Danny Maloney on AI’s Leap to Marketing Mastery
In this episode, Danny Maloney is going to lead you on a journey to explore how artificial intelligence is not just a fleeting novelty but a tool of immense utility that’s changing the playing field for individuals and small businesses alike. Danny brings his passion for algorithmic innovation from his experiences as a data-loving youth to his leadership role at Tailwind, where they leverage AI to level the marketing playing field for small businesses.
Show Notes
06:16 Early days of prevalent AI models, feedback loop.
08:37 Small businesses struggle with limited resources for marketing.
12:31 AI guides marketing decisions for faster success.
16:34 AI leader initiating internal discussions on AI’s impact.
19:18 Experts experimenting, varying responses to AI capabilities.
23:43 Early phase of tech development and impact.
26:35 Tool Dingo ported from C# to Python.
29:47 Making prompt engineering unnecessary for average users.
31:39 Requested a specific image prompt and tested.
35:03 OpenAI developing GPT-5, creating internet frenzy.
38:58 Helping users personalize and develop voice technology.
43:24 Retro tech culture and its work ethic.
45:47 Chat GPT upsets media writers, AI creativity.
48:57 Digital journey ends with gratitude and encouragement.
Transcript
Ladies and gentlemen, boys and girls, and AI entities of
Speaker:all computational capacities, welcome to another riveting
Speaker:episode of the data driven podcast. Today,
Speaker:we embark on a journey through the digital landscape where data
Speaker:isn't just numbers. It's the very fabric of our
Speaker:digital existence. With the ever charming Frank steering
Speaker:the ship solo, Andy was off gallivanting at Busch Gardens when this was
Speaker:recorded. We delve into the world of generative AI and
Speaker:marketing marvels with the illustrious Danny Maloney, the brain behind
Speaker:Tailwind. Picture this. A
Speaker:world where small businesses wield the power of giants, thanks to
Speaker:Tailwind's arsenal of automated marketing tools.
Speaker:From the nostalgic lanes of Google Street View to the cutting edge frontiers of
Speaker:AI driven marketing strategies, we're in for a treat.
Speaker:And since today is February 29th, leap day, we
Speaker:figured, why not leap into another episode? So
Speaker:adjust your antennas, polish your circuits, and prepare for an
Speaker:electrifying discourse on how Tailwind is reshaping the marketing
Speaker:cosmos, one algorithm at a time. Buckle
Speaker:up. It's going to be a data driven ride.
Speaker:Hello and welcome to Data Driven, the podcast where we explore the emerging fields
Speaker:of data, artificial intelligence, and the ever present world of data engineering.
Speaker:But speaking of data engineering, my co host, Andy, is not
Speaker:here today. He's having a fun family day at Busch Gardens.
Speaker:So, we wish him good weather and,
Speaker:short lines. So with me today
Speaker:is Danny Maloney, a successful Internet entrepreneur,
Speaker:CEO and cofounder of Tailwind, a software platform that
Speaker:provides small businesses with the marketing tools they need to compete. With
Speaker:over 1,000,000 users, Tailwind is a leader in
Speaker:generative AI and providing automated marketing plan
Speaker:creation, visual content design, copywriting, email campaign
Speaker:building, ad optimization, and more. But before starting his
Speaker:own company, he worked at Google, YouTube, and AOL,
Speaker:and is part of it apparently had, worked on something I'm very
Speaker:fond of is, Street View and Google Maps. And, maybe we
Speaker:could talk a little bit more about that. So welcome to the show, Danny.
Speaker:Great. Thanks for having me, Frank. Glad to be here. Yeah. Good to have you.
Speaker:Good to have you. So generative AI, right, as
Speaker:probably since, you know,
Speaker:last November has been on everybody's
Speaker:minds. What you've been how long
Speaker:ago did you start Tailwind? Yeah. So,
Speaker:Tailwind itself came to market in 2015. Okay. So we've
Speaker:been around for a while. You know, the way we tell our story is
Speaker:really in a couple of chapters. So there was kind of tailwind v 1,
Speaker:and there's tailwind 2 point o as we call it. And so for
Speaker:us, that was kinda 2015 to 2019 and then 2019 to the present
Speaker:day. But going down the path of,
Speaker:generation, for clients and really thinking about how we
Speaker:do more of the work of marketing for them, was
Speaker:the real theme of this second chapter. So we started on that path
Speaker:back in 2019, believe it or not, even though, you know, the market is
Speaker:really kinda waking up and starting to pay attention to your generative AI today,
Speaker:there was a lot to learn and a lot to figure out in terms of
Speaker:how to go about it the right way. And so it's been a pretty fun
Speaker:past few years for us going down that road. Interesting. So what
Speaker:does Tailwind generate specifically? Is it NLP? Is it
Speaker:LLM? Like, what is, image generation?
Speaker:Yeah. So there's actually multiple components. So like you said, we're trying to
Speaker:give small businesses the tools that they need to compete as marketers.
Speaker:So for us, that's not just one use case, but it's actually thinking
Speaker:about the entire marketing life cycle. And so we look at,
Speaker:today, 4 core components of what we help generate.
Speaker:One piece is actually the marketing plan itself. And that so that's our own
Speaker:IP, our own technology that we built that builds and
Speaker:extends and evolves marketing plans based on the specifics of the
Speaker:business. Once you have that plan, then you've gotta actually execute it. So
Speaker:now we get into the next couple components. So one is a tool called Tailwind
Speaker:Create, which is about visual generation. So we actually
Speaker:create the images that someone needs and, you know, don't think
Speaker:necessarily about a tool like a Canva or a PicMonkey in that
Speaker:sense of, you know, I'm just going through templates, and I'm designing it
Speaker:myself. What we do is we actually take in the assets and the
Speaker:inputs from a brand, and then we generate a large array of
Speaker:designs that they can scroll through and choose from instead of
Speaker:having to do the design work. So it kinda takes the design process from
Speaker:45 minutes to about 2 minutes, typically, and we generate it
Speaker:in all the formats they need for their different marketing assets. So
Speaker:visual design is the second part. The third part is copywriting.
Speaker:And so there, it's a mix of our IP and also
Speaker:leveraging third party tools that are out there. So leveraging LMS,
Speaker:for example. But that's about, you know, being
Speaker:able to help people write the copy that's gonna be convincing, that's gonna
Speaker:help communicate with their audience. And then the 4th component
Speaker:is ads. So we actually acquired a company last year
Speaker:called Nectar 9, who they themselves had been building out
Speaker:an AI driven ad management platform for about 5 years prior
Speaker:to that. And, we're now integrating that into
Speaker:Tailwind so that our users will be able to mostly automate
Speaker:the process of paid advertising also. But we look at that as a
Speaker:complete cycle. Right? So I've got a plan. I can create the content for the
Speaker:plan, then I can distribute it and get it out to my audience.
Speaker:And the data from that cycle should inform my
Speaker:plan so I can get better and get smarter into the future.
Speaker:Interesting. That's one of the things that I I I
Speaker:wonder about these these generative tools. Like, do they take the feedback
Speaker:and that, like, you know so eventually, it would get better. Ideally, it
Speaker:would eventually get better. And it sounds like you do include that in your feedback
Speaker:loop. There are some. I mean, honestly, I think we're in early days of
Speaker:that. Right? If you look at some of the more prevalent models that are
Speaker:out there, you know, the the various OpenAI models, for
Speaker:instance, I think there is an inherent feedback loop in how
Speaker:those operate as they continue to train on more data.
Speaker:Right? When you talk about things from a marketing context in
Speaker:particular, the feedback loop for us has been around
Speaker:things like, how do we train, you know,
Speaker:our models, our products, or third party models
Speaker:around specific use cases. Right? So, looking at the
Speaker:corpus of data we have around what works, what doesn't work,
Speaker:enabling it to be more tailored to the use case that the user
Speaker:is trying to solve. So, you know, when you're designing a
Speaker:pin for Pinterest, that should be different than a, you know,
Speaker:feed post on Instagram. Right? A feed post on Instagram is different from
Speaker:writing a script for a reels, right, or a TikTok. So,
Speaker:for each of those use cases, we found you can start generating much stronger
Speaker:results with more specific training.
Speaker:Interesting. Yeah. I I I wonder, like, what
Speaker:what the future holds for this type of work. You know, there's definitely a lot
Speaker:of there's a lot of, you know,
Speaker:grinding and gnashing of teeth and, you know, prog you know,
Speaker:prognostic that's that's my fault for trying to use a
Speaker:a big word on, basically, what is a kind of a holiday
Speaker:week. But, you know, but I
Speaker:think that it's interesting because it's it's probably opened up the opportunity for a
Speaker:company like you, to step in and kinda build these
Speaker:tools that really empower the smaller businesses because that seems like it's a
Speaker:pretty large, audience of folks.
Speaker:Yep. Yeah. It definitely is. And, honestly, I think it's the audience that
Speaker:needs it most. Right? Because when you look at very large
Speaker:enterprises out there, and they've got world class marketing teams and world
Speaker:class agencies that they're working with, tons of data, tons of
Speaker:resources. They've got access to the best tools. And a
Speaker:lot of small business owners don't realize that in today's world of
Speaker:marketing, when you start a small business, you're actually competing with the big
Speaker:guys from very early on. Right? There's only so much
Speaker:consumer attention out there. There are only so many eyeballs, so many minutes
Speaker:to compete for. And small businesses are
Speaker:drastically under resourced and under tooled compared to
Speaker:large enterprises. And then on top of that, it's usually the
Speaker:founder trying to figure out the marketing on their own, or maybe they can hire
Speaker:a 1 person team or a 2 person team in the early days.
Speaker:But I think, you know, you you look at this type of generative AI
Speaker:evolution that's happening right now. I think it's that
Speaker:smaller, business, and I think it's, people
Speaker:who have been on the fringes of really having access to having
Speaker:their voice heard that this potentially helps the most. So
Speaker:just an example of that, a pocket of users we've seen
Speaker:on our ghostwriter capability, which is the the copy
Speaker:generation, part of Tailwind, a
Speaker:pocket of users we've seen who are really passionate about it are people who are
Speaker:non native English speakers. Right? Right. And you
Speaker:go through user research, and you're watching some of
Speaker:the interviews, and we've literally had people in tears of,
Speaker:you know, how happy they were to be able to clear some
Speaker:really big emotional blockers for them around
Speaker:anxiety and fear of having to communicate in a non native language and
Speaker:knowing now that they can produce work at a higher
Speaker:level is is life changing. Right? So, yeah, I I think of this, and
Speaker:I look back to the evolution of the Internet and think of,
Speaker:you know, I was like you said, I was at YouTube before. And we
Speaker:had similar stories at YouTube of,
Speaker:you know, the teenager in Africa who teaches himself calculus
Speaker:through YouTube videos. Right? And Right. Right. Right. Yeah,
Speaker:or software development. Right? Who otherwise would not
Speaker:have had access to that level of education and
Speaker:couldn't have dreamt of it in some cases, but those were very real stories.
Speaker:And and it's a reminder for me of the power of technology to
Speaker:democratize access. Right? And so I think this next wave is gonna
Speaker:do that as well. And hopefully, in the long term, it just leads to, you
Speaker:know, the best ideas, the best content, the best product winning out.
Speaker:Interesting. So what what did the and, you know, what did the first
Speaker:version of Alwin look like before the generative AI? Right? Were
Speaker:you you know, what did that touch AI in any way, or
Speaker:was it kind of sort of not? Yeah. It really
Speaker:didn't, but there was a common underpinning in a
Speaker:way. So one of the things we heard very early when we
Speaker:were starting and building up the company from especially the small business
Speaker:audience was, you know, everyone's giving us tools,
Speaker:and these tools give us data and analytics. And
Speaker:we're supposed to have time and energy to go through that data and
Speaker:figure out what it's telling us. But, frankly, I don't wanna do that. Right? I
Speaker:I didn't start my business to dig through reports. I wanna spend
Speaker:time designing products or engaging with my customers or creating content,
Speaker:not doing data analytics. Right? And so what we heard was this common
Speaker:theme of, I wish I wish someone would just give me the
Speaker:tool that tells me what to do instead
Speaker:of making me figure it out on myself. Like, pushing
Speaker:bits from point a to point b is fine. Right? There's value in that. There's
Speaker:value in things like scheduling content and
Speaker:email automation. Sure. That helps scale.
Speaker:But the real part people were struggling with was knowing what to do in the
Speaker:first place. And so as we built the first version of
Speaker:Tailwind, which was largely focused on social media scheduling and
Speaker:publishing, What we found was
Speaker:certain features that we built that were more predictive
Speaker:in nature. So for example, the best time to post,
Speaker:for your audience as an example, or which you know, we have a
Speaker:hashtag finder feature from the first version. You know, which hashtag should I use
Speaker:on this post? Those were some of the features that people
Speaker:got most excited about and that really generated a lot of energy, a
Speaker:lot of curiosity around. And so I see a common thread there.
Speaker:Right? Because it was taking away those psychological
Speaker:barriers and helping people clear that question of, am I
Speaker:doing the right thing, right, by giving the recommendation. Okay. I can
Speaker:follow the recommendation, follow the doctor's orders, so to speak, and
Speaker:now I am unblocked. And I'm gonna move faster, and I'm gonna do more
Speaker:marketing, which is gonna help me build my business. And so, yeah, I
Speaker:think this generative AI wave now is the
Speaker:next level up of that concept of being
Speaker:able to guide people in an
Speaker:opinionated but data informed fashion on
Speaker:what they should be doing in far more points of their marketing journey
Speaker:in a way where beginners can get up to speed, you know, much faster than
Speaker:they otherwise would. Interesting.
Speaker:Are people comfortable with being told what
Speaker:to do, or it depends on the the audience. Right? Because that was the first
Speaker:thing that's the first thing that you said that kinda made me, like, I wonder
Speaker:how people feel about that. Yeah. It definitely depends on the
Speaker:audience. I I think the if I had to observe the divide
Speaker:there that I've seen over time, experts are less comfortable being
Speaker:told what to do because they're experts. They've earned that
Speaker:expertise. They've taken years of learning and labor to get
Speaker:there. Beginners are thrilled to be told what to
Speaker:do in the vast majority of cases. And and, you know, we we usually don't
Speaker:frame it that way. It's more of, like, guiding and recommending for that,
Speaker:then, you know you don't want it to sound overbearing. But but the
Speaker:reality is when you're at that early stage and someone's there to help
Speaker:you out and you don't have to drop, you know, $1500 a month on a
Speaker:marketing consultant, you know, who's working part time for you and 10 other
Speaker:brands. Right? Like, that's that's a breakthrough for a small business
Speaker:owner. So they tend to be thrilled, and then you kinda have
Speaker:the messy middle, so to speak, of people who are kind of becoming experts
Speaker:not quite there. Or maybe they're expert in a given area,
Speaker:and they don't feel like they need the advice there, but they don't feel as
Speaker:expert in the next area that they're trying to learn or expand into.
Speaker:And so, you know, I think that's that audience is where
Speaker:we kind of stretch up to serving. Right? And so
Speaker:we've had to build the interface and the system in a way
Speaker:where people can opt in or opt out of the advice at each
Speaker:individual point. Right? You don't have to follow it. So we
Speaker:don't fully do it for you because, you
Speaker:know, I I think even from an ethical perspective, that that crosses the line,
Speaker:especially with AI. Right. But you don't wanna be like
Speaker:Clippy either, where it's like, hey, it looks like you're hey, it looks like you're
Speaker:writing a letter. Hey, it looks like you're writing a marketing plan. Like, that you
Speaker:know? And you want there to be a cycle of learning and
Speaker:oversight there because, you know, every brand should be unique. Every
Speaker:voice should be unique. You do get into
Speaker:industries where facts really matter. Right? Like, you you need the
Speaker:marketing to be factual, and, you know, there are certainly been observed cases
Speaker:where LLMs don't always do so well at, you know,
Speaker:finding the factual information versus filling in the gaps themselves.
Speaker:But in the vast majority of cases, they do a pretty good job.
Speaker:So, you know, I think that's where you get into some AI ethics
Speaker:conversations, and there's a line we don't wanna cross. Right?
Speaker:There should still be a human involved, but we can
Speaker:empower that human to do a lot more and to do it in less time.
Speaker:Well, I like that. I like that you're you're thinking about the ethical concerns here
Speaker:because a lot of a lot of businesses
Speaker:don't really think about that upfront. Yeah.
Speaker:And so you you you mentioned the line. You don't wanna cross it. Do
Speaker:you do you have you found yourself in situation where you're kinda getting close to
Speaker:the line? Yeah. I think that's a really interesting
Speaker:question. It's something we debate, and and we're
Speaker:actually, you know, Greg Starling, who's been leading up the
Speaker:AI initiatives for us from early on, he's actually
Speaker:starting an internal kind of, like, lunch talk
Speaker:series with the team to dive into those questions
Speaker:collaboratively. Right? Like, where are people struggling with this individually?
Speaker:Where are they seeing potential conflicts in their work? Or where are they running
Speaker:up to barriers that they don't know what the answer should be?
Speaker:You know, I I think, the reality is there's
Speaker:not a big situation I can point to today where I
Speaker:can say, like, yeah. There's this really compelling example, but we know it's
Speaker:out there on the horizon. Right? And so the things that we
Speaker:worry about are, ways that AI can
Speaker:be misused, right, to spread misinformation, to do
Speaker:harm to society, to, do harm to other people.
Speaker:And we think a lot about protecting against that
Speaker:type of abuse and trying to figure out, you know, how do we even
Speaker:detect and observe that type of abuse in the first place if it is
Speaker:happening so that we can then help protect against it.
Speaker:Interesting. Interesting. What
Speaker:do you see what do the experts think of this? Do they view
Speaker:this as a threat or are they just kind of because you're you're smart
Speaker:by going after the people who are not below the messy
Speaker:middle. Right? I think that's smart because if someone wants to
Speaker:start a burger stand, they don't get into the burger stand
Speaker:business because they wanna rock social media. Right. Right? They don't
Speaker:do it because they wanna make marketing plans. They want it because they wanna cook
Speaker:burgers. Right? Like Yeah.
Speaker:In in the virtual green room, we were talking about Sonic and how they're headquartered.
Speaker:You're in Oklahoma City and how Sonic is headquartered. So now I'm thinking
Speaker:about, like, fast food. But,
Speaker:I mean, like so, I mean, is it sounds like you found a
Speaker:receptive audience there. But what do the experts say? Do you like
Speaker:they kind of, you know, are they they
Speaker:they probably brush it off, but are they brushing it off in a
Speaker:legitimate way, or is it a little bit of, you know, jealousy or
Speaker:fear? I think there's a mix. I mean, realistically, just over that large
Speaker:of a population. Of what I observe.
Speaker:Yeah. In the grand scheme of things, I'd say we're in the
Speaker:early adopter phase of generative AI. Right? So we are nowhere
Speaker:near mass market adoption yet as fast as it seems like
Speaker:certain tools have taken off. Right? Like, we're we're nowhere near mass
Speaker:market adoption yet. That's gonna be years down the road.
Speaker:And so, you know, you're gonna get a variety
Speaker:of responses. I'd say they range from, you know, some
Speaker:experts who have leaned in whole hog and are experimenting
Speaker:as much as they possibly can with all the different platforms and are publicly
Speaker:publishing those experiments on LinkedIn or Twitter or wherever they might
Speaker:be doing it, so others can learn from it as well.
Speaker:You've got, folks at the opposite end of the spectrum who I think
Speaker:are often reacting with denial or fear.
Speaker:Right? And I think you also have some just
Speaker:very correct observations around things
Speaker:that AI is not yet going to do well. Right?
Speaker:So, yeah, I I observed this especially
Speaker:within the community of kinda, like, professional SEO
Speaker:folks. Right? And, yeah, there are systems and
Speaker:philosophies of SEO and processes that have been developed that are highly
Speaker:specialized that a chat gpt is
Speaker:not designed for today. Right? And so you'll see
Speaker:these threads where someone goes to chat gpt, and they say, hey. Give me an
Speaker:idea for 10 keywords that I can target on this blog post that I'm
Speaker:writing, and it spits out an answer. Right? And the professional SEO will look
Speaker:at that and say, well, are you using data to analyze
Speaker:whether or not these are actually the right terms to be targeting based on
Speaker:where people are searching and, what you can actually win
Speaker:and so forth. And that's a very fair criticism.
Speaker:I think we're not far from the point where some tool, and
Speaker:maybe it's the tools who are already deep in the SEO space, is
Speaker:going to marry that data to generative AI. Right?
Speaker:And so now you will have the response being informed
Speaker:by data. Right? And I think that's where it becomes a lot
Speaker:a lot more frightening because then people start asking, like, okay.
Speaker:What is my job now? And that's really the deeper
Speaker:conversation I think needs to happen because what should result
Speaker:here is that people are leveled up. Right? People are leveled up to
Speaker:more strategic thinking and more strategic work, And they
Speaker:don't have to do as much of the rote processing that happens in
Speaker:a lot of jobs today. But when you're spending a lot of time on
Speaker:that today, it's a really scary thought. Right? So,
Speaker:as a society, we're gonna have to navigate that. And we're gonna have to
Speaker:figure out training paths for people to
Speaker:find their way to that next, next definition of
Speaker:their job. But, you know, I I think it's gonna
Speaker:play out over years. It's not gonna play out over, you know, months or quarters.
Speaker:Okay. Interesting.
Speaker:What, what do you think the next step is
Speaker:in this? I know you kinda mentioned it, but, like, you you did you touched
Speaker:on some of that. It
Speaker:it's interesting. It's interesting to me how that
Speaker:how this is gonna unfold. And I know it's hard to predict the future, but
Speaker:what are your thoughts on kind of, like, of this? So, I mean, we
Speaker:are early in the phase. I mean, I think a lot of people I think
Speaker:chat gpt got so much wind behind it
Speaker:because I I don't think anyone in the field expected that chat g p
Speaker:t would be as good as it was. I I I knew that we would
Speaker:come up with something like it, but I thought we were still, you know, 3
Speaker:to 5 years out from something that good. Right. And I think
Speaker:that because it kinda leapfrog people's expectations, I think that
Speaker:really boosted it.
Speaker:Yeah. You know,
Speaker:I think has really ignited people's imaginations both in good ways and
Speaker:bad ways. Yeah. I I think you're dead on there.
Speaker:And I think we're in the experiments the experimentation early
Speaker:adopter phase. Right? So if I think back to other ecosystems,
Speaker:like, you know, the early Facebook API and Twitter API
Speaker:or the early mobile app ecosystems, right, What we saw
Speaker:was this almost gold rush mentality of
Speaker:tons of experimentation, a lot of independent developers coming in and just
Speaker:building things to see what is possible. And then you fast forward 3
Speaker:years, and 90% of those projects are dead. Right? Right.
Speaker:Not Because it didn't pan out or it wasn't quite
Speaker:impactful enough or the developer lost interest. So I think we're
Speaker:in that phase right now. You know, the the good news is the 10%
Speaker:that survive that phase can end up being really impactful
Speaker:applications and really impactful companies.
Speaker:What's interesting and maybe different this time around
Speaker:is I see a lot of incumbents and established companies jumping in
Speaker:the game early. Right? Right. And I consider this still relatively
Speaker:early. You know, even though we've been at it for a while on our road
Speaker:map, That was maybe too early or, you know, super
Speaker:early. But I think a lot of companies right now
Speaker:are asking themselves, what does this mean for us and what does
Speaker:this mean for our user? Because a lot of software
Speaker:that's been built will need to be rebuilt
Speaker:and rethought. A lot of processes will need to be rethought. So, you know, a
Speaker:good, example I like is what HubSpot put out where they
Speaker:said, you know, with the simple example of using a chatbot to do things
Speaker:like update a CRM record. Right? Right. And it's
Speaker:like, right, you have to you know, before it was input, output, input, output, input,
Speaker:output, input, output, and then eventually, the record was updated.
Speaker:Right? Right. And the user had to click all those buttons and provide all those
Speaker:inputs. And now it could just be a one line
Speaker:chat input, and it's updated. Right?
Speaker:So it's rethinking interfaces. It's rethinking
Speaker:what software can actually do in someone's life. And and I think that's
Speaker:gonna be a multiyear cycle, because companies are
Speaker:gonna have to experiment. Some are gonna hit on big
Speaker:innovations. Some are gonna fail.
Speaker:But I think that's the next chapter. Right? Can can we
Speaker:take chat chat gpt and similar concepts, similar
Speaker:models from novelty to
Speaker:broad utility in a way that makes sense for people. No. That's
Speaker:a great point. And if you you kinda watch I grew up watching, you
Speaker:know, Star Trek the next generation and d space 9. And if
Speaker:you watch those shows, there's always this this
Speaker:thing where the computer becomes a character in the story in the sense that
Speaker:they say, computer, extrapolate all possibilities of the warp drive
Speaker:going, like, whatever. Yep. You know? And
Speaker:you even saw this in The Expanse, which is Andy and I's probably
Speaker:our favorite show. And there was a there's a scene where one
Speaker:of the characters is interacting, trying to find an
Speaker:ideal orbit around, getting around all these warships
Speaker:and the whole thing. But he goes, you know, what
Speaker:if I did it with minimal thrust? And he was asking you all these questions,
Speaker:and it was basically computing all of these kind of parameters. When I
Speaker:use chat GPT, I kinda feel like I'm doing that.
Speaker:Yeah. You know? You know, why one of the things that that that that
Speaker:I have for for my blog is I have a tool called Dingo,
Speaker:and it kind of helps me produce content and all that.
Speaker:And I originally wrote it in in c sharp, but I ported
Speaker:it over to Python with the
Speaker:help of chat gpt. I basically asked it, What if I wanted a pro
Speaker:I described the program. How would I write that in Python?
Speaker:And, you know, I was able and the code wasn't
Speaker:perfect to your point that, you know, they're not always factual. Right? The
Speaker:code, if I copied and pasted it, had issues, but those were
Speaker:not insurmountable issues. So I I I I poked around with
Speaker:it. Long story short, original version of Dango took about 3,
Speaker:4 weeks. The Python ported of Dingo
Speaker:took about a day and a half to get feature parity
Speaker:Interesting. Which isn't yeah. And obviously pretty
Speaker:revolutionary. And and you can just skeptic in me. It's like, yeah. But you already
Speaker:wrote it. Right? So you kinda already thought through a lot of these problems. However,
Speaker:see you know, it just seems like it's a much faster process. Because I also
Speaker:think too, it's also a pretty patient mentor. You know what I mean? Where you
Speaker:can ask it dumb questions. And as long as the server is still
Speaker:running and it's not overloaded, it's always happy to answer your
Speaker:questions. Right. Which which I think I think is really kind of the
Speaker:I saw a video the other day how it's gonna change language learning and this
Speaker:guy was talking about, you know, hey, you know, I'm intermediate Spanish speaker
Speaker:and I wanna to go to the next level. I can ask chat
Speaker:gpt to create this material that is
Speaker:custom tailored to me. And I think that is that's a fascinating use case
Speaker:because I don't think anyone thought that. That's not a use case that
Speaker:you ordinarily would do it. So I find it interesting that I never would have
Speaker:thought marketing campaigns either, you know. Although I will admit, I have written, like, a
Speaker:YouTube video description, and I'm, like, make it more exciting, make it more, you
Speaker:know, dynamic. Right? Yeah. And it does. It does. It's it
Speaker:it, you know, it's not always something I would say, but that's that's
Speaker:for me, the human to go in and kinda edit it. Yeah. I gotta hook
Speaker:you up with a ghostwriter account so you can give us feedback on that. Oh,
Speaker:sure, man. That'd be awesome. Because we've got things like, you know, YouTube
Speaker:description generation and and Right. Summarizing and cleaning up my
Speaker:copy. But but it's interesting because what you hit on there, which is you
Speaker:had already spent the time thinking through the issues and architecting it and
Speaker:and figuring out what the challenges were, I see
Speaker:a direct parallel to what we've been working on in applying this tech and marketing.
Speaker:Because I I'm taking Ghost Rider as an example, Where a
Speaker:lot of the work has come is actually in prompt
Speaker:engineering, testing, quality control, and iteration.
Speaker:And it's not a once and done process. Like, we're actually
Speaker:tracking success of various prompts and various use cases and
Speaker:going back and improving them over time because,
Speaker:I mean, first, the underlying models are changing. Right? And so, you know, that's a
Speaker:continuous force of change. But, also, more users are now
Speaker:using them, and so we have more data. But
Speaker:but I think part of the key there is you look at this field of
Speaker:prompt engineering that has now exploded all of a sudden. And while there are
Speaker:a lot of early adopters diving in there and getting really excited about
Speaker:it, the reality is the average person should never have
Speaker:to learn prompt engineering as a new skill. Right?
Speaker:Like, that is that's not where we end up from a user interface
Speaker:perspective here. And so we're kind of baking that into our
Speaker:solution where we say, okay. Part of the value we bring is that
Speaker:we have a series of
Speaker:expert created and groomed and tested
Speaker:prompts, that are trained on real data that, you
Speaker:know, helps perform an improvement or it helps improve performance.
Speaker:And that makes the technology now accessible to people who
Speaker:are not going to spend time on learning prompt engineering. And so,
Speaker:I I think we'll see those types of evolutions here. And maybe,
Speaker:eventually, prompt engineering won't be as necessary as it is at
Speaker:the moment as the models get better, But that might be a longer time line
Speaker:until you can really get to that point where you don't need
Speaker:that type of work at least going on in the background. Right. I I do
Speaker:wonder, like, is prompt engineering the next hot job title?
Speaker:Yeah. Or is it gonna be more as these models improve, like
Speaker:you say, it'll be more prompt optimization. Right? Or and I'm
Speaker:sure I'm sure we'll come up with a fancy acronym for that because we always
Speaker:do. But probably start using AI to optimize the prompts
Speaker:also, right, to to monitor and measure and and
Speaker:No no joke. I've done that. I've done that with DALL E with Dolly. So
Speaker:so last year was an interesting year in AI. Right? Because Dolly alone and the
Speaker:work that was done in image generation would have been the headline story. But at
Speaker:the, you know, the last minute, you know, chat
Speaker:gpt kinda took all the oxygen out of the room. I don't think people realize
Speaker:that. So there's actually I've actually, like, done it where I if I
Speaker:I I give it a prompt. Yep. That I wanna generate an image with.
Speaker:And I tested this, and I'll let me do a blog post on this, where
Speaker:I say, give me a picture or painting of a doctor in the style of
Speaker:Rembrandt. Right? And then I asked
Speaker:chatty Pete, hey. How would you write this as a prompt to get the
Speaker:best output for DALL E 2? And it came up with a paragraph.
Speaker:Mhmm. You know, use this type of lighting, this type of paint. I mean, it
Speaker:was just stuff I never would have thought of. And then just for grins, I
Speaker:put pasted that in and it the the the
Speaker:obviously, art is subjective, but I would say that the
Speaker:quality of that improved pump, the output was an order of
Speaker:magnitude better. That's really it. And
Speaker:that's cool because anyone can kind of experiment with that. Right? That's like a simple
Speaker:thing anyone can do, you know. You give it a basic prompt and you ask
Speaker:it to get make it make it more for, I think you're probably doing it
Speaker:now. But, but, like, it's
Speaker:fascinating. Like, it comes up with a much better quality. And and I I think
Speaker:that that's interesting for a number of reasons. One, we're using AI to talk to
Speaker:AI. Right? Which is kind of a, I don't know, like, that sounds like the
Speaker:start of every bad sci fi movie. And the
Speaker:other thing is is that that potential to create that
Speaker:better image was always in that model.
Speaker:Yeah. We're just using the prompt to draw it out. I
Speaker:I I I think there's something very
Speaker:curious and interesting about that. Right? That that model is they the model
Speaker:is maybe more capable than we're aware of Yep.
Speaker:Which blows my mind. I was looking up a
Speaker:thread that you know, pretty recent. If you look up
Speaker:Nick Floats on Twitter, at Nick Floats,
Speaker:so Midjourney came out with their new described
Speaker:endpoint. And so he's basically doing that where he's he's
Speaker:generating an image, asking it to describe
Speaker:the image as a prompt, and then rerunning that prompt to see
Speaker:what the output looks like after rerunning the prompt that the tool actually generates
Speaker:for an existing output. And it's fascinating to see,
Speaker:both how Midjourney actually describes, right, in the in the
Speaker:first place, Right. The piece of content, but then also,
Speaker:the difference in in the two outputs. Right?
Speaker:Yeah. I'm looking at his I'm looking at his Twitter feed now, and, like, I
Speaker:see the pictures and, you know, not that long ago, I would have said,
Speaker:oh, wow. You must have an artist friend or he must be an artist himself.
Speaker:Yeah. But now it's like and it's funny. I don't know. Maybe it's my
Speaker:imagination, but I can look at an image. I'm like, oh, that looks like a
Speaker:stable diffusion. That looks like a DALL E. That looks like a,
Speaker:Midjourney. Like, it's interesting how
Speaker:those models have a certain style. Yeah. I did that
Speaker:for fun last year with DALL E. I, you
Speaker:know, generated a virtual piece of artwork and then posted it on my
Speaker:Instagram. And, I think I put the caption something like, you
Speaker:know, so can I say I can paint now or something? But I didn't include
Speaker:any other context. And I got comments back from some of my friends
Speaker:like, you made that. Oh my goodness. Right?
Speaker:But Yeah. And and and what a shift there's been. Because if you did that
Speaker:today, people would be like, oh, you're using AI for that. Yeah. And You
Speaker:know? That couldn't have been more than 6 months ago that that happened. So
Speaker:It's really it's really gone fast. And you think about, like, GPT 4 is
Speaker:out and Yeah. GPT,
Speaker:5 is in the works. And
Speaker:if nothing else, obviously, they have smart people working there, but the people who do
Speaker:the marketing at OpenAI are top notch because
Speaker:they already had GPT 4 kind of waiting in the wings when
Speaker:they released g chat GPT. Mhmm. And then once
Speaker:that once everybody got all crazy over that, then they released that. And apparently,
Speaker:4 was being worked on. I mean, 5 was being worked
Speaker:on even as that was happening. So it's just it's fascinating to see
Speaker:how this is going and it you're right, though. It's I
Speaker:haven't seen people kinda go this crazy since the beginning of the Internet.
Speaker:Yep. It's just in terms of, you
Speaker:know, we are in a not an ideal economic environment. There are banks
Speaker:collapsing all that, but the investment is not dried up in
Speaker:specifically AI. Big tech tech has obviously taken a hit,
Speaker:but, you know and you're right. Yeah. Even big players are jumping in with both
Speaker:feet, You know? Although, I don't know if you played with Bard. I I haven't
Speaker:yet, personally. I I I don't wanna I know that they're
Speaker:working feverishly on it, but I was not impressed because I
Speaker:asked it. So if you ask if you ask chat g p t, you know,
Speaker:hey. Write a script that goes infectious weather data. Right? That's, like, my
Speaker:hello world. Right? Just to test it out. Right? Check CPT will
Speaker:happily give you a whole thing, talk about it, give you code,
Speaker:You can copy and paste it. It mostly works. If
Speaker:you ask Bard, Bard actually says, I'm I'm a language model.
Speaker:I I can't do that. Interesting. Like, now, again, that
Speaker:was a week ago. This is a fast moving field. Yeah. But it it's
Speaker:kind of funny how I don't know, I think if you
Speaker:can kinda sense the style that's different in visual mediums,
Speaker:like, you know, the mid journeys and the, the dollies and,
Speaker:the stable diffusions. Is there going to be kind of a
Speaker:similar style difference,
Speaker:you know, in the text generation ones. I suspect there will be.
Speaker:And it's interesting what g p t 35 knows versus g p
Speaker:t 4. Like Yeah. It it knows about the articles
Speaker:I've written. Right? So I can ask it to write an article in
Speaker:the style of Frank Lavinia. Right? And and and
Speaker:and it and it did. And I I read it and I'm like, yeah, it
Speaker:does look something like I would write. You know? Yeah. And I
Speaker:when I was writing for MSDN Magazine, well, I coulda I I could wholly coulda
Speaker:used that. But it's interesting. When I asked g p t 4, GPT
Speaker:4 has no idea who I am, which I'm not sure if I should
Speaker:be happy with that or a little upset at that.
Speaker:It just hasn't found you. Yes. Right. It hasn't found me yet. But it it
Speaker:it's it's also telling that there's a lot of motion here of
Speaker:people taking this offline. Right? So you wanna train your own
Speaker:model and, like, you know, the the amount of I think you're right.
Speaker:We're only at the beginning of the innovation curve on this one. Yeah. You
Speaker:know, like, when the Internet first came out, who would have thought
Speaker:of something like Uber or Lyft. Right? Or DoorDash. Right?
Speaker:Like that or, you know, or I don't think we can we're so
Speaker:early in it. We can't really predict the future beyond the next couple of
Speaker:weeks. Yeah. Yeah. And and it's interesting you bring up that
Speaker:example because that's something we're working on in real time in our context. Right.
Speaker:We have a a large user base. And, again, there's the the experts down to
Speaker:the beginners and all different levels of experience in between.
Speaker:And so within our member
Speaker:base, we have people who have incredibly well defined
Speaker:brand voices and styles where they do
Speaker:have enough you know, 1, have enough seed content to train
Speaker:personalized versions of the model on their voice,
Speaker:and to want that. Right? Like, they they wanna maintain their
Speaker:voice, and they don't wanna sound like everyone else. And then you have folks at
Speaker:the other other end of the spectrum who might need help even
Speaker:being introduced to the concept of developing your voice and working through what
Speaker:your voice should be and testing iterations, off of
Speaker:different sample datasets. And so, we're
Speaker:basically working through from a, you know, how does this technology come to market
Speaker:perspective, solving that exact problem now of, you know,
Speaker:for a large diverse user base, how can we give people
Speaker:the ability to tailor outputs to their voice and their style if they
Speaker:know what that is and simultaneously help people develop that if they
Speaker:don't know what it is. So that's yeah. Hopefully, we'll
Speaker:have something there in market pretty soon. I I think we're not that far away,
Speaker:but I'm sure that's something we'll need to, you know, iterate on and learn learn
Speaker:about over time too. Well, cool. So now
Speaker:we'll switch over to the pre canned questions. Okay.
Speaker:Which I pasted in the chat. Not they're not, real brain teasers. They're
Speaker:just kind of, general stuff.
Speaker:Yeah. How did you find your way into data and AI?
Speaker:Did did data find you, or did you find data?
Speaker:I always loved math. Always, always, always loved math. If I had
Speaker:to really answer this, I'd say, you know, the 2 earliest examples that come to
Speaker:mind, I was nuts about baseball growing up and
Speaker:baseball statistics. Oh, cool. That was probably
Speaker:one of the first ways that data found me. And then a little bit later
Speaker:than that, the the first company I ever tried starting when I was in
Speaker:college was, essentially arbitraging the collectibles
Speaker:market. So think about things like Magic the Gathering cards. Right?
Speaker:And so, yeah, I I began tracking
Speaker:prices that different collectibles were selling at on eBay and other
Speaker:platforms. Yeah. This is going back,
Speaker:over 20 years at this point. But Right. Right.
Speaker:Tracking prices to learn what was a good price and what was a bad price
Speaker:before we had that information readily accessible and then buying and selling
Speaker:against it. Those are the or 2 of the examples,
Speaker:I think. But, yeah, data just always spoke to me. I I've
Speaker:always loved math, and so it was symbiotic. Yeah.
Speaker:That's funny. I was also a huge baseball fan growing up, and
Speaker:it's one of those I mean, if you're if you're if you're a
Speaker:baseball fan, statistics is a natural
Speaker:field for you to study because you've already Mhmm. You've already done a lot
Speaker:of it. Right? So it's it's that's it's it's interesting.
Speaker:We'll have to figure out what who who which team you root for.
Speaker:But, I don't know if I wanna admit that.
Speaker:Well, my my Xbox gamer tank
Speaker:was Frankie Bronx, so you could probably figure out that I'm a Yankee fan.
Speaker:That I grew up mostly. I'm sorry. No. I expect the
Speaker:hate mail to come in. Yeah. I I grew up mostly in South
Speaker:Florida, and the Marlins came about. So when I was very young, we were a
Speaker:mess household. Oh, okay. And then, the Marlins came
Speaker:about, and I became a Marlins fan. So yeah. Cool. Being a Marlins fan, it's
Speaker:been a few years that were really fun and a whole bunch of misery.
Speaker:I I I've noticed that. I'm kinda surprised at that, but I
Speaker:guess that's what it that's, you know
Speaker:although you went from being a Mets fan. I know Mets Mets have good years
Speaker:and bad years. Some would say good decades and bad decades,
Speaker:but Yep.
Speaker:The nice thing about baseball is that there's always another game and another season, you
Speaker:know, you can you can always hold out hope. Yeah. Alright.
Speaker:So on to the next question. What's the favorite part of your current
Speaker:gig? Yeah. I think it's it's honestly this period of
Speaker:innovation we're in. It's fun. It's new. The answers
Speaker:are unknown. There are so many different paths it could take, and I
Speaker:think there's a lot of good that can be created. So it it reminds me
Speaker:in a lot of ways of the early ideological days of the
Speaker:Internet. And, you know, I think we need to learn
Speaker:from that chapter in terms of things that weren't
Speaker:regulated or managed well as the Internet grew
Speaker:and make sure we don't make those same mistakes again. But that that
Speaker:excitement's back for me personally. I think we're experiencing a pace of
Speaker:innovation all of a sudden in software that we really haven't seen,
Speaker:you know, in 10 or 20 years. Yeah. That is a really
Speaker:I that is a that is an excellent point. I think the fact that
Speaker:they kept pushing out models every few weeks publicly
Speaker:and being
Speaker:able to keep up with the met the demand more or less,
Speaker:as as does take me back to those days where people were, like,
Speaker:I'll never forget. It was an ad I saw in a magazine, which one of
Speaker:them one of the millions of web development magazines that came out in 96.
Speaker:Right? Yep. And, and it was, like, you know,
Speaker:they show a picture of somebody with a with a sleeping,
Speaker:bag under their desk and somebody checking their email right away,
Speaker:like like, half in the bag, half well, that sounds bad. Half in the sleeping
Speaker:bag and half on the computer. And they were saying, you know, something
Speaker:like and the caption was, what did you miss? Like, it was like it was
Speaker:like really that type of mentality that you're right. I haven't
Speaker:seen this in a while.
Speaker:So the we have 3 complete this sentence, questions. The first
Speaker:one is, when I'm not working, I enjoy blank.
Speaker:Being a dad. That's that's it for me.
Speaker:You know, between founder life, that's about all there is time for, honestly.
Speaker:Yeah. But, our daughter is at a really fun
Speaker:age. She's gonna turn 8 pretty soon here, and,
Speaker:I just enjoy being able to be goofy and have fun and
Speaker:play. And, yeah, she's wonderful.
Speaker:That's cool. That is a fun age. My youngest is 8. So
Speaker:Yeah. The, the next
Speaker:question is, complete the sentence.
Speaker:I think the coolest thing in technology today is blank.
Speaker:Well, I guess I gotta say a tailwind. Right? What we're doing here.
Speaker:Right. I really do think a lot of what we're doing is cool, but but
Speaker:more broadly, I'd say, yeah, I think
Speaker:the coolest thing is that our
Speaker:perceptions of what technology is capable of are changing
Speaker:very quickly, and, that's fun. Yeah.
Speaker:Absolutely. I mean, one of the things you would hear was that
Speaker:creative jobs were safe for a long time. And I think
Speaker:that, if anything, we've learned in the last 3 to 6 months,
Speaker:that's not necessarily the case.
Speaker:Yeah. So I think that's also a part of, you know it's it's probably not
Speaker:a coincidence that, you know, Chat GPT upset people more than the image
Speaker:ones because the pack media,
Speaker:the people who write those articles, their jobs are, I think, are in I wouldn't
Speaker:say imminent jeopardy, but they are definitely on the firing line. Yeah.
Speaker:Whereas, you know, a year ago, oh, no. No one AI can never be
Speaker:creative. And yet, here we are. Mhmm.
Speaker:Alright. The last, complete the sentence. I look forward to the day
Speaker:technology I can use technology to blank. Teleport.
Speaker:Yes. How's I like that. I
Speaker:like that answer. Yeah. That's even better than self driving cars because
Speaker:you wouldn't waste time in the car at all. Exactly. I love
Speaker:seeing new places. I hate getting there. Yeah.
Speaker:And we should probably ask chat g p t if there's a way we can
Speaker:make air travel less awful.
Speaker:We'll have the gpt airlines soon. Too. GPT
Speaker:Airlines. That's funny.
Speaker:Alright. So share something different about
Speaker:yourself, but, we like our clean Itunes, rating.
Speaker:So keep it keep it within those parameters.
Speaker:Oh, this is a tough one. I I never know how to answer this, honestly.
Speaker:Something different about myself. Like, you know, I I don't know if it's super different,
Speaker:but I'll just say I'm I'm a huge strategy game nerd. Oh,
Speaker:interesting. And so, yeah, I mentioned with the magic cards before, it's
Speaker:yeah. I still love magic. I love settlers, risk,
Speaker:chess. Throw any strategy game at me, and, I
Speaker:could lose myself for days. That's cool. That's cool.
Speaker:So, Audible sponsors data driven.
Speaker:Can you recommend a good audiobook if you do audiobooks? If you don't do audiobooks,
Speaker:just recommend a good, regular old dead tree book.
Speaker:Yeah. So, I'll recommend oh, I'm, like, staring at my
Speaker:bookshelf here. I think BoomTown
Speaker:is a really interesting one. So,
Speaker:before I moved to Oklahoma, I knew nothing about Oklahoma.
Speaker:Uh-huh. And, its history is absolutely fascinating.
Speaker:So, I think that's a really cool one, for people who
Speaker:have never been here before and just wanna learn about a new place
Speaker:and a new place and time. It might challenge a lot of perceptions,
Speaker:but, it's a really good read.
Speaker:K. I'll have to check that out. And where can folks find
Speaker:out more about you and what you're up to? Yeah. Absolutely. So,
Speaker:for Tailwind, we are tailwindapp.com.
Speaker:That's our website. And you can follow us and and find us pretty easily on
Speaker:all the different social platforms and blog. For me personally,
Speaker:you know, probably just look me up on LinkedIn. I'm Daniel Maloney.
Speaker:I think Daniel p Maloney is my handle. But that's probably one of the
Speaker:platforms I'm more active on these days, in terms of
Speaker:wanting to connect. Well, that's awesome. Well, thanks for joining
Speaker:us, and, I'll let Bailey finish the show. And just
Speaker:like that, we've reached the end of today's digital odyssey.
Speaker:A huge thanks to our phenomenal guest, Danny Maloney, for
Speaker:sharing his insights and to Tailwind for redefining the marketing
Speaker:landscape. To our listeners, your curiosity
Speaker:fuels this journey, and we're immensely grateful for your companionship.
Speaker:If today's episode sparked a bit of that data driven wonder in you,
Speaker:why not share the love? Like, share, and
Speaker:subscribe to keep this conversation going and to ensure you never miss an
Speaker:episode. Until next time, keep those circuits
Speaker:buzzing and your data flowing. Cheerio.
About the author, Frank
Frank La Vigne is a software engineer and UX geek who saw the light about Data Science at an internal Microsoft Data Science Summit in 2016. Now, he wants to share his passion for the Data Arts with the world.
He blogs regularly at FranksWorld.com and has a YouTube channel called Frank's World TV. (www.FranksWorld.TV). Frank has extensive experience in web and application development. He is also an expert in mobile and tablet engineering. You can find him on Twitter at @tableteer.