Discover how Publicis Groupe's massive $326 million AI investment is set to revolutionize ad agency operations on Episode 81 of The Marketing AI Show with hosts Paul Roetzer and Mike Kaput. We also discuss Google Bard's leap over GPT-4 in AI model rankings and introduce our mission to enhance AI literacy for all with our latest educational courses. Tune in for even more insightful analysis on these significant AI advancements!
Listen or watch below—and see below for show notes and the transcript.
This episode is brought to you by our sponsors:
Many marketers use ChatGPT to create marketing content, but that's just the beginning. When we sat down with the BrandOps team, we were impressed by their complete views of brand marketing performance across channels. Now you can bring BrandOps data into ChatGPT to answer your toughest marketing questions. Use BrandOps data to drive unique AI content based on what works in your industry. Visit brandops.io/marketingaishow to learn more and see BrandOps in action.
Today’s episode is also brought to you by Marketing AI Institute’s AI for Writers Summit presented by Jasper, happening virtually on Wednesday, March 6 from 12pm - 4pm Eastern Time.
Following the tremendous success of the inaugural AI for Writers Summit in March 2023, which drew in 4,000 writers, editors, and content marketers, we are excited to present the second edition of the event, featuring expanded topics and even more valuable insights.
During this year’s Summit, you’ll:
- Discover the current state of AI writing technologies.
- Uncover how generative AI can make writers and content teams more efficient and creative.
- Learn about dozens of AI writing use cases and tools.
- Consider emerging career paths that blend human + machine capabilities.
- Explore the potential negative effects of AI on writers.
- Plan for how you and your company will evolve in 2024 and beyond.
The best part? Thanks to Jasper, there are free ticket options available!
To register, go to AIwritersummit.com
Listen Now
Watch the Video
Timestamps
00:05:07 — Publicis to invest 300 mln euros in AI plan over next three years
00:22:14 — Bard with Gemini Pro surpasses GPT-4 on top LLM evaluation leaderboard
00:29:49 — Introducing AI Literacy for All
00:42:50 — The career trajectories of the researchers behind one of AI’s most important papers
00:47:04 — Google announces AI video generator Lumiere.
00:51:21 — Chrome is getting a new experimental AI writing feature.
00:55:04 — Biden robocall tells NH Democrats not to vote, ElevenLabs tech is likely behind it.
00:58:09 — White House calls for legislation to stop Taylor Swift AI fakes
01:01:22 — What AI can’t—and shouldn’t—do
01:04:39 — One easy way to apply AI throughout your day
Summary
Publicis to invest 300 mln euros in AI plan over next three years
We just received big AI news in the agency world… The world's largest ad group is spending $326 million on AI in the next 3 years:
Publicis Groupe is building "the industry's first AI-powered intelligent system." The system is called Core AI and it is a central AI system that will be accessible by all 100,000+ of the group's employees.
“We are bringing together all the data of the group, all the knowledge of the group,” the company's CEO told ADWEEK.
The publication says CoreAI is being built and trained in-house by a whopping 45,000 engineers and data scientists. And it's trained on the firm's trillions of data points, billions of personal profiles and daily bid impressions, and millions of creative assets.
The vision is to have each employee plug into this centralized intelligence anytime they serve clients, to instantly converse with the company's decades-worth of proprietary data and expertise.
In other words, they are essentially building an ultra-intelligent machine brain that sits at the center of the company and augments each human employee.
Bard with Gemini Pro surpasses GPT-4 on top LLM evaluation leaderboard
Google Bard just made a stunning leap in capabilities, it just beat GPT-4 on a top leaderboard that evaluates AI models.
The leaderboard, from the Large Model Systems Organization, shows Bard (using Google's Gemini Pro model) now in 2nd place in terms of performance.
The leaderboard takes into account 200,000+ human votes on which models users prefer. It also assigns an "Elo" rating to each model, which is a method of calculating how good players are at zero-sum games like chess.
Bard still trails behind GPT-4 Turbo, but now surpasses other versions of GPT-4 and other popular models like Claude and Mistral.
At the same time, there have also been indications online that Google may soon release a paid advanced version of Bard. (However, this is unconfirmed at the moment.)
Introducing AI Literacy for All
In the last year, we here at Marketing AI Institute have presented to and talked with thousands of marketers and business leaders.
We have seen firsthand how executives are scrambling to adapt and devise AI roadmaps, while facing complex challenges, including a lack of AI-savvy talent, legacy tech stacks, a rapidly expanding AI tech landscape, fear of change from staff, industry regulations, privacy, and security concerns, and mounting competitive pressure.
What has become clear from all of this is that our mission must evolve to pursue a north star of "accelerating AI literacy for all."
We believe you can build a smarter version of any business through a responsible, human-centered approach to AI, but success requires a commitment to AI education and training across the organization.
We hope to play a part in moving human-centered AI forward across industries.
The first step in accelerating AI literacy is a significant expansion of our online education programs with the announcement today of two new AI course series—Piloting AI 2024 and Scaling AI 2024—plus a completely re-imagined AI Mastery Membership program.
Marketing AI Institute, and the marketing industry, are only the beginning.
Links Referenced in the Show
- Publicis to invest 300 mln euros in AI plan over next three years
- Bard with Gemini Pro surpasses GPT-4 on top LLM evaluation leaderboard
- Introducing AI Literacy for All
- Where are they now? The career trajectories of the researchers behind one of AI’s most important papers
- Paul Roetzer LinkedIn Post
- More DeepMind staff depart to launch startups
- Aidan Gomez + Cohere’s Coral enterprise chatbot
- Paul Roetzer LinkedIn Post
- Google announces AI video generator Lumiere.
- Chrome is getting 3 new generative AI features
- Fake Joe Biden robocall tells New Hampshire Democrats not to vote—and researchers say ElevenLabs technology was likely behind it.
- Fake Joe Biden robocall tells New Hampshire Democrats not to vote Tuesday - NBC News
- Researchers Say the Deepfake Biden Robocall Was Likely Made With Tools From AI Startup ElevenLabs - Wired
- The Biden Deepfake Robocall Is Only the Beginning - Wired
- AI voice generation company ElevenLabs becomes a unicorn with $80M Series B.
- White House calls for legislation to stop Taylor Swift AI fakes
- What AI can’t—and shouldn’t—do
- One easy way to apply AI throughout your day
Read the Transcription
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: We have to get companies prepared. We have to get governments prepared. We have to get nonprofits and associations and like school systems. Like all these people need to understand AI at a deeper level so they can figure out what it means to their domain, their industry, their business. And that's going to take a whole community of people moving in the same direction.
[00:00:21] Paul Roetzer: Welcome to the Marketing AI Show, the podcast that helps your business grow smarter by making artificial intelligence approachable and actionable. You'll hear from top authors, entrepreneurs, researchers, and executives as they share case studies, strategies, and technologies that have the power to transform your business and your career.
[00:00:41] Paul Roetzer: My name is Paul Roetzer. I'm the founder of Marketing AI Institute, and I'm your host.
[00:00:51] Paul Roetzer: Welcome to episode 81 of the Marketing AI Show. I am your host, Paul Roetzer, along with my co host, Mike Kaput. Good morning, Mike. [00:01:00] Morning, Paul. We are Both in Cleveland at the moment, I will not be for long. It is January 29th, 9am. I actually get to go tomorrow back to Ohio University, which is, as you know, where I graduated from.
[00:01:15] Paul Roetzer: To lead a workshop for a group of executives from Brazil, which should be really cool. And then I am on a flight to Arizona. Um. to spend some time, talking with lawyers about the impact of AI. So it should be a fun week. And actually I am in the midst of recording my piloting AI 2024 courses, which I know you have already done your part.
[00:01:39] Paul Roetzer: But we'll, we'll talk a little bit about what we're doing with AI education in a few minutes as one of the topics today. But, the big thing we're doing initially is this piloting AI series, which we launched last year. and had over 800 people go through that and get certifications in piling AI. So Mike and I are doing a full refresh and, [00:02:00] Mike did his part and I am, I am, getting my part done as well with those are going to be launching here in a couple of weeks.
[00:02:06] Paul Roetzer: So I am in the midst this week of also recording. I think I have nine courses I'm, I'm doing or something like that with the, with the welcome. So it's a, it's a busy week, but it's a fun week. It's cool to be able to do this kind of stuff. Mike and I think we're both excited for this episode. You know, we usually in our zoom sandbox, like Mike and I kind of share links throughout the week of things to talk about.
[00:02:29] Paul Roetzer: And in a normal week, it's, it's probably like maybe 20, 25 links this week. I think we might've hit 40. Like there was just so much happened last week and it was kind of hard to curate it down. But Mike does a great job of kind of getting us focused in on the key topics and then the rapid fire items. So we've, we have a ton to talk about today and some topics that I'm personally really excited to get into.
[00:02:51] Paul Roetzer: So with that being said, let's, first thanks BrandOps, our sponsor for today. Many marketers use ChatGPT to create. Marketing [00:03:00] content, but that's just the beginning. When we sat down with the BrandOps team, we were impressed by their complete views of brand marketing performance across channels.
[00:03:08] Paul Roetzer: Now, you can bring BrandOps data into ChatGPT to answer your toughest marketing questions. Use BrandOps data to drive unique AI content based on what works in your industry. Visit brandops.io/marketingAIshow to learn more and see BrandOps in action. And this episode is also brought to us by Marketing AI Institute's AI for Writers Summit, which is coming in hot.
[00:03:34] Paul Roetzer: It's March 6th. That's only like Five weeks away or something. And February is a short month. I want to think about this because this is like, once I get through this week of all these presentations and the courses, this is like my, my next thing is getting ready for writer's summit. So, registration looks phenomenal.
[00:03:52] Paul Roetzer: I think we're over a thousand already registered. We hit, 4, 000 registered in 2023 at the inaugural virtual summit. So this [00:04:00] is a half day virtual event. There is a free registration option. It is from noon to four Eastern time. if you are international and can't make that, there is an on demand option as well.
[00:04:11] Paul Roetzer: so again, yeah, it's just incredible. We've got talks on, sort of state of for writers. Mike's going to talk about writer's tools. I'm, going to do an interview, with an IP attorney about the impacts of copyright and intellectual property, related to generative AI. We're going to have an awesome panel on the adoption of AI writing platforms within enterprises.
[00:04:32] Paul Roetzer: There's going to be a demo day style thing we're going to go through. We haven't finalized how many, but we're going to try and do like five to seven demos of Gen AI tools that are valuable for writers and editors and content marketers. And then wrap it up with a big, ask me anything session with like four or five of the presenters from the day.
[00:04:49] Paul Roetzer: So it's going to be a ton packed in, but you can go check out AIwritersummit. com. If you want to go direct to it, you can also find it on the Marketing Institute site under events, if you're there. Okay. [00:05:00] so that is the, presented by part of the session and now Mike, let's get into the big topics for the day.
Publicis to invest 300 mln euros in AI plan over next three years
[00:05:07] Mike Kaput: All right, Paul. So first up, we just got some very big AI news in the agency world. the world's largest ad group is actually spending 326 million on AI in the next three years. Now, HopelessSys Group is building what they call the industry's first AI powered intelligent system, and it is called Core AI.
[00:05:33] Mike Kaput: This is a central AI system that's going to be accessible by all 100, 000 plus of this group's employees. Now, the company's CEO told Adweek, quote, We are bringing together all the data of the group, all the knowledge of the group. And the publication actually says that Core AI. is being built and trained in house by a whopping 45, 000 engineers and data [00:06:00] scientists.
[00:06:00] Mike Kaput: It's being trained on the firm's trillions of data points, billions of personal profiles and daily bid impressions, and millions of creative assets. So basically the vision that Publicis has is having each employee plug into this centralized intelligence anytime they are doing work for clients. So. For instance, you could instantly converse with the company's decades worth of proprietary data, knowledge, and expertise.
[00:06:31] Mike Kaput: In other words, it kind of sounds to me like they're building an ultra intelligent machine brain that sits at the center of the company and augments each human employee. So, Paul, you obviously have some deep experience in the agency world, having been an agency owner for much of your career so far. What did you make of this announcement?
[00:06:51] Paul Roetzer: Yeah, this one, it got a lot of views and shares within my network. I had multiple friends text me like as soon as [00:07:00] this video hit, to make sure I saw it. And I did finally sit down, Sunday night and watch this and took a lot of notes. so the first thing I want to touch on is, What they presented is a pretty remarkable vision.
[00:07:18] Paul Roetzer: It is not reality at the moment. So the things they're showing aren't necessarily the capabilities of the current system. It's, it's kind of like what they're building. That being said, the way that it was presented, there was one moment, Mike, where you, I don't know if you caught it, but Artur, the CEO of Pulis group, was talking about this kind of like intelligence engine they were building, this intelligence system within the company, the core AI you mentioned, and I thought, oh my god, like that is the closest vision I've heard yet.
[00:07:53] Paul Roetzer: To the thing that started our pursuit of all of this. So, quick origin story for anyone who [00:08:00] hasn't heard it yet. the way Mike and I got into AI goes back to 2011 when IBM Watson went on Jeopardy. And so what happened was I was, you know, I had my agency at the time, PR 2020 where Mike was working as a content strategist and specialist and, you know, helping run accounts for clients.
[00:08:20] Paul Roetzer: And so when Watson won, I was writing, my first book, The Marketing Agency Blueprint. And after I finished that manuscript, I started trying to figure out what was Watson, and how did it work, and how did it make its predictions, and how did it answer questions on Jeopardy, and things like that. And so I started kind of researching what was Watson and could we eventually apply it to marketing strategy for clients.
[00:08:44] Paul Roetzer: So at the agency, clients would come to us and they would want to grow their audience or generate more leads or increase the quality of leads or increase conversion rates or drive traffic to the site or whatever their goal was. And then we would try and build strategies and allocate budgets to do it.
[00:08:59] Paul Roetzer: So a [00:09:00] client would come in and say, Hey, we have 100, 000, we have a million dollars, whatever it is. This is the budget. This is the goal we want to achieve. And so by 2011, I had become convinced that the human mind was actually incapable of doing that in an optimal way. That when I got out of college in 2000, there was like 10 ways to spend marketing dollars.
[00:09:21] Paul Roetzer: You know, you had direct mail and PR and advertising and communications and all these things. But it was pretty basic. formula to figure out how to, how to allocate that, those dollars. By 2011, when Mike and I were working with companies in healthcare and in SAS, and I don't know, my, what else, professional services, like there's all these industries we're working in and someone would come to us and want a strategy.
[00:09:42] Paul Roetzer: It's like, shoot, man, there's like 10, 000 ways to spend this money. And how are we supposed to know that? Even though I consider myself a pretty decent marketing strategist at the time. So my vision was, well, what if this Watson thing can do this? Like, what if, what [00:10:00] if AI is actually the way to solve strategy?
[00:10:03] Paul Roetzer: That the AI isn't limited in what it can do. So right after I finished the first book, I ended up reading this book called Automate This by Christopher Steiner in 2012. And that book told the story of intelligent algorithms being applied to industries like logistics and finance, talked about trading on Wall Street, and how they were applying, at that time it was machine learning, like making predictions basically, to disrupt industries.
[00:10:32] Paul Roetzer: And so I thought, well, that's going to come to marketing, right? Like, that's Advertising is absolutely like a prime way to do that. So, two years later, in 2014, I wrote my second book called The Marketing Performance Blueprint. Mike helped me research a section where we've, it was the first time where I wrote about AI.
[00:10:52] Paul Roetzer: And so, we did research to try and figure out what was going on right now. Was Watson or was any form of AI [00:11:00] actually being applied in businesses outside of what we were reading about in Automate This. So by that point in 2014, I sort of developed this idea that I was calling an intelligence engine. So in the 2014 book, you can go, go read it.
[00:11:14] Paul Roetzer: It is the only part about AI in that entire book. So 50, 000 word manuscript. And there is one section of one chapter about it. And I'm going to read it to you because I think it's extremely relevant to what Hublisys is now doing. So this again is an excerpt from Marketing Performance Blueprint in 2014.
[00:11:31] Paul Roetzer: The header is Origins of the Intelligence Engine. Marketing automation platforms save time, improve efficiency, increase productivity, and help manage big data. They give companies unprecedented abilities to understand buyers, identify opportunities, track campaign performance, and link marketing activities to business outcomes.
[00:11:52] Paul Roetzer: But they do not provide insight into the billions of bits of data being created as consumers move from screen to screen and interact [00:12:00] online and offline with brands. According to IBM, again this is back in 2014, 90 percent of all data in the world is less than 2 years old. Humans are not programmed to keep up, and yet turning data into intelligence and intelligence into strategy And strategy into action remains largely human powered.
[00:12:22] Paul Roetzer: What inevitably comes next are marketing intelligence engines that process data and recommend actions to improve performance based on probabilities of success. Think about it. Are we really that far off from an automated marketing strategy in which the marketer's primary role is to curate and enhance algorithm based recommendations rather than devise them?
[00:12:47] Paul Roetzer: Humans are limited by their biases, beliefs, education, experiences, knowledge, and brainpower. All of these things contribute to our finite ability to process information, build strategies, and [00:13:00] achieve performance potential. Algorithms, in contrast, have an almost infinite ability to process information. They possess the power to understand natural language queries, identify patterns and anomalies, and parse massive datasets to deliver recommendations better, faster, and cheaper than people can.
[00:13:19] Paul Roetzer: They already do it in healthcare, financial services, and customer service, and it will not be long before bots, multiple linked algorithms aimed at performing one task, descend on the marketing industry. As Steiner says in Automate This, The next field to be invaded by bots is the sum of two simple functions, the potential to disrupt plus the reward for disruption.
[00:13:40] Paul Roetzer: So I went back last night and re read that because that is like, it took 10 years, I said not long, it was 10 years, a decade, before I actually heard a company that is pursuing that vision, that seems to actually have all the components of putting [00:14:00] that together. They're combining knowledge and experience of their own company, the proprietary data of 35 years of digital transformation strategy.
[00:14:07] Paul Roetzer: They have Epsilon within their Publicis group, so they have all this data. They said what, I think Artur said, 2. 3 billion profiles on people around the world, and thousands of attributes. And they talk about data as the superpower for their people. So that was, again, I don't know if you thought about this Mike, but that was like, as soon as he said this, I was like, Oh my God, like, this is it.
[00:14:29] Paul Roetzer: Like 10 years later, here, here we are. We're finally building the intelligence engine, at least within advertising. It's still not a global vision for like all marketing dollars, but at least within advertising, it seemed like it.
[00:14:40] Mike Kaput: Yeah, that was actually the number one reason I wanted to focus on this as a main topic is it's literally why the Institute Came into being.
[00:14:49] Mike Kaput: This is the first thing I thought of and also notably at least from my perspective This is one of the first if not the first like Pretty clear [00:15:00] vision I've seen from a major agency. Plenty of people are doing plenty of things and I've seen really cool examples, but I don't know if I've seen one that has this like breadth of vision, in the agency world.
[00:15:13] Mike Kaput: Did, was that kind of your perspective or have you seen? More like this from other, you know, at least among the big handful of agency groups out there.
[00:15:22] Paul Roetzer: Yeah, honestly, like it's, it's definitely in the agency world is the first I've seen that clearly has kind of laid this vision out. Again, we're not saying it's all reality.
[00:15:33] Paul Roetzer: We're not saying they're going to even like nail it all, but we're saying like to lay out the vision saying we're going for it. It's the first I've seen. And, you know, we've talked about this idea that. In the future, all businesses, there'll be three, three type in every industry, AI native, AI emergent, and obsolete.
[00:15:49] Paul Roetzer: So the AI native is you just build a smarter company from the ground up, just infuse AI and you just build a smarter version, needs fewer human resources, fewer financial resources, because you're going to infuse AI in everything you do. [00:16:00] AI emergent is an existing company that has legacy tech, legacy data, legacy clients, existing employee base.
[00:16:06] Paul Roetzer: And they create a vision to evolve and become a smarter organization. This is it. Like this is the AI emergent company that we envisioned, you know, last year when I wrote the future of all businesses, AI are obsolete. so it's, it's one of the best. overall visions I have heard, for what an organization become.
[00:16:28] Paul Roetzer: And there was a few things that jumped out to me. and I'd love to hear Mike, if there was any that, you know, you really, saw as significant, but a couple of things they said is. AI will never replace great creative minds, but it will push the boundaries further. So I like the fact that they were addressing creativity, which is obviously what they do.
[00:16:47] Paul Roetzer: They said the approach to being data led, but human first. So I liked the idea that they were kind of looking at that. They laid out five key areas where they were going to apply this core AI to supercharge their people. [00:17:00] And throughout each of the five, so it was insight, media, creative, software, and operations, they, they bullet pointed out like the three elements of each of those.
[00:17:07] Paul Roetzer: And it was all related to faster, more accurate, more innovative. So in each of those areas, they were basically saying with insights, with media, with creative, how can we be faster? How can we be more accurate? How can we be more innovative? Do things that haven't been done. they talked about a hundred trained micro agents that are like performing tasks within their system, which I thought was interesting.
[00:17:28] Paul Roetzer: The 330 million euros over the next three years, it was fascinating to me. They said they're financing it through efficiency gains. So they're this again, this AI emergent company, like what should it be? They're not taking the gains from AI and cutting staff. At least that's what they're saying, they're not doing.
[00:17:45] Paul Roetzer: They're taking the financial gains and reinvesting it in two things, their people and their technology. So I love this idea of reskilling, upskilling people with the money being generated from the efficiency [00:18:00] gains of AI. they said it'll allow our people to do things tomorrow that they can't do today.
[00:18:05] Paul Roetzer: And so my overall take is this is the vision for an AI emerging company. This is what a CEO should be doing right now. So if you're, if you are a CEO listening, or if you work for a company, this is the kind of thing you should be expecting in 2024 from the CEO of a company to lay out a clear vision.
[00:18:25] Paul Roetzer: with their leaders for what they can become. They acknowledge that it won't be easy. They acknowledge that it'll be experimentation iteration, but it's a worthy pursuit. And so my feeling is like, it's not going to be a straight line. There's going to be bumps. there's going to be lots of ways it can go wrong that the tech doesn't deliver on what they're showing in the video, but resiliency and leadership and a clear vision and like a wheel to see that vision through.
[00:18:51] Paul Roetzer: can get them further along than their peers. And that's what it's all about. Like, how can you out innovate your peers right now? And this is a really [00:19:00] big step, it seems, for them to do that.
[00:19:03] Mike Kaput: Yeah. The, ect about how this is going to be used to augment Their people is really, I think, subtle but important here in the, at least in the agency world.
[00:19:17] Mike Kaput: I mean, I remember countless times where we did a very good job of trying to arm our people with processes, and resources to serve clients on their own. But the moment there's not a process, there's not a resource. You're kind of trying to figure it out on your own, how to best serve a client. And that comes down to the experience, the skills and the seniority, perhaps of the person doing the work.
[00:19:45] Mike Kaput: So you get varying outcomes. And even the best person within an agency has no ability to understand what happened. Two decades ago that might have been super relevant to the current challenge. So I think having almost [00:20:00] Personifying the agency as this centralized brain in the heart of the company is really really interesting to me And I think it's a very broad blueprint perhaps for how any company should be thinking of like what makes your company your company That should be What you are building these centralized systems to embody.
[00:20:23] Paul Roetzer: Yeah, and the other thing that jumps out to me is strategy for me as the CEO of an agency, you know, for 16 years I, you know, ran a professional services firm, strategy is absolutely, in my opinion, the hardest thing to teach. So, When I would interview people straight out of college, I was trying to assess their strategic ceiling.
[00:20:48] Paul Roetzer: Like, were they a strategic thinker? Could they connect the dots? Did they seek information, like cross discipline information, so that they could actually understand something happened in the economy? Here's what it means to my client in the manufacturing industry. [00:21:00] Strategic thinking is a really, really hard thing, and it's a really valuable thing.
[00:21:05] Paul Roetzer: It's also an ephemeral thing. What I mean by that is, Mike, let's say like you were running four manufacturing clients at the agency, and you left, and you were the lead strategist. The knowledge of how to build those campaigns and run those accounts is gone. Like, the strategist, like, So that, to me, is one of the potential values of what they're doing here and what the value of AI is, is being able to codify strategic thinking and institutional knowledge.
[00:21:34] Paul Roetzer: So that if someone leaves a company or moves to another area, or if I'm working late at night and I need that knowledge, that I can have a conversation with an internal chat bot about it versus like, Oh, Mike's not here. Mike's on PTO this week, like whatever it is. So strategy often is centralized within the brains of the few talented people who can do it within corporations.
[00:21:57] Paul Roetzer: And so that's a really. hard [00:22:00] thing to scale. And so I think it's a really fascinating pursuit. And that's why I'd like to me, the origin of the intelligence engine was what led me down the AI path, whatever that was 12, 13 years ago now.
Bard with Gemini Pro surpasses GPT-4 on top LLM evaluation leaderboard
[00:22:14] Mike Kaput: So in our second big topic today, we just saw Google Bard make a stunning leap in capabilities. it actually just beat out a version of GPT-4 on a top leaderboard that evaluates AI models. Now this leaderboard comes from an organization called the Large Model Systems Organization, and it shows that Bard.
[00:22:36] Mike Kaput: Using Google's Gemini Pro model is now in second place in terms of performance when compared to a range of popular AI models. Now this leaderboard in particular takes into account, it says 200, 000 plus human votes on which models users prefer. And it also assigns what's called an ELO rating to each model, which is a [00:23:00] widely used method of calculating how good players are at zero sum games like chess.
[00:23:05] Mike Kaput: It's pretty popular in rating chess players. Now Bard still trails behind GPT 4 Turbo, but it now surpasses other versions of GPT 4 and other popular models like Claude and Mistral. Now, at the same time, we've also seen some rumblings online that Google may be soon releasing an advanced version of BARD that you have to pay for.
[00:23:29] Mike Kaput: Now, this is unconfirmed at the moment, but it sounds like Google is hard at work kind of recapturing some of the lead when it comes to the AI model arms race. Now, Paul, first up, can you talk to us a bit about the legitimacy of the Large Model Systems Organizations Leaderboard? It is one of the leading benchmarks out there for AI models.
[00:23:53] Mike Kaput: Like, why is that?
[00:23:55] Paul Roetzer: It's a, it's actually a site I wasn't familiar with until recently, [00:24:00] honestly. So the way this works is, to see how powerful these different models are. Usually we wait around for a research paper that comes out and does evaluations in comparison to some human standard of whatever test it is that they're evaluating against.
[00:24:14] Paul Roetzer: Those can be biased. It can be affected by how the prompting was done. Like there's lots of different reasons why those are hard and often they're outdated. So you'll see a big research that comes out and says. These models can or can't do something. And then you read the fine print. It's like, oh, they did this research in July of 2023 when they weren't even using the most powerful model that's out today kind of thing.
[00:24:35] Paul Roetzer: So this leaderboard it's called the chatbot arena, is open source research project developed by you mentioned L LMSYS, but also UC Berkeley sky lab. So they say their mission is to build an open source platform. Um. to collect human feedback and evaluate LLMs under real world scenarios. And I actually went to the site.
[00:24:55] Paul Roetzer: It's, it's fun. So we'll put the link in the show notes, but it's [00:25:00] arena. lmsys. org. And so they have arena battles and what it does, like I went in and played around with it last night. You can put, it'll blind which two models you're going to use, but you give it a prompt and then it gives you two outputs and then you rank.
[00:25:15] Paul Roetzer: Which one is better, whether it's a tie, whether they're both bad or which one's better. And that's how this is done. It's actually human rating of outputs. And so you can do an arena battle where you just like, you don't know which model is producing, which thing you just say, which one you like. And then it reveals for you, which model was which, or you can do a side by side.
[00:25:35] Paul Roetzer: So if you want to go in and see Gemini Pro versus GPT 4 Turbo, you can give it the same prompt and it gives you the output. And then there's also a leader leaderboard tab. So again, really cool site. The reason it surfaced for me was two tweets last week. So this new, The leaderboard came out January 26, so there was a lot of buzz in my kind of alerts feed, all the AI influencers [00:26:00] that I follow and trust.
[00:26:01] Paul Roetzer: So one was Oriol Vinales, I think. He works at Google DeepMind. And so he had tweeted that evaluation of LLMs is very hard and nuanced, especially academic evals, which are leaked massively. Evils that rely on human judgment are far superior, so it feels good that BarGeminiPro, the free tier, climbed pretty high, on this leaderboard.
[00:26:24] Paul Roetzer: And then he actually teased, looking forward to Gemini Ultra release. Ultra is coming soon, hashtag. That was the first thing when I saw this, I was like, oh, this is fascinating. Like, not only did it jump to number two, this isn't the most powerful version of Gemini. Like, we, we know Gemini Ultra is pending release.
[00:26:42] Paul Roetzer: And so you gotta think that. That's probably going to maybe take that leap and get ahead of GPT 4, which was the big question mark. the other one that caught my attention was our, our, our guy, Andrej Karpathy, who we talk about a lot, OpenAI researcher, former head of AI at Tesla, [00:27:00] and he said, in a reply to someone else's tweet, he said, I pretty much only trust two LLM evals right now.
[00:27:08] Paul Roetzer: Chatbot Arena. And, there was a Reddit one. Local Llama comments section. So, if Andre says Chatbot Arena is the place to look at, then I trust that this is a extremely legitimate LLM ranking system. So, worth paying attention to. My main takeaway is, I gotta try BARD again. Because, the, I went to look and say, well, which version is this?
[00:27:32] Paul Roetzer: And so, on the BARD. google. com site, there's a slash updates page. Again, we'll put in the show notes. And so, this version of BARD, Gemini Pro, went into BARD on, looks like December 6th. So, they said BARD is getting its biggest upgrade yet with Gemini Pro. Starting today, we're introducing Gemini Pro and BARD for BARD's biggest update upgrade yet.
[00:27:54] Paul Roetzer: We specifically tuned Gemini Pro to be far more capable of things like understanding, [00:28:00] summarizing, reasoning, coding, and planning. so, yeah, I mean, I think if anything, it just, we've always said, like, I've been hard on Bard personally, it's, it's not been on par with ChatGPT, but I've also said I wouldn't bet against Google ever in this arena.
[00:28:18] Paul Roetzer: And I'm very anxious to not only try it again now, but to see what happens when Gemini Ultra becomes a thing.
[00:28:28] Mike Kaput: So, what kind of steps should business leaders and professionals be taking in light of this news? Like, should people be switching to BARD now?
[00:28:38] Paul Roetzer: Yeah, I mean, the way you and I approach this, and I think the way we always teach this, is you have to constantly be testing.
[00:28:44] Paul Roetzer: And this is why it is so hard to make bets on which platform to use and which ones to integrate into your workflows because they keep evolving as to which is best for which use cases. so I think you have to have a culture of always testing. And [00:29:00] again, not everyone in your company needs to be testing.
[00:29:02] Paul Roetzer: Like at some point you have to say, okay, you're all using this system for these use cases. Here's the templates. Here's the sample prompts, like go do your thing, but you need to have a component of your team. Maybe it's within your AI council who are regularly testing Claude, Bard, ChatGPT, whatever the tools are against the common use cases within your organization.
[00:29:24] Paul Roetzer: So if you're using it for script writing and blog posts and summarization and like whatever your common use cases are, someone on your team needs to every 30 days or so or 90 days or whenever the leaderboard changes, go in and run those use case tests against the different systems. and see if someone has made a leap forward that changes the kind of technology the rest of your team should be using.
Introducing AI Literacy for All
[00:29:49] Mike Kaput: All right. So in our third big topic today, we have some big news at Marketing AI Institute and some news that has some kind of wider implications for the marketing industry and business [00:30:00] as a whole. So in the last year here, we at Marketing AI Institute have presented to and talked with thousands of marketers and business leaders.
[00:30:10] Mike Kaput: And Paul, I know you and I have seen this firsthand, like how executives are scrambling to adapt and devise AI roadmaps, and they face all these complex challenges to adopting AI. Things like lack of talent, legacy tech stacks, crazy expanding AI tech landscape, fear of change, industry regulations, privacy, security concerns, mounting competitive pressure, and more.
[00:30:36] Mike Kaput: Now what's become clear from all of this is that Our mission here at Marketing AI Institute needs to evolve to pursue what we're calling this North Star of accelerating AI literacy for all. I believe that's how you put it in a recent post, Paul. And basically what this means is that We at Marketing AI Institute believe that you can build a [00:31:00] smarter version of any business through a responsible human centered approach to AI, but that success requires commitment to AI education and training across the organization.
[00:31:11] Mike Kaput: Now, we at Marketing AI Institute hope to play a part in this in moving human centered AI forward across industries. And the first step, which we announced last week, is accelerating AI literacy through a significant expansion of our online education programs. this includes two new AI course series, Piloting AI 2024 and Scaling AI 2024, plus a completely reimagined AI Mastery membership program.
[00:31:43] Mike Kaput: Now, Paul, before we get to kind of the why here, let's talk about the what. Can you Talk to us a little bit more about these new courses and the membership that are about to be made available to our audience.
[00:31:57] Paul Roetzer: I mean, it's sort of a [00:32:00] continuation of what we set out to do. So, you know, I told the story up front about how we got started in AI, like out of a curiosity and a, like a specific problem we were looking to solve for, for clients and for the agency.
[00:32:14] Paul Roetzer: but by 2016, when I, when I started the Marketing Institute, like, The way I looked at it was, we weren't uniquely capable of building AI, we weren't uniquely capable of necessarily building services around it because we didn't have highly technical people and engineers and data scientists within our agency, but we were storytellers by trade, and that's what we did for clients, it's what we were trained to do, it's what we went to school for, and so, The institute initiated as, let's tell the story of AI, and let's see if other people are intrigued by it, and if they are, we'll, we'll build something around this, we'll figure out how to like, help people along.
[00:32:51] Paul Roetzer: So that whole idea of approachable and actionable was the original mission. How do we make it accessible to people? That then led to November of [00:33:00] 2021, so a year prior to ChatGPT, I introduced this thing called the Million Marketer Challenge, where we were like, okay, there's 13 million marketers worldwide, let's get a million of them aware of AI.
[00:33:13] Paul Roetzer: Like, let's try and, like, get people aware of it to take some first step. Take a class, read a book, whatever it was. Didn't have to be our stuff, just anything. So we introduced Intro to AI for Marketers and I started teaching this free 30 minute class every three weeks on Zoom. So that class became our like entry point to help people understand this.
[00:33:35] Paul Roetzer: To date, I think we've had over 17, 000 people register for that class. We've had done I think I've done 33 of those. And I've told this story before, like to me, it was kind of like building a band, like you're going to show up and there's going to be 10 people there. And then sometimes you'll show up and there's a hundred people there.
[00:33:53] Paul Roetzer: And then all of a sudden ChatGPT shows up and we have a thousand people there. And so that whole million marketer challenge [00:34:00] was sort of like, became unnecessary, as a pursuit, because ChatGPT took care of that for us. Because a million marketers tried ChatGPT in like the first 72 hours. So, for us, it's always been about how do we make this information accessible?
[00:34:15] Paul Roetzer: But what happened is, after ChatGPT, it wasn't just marketers calling. Everyone started calling. So I was talking to heads of VC firms, presidents of universities, provosts at universities. heads of school systems, like, you know, K through 12 school systems, government leaders, association leaders, like everyone, started calling asking, can you come talk to us about this?
[00:34:40] Paul Roetzer: And that's what led you and I, Mike, last year to spend a good portion of our year on the road meeting with all kinds of people, not just marketers, like everything, lawyers, accountants, CEOs. And so in that process, we started to, to really realize. how unprepared the [00:35:00] world is for what is happening. And not again, not just marketing leaders, but CEOs of major corporations.
[00:35:06] Paul Roetzer: I mean, we have done talks. I'm not going to name drop here, but like we have done talks to some of the biggest brands in the world, like in the, in the last 12 months, we have sat down with executives from some of these major brands. And again, truly firsthand heard. The questions they have, the challenges they are facing, the fears that they and their employees have around this stuff, the uncertainty about what comes next.
[00:35:33] Paul Roetzer: And so that led to, okay, we need to keep doing the intro thing, but we need to expand it. We need to like, it can't just be for marketers anymore. We have to really start thinking about how do we bring this to more people? and the piloting AI course that we launched last year, I mentioned at the opening, we had over 800 people took that course in 2023.
[00:35:55] Paul Roetzer: So it's like, okay, let's. Let's get that updated because by like fall of [00:36:00] 2023, we were getting the questions. Well, when were these courses recorded? Like, are they still relevant? So it's like, okay, we have to, every year let's update piloting. Like we know that's a really helpful thing to get people started.
[00:36:11] Paul Roetzer: So you take the intro course. Then what do I do? Okay. Here's the piloting series. It's 18 on demand courses in eight hours. You can go get a professional certificate. and you can be advanced. We have also always known there needed to be something beyond piloting, which was scaling AI. And so I announced plans for that I think at MAICON 2022.
[00:36:32] Paul Roetzer: I think I said we're going to build this. It just took a year. And so we're now in the midst of building the scaling AI course, which is for, kind of director level and above, or people within AI councils. It's designed for the people actually going to do this stuff. So it gets into how to build an AI council, building your AI roadmap, generative AI policies and responsible AI principles, state of AI.
[00:36:55] Paul Roetzer: It's, it's really more for the leadership level. And then the final piece of what [00:37:00] we announced, last week is this idea of AI mastery as a membership. Because what we'll often hear is, Hey, we took piloting. It was awesome. How do we stay on top of this? How do we actually like keep applying this knowledge?
[00:37:13] Paul Roetzer: And so I realized like we had to reinvent what our membership model was to be much more about competency and mastery of these topics. And that's where the idea came like, okay, let's start doing like. quarterly demo days where we're showing the technology being applied to specific use cases. Let's do ask me anything sessions every quarter where people have a chance to actually get in and learn this.
[00:37:36] Paul Roetzer: Let's do a state of each quarter where like, what are the trends that matter this quarter? and so that's really what the education program became. And then in the process, as you were saying, this AI literacy for all, it's like, I mean, I just, I journal stuff. So I don't know if other people work like this.
[00:37:54] Paul Roetzer: This is how my mind works. Like I'll, I'll go give a talk, you know, to a, like there was [00:38:00] one in particular, it was a, a, a major, technology company, like a fortune 100 technology company, and I'm like on the flight back thinking. that was like wild, like to, to hear a company that has doing any AI for 20 years, like at least 20 years, but not within the operations and the marketing and the service and the, and the sales of the company.
[00:38:23] Paul Roetzer: They were doing it in the technical side and the product side. And to realize even a company like that. doesn't have the knowledge they need. And then like the next day you go meet with the president of a university. It's like, they're, they don't know how to integrate this into classrooms and how to find professors to even teach this stuff.
[00:38:40] Paul Roetzer: And then you go talk to like a nonprofit, like a board I'm on, and they're like, how do we use AI? And so. What I realized is like, we have this kind of broad knowledge. The listener base and the, you know, the audience for the podcast is not all marketers, not even close. Like we hear from people all the time on LinkedIn, reach out to you and me, Mike, who are not marketers.
[00:38:59] Paul Roetzer: And so [00:39:00] it really started to evolve as I was journaling my thoughts after a different meetings or random times, this AI literacy for all just really kind of kept coming back to me of like, we need to do more. Like there's such an urgency right now to help people understand this and to help prepare other professionals to go out and educate.
[00:39:19] Paul Roetzer: Like it can't be all us. Like there's nothing about our model that has ever been, Hey, we're going to like keep this all in house. And no, we're not going to share what we know. and, you know, we want to make it proprietary so we can make as much money as we want. I don't, I don't care about that stuff.
[00:39:32] Paul Roetzer: Like we're building a really good business. It's a very strong business now. It wasn't a year ago, but like, we're in a very good place from a business perspective. And my feeling is. We, we don't focus on our revenue, our growth. We focus on telling the story to have the greatest impact and all that other stuff takes care of itself.
[00:39:52] Paul Roetzer: Now, don't get that mixed up like I don't like run the business, you know, to be a profitable business. We do. But it is [00:40:00] not why we do what we do. and so I think that, you know, all the time we're spending on the road, all the time we're investing in free education and training through like our intro course and like speaking series.
[00:40:12] Paul Roetzer: It's just like where we're going. And the thing I said in that post was Marketing Institute and the marketing industry are just the beginning. Now I'm not going to announce like other stuff that's coming right now. It's not the time for it, but there are other things that we're going to be announcing probably in the next, like, I don't know, like two months or so that'll make it.
[00:40:33] Paul Roetzer: a little bit more clear and tangible what, what AI literacy for all really is going to mean. But I just think it's really important that we think about that as a North Star. But if you're like a listener to this show, maybe you think about what role you can play in that too, because We have to get companies prepared.
[00:40:55] Paul Roetzer: We have to get governments prepared. We have to get nonprofits and associations and like [00:41:00] school systems. Like all these people need to understand AI at a deeper level so they can figure out what it means to their domain, their industry, their business. And that's going to take a whole like community of people moving in the same direction.
[00:41:16] Mike Kaput: That is a great rundown. And I would just, the final note I would add to that is that if you do want to explore some of these options for yourself and your team, they're all right on the website. If you go to marketingainstitute. com, click on education, you can see intro to AI, piloting AI, scaling AI, and the mastery membership all right there.
[00:41:35] Mike Kaput: So you can learn a little more about everything that Paul just ran through.
[00:41:39] Paul Roetzer: And the other thing I'll say is like, we have a marketing AI conference MAICON coming up September 10th to the 12th. And I would really encourage you to like, be there, like if you want to be a part of this, like this kind of broader vision of like, okay, let's spread AI literacy as fast and far as we can.
[00:41:58] Paul Roetzer: That's probably like the [00:42:00] next point where we're going to get. Those kind of people together and that's part of the value for me and like why I love running the event is to see the community sort of like emerge and spread with us just facilitating the meeting of people. And I've heard so many stories about like businesses being started from people who've met at MAICON, the community you see like Slack community now is over 5, 000 people, just like the way people are interacting and starting to kind of take it and run and make it their own.
[00:42:31] Paul Roetzer: And so I would say like, you know, if you can be at MAICON, do that. And then as we announced some of these other things, like hopefully there, you know, you'll find a spot that you can get involved and start, you know, really being a part of expanding AI literacy within your network and your community.
The career trajectories of the researchers behind one of AI’s most important papers
[00:42:50] Mike Kaput: All right, let's dive into this week's rapid fire topics. So first up, we have, an interesting breakdown of the career [00:43:00] trajectory of some really important people in AI. So, back in 2017, a team of eight researchers on the Google Brain team, which is now part of Google DeepMind, published a research paper titled attention is all you need.
[00:43:13] Mike Kaput: Now this paper invented the transformer architecture that became the basis for ChatGPT from OpenAI. GPT stands for generative pre trained transformer and basically led to the acceleration of AI technology that we're experiencing today. Now Paul, you had mentioned at the top of this episode You're working on some new AI courses, and we just went through some of what those are.
[00:43:40] Mike Kaput: you recently revisited the paper as part of your work there, and you tracked where the eight authors are now. And what you found is pretty incredible. All of them have left Google since the paper. Seven of them have founded AI companies, and those AI companies have raised a combined 1. 3 billion [00:44:00] to date, and they're sure to raise more in the future.
[00:44:03] Mike Kaput: Can you walk us through this a little more in depth? Like why is this small group of people so important and how are they shaping the future of the AI we're going to see tomorrow?
[00:44:13] Paul Roetzer: I just think they're Fascinating to track. Like, so what I was doing is the second course in the piloting AI series is a new generative AI 101 course.
[00:44:23] Paul Roetzer: And so as I was building that, well, I guess it was last week or two weeks ago. I was looking back at that attention is all you need, all you need paper. And we've talked about that paper on the podcast before, if you've been a listener for a while, you've probably heard about that paper in 2017. And we've also talked about.
[00:44:39] Paul Roetzer: a number of those people who were authors on that paper who have since left. So like, Noam Shazier, who's the founder and CEO of character. ai. They're series A, 150 million. We've talked about Aiden Gomez, co founder and CEO of Cohere. series C, 445 million and rumored to be raising another, you know, half a billion to a billion.
[00:44:59] Paul Roetzer: So we've [00:45:00] mentioned a few of these people, but I had never really dug into who were all of these people. So yeah, I just kind of stopped out of curiosity and There are, there's a co founder and CEO of Essential AI that's raised 64 million. another co founder of Essential AI. then we had Noam I mentioned, Jakob, who's, I don't know how to say Jakob's, Uzcuret, I think maybe, Inceptiv, they've raised 120 million.
[00:45:25] Paul Roetzer: Leon Jones, Sakana AI's 30 million. Lukas Kaiser is at OpenAI, he's a researcher now. And then Ilya Poluskin. has near protocol, which is like a web three AI company, 533 million. So I don't know. I think I've talked about it on the show before, like one of the ways we stay in the loop on what is happening and what's coming next is by reading the research papers.
[00:45:48] Paul Roetzer: The irony of this paper in 2017, as I think we've talked about before, is it wasn't even perceived as a big deal within Google. Like it wasn't until like a year later that Google actually started [00:46:00] building around this stuff and realized like, Oh, this, we might be onto something with this, but they opened up that paper and gave it out.
[00:46:07] Paul Roetzer: And that's what led to ChatGPT being built on it. So just the fact that it was so influential, it was so under, I don't know, appreciated, but like, They just didn't get it like the significance at the time and that's what often happens with these papers and is like in the rearview mirror It's like, wow, that was a big deal.
[00:46:26] Paul Roetzer: Like, how did we miss that being a huge deal? And so it's fascinating to me just to kind of keep up with this. And then what I'll do is when I see a significant paper, I will go follow all of the authors on Twitter and add them to a list. And then I keep track of who those people are. And then how we kind of like follow AI industry is like, well, who are they connected to?
[00:46:46] Paul Roetzer: Who do they talk to? Who do they retweet? And that's how I build like my network of influence of who, when something happens, like the chatbot arena, are they talking about it? And if they're all talking about it, then it's obviously significant. So that's kind of our path [00:47:00] of. Discovery, in a way.
Google announces AI video generator Lumiere.
[00:47:04] Mike Kaput: So next up, Google just announced a new AI video generator.
[00:47:10] Mike Kaput: It's called Lumiere, and it uses AI to generate five second videos. Now, five seconds does not sound like a lot, but this is a bit of a big deal. The publication Ars Technica laid this out pretty well. They said that Google's tech is designed to handle both the space, where things are in the video, and time, how things move and change throughout the video.
[00:47:32] Mike Kaput: aspects simultaneously. So instead of making a video by putting together many small parts or frames, Lumiere can create the entire video from start to finish in one smooth process. So basically, if you've used any type of AI video generator, they're pretty cool, but you do see a lot of this kind of weird stuttering and shifting in AI generated videos, and that's because they're made in this way of putting together, you know, a bunch of smaller frames.
[00:47:58] Mike Kaput: But Lumiere [00:48:00] doesn't seem to do that nearly as much thanks to its unique approach to video generation. we only have demo videos to kind of go off at the moment, but they do look pretty smooth, seamless, and convincing. Now Google has not revealed when or even if this tool will be made available to the general public, but it does seem to represent a A step forward when it comes to AI for video.
[00:48:24] Mike Kaput: So Paul, I kind of wanted to ask you a big picture question of kind of what does Lumiere mean for this next stage of AI for video that we're going to see in the near future?
[00:48:34] Paul Roetzer: And this is one where I'd turn to our favorite new tool, Perplexity, because I read like the research paper and like the information on it as I was, it was really technical.
[00:48:44] Paul Roetzer: So I actually asked Perplexity, what makes Google's linear video diffusion model so significant? How is it different from a runway PICA 11 labs are doing? And so just the quick was a significant advancement in the field of video synthesis technology due to its ability to create realistic, [00:49:00] diverse, and coherent motion in videos.
[00:49:02] Paul Roetzer: It's achieved through the novel space time U net architecture that you mentioned, which is a departure from traditional. It synthesizes videos in a single pass, unlike most existing models that use cascading approach resulting in better temporal contemp Temporal, Consistency, and Motion Quality. Now, temporal consistency is a term you hear a lot.
[00:49:21] Paul Roetzer: Runway talks about that all the time. so, best explanation for that is, refers to the model's ability to maintain a consistent and smooth transition of frames over time. contributing to realism and the quality of the synthesized video. So if you take one image and like frame to frame and it starts like distort a little bit, the temporal consistency is like photo quality.
[00:49:42] Paul Roetzer: Like it's the same, you know, as you move from frame to frame. So it seems like whatever they've done here is an advancement in that way. My main takeaway at a really comprehensible level is 2024 is the year of AI video. So like last year was the year of. Text, for [00:50:00] sure. Like major leaps forward in that with GPT 4, the language model.
[00:50:03] Paul Roetzer: We started seeing some major innovations with Runway, back in March and April when they started announcing things with Gen 1 and then Gen 2. We saw Pika emerge recently, 11 Labs, HeyGen, Google, Meta, they're all pouring tons of money in and it seems like we've now had some of the breakthroughs needed for video to become a massive thing.
[00:50:25] Paul Roetzer: So I think, you know, if last year was kind of images and text. This year's is more in the video and audio. We're just going to see a lot of innovation and a lot of like practical applications that the average person can go use, much like you can now go use an image generation tool, at any point. And so I, the other thing I think is you have to look at all these individual advancements from Google and assume at some point they all come together under Gemini.
[00:50:52] Paul Roetzer: So if you want to imagine like what can BARD be, imagine being able to create any kind of quality image you want. Obviously [00:51:00] the text piece, audio, video, all of it. It can understand it can create it, code. I think that's where Google's going is whatever we see here, assume it all eventually is under one multimodal model and that we will have the ability to do all of these things in one centralized chat interface.
Chrome is getting a new experimental AI writing feature.
[00:51:21] Mike Kaput: So also in some Google news, they are now baking what they're calling, quote, experimental writing feature into Chrome. So the company says this feature will help you write with more confidence on the web. And to use it, you'll right click on a text box or a field on a site when you're using the Chrome web browser and click Help Me Write.
[00:51:41] Mike Kaput: Now this small, sounds like a small change, but it could have a pretty big impact given that a whopping 63 percent of all internet users use Chrome. Now, Paul, you posted on LinkedIn that the proliferation of AI writing tools is hard to comprehend and [00:52:00] creating an increasingly confusing tech landscape.
[00:52:03] Mike Kaput: Can you tell us what you mean by that? And how does Chrome fit into that landscape?
[00:52:08] Paul Roetzer: Yeah, I just highlighted, like, because I was thinking about it, like I'm reading this Chrome, and I was like, oh, that sounds kind of cool. And their use cases, they say, writing on the web can be daunting, especially if you want to articulate your thoughts in public spaces.
[00:52:19] Paul Roetzer: so it'll help you write more confidence on the web, including well written reviews, writing a friendly, crafting a friendly RSVP to a party, or making a formal inquiry about apartment or rental. Well, if I have a Chrome extension, like Grammarly, HyperWrite, whatever, I already can do that. Like, do I, do I need?
[00:52:37] Paul Roetzer: Chrome's actual Chrome built in tool to do it if I already have an extension that can do it, which then led me to like, well, this is obviously just a pilot for like broader expansion of writing capabilities within Chrome. So what if I assume what's going to happen is Google Barge is going to be built into Chrome.
[00:52:54] Paul Roetzer: And now if I'm in Google Docs or if I'm in LinkedIn or if I'm in HubSpot, [00:53:00] I'm just going to be able to use Google Barge right within Chrome, 63 percent of the world, like two and a half billion people on the internet. We'll just have BARD built into Chrome. And so now all of a sudden you have this massive distribution of maybe the most powerful model in the world, if Gemini Altra becomes that, where I can already write everywhere.
[00:53:19] Paul Roetzer: So then, what do I, what do I need? So if I went through my own list, I have Google BARD already for free, we have Google Workspace Duet AI internally that we're testing. I have Anthropic Claude, which is great. I have InflectionPy, I have ChatGPT personally, and we have the team license for ChatGPT I have Writer, I have Jasper, Perplexity, Writer, which I probably won't need if Chrome can do it for me.
[00:53:42] Paul Roetzer: HubSpot AI Writer, which I won't need if Chrome can do it for me. So, this isn't even a complete list. I have all of these tools that I can use to write content. It is a really complicated environment for you and I who live this stuff every day. Like, [00:54:00] I don't, I don't even know which one I should use anymore.
[00:54:02] Paul Roetzer: Right. And so if Chrome, if Google. does what we were just talking about and builds this truly powerful multimodal model and they inject it directly into Chrome. So it's super easy to use. What do I need anything else for? So now it just becomes this like, how do, how do I ever decide what to scale my company on when it's like constantly moving piece?
[00:54:24] Paul Roetzer: So I was more just saying, like, I empathize with. The people that are having to make these decisions right now. It's why we're going to have a session on this at the AI for Writers Summit is like, what do you do? Like, if you're the CMO and your job is to figure out what AI platform to train your team on, and there are literally dozens.
[00:54:42] Paul Roetzer: And it seems like next two months, Google may be the best option again until OpenAI does GPT 5 and then maybe it's them again, like. How do you decide this stuff? And I don't actually have a great answer. Like, I was just more like posing it as like, Hey, I'm with you. Like if you're feeling overwhelmed by this, like.
[00:54:59] Paul Roetzer: So are [00:55:00] Mike and I, and we do it for a living.
Fake Joe Biden robocall tells New Hampshire Democrats not to vote
[00:55:04] Mike Kaput: All right, so buckle up because in the U. S. we are starting to see AI deepfakes that are designed to influence the upcoming presidential election. We just saw news that a robocall that made the rounds in New Hampshire is now being investigated because it appears to have impersonated President Joe Biden using an AI generated deepfake of his voice.
[00:55:30] Mike Kaput: This fake phone message encouraged voters not to cast ballots in the New Hampshire primary and instead save their vote for the general election in November. What's more, Wired has published a story that quotes two experts who claim that the DeepFake audio was made with technology from Eleven Labs, which is a major leading AI voice generation startup.
[00:55:54] Mike Kaput: Eleven Labs actually just became a unicorn as well with its recent Series B [00:56:00] fundraise of 80 million. Now, just to clarify, Eleven Labs technology was used by some party to create this. The company itself was not the one creating this DeepFake. Now, Paul, this is not the first and certainly won't be the last deepfake we see trying to influence the election.
[00:56:18] Mike Kaput: Can you kind of talk to us a bit about your perspective on 11, how 11 labs fits in here? And maybe give us a sense of just how easy or hard it is to create these types of deepfakes.
[00:56:29] Paul Roetzer: I mean, 11 labs is just one of the probably dozens or hundreds of tools that could create this like it's, it's, you know, it's going to come from everywhere.
[00:56:40] Paul Roetzer: Like we've talked about this before. I mean, someone could run a podcast just on deep fakes and just talk about the 10 biggest deep fakes each week. Like it's going to be that prevalent where it's, you know, my concern is it's going to be so commonplace. That people are just going to tune it out and not realize what a massive deal it is to [00:57:00] society that we just don't know what's real anymore.
[00:57:03] Paul Roetzer: And this isn't just audio, it's going to happen with video, you know, it's certainly going to happen with images, it's going to happen with text. you know, it's going to happen where politicians are going to claim that things they actually said were just deepfakes, like that's inevitable, deepfake news, we'll probably call it.
[00:57:21] Paul Roetzer: it's just going to be a train wreck. Like, again, I've said this before, like, I wish I had a, an optimistic take on this and that there's a solution around the corner. There isn't. Regulation isn't going to fix this. The problem right now, as we'll talk about in the next topic, is like, The social media platforms.
[00:57:41] Paul Roetzer: You know, they're the ones that are supposed to kind of gatekeep this a little bit, but in reality, this kind of stuff generates tens of millions of views or listens or whatever. And so monetarily, there's not massive like incentive for them to actually solve this problem. I [00:58:00] just, I have nothing.
[00:58:02] Paul Roetzer: hopeful to say. I really, really wish I did. It's just going to be, it's going to be really bad.
White House calls for legislation to stop Taylor Swift AI fakes
[00:58:09] Mike Kaput: Well, unfortunately, our next topic doesn't have a lot of hope either here because, the White House. just released a statement calling for legislation to protect people from fake sexual images generated by AI.
[00:58:23] Mike Kaput: And unfortunately, this statement came in response to very highly publicized fake sexualized photos of the singer Taylor Swift that have proliferated on social media recently. These images appear to have spread primarily on X, and one of them got up to 45 million views, according to The Verge. X was slow to take down the images and they still appear to exist in other places on the platform, though I think they've taken action against some of the main accounts.
[00:58:52] Mike Kaput: Now, Paul, obviously this is like an uncomfortable topic and a nightmare for anyone it happens to whether or not they're famous, [00:59:00] but the high profile nature of this incident is drawing attention, like If one of the most famous people in the world can have this happen to them, and there's nothing to really do, like, what, if anything, can or should be done about this?
[00:59:14] Paul Roetzer: Yeah, I mean, a few thoughts, I guess. Like, one, this goes to the point of what I was saying, like, X, you know, 45 million views. Like, the way X's business model works is it pays people to generate Views and clicks and engagements. So there are people spreading these images who are getting paid by X to spread these images.
[00:59:37] Paul Roetzer: You get a million views. I don't know how much money you make. I don't know how, like how exactly the model works. Creators are paid. to spread conspiracies, deepfakes, synthetic content on X. It is the business model. I don't know how you fix that. The way they solved it, as of yesterday I saw, was they, they basically shut down the ability to search for the content.
[00:59:59] Paul Roetzer: So if you go [01:00:00] try and search for it right now, It's probably not going to turn any results because they basically turned off that capability for one person. Right. Because it's Taylor Swift, they can do it. But if it is the average person or, you know, a B list or C list celebrity, they're not shutting off search at X for it.
[01:00:17] Paul Roetzer: They don't care. Like, it's just not, they're not solving for it. Um. So that's one. Two, the only repercussion we have right now is existing laws. So obviously, this is illegal if they could ever figure out who did it, but they don't have the resources to go track these people down. And it spreads so fast, it's hard to find the origin point.
[01:00:35] Paul Roetzer: So that's not a solution. My bigger concern is what you alluded to, like, This is just a high profile example of this. This can be done to anyone, anywhere, in schools, bullying through this kind of stuff. Like this is when I get asked the question, what do you lose sleep over? This is the stuff I lose sleep over.
[01:00:55] Paul Roetzer: And so I don't know what we can do other than awareness. And [01:01:00] I don't know, maybe part of what we're, you know, our mission moving forward is to try and find some way to. create some positive outlook for where this can go and how we can solve for it. But it's a really messy problem and there's no obvious solution sitting in front of us right now.
What AI can’t—and shouldn’t—do
[01:01:22] Mike Kaput: So Paul, in other news, you recently posted about the things that AI can't or shouldn't do in response to some content posted by Greg Brockman, who's a leader at OpenAI. And you wrote as part of this, the most powerful models we have aren't replacements for humans thinking deeply about challenges, putting in the work and taking ownership of outcomes, no matter how much AI assists along the way.
[01:01:47] Mike Kaput: This seems like a really important reminder. I'd like you to tell us a little bit more about and what prompted it. Cause we do spend a lot of time encouraging people to find new use cases for AI, but it sounds like there are some things [01:02:00] they shouldn't be. trying to apply AI to in their daily work or life.
[01:02:05] Paul Roetzer: It was a, like an optimistic look. Cause you know, again, coming off of like the last couple of topics, I think it's good to switch gears and head towards the optimist side. And I felt like, so it was the tweet from Greg Brockman was on January 23rd. We'll put the link in the notes said, it's hard to describe how much you learn by actually doing, by carefully considering all factors, making a decision, and then taking responsibility for the outcome unlocks wisdom that cannot be arrived at.
[01:02:28] Paul Roetzer: Any other way. So the reason I thought it was significant was because it's Greg Brockman. Like, this is the co founder and president of OpenAI. And so my thought here is, like, he probably has access to more powerful versions of ChatGPT than we have. I think I talked about this maybe on last week's show.
[01:02:49] Paul Roetzer: He likely has access to whatever GPT 5 is going to be. And if even having the most powerful models in the world, he is still acknowledging [01:03:00] the human element of all of this, I think that's significant because he's looking out and saying, like, there, there are things we're going to need humans for. And so the things I had called out was.
[01:03:13] Paul Roetzer: you know, the value of experience, of having been through things, of seeing things, of having domain expertise, to know if what the AI does is any good. I can ask AI to generate an image, but I'm not a graphic designer. I don't know if the image is any good. I can ask it to create a video for me with Lumiere or Runway or Pika.
[01:03:30] Paul Roetzer: I don't know if the video is any good. I'm not a video producer. Like, I'm not a domain expert. I just have the ability to kind of do these things. And so I think the mix of experience, domain expertise, human intuition, Those seem to be defensible moats for humanity. Like, they seem to be things that, no matter how much the AI assists us moving forward, the humans still need to think, and we still need to have agency over key actions and decisions.
[01:03:58] Paul Roetzer: And if that's not gone away for [01:04:00] Greg I don't think it's going away for any of us anytime soon. So that is not saying AI is not going to play an increasing role in strategy. It's not going to play an increasing role in everything we do. But I do think that there is a significant place for experience, expertise, intuition, common sense.
[01:04:19] Paul Roetzer: Those are things that we stick to. Now, what that means for your job, you know, for within your company or what you do next in your career path. I don't know. Like I can't tell you five years from now that's applied to, but I think those are relatively safe things when I look out and say, what are humans still going to be doing in three years, five years, 10 years.
One easy way to apply AI throughout your day
[01:04:39] Mike Kaput: Alright, last but not least, Paul, you posted a really helpful tip on one way to easily apply AI throughout your day that basically anyone can take advantage of. Can you walk us through what that is?
[01:04:52] Paul Roetzer: Yeah, so this was a fun one for me. So I use on my phone, like probably many of you, I use dictation, like just voice memos, whatever it is, [01:05:00] like as I'm driving the car, as I'm walking, as I'm responding to emails, like it's just way easier than typing out on my phone.
[01:05:05] Paul Roetzer: So I use voice memos. all the time on my phone. I don't think to do it on my computer all the time because it's not as like natural. so what I put up there was one potentially overlooked way to use AI every day that can save you a reasonable amount of time and reduce the stress of your fingers and wrists for those who type all day is to use dictation capabilities built right into your computer.
[01:05:27] Paul Roetzer: So Mike and I have Mac books. I put up how to do that. You just go to settings, click keyboard, turn on dictation, and then pick the shortcut you want. I use. Press the control key twice. So as I'm sitting here, if I'm going to go, you know, type an email, or if I'm going to put a LinkedIn post up, I can just click the control button twice and just.
[01:05:44] Paul Roetzer: And then I can edit that and it's way easier than actually typing out the whole thing and way faster, like we speak faster than we type, the vast majority of us. and the thing I think has been a bit of a breakthrough is the accuracy and speed, at least on the Apple, has gotten way better. Now if [01:06:00] you're a Apple, Macbook, or, You know, computer user, but also a Google Workspace user.
[01:06:05] Paul Roetzer: What I have noticed is it doesn't work in Google Docs, but you can actually go up to tools, voice typing within Google Docs and do the same thing right within your Google Doc. So my whole point here was, I guess, a lot of times I personally overlook this, and I assume other people do as well, to use the same voice efficiency you gain on your mobile device.
[01:06:25] Paul Roetzer: on your computer. The capabilities are built into there. They're good and getting better. and so yeah, it was just kind of like a quick hack, fun way to apply AI every day that, I mean, it's saving me, you know, maybe it's 10 minutes a day, 20 minutes a day. I don't know, like. But it saves time and it, it's a really nice kind of fun way to, to use it.
[01:06:46] Paul Roetzer: Awesome.
[01:06:46] Mike Kaput: Well, Paul, thank you for walking us through everything happening in the world of AI. I want to remind our listeners this week and every week that we have a weekly newsletter that we also [01:07:00] publish to our audience to help you stay Up to date on what's going on this week in AI. It includes not only what we've talked about today, along with some more in depth analysis, but also all the this week in AI topics that we didn't get to cover.
[01:07:15] Mike Kaput: And we literally have, you know, 10 15 of them every single week that we can't possibly get to in a single episode. So if you are interested in staying up to date in a single newsletter and what's going on in AI for marketing and business, go to marketing ai institute.com. Click on resources and click on newsletter to subscribe.
[01:07:37] Mike Kaput: Thanks again.
[01:07:38] Paul Roetzer: Yeah, thanks Mike. And if a reminder, you know, definitely subscribe to the podcast if you aren't ready. But leave a rating and review. Like we really appreciate hearing from listeners, you know, what you find valuable. so you know, please do leave a rating and review on your favorite podcast network and keep sharing it.
[01:07:54] Paul Roetzer: We hear all the time with, oh, I send this to so and so, and. We love that. So, you know, definitely pass it along to coworkers and friends if you [01:08:00] find the show valuable. And we will be back, I think next week is a regularly scheduled week, so we'll be back next week. everyone have a great week. Thanks for being with us. Thanks for listening to the Marketing AI Show. If you like what you heard, you can subscribe on your favorite podcast app, and if you're ready to continue your learning, head over to www.marketingaiinstitute.com.
[01:08:22] Paul Roetzer: Be sure to subscribe to our weekly newsletter, check out our free monthly webinars, and explore dozens of online courses and professional certifications.
[01:08:30] Paul Roetzer: Until next time, stay curious and explore AI.
Claire Prudhomme
Claire Prudhomme is the Marketing Manager of Media and Content at the Marketing AI Institute. With a background in content marketing, video production and a deep interest in AI public policy, Claire brings a broad skill set to her role. Claire combines her skills, passion for storytelling, and dedication to lifelong learning to drive the Marketing AI Institute's mission forward.