuff that's already out there. And I think that's something that, people may or may not realize has been really, frankly, a In AI for businesses and really a great overview.

[00:23:23] Hector: If people want to go deeper, get into your world, learn more. I think you even are offering some support for people. Where can people go and find out about that?

[00:23:31] Frank: Yeah, appreciate that, Hector. First of all, I would love to hear about people's AI journey, how they're approaching it and where they're going. You can reach me at frank at catalytics dot AI. That's my email address. Or linked in. You can find me on linked in frank Mendoza. But Hector, I would love to offer your audience a free 30 minute strategy session.

[00:23:52] Frank: No ties, nothing it's truly free to just hear about your journey, the challenges that you're exploring. And I'll provide you a link if you're open to sharing that on your notes as well.

[00:24:03] Hector: We appreciate you for being here. If you made it this far, we would love rating, a review wherever you get your podcast. And if you know someone who is scaling the business in the midst of that journey or thinking about starting one, send him this episode. Let's let's grow together and go and get connected with Frank, myself on LinkedIn as well.

[00:24:21] Hector: Like I said, we appreciate you being here and being part of the tribe. We'll see you on the next one. Later, y'all.