Suresh and Dave chat with Adam Buehler SVP, Digitas NA

In this episode of the Web3 Marketing Association Podcast, hosts Suresh and Dave Wallace are joined by Adam Buehler, Senior Vice President of Creative Technology at Digitas North America. Together, they dive into the fast-moving worlds of generative AI and Web3, exploring how these two transformative technologies are beginning to converge.

Adam outlines the rapid rise of generative AI—its “iPhone moment”—and explains why tools like ChatGPT and MidJourney are changing the way marketers, brands, and creatives think about content. From automating everyday tasks to reimagining search, art, and design, generative AI is no longer speculative: it is disrupting industries in real time.

The discussion also touches on the risks: deepfakes, hallucinations, and the erosion of trust in digital content. Adam explains how blockchain and Web3 technologies could provide solutions, particularly through cryptographic provenance and content authentication standards such as C2PA, which could safeguard society against a collapse of shared reality.

Throughout, the trio debate how AI might reshape the agency model, whether brands should set new guardrails on creative content, and how marketers can responsibly harness these tools. A recurring theme is partnership: machines won’t replace humans, but humans who use machines will outpace those who don’t.

This is a thought-provoking conversation for marketers, creatives, and technologists alike—highlighting both the promise and peril of the coming collision between AI and Web3.

Transcript

Speaker 00: 0:00

For all the talk about chat GPT, there is a future, and we should say a near future, where artificial intelligence and machine-created reality will be so compelling, so authentic, that humans won't be able to tell the difference. Now, this raises all sorts of issues when it comes to creativity and marketing, and today on What the Web 3, Suresh and Dave have Adam Buehler, the Senior Vice President of Creative Technology at Digitus to walk us through the hope and the dangers just ahead of us.

Speaker 02: 0:51

Hello.

Suresh: 1:11

Good morning, good afternoon, good evening to all our listeners. Welcome to yet another episode of Web3 Marketing Association's official podcast called What the Web3. Once again, it's Dave and I speaking with someone cool in the emerging technology space. This time around, we have with us Adam Buehler calling in from Boston, Massachusetts. And of course, Adam, for some of you marketers, you probably already know him. Some of you who don't know him, we'll let Adam introduce himself properly, but Adam is the Senior Vice President for Creative Technology at Digitas in North America. Previously, he's done all sorts of innovation technology stuff, his backgrounds in experiential marketing backgrounds and all sorts of stuff from in data technology and creativity. So really pleased to have Adam on board. In today's episode, we will have a deep conversation about generative AI, something that Adam's been leading and in the forefront of talking to his clients, getting deep into it, you know, in the leading edge of the sword there from a marketing perspective as well as we can't let go of him without having a conversation about web3 and how ai and web3 come together both of our passion points it's good to have adam here speaking about two of these big tsunamis which are going to collide at some point and create some magic so we will pick his brains on those without further ado adam welcome to this call good morning

Adam: 2:37

good morning it's really exciting to be here i love to nerd out with fellow Hello, nerds.

Suresh: 2:43

That's exactly what you're going to get. So let's nerd out. The first question for you, and this is the first question that Dave and I ask everyone on this podcast, what the Web3 is going on with you right now?

Adam: 2:55

I'm going to interpret that question to mean specifically, like, where's my head at in the Web3-verse? That's right. Wow, there's so many different things to talk about. We have a really vibrant internal chat group at Digitas and Publicis Group, the holding company of the Digitas is part of that is talking about web three from morning till night and when i wake up in the morning i have to read back through the last eight hours of conversation so for me the really exciting stuff that I'm seeing in Web3 right now, of course, are the branching out into loyalty programs the way that Starbucks is doing with Odyssey. And that's getting a lot of internal discussion for us because we see that as a break into the mainstream to bring the technology to the big screen while obscuring a lot of the complexity of Web3 for the average user to make the onboarding really simple. So that could be... a breakthrough moment where Web3 gets a lot more traction with the average person. And we're watching that really closely. Or is your question more about what I'm personally doing in Web3?

Suresh: 4:02

That's great. And tell us what you're personally doing in Web3. And also tell us about what you're personally doing in AI.

Adam: 4:09

I was hoping today that we could talk about the places where Web3 and generative AI are starting to melt together. But Maybe first we should talk about Gen AI for people who aren't familiar with what it is.

Suresh: 4:25

Thank you.

Adam: 4:26

Don't mind if I'm just kind of meandering on the path here. So If you have been reading the trades lately and following what's going on in the industry, you've probably noticed that AI is having a bit of an iPhone moment where it's suddenly everywhere, right? And people are waking up overnight to the fact that it is threatening a lot of long dominant players in the space. Google, for example, are having what they're calling a code red moment over the potential for generative AI to undo a lot of the monopolistic chokehold that Google's had on the space. So what is behind this iPhone moment? Why is it suddenly having millions of users who are accelerating and enhancing their own work streams, 6X, 10X with these tools overnight? And that's not an exaggeration. Something like Microsoft's chat GPT-based co-pilot, which is a tool that automatically writes lines of codes for engineers, and it's also purely generative AI, is turning out to be like a competitive edge for programmers and developers who use it that's vast over those who don't. So we're here to talk about real impact of these tools. We're leaving the realm of science fiction and speculation and getting into quotidian, daily, you wake up in the morning and go to the office and have your coffee and you sit down and you start working and suddenly you're jet propelled right that's what a lot of this stuff is about and why people are so excited about it so I think of it, generative AI, as we saw the first wave of AI, those kind of crappy chatbots that Facebook tried to foist on us, the messenger bots, and everybody thought that that was going to be the future of e-commerce. Those bots did indeed use artificial intelligence, but only so far as to be able to interpret and understand the natural language the open speaking that people were bringing to them and then figuring out the user's intent so there is ai there i am understanding language which in the 60s and 70s entire books were written about how that was never going to be possible for a computer to understand spoken speech but here we are big breakthrough 2015 2016 became commonplace with alexa and siri and so on so the hatch was one Once it interpreted your intent, which is miraculous for that time, it then handed itself off to a set of very rigid services and APIs. From that point on, you were handed off to a decision tree that was very brittle and could only respond with what it had already been explicitly set up to respond. If you put input in that it couldn't recognize, it couldn't handle it. not a good user experience for the average human and a lot of people did not fall in love with this first wave of chat bots and so there was a hype cycle and there was excitement and everybody thought that automated customer service was about to come into its own and that largely didn't happen those bots came and went and we hadn't been talking about them for a few years generative ai is not the whole package yet it is not equivalent to three range human agency but it's a huge leap forward from those decision tree conversational tree you have to stay on the path chat bots that we used to have now we can understand and interpret and parse your natural language input which is great we know what you're trying to say and what you're asking and now It can go into this vast, vast bucket of quote-unquote knowledge that it's been trained on, where let's take one of these tools, for example, GPT, made by OpenAI from San Francisco, and they trained this tool. They fed it all of Wikipedia. all of social media up till 2019, 2020, billions and billions and billions of words, and began to teach these systems about the statistical relationships between each of these words, sentences, fragments of thought, so that the system gradually started to get a sense for what the next best word would be, what the next best phrase would be, what the next best fully formed thought would be, and voila, you have a system that can respond with making inferences from the vast body of knowledge that has been trained on to speak to you and to provide you information in a very, very human way, a way that is so human that it freaks people out. I've been freaked out multiple times. There was one time I was having a conversation with GPT and I was so unnerved that I had to pull my car off to the side of the road and just sit there and calm myself down because, you You know, I grew up on a steady diet of science fiction and this is becoming reality. So these tools now There's an interesting thing that happened when you train a system on a vast corpus of knowledge and you give it enough juice and the data set is large enough. These are called large language models. It's a crude name for them, but that's how they work. They're so big and there's so much information for it to fill in blanks from that eventually you get magic emergent properties that weren't expected if i train gpt on all the human thought of the you know last 10 years A funny thing happens. What if I now, you know, I can ask it, describe strawberries to me, and it'll give me a paragraph on strawberries that looks like it came out of a dictionary or an encyclopedia. That's great. And then if I twist the prompt to tickle other neurons in its system, and I say, instead of describe strawberries to me, I say, describe strawberries to me from the perspective of a scientific researcher. Now, that spin that I've put on the prompt on the question that I'm asking GPT, now it comes back in a very, very formal, looks like a scientific paper, looks like a paper that's been submitted for peer review. And it gets into the details of the inner structure of the strawberry and how it's composed. And it responds in that way. That's great. But there's an emergent property beyond that. It so happened that if I asked GPT in English, what's two plus two? It answers four. It can do math. It was not taught how to do math and it was not trained how to do math, but because it is looking for the statistical probability of the next best answer based on everything that is read and it's read the equivalent of what we would spend millions of hours consuming, it has answers that we didn't explicitly give it. And that's powerful and makes the hair stand up on the back of your neck.

Dave: 11:56

I remember there's the Turing test, which is basically it kind of responds in a way that it would seem to be human. And I guess actually responding to questions or responding in a way that it's answering a question you haven't asked sort of feels very human. So how near to passing the Turing test is GPT?

Adam: 12:18

That's an overlooked question. that I think is really important. So the Turing test is I sit down in front of a computer and I have a conversation that's carried out via typing. And I can't see if the person on the other side of the screen is a flesh and blood human or is an automated system. Then the Turing test, the proposition Alan Turing said, when the day comes that you cannot fool the system, you cannot tell which one it is. You simply can't figure out which one you're talking to. Then we have entered a new epoch of of AI. But David, what's interesting is we have blown so far past that now, the test is no longer applicable almost. Fascinating. We need a new test. We need a new test. Yeah. But the thing I want to convey to the listeners of why is this useful or interesting or like, what does it mean to them? Even just bringing it down to the level of marketing is you can quickly make a leap here. toward not just efficiencies in certain tasks that we're doing in the agency and brand world every day but you can look at entire categories of disruption that are exploding all around us right now it's a very exciting time web search just to pull one out of the air for the first time I personally, you know, I have never imagined that anybody could dethrone Google for web search. And they are just so damn good at what they do. And they have been doing it for so long. I mean, do the thought experiment, right? Can you picture a world where Google does not have 92% of all incoming search query, which is the stat that they currently do have? It's hard to picture. But now I can see a path to that because if I ask Google, tell me about the Battle of Hastings, I will get back 10,000 ads Then I will get maybe one extracted summary sentence that's not quite enough for me to go on if I'm doing a scholarly research paper. And then I will get dozens and dozens of links that put the onus on me to pick which one is going to be most applicable, and then I have to go in there and dissect it. It's not the optimal user experience. If I ask chat GPT, tell me about the Battle of Hastings, I get a perfectly conversational, to the point, no fat response written in perfect English, as long as I want one paragraph, 10 paragraphs that does everything for me and does not come with any cruft or extraneous junk on it. And the first time you ask a question to chat TPT, and it comes back in this neat package, as if you've just asked your best friend who happens to be an expert on the battle of Hastings to give you a summary paragraph, you go, ah, why would I need Google for this anymore? Now they're a lot of caveats there's a lot of work yet to be done for that to be the right paradigm but for the first time we can imagine a better search experience so my clients are all over me to tell them do i need to cancel all my spend with google right like what should i do and there's a lot of nuance to be unpacked there in what to tell them because it is a new age that's what this generative ai tooling is bringing

Dave: 15:50

One of the rebuttals to that is actually the knowledge of GPT starts from 21 backwards. So things, as far as I can see, are not up to date. Whereas Google, as soon as you think it, it seems to be in the search engines ranking. So the question which I've got, which is related to that, but the difference between a previous version of GPT and the latest version, the difference is just so incredibly big. A few weeks ago, So I was writing an article myself on generative AI. This was before GPT-3 came out and I used a tool called Jasper, which kind of helps with writing and copy and search engine optimization, copy and all that. It was good, but it wasn't brilliant. And then, you know, I repeated the same thing with GPT-3 and I was like, wow, this is just like incredible. So the next version of GPT, is that going to address that sort of issue? of timeliness.

Adam: 16:51

That's the joke that we have inside my agency, which is anything that you think is a couple of years off probably happened last week and you just haven't heard about it yet. And that's applicable to GPT because GPT-4 is supposed to be coming this quarter. And that is an order of magnitude more brain melting than what you're seeing now. The compression of time between major breakthroughs is accelerating. And I find that both exhilarating and terrifying in equal measure I do want to say one thing before we move on that you did correctly point out that these large language models are trained up to a point, and then they're ignorant of anything that happens after that date and time when their model was created. You're absolutely right about that. And there are a lot of very smart people who are working on ways to wire in the latest stock information, for example, that's up to the minute to give GPT and the systems like it access to real-time data. And there's two schools of thought, one school of thought says there's an Oracle problem and these systems will never be able to be smart enough to be real time and the other school of thought says hold my beer I'm going to take care of that problem for you right now you know philosophically depends which way you're going but to your point if the system let's say it you know stopped being trained in 2019 but you ask it you know who was president in 2020 the problem with large language models is it will happily peacefully with a beatific smile on its face imaginary face it will hallucinate and lie to you and tell you whatever it thinks is close enough because these systems are giving you what it thinks is the next best piece of vocabulary, not the next best piece of knowledge. That is the Achilles heel of all of these large language model systems. They are not smart but they are an incredible simulation of smartness and in most cases the simulation is good enough that it is immaterial it doesn't matter but there are plenty of instances where it will happily hallucinate an answer to you and you might never know the difference and that's yet another group of people are working on ways to surface citations within these responses so that you can go back and check to make sure you haven't just been lied to in a happy way so there's this endless untangling of issues that we're going to get to watch play out over the next few years and it's going to be very interesting

Suresh: 19:44

i love it large language models and the language could be anything right it's not just english oh yeah it could be music it could be a whole host of things the couple of things that fascinate me definitely i've been on the discord for mid journey and i see people create to create generative AI images. An artist is creating images. I saw someone who asked Midjourney to create, he was an interior designer and he said, here are the elements that I want and this is the age of my daughter and I want you to develop a design for her bedroom. And what came out of Midjourney was just stuff that we could have never imagined. And the number of variables you can see and the kind of things that you can put together and the amount of creativity can unleash without all the biases in them is what is mind-blowing. And in the context of brands, do you see you are the head of creative technology, your business was born out of the creative agency model, will a lot of what is happening in the agency model, or even the modern agency model, are they ripe for disruption? Everything from music to art to commercial advertising stuff, copy to So will all of that get disrupted?

Adam: 21:01

The short answer, in my opinion, is absolutely. But the more nuanced answer is a lot of artists and creative types are rightfully concerned that if these tools get good enough, you could imagine a near future where the human element may be less necessary. I would argue, instead, I am old enough to remember when the portable handheld calculator first became widespread. And there was a moral panic at that time. This was 40, 50 years ago. And people were saying that the introduction of the calculator was going to cause everybody to never know how to do math again, and that humans would lose that capability. Instead, what happened was the calculator became an essential tool to boost the productivity of the human. I think that's the analogy that I want to go for here with the creative endeavors. We are seeing these kinds of generative images, and I love mid-journey. I've spent hours going crazy with it. Right now, people are being kind of tentative in its usage. We are using it internally to start the thought process, to create storyboards, to do mood boards for a campaign, but we are not yet tripping the wire and using the generative assets in the wild yet. Maybe other agencies are, but mine is not. First, the reason is because There's a bit of controversy. These generated images are generated from original pieces that were made by humans at scale. We're talking about millions and millions of images, but There have been instances where you can recognize what inspired the AI when it was imagining up the beautiful artwork that it gave to you. And especially in the case where you say, and this is something you can do with generative AI, especially when you say, give me a picture of a space battle in the style of this famous contemporary artist. And then it will come back and be, absolutely you couldn't tell that it wasn't made by that person which is amazing in and of itself understandably that person is not happy that contemporary artist you know so you can say picasso you can say van gogh but if you start to say somebody who's still alive things get messy so what we are finding is necessary for us to do is build our own internal toolkits that allow us to take a piece of generated art and trace it back and figure out if it is a certain percentage of similarity to an existing piece of work which is in itself an ai based task there's a website called have i been trained.com where you an artist can put in your artwork and see if it was part of one of the training sets that Google or OpenAI or Microsoft or whomever used to make their model that informs their AI. So that's one of the speed bumps, but there are other uses. We don't have to use generated AI explicitly to make visual content for ads. Another way you can think about these tools is banner advertisements and rich media on the web, which you need to make dozens and dozens of form factors and aspect ratios for different websites and different use cases. This is an extremely time-intensive task for humans. I give you a banner image and I want you to apply your creative thinking to reflow it into a tall rectangle, into a long horizontal bar. I can tell generative AI to do that for me in less than a minute. Thousands of them, any aspect ratios I want, nothing simpler. based off the original image. So there are many ways to peel this onion.

Dave: 25:24

Going back to original content and things, I mean, one thing I've suddenly thought of is our brand's going to have to start being quite careful about what their agencies are giving them in terms of content and what are the guardrails. I'll just give you an example. My son had an essay to write before Christmas, and I was like, oh, just go to this thing called ChatGPT. and use that. And he said... Great parenting, great parenting. Well, he turned around and it's, who's the adder? He turned around and said, no, that's cheating. And he said, by the way, have a look at this TikTok, which was someone reverse engineering a piece of generative AI and saying, well, you can tell it's kind of generative. But it does make me think that actually, you know, for brands, there is going to be a need to be quite careful about where things have come from.

Adam: 26:16

That's an interesting way of coming at it. Like, we know that the chat GPT is right on the cusp, like standing teetering on the edge of being able to pass the bar exam, for example, which human lawyers in training don't. can tell you how difficult that is. And schools around the country are banning the use of chat GPT. China, as a country, has started furiously passing laws to stamp out generative AI almost entirely because they see the threat of inauthentic content and manipulated content. Right now, David, if you are a brand and an agency is giving you copy we're still in the last innings the last few minutes historically of being able to tell that it was generated there are some extremely subtle telltale signs that ironically enough another ai can detect there are ais now that are standing up to detect ai generated content it's a bit of an arms race but soon The only way we will be able to trust the origin, provenance, and authenticity of any piece of content, written, visual, or even video, and here's where we get our Web3 kicks in, it is not far off that the only way that we're going to be able to tell where the content came from is if it has been watermarked using asymmetric encryption which is the tool that powers blockchains the only way you'll be able to tell if you know your agency is giving you something legitimate or at least that it came from your agency and nowhere else or that if a photograph came up from a particular photographer and hasn't been altered is its cryptographic signature derived from that image is going to need to match the key the public key of the creator and this is as you know the same mechanism that allows us to hold our crypto and allows us to verify content on the blockchain and use math to show that it hasn't been altered and to follow a chain back to its genesis so this is where AI and Web3 really go hand in glove. Adobe is leading a consortium of like 800 other companies to make something called the, I had to write this down because I can never remember it, the Coalition for Content Provenance and Authenticity, C2PA. This is a standard that uses asymmetric encryption, the tool that makes the blockchain possible, that makes crypto possible, that makes it possible for you to submit credit card numbers over the internet without somebody intercepting them and decrypting them that same tool is going to be built into hardware cameras video cameras and i'm going to establish that this camera belongs to me because only i know the seed phrase the code this will sound familiar to people in the blockchain world and i'm going to use that public key to prove that this image this set of images are mine and if the math doesn't add up then somebody has messed with this image and made this politician look like they're saying something or doing something that they did not actually say. That is the only way that I can see that we're going to avoid a mass societal calamity where there's no more consensus reality, which is what's going to happen if tools like this don't get traction. That's a conversation in itself.

Dave: 30:18

Wow. I mean, that's such an interesting use case. I can absolutely see it. It's already started. Yeah,

Adam: 30:24

I

Dave: 30:25

think it makes perfect sense. I mean, this is where I sort of feel like Web3 technology is really going to kick in, you know, because this is going to be a massive problem, isn't it? I mean, already you watch videos of people who are sort of spoofs of other people, and you go, it's not far before that's going to create real, real problems, isn't it?

Adam: 30:47

Yeah, do the thought experiment and really sit and marinate in it. Imagine a world where you cannot trust anything you see or read on TV, any content. You just won't know. You're talking about UK politics, I think. But here in the States, right, we've had some first crude, Attempts at this, a year or two ago, there was a video of Nancy Pelosi that was slowed down to make it seem like she was slurring her speech and that she was drunk. And it was an obvious fake, but what if it hadn't been an obvious fake? What if somebody did a good job with that, right? That's very destructive. And how will you know? Eventually, the generative AI toolkit is going to be impervious to detection. Not eventually, there are instances now where you just cannot detect. well and so we need to be thinking about very actively thinking about tools to prevent civilization from ending from the threads that hold us together right now. The few shared glimmers of cohesive truth that we all still believe in from blowing off into the wind. And technology is going to have to be both the problem and the solution in this instance.

Suresh: 32:04

Agreed, agreed. And I think the way that I look at it is in some ways, AI is trying to make machines look like humans and become more human. And on the other end, Web3 is trying to make the world that we live in much more real through digital asset ownership. Right. So it's sort of the point that you make, which is technology is the solution for the potential evils of the technology and the panacea for the evils that might come from the technology in terms of creating challenges for humans that we don't know exist. And someone said, you know, in a recent conversation that I had about this, someone said, it's not going to be humans versus machines. It's going to be humans and machines. So the point you made about the calculator, the other point that the person spoke was they said that it's going to be the race is going to be between humans who have the machines versus humans who don't have the machines.

Dave: 32:57

Yes. Fantastic. Well, listen, thank you so much. That's been so, so, so instructional and really enjoyed it. Real food for thought. I do appreciate you joining us today. So thank you so much again.

Adam: 33:12

Anytime. Let's do it again.

Speaker 02: 33:19

You have been listening to W3MA, an NMD Plus production.