Marketing AI Institute | Blog

[The Marketing AI Show Episode 65]: DALL-E 3 from OpenAI, Generative AI’s Second Act, How AI Could Solve Labor Shortages, and Microsoft Copilot Release

Written by Cathy McPhillips | Sep 26, 2023 12:03:24 PM

We’re back with an exciting week of AI news! This episode was recorded Sunday, Sept. 24, only to wake up to Anthropic/Amazon news, and ChatGPT updates! But more on that next week. In the meantime, Paul and Mike break down some exciting updates from OpenAI, Sequoia Capital, Microsoft, and more.

Listen or watch below—and see below for show notes and the transcript.

This episode is brought to you by AiAdvertising. Start winning with AiAdvertising’s innovative approach to maximizing budget and performance. Use AI to optimize campaigns by gaining deep customer insights, drawing out motivations and behaviors, enabling intelligent targeting and ensuring messages hit the mark. Stop wasting time, money, and resources. Let AiAdvertising lead while you take the credit! Visit their website to learn more!

Listen Now

Watch the Video

Timestamps

00:02:44 — DALL-E 3

00:08:12 — Sequoia Capital’s Act 2 of their Gen AI Market Map

00:18:45 — AI and the accounting industry

00:30:06 — Bard and the Google apps connection

00:33:29 — Microsoft announcements at its Surface and AI event

00:37:56 — EU Act implementation

00:42:28 — Jim Thune and Amy Klobuchar unveil major AI bill

00:43:00 — Famous writers join Authors Guild in class action lawsuit

00:46:17 — TikTok rolls out new label for AI-generated content

00:48:01 — Andy Crestodina’s AI for SEO blog post

00:49:58 — Twitter hides X posts with external links

Summary

OpenAI announces DALL-E 3

OpenAI has announced DALL-E 3, the latest version of its image generation model. According to OpenAI, “DALL-E 3 understands significantly more nuance and detail than our previous systems, allowing you to easily translate your ideas into exceptionally accurate images.” What’s more, starting in October, DALL-E 3 will be directly integrated into ChatGPT, which means you’ll be able to generate images right within ChatGPT if you’re a ChatGPT Plus subscriber. OpenAI says, “DALL-E 3 is designed to decline requests that ask for an image in the style of a living artist. Creators can now also opt their images out from training of our future image generation models”

Sequoia releases Act 2 of their Generative AI Market Map

After an initial explosion of hype and novelty apps, generative AI is entering its next phase focused on real customer value, according to a new breakdown from famed venture capital firm Sequoia Capital. In it, Sequoia says we're transitioning from Act 1 technology demos to Act 2 solutions solving real human problems. Sequoia states that Act 1 was “technology-out,” while Act 2 is “customer-back,” solving human problems. Sequoia notes that early winners in this space developed strong product-market fit like ChatGPT for developers and Midjourney for creators, but that overall user engagement/retention remains low, signaling issues with sustained value. The path forward is more complex AI systems, new interfaces beyond chatbots, and workflows optimizing entire systems - not just individual users. The map and article are ones to bookmark.

How AI can (and will) impact the accounting industry

More accountants are quitting, says the Wall Street Journal, due to monotonous work, limited career growth, and burnout from long hours. Others are weighing a switch as new AI threatens their roles. According to the article, among other factors, accountants increasingly face the risk of generative artificial intelligence endangering their roles, fueling their frustration. “As technology gets further and further ahead, we’re stuck in place and the gap is getting wider,” said Ben Wann, who runs an accounting-education company. Rather than the worry of AI taking jobs, AI may benefit this industry because of the labor shortage.

It’s another exciting week in the world of AI. Listen, subscribe, and we’d love your review!

Links Referenced in the Show

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: everything we always teach is use case based and problem based. And as long as you're doing those two frameworks, like you're finding use cases that fit what you already do, improve workflows, drive efficiencies in your work, increase productivity, or solving problems, more intelligent, like customer churn or audience growth or revenue acceleration, whatever your goals are, like, as long as you align the technology you're getting with those, and you properly implement them, you're going to get value.

[00:00:27] 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.

[00:00:47] Paul Roetzer: My name is Paul Roetzer. I'm the founder of Marketing AI Institute, and I'm your host.

[00:00:57] Paul Roetzer: Welcome to episode 65 of the marketing AI show. I'm your host, Paul Roetzer, along with my cohost, Mike Kaput with a special Sunday night, September 24th recording.

[00:01:06] Paul Roetzer: We usually record this on Monday mornings, but we are. Coming to you live at 9 15 p. m. Eastern time on Sunday night due to travel schedule craziness and Mike's got talks. I got talks. I don't you you're back. I'm leaving like, but here we are Sunday night doing this, for episode 65. So This episode is brought to us by AI Advertising.

[00:01:33] Paul Roetzer: Start winning with AI Advertising's innovative approach to maximize budget and performance. Use AI to optimize campaigns by gaining deep customer insights. Drawing out motivations of behaviors, enabling intelligent targeting, and ensuring messages hit the mark. Stop wasting time, money, and resources. Let AI advertising lead while you take the credit.

[00:01:56] Paul Roetzer: Visit aiadvertising. com slash AIpod to learn more. That's one of those, Mike, like we get asked all the time about like advertising use cases for AI. So if you've been like pondering ways you can use

[00:02:12] Paul Roetzer: AI in advertising, check out their site, AIadvertising. com slash AIpod, and just get an idea of some of the use cases that are possible in AI. So it's a good partner to check out. We appreciate them sponsoring the show. So with that, Mike, let's get into it. We actually had to cut a few things right before we started recording.

[00:02:28] Paul Roetzer: There was again, another week where it just didn't feel like there was a ton. And then all of a sudden Thursday, Friday hit. And I realized like we had like. 30 topics in the sandbox and we can only do like 10. So let's jump into the ones we picked for the week. Mike sounds

[00:02:43] Mike Kaput: good. So first up open AI has announced Dolly three, which is the latest version of their image generation model.

[00:02:52] Mike Kaput: So according to open AI, Dolly three understands significantly more nuance and detail than our previous systems, allowing you to easily translate your ideas into exceptionally accurate images. So you're able to do even better image generation using Dolly three. And what's more starting in October, this model Dolly three will be directly integrated into chat GPT, which means you'll be able to generate images right within it.

[00:03:24] Mike Kaput: If you're a chat GPT plus subscriber, interestingly, open AI says that Dolly three is designed to decline requests that ask for an image in the style of a living artist. So creators can also now opt their images out from training of our future image generation models. So we've got this super. High powered image generation software that's going to be built right into chat GPT.

[00:03:53] Mike Kaput: Paul, how big a deal is it that we have dolly three now right within chat

[00:03:57] Paul Roetzer: GPT? It's certainly going to help with adoption. I, so personally for me, I think it's safe to say mid journeys. Probably the most powerful model that's out there. Yeah. Most powerful tool. Seems like most of the examples you see of the most impressive stuff comes from MidJourney.

[00:04:14] Paul Roetzer: But to use MidJourney, I believe you have to go into Discord, right? So I've never done it. I don't want to go into Discord and create something. So, for me, I will use DALI 3, I assume significantly more. I have used DALI 2 many times since it came out last spring. So that came out in spring of 22, if people aren't familiar.

[00:04:36] Paul Roetzer: And mid journey actually came out right around that time. So yeah, the ability to have it integrated right into what you're doing is going to be, I think, huge for adoption and the early examples they were showing were certainly impressive to see. So I'm, I'm pretty excited about this. I thought this was going to come in November and we, you know, we've known this was coming and I haven't said a thing about Dolly, but they had, they'd announced a big developer conference.

[00:05:00] Paul Roetzer: I think it was November 6th or something. And so in my mind, I was kind of assuming we were going to hear something about Dolly. Plus when Bard announced that they had the ability to have images baked into Bard, you knew OpenAI was going to, you know, have that capability too. So yeah, I think this is going to be a big deal.

[00:05:16] Paul Roetzer: And if it works the way it appears in the demo, I think they're going to get a ton of use, from consumers in it.

[00:05:24] Mike Kaput: So like you mentioned, MidJourney is a huge leader in this space, you know, I've used it in Discord to your point, it's not exactly easy to set up and get started with, but how do you see this stacking up against tools like MidJourney that are leading the way right now in image

[00:05:43] Paul Roetzer: generation?

[00:05:44] Paul Roetzer: The way they were positioning it and its capabilities and its ability to handle like words and nuance, and, and also the ability, we've talked about how prompting isn't going to be as critical in the future that the AI is going to, in essence, do the hard work for you behind the scenes. They talked a lot about that, that the thing's going to basically understand a lot more of the context of what you actually are looking for.

[00:06:08] Paul Roetzer: So you don't have to be as good at prompting to get, you know, high quality results. So I think we're just going to continue to see that and obviously OpenAI is continuing to commit to image generation as part of their play moving forward. So

[00:06:25] Mike Kaput: if I'm a marketer or a business leader at this stage, like, should I be taking the plunge and getting ChatGPT Plus in anticipation of having access

[00:06:36] Paul Roetzer: to this?

[00:06:38] Paul Roetzer: Like we've said on the show before, I think everybody should probably have chat GPT plus at this point. Chat GPT enterprise, I don't know, that's a whole nother story, but the 20 bucks a month to play around with this, certainly. And then I think they're opening up to the API as well, right? Like they're going to let people build on top of it.

[00:06:56] Paul Roetzer: Yeah. So I would assume like, you know, I know. What like HubSpot, I think is using APIs for Dolly too. So, you know, if you're in HubSpot customer, I think you can build images and they're using Dolly. Yeah. I would imagine you're going to see this stuff proliferate through not just the use of chat GPT, but the ability to have, image generation and through, through the API.

[00:07:17] Paul Roetzer: So, yeah, I think that if it does what it, again, what it's showing, it's going to be able to do, it's probably going to wide scale adoption. By non designers like me, it's, that's the beauty of it is like, I have no design capability whatsoever. And I hate looking for stock photos. So if we have the capability to go in and use a tool like this, in social posts and blog posts and proposals and keynote presentations, like that was for me, you know, I do so many talks and you're doing a ton of talks now too.

[00:07:48] Paul Roetzer: Finding images for keynote presentations alone, I would pay 20 bucks a month to be able to just like have images on demand. So, yeah, I think it's just going to be a huge value, again, if it works the way it's supposed to. And, Dolly 2 is pretty good, so I would expect this is going to be pretty powerful.

[00:08:08] Mike Kaput: So in another major news story this past week, so generative AI is actually entering its next phase. And this phase is focused on creating real customer value according to a breakdown from venture capital firm Sequoia Capital. And in this article, which they published on their website, which we'll link to in the show notes, Sequoia says we're essentially transitioning from Act 1 To act 2 of generative AI act 1 is more proving out the technology act 2 is actually solving real human and therefore customer problems.

[00:08:48] Mike Kaput: So they say quote generative AI's 1st year out the gate act 1. Came from the technology out, so to speak. We discovered a new quote unquote hammer foundation models and unleashed a wave of novelty apps that were lightweight demonstrations of cool new technology. We now believe that the market is entering act two.

[00:09:11] Mike Kaput: Which will be from the customer back. Act 2 will solve human problems end to end. These applications are different in nature than the first apps out of the gate. They tend to use foundation models as a piece of a more comprehensive solution, rather than the entire solution. They introduce workflows stickier and the outputs better.

[00:09:36] Mike Kaput: They are often multimodal. So Sequoia has basically outlined that the early winners in this space include people like ChatGPT and MidJourney who have found product market fit. But overall, the user engagement and retention of a lot of these GenAI apps and tools remains low. So they're really focused on how do these companies and these tools evolve to actually Retain customers.

[00:10:02] Mike Kaput: So there's a lot to unpack here, but Sequoia seems to be saying that essentially the gen AI market is here to stay and it's moving faster than even the biggest champions of generative AI have anticipated, but there's still some roadblocks to figure out here from a consumer app facing perspective. So Paul, as you're reading this, what is your take on where we're currently at in the generative AI

[00:10:28] Paul Roetzer: market?

[00:10:30] Paul Roetzer: I think overall, I agree with their premise here, but I also think a lot of times what happens with not just Sequoia, but the venture capital firms, Silicon Valley in general, is they kind of live in this bubble where they're seeing adoption rates at a certain clip, they're having conversations with really forward thinking organizations, and you know, I think sometimes they're just too far out ahead of The market a bit in their perception of what's happening.

[00:11:04] Paul Roetzer: So I don't disagree at all that, that act one, you know, as we've got these models and now we need to move on and like make them super value based and active, like I get that, but I also think that the existing, like act one in their terms are still almost largely untapped, like there's. There's so many use cases people could be doing that they just don't know to do.

[00:11:26] Paul Roetzer: There's so many professionals who aren't experimenting with this stuff. So many CEOs who still don't really know. We talked about this on the last episode. There's the adoption is so low still. And, you know, I commonly see, you know, tweets from like an Ethan Mollick or just like, I don't think people realize like the capabilities these things have.

[00:11:44] Paul Roetzer: And if you just learn to prompt a little bit better, what you can unlock with them. But I do think that there's also this. Challenge of Models aren't normal software like an enterprise and a business. They want things to work like software works. Like you buy a thing to do a very specific task and you want it to do it right every time.

[00:12:04] Paul Roetzer: And that's not how these things work. And so I just, I feel like, you know, we keep coming back to this idea of AI literacy and people really understanding what this stuff is and how it works. And I don't care if we're on act five, like until we get. Literacy, right? And until people understand at a broader level what this stuff is and what it's capable of, it's just, it's not going to matter as much.

[00:12:26] Paul Roetzer: So, I don't know. I think overall again, I, I get where they're going. I think building vertical solutions, building AI agents that take more actions that are more, you know, tangible in terms of their value. That makes a lot of sense. But I also come back to like, none of these things even have user guides.

[00:12:43] Paul Roetzer: Like we're just, we're putting out these powerful things in the world, like a GPT four and no one teaches you how to use it. It's like five months later, open AI releases some guide of, you know, here's. You know, best practices of prompting, but it's just not like normal software. And so the adoption curves are going to be the same.

[00:13:00] Paul Roetzer: The value extraction is going to be the same. Like they're using all these analogies about what standard retention rates are and things like that across software. And it's like, I don't, I don't know, I don't know that's comparable. But I did, there was one thing that jumped out to me. When I was reading it that, you know, people think this all just happened overnight.

[00:13:19] Paul Roetzer: Like, I always come back to this ChatGPT moment at November 30th, 2022. We're almost coming up on the one year anniversary of that day. And how much it changed things and how so many people think AI. Was like this thing that just sort of showed up 2022, but they had a paragraph in here that said this moment has been decades in the making six decades of Moore's law have given us compute horsepower to process X of flops of data.

[00:13:46] Paul Roetzer: Four decades of the internet have given us trillions of tokens worth of training data. Two decades of mobile and cloud computing have given every human a supercomputer in the palm of their hands. In other words, decades of technological progress have accumulated to create the necessary conditions for generative AI to take flight.

[00:14:03] Paul Roetzer: So I thought that was a cool frame. And then the other thing that jumped out to me was toward the end, they go into when they published a year ago, they're kind of like generative AI. They were revisiting their thesis and I respect that they were straight up like, here's everything we got wrong, but the things they got wrong are really significant.

[00:14:25] Paul Roetzer: Like, so it kind of goes to show. You know, we say all the time, like, nobody has a clue what's going to happen. So their theory of Act 2 is as good as any theory I've seen. I have no idea if it's right or not. Like, it makes sense in some cases. But listen to this. They said, the first thing, what we got wrong, things happened quickly.

[00:14:44] Paul Roetzer: So keep in mind, they published this, what, about 45 days or so before ChatGPT came out last year. They said, last year, we anticipated it would be nearly a decade before we had intern level code generation. Hollywood quality videos or human quality speech that didn't sound mechanical. All of those things happen in like six months after they wrote this thing.

[00:15:07] Paul Roetzer: But a quick listen to 11 labs voice on TikTok or runways AI film festival. Makes it clear the future arrived at warp speed. They didn't anticipate the lack of supply of chips. They did. Like they thought verticals would take off sooner. So just, I guess my point is like, take all of this, like we say with everything else, you have to absorb this information.

[00:15:28] Paul Roetzer: You can listen to us talk about it. You can go listen to some other podcasts, go read some books, whatever it is, which you have to kind of like form your own thesis and then figure out what that means to your business and your career, because even the smartest people in Sequoia is brilliant. The people doing this stuff, they're the top of their field and have been for decades, one of the most respected firms in the industry, and they get major things wrong in their own thesis.

[00:15:53] Paul Roetzer: So it's just that things are moving so quickly. It's really hard to. To look around the corners, like we say, yeah, that was kind of a few of my initial thoughts on this piece. So,

[00:16:07] Mike Kaput: to that point about trying to look around the corners, like, You started to kind of give some advice here, like if I'm a business or a business leader or a marketing or business practitioner, you know, trying to wrap my head around this stuff, like, given what Sequoia is talking about here, given what we've just discussed, like, what should I be thinking about when it comes to approaching generative AI tools?

[00:16:33] Paul Roetzer: Well, they, toward the end, they said, in short, generative AI's biggest problem is not finding use cases or demand or distribution in proving its value. This is like everything we teach corporations all the time. You know, marketers. Business leaders find use cases that are valuable to you. Like if you're going to get a language model, you're going to get an AI writing tool, are you spending 50 hours a month doing transcription of podcasts and summarization of those into blog posts?

[00:17:01] Paul Roetzer: Like you can get value right now with act one of generative AI, you don't need act two to get a ton of value. So everything we always teach is use case based and problem based. And as long as you're doing those two frameworks, like you're finding use cases that fit what you already do, improve workflows, drive efficiencies in your work, increase productivity, or solving problems, more intelligent, like customer churn or audience growth or revenue acceleration, whatever your goals are, like, as long as you align the technology you're getting with those, and you properly implement them, you're going to get value.

[00:17:37] Paul Roetzer: You don't have to wait around for GPT 6 to get value out of this stuff. We, again, we talk to companies all the time that are doing nothing with this stuff. Like they're just sitting there almost like paralyzed by like too many things, too many opportunities just pick something. And like, like we use, we've talked about the Institute just writing tools alone.

[00:17:57] Paul Roetzer: It's so transformational to our blogging, our podcasting, our content strategy, our social media, like. As a team of now seven. It's gotta save us a hundred plus hours a month just across seven people. So that again, like awesome. I love the thesis. Love like the big picture thinking from Sequoia about where this is going.

[00:18:17] Paul Roetzer: Don't. As a listener to this show, get caught up in thinking you need to understand whatever the heck act two is, and that like, you're already behind because you're, you didn't do act one, right? Like, no, it's great information, process it, think about it, and then just go find use cases and problems to solve today with the stuff we already have, because it's already awesome.

[00:18:42] Mike Kaput: That is good advice. So our third big topic is on the surface about the accounting industry, but it's really about much, much more. And so if you're not an accountant, trust me, this will have value to you, but. We saw a report from the Wall Street Journal recently that said more accountants are quitting due to some of the work factors like monotonous work, limited career growth, and burnout from long hours.

[00:19:11] Mike Kaput: However, a big factor underneath that is also that people are Weighing, switching jobs due to AI threatening their role. Now, we see these accounting firms actually increasingly adopting generative AI, and accountants are increasingly facing this kind of risk where gen AI is endangering their roles and kind of fueling some of the frustration they have about their jobs.

[00:19:39] Mike Kaput: In the Wall Street Journal article, one of the people who runs an accounting education company said, As technology gets further and further ahead, we're stuck in place and the gap is getting wider, essentially between what is expected of them and their actual job satisfaction. So we're talking about accounting here, but really this is a bigger picture because, Paul, you posted about this on LinkedIn in the sense that we are projecting.

[00:20:08] Mike Kaput: That 80 percent of what we do as knowledge workers generally, that includes accountants, will be AI assisted to some degree in the next couple of years. And this isn't just a story about jobs being lost or people being worried about AI. It's also a story about people like accounting firms not being able to find the right workers, the right level of talent.

[00:20:36] Mike Kaput: Can you unpack for us what you kind of found interesting about this?

[00:20:39] Paul Roetzer: Yeah, so, if I had to rank the things I spend the most time thinking about related to AI would imagine the impact on jobs is probably number one. So, I think about this topic non stop, and I keep I wish I had like two weeks to just really dive into this and formulate all my thinking around it.

[00:21:00] Paul Roetzer: But in essence, what I'm trying to do is take a look out and project what is the real impact going to be. Net gain, net loss of jobs over 12 months, over five years, like really trying to analyze this. And so often when this conversation comes up, because I get asked it almost every talk I do about job loss, And I don't always have like the best answers, but I've done like, I just did insurance industry.

[00:21:29] Paul Roetzer: I've, I've talked with a banking conference week. I've talked with auditors of universities. Like there are all these different audiences. I talked with students last week. So I did a talk at a university where there was a bunch of college students who asked maybe some of the best questions I've heard about AI, just brilliant questions.

[00:21:44] Paul Roetzer: But when we talk about it, what we often think about is, okay, we have writers, we have marketers, we have SEO professionals, we have accountants, we have lawyers, we have all these knowledge workers. And you look out and you say, okay, you know, if we take 10 lawyers, and this is what they do all day. And the, I can help do that.

[00:22:06] Paul Roetzer: Do we need as many lawyers in the future? Not that we don't need lawyers, but do we need as many? So the, it always comes down to this, like. Well, do we just need as many humans doing the thing? And so that's the main lens we look through. So the example I gave on LinkedIn, I said, you know, the change in knowledge work is being accelerated because Google, Microsoft, Salesforce, HubSpot, Adobe, Oracle, all the software we use in business is infusing generative AI into it.

[00:22:30] Paul Roetzer: So you're not going to be able to get away from generative AI two years from now. So I'm fairly confident in this 80 percent plus number. Will be, you know, work will be assisted by AI because it's going to be in every piece of software we use. So it's going to be everywhere. So if that is true, then you have to assume you're going to create efficiencies in the production of products and services and that that will likely reduce the amount of people needed to do the work.

[00:22:59] Paul Roetzer: So again, the AI can't do the job of a lawyer. It can't do the job of a writer or an SEO professional or email marketer or like any, any knowledge where an accountant, you can't do their job. But it can make their job way more efficient. So you just need fewer of those people. So the thing I said in here is like the two variables about this part of the equation is, is there demand?

[00:23:18] Paul Roetzer: And we talked about this before, but I'll recap in case people missed it on the podcast, is there demand for increased output? So AI helps drive efficiency for marketers or accountants. And is there demand for increased output? In other words, if you can produce X, whatever that is, whatever product or service or asset that is in 30 percent less time, is there demand to produce more of that with the hour saved if there isn't demand for more of it, then you.

[00:23:44] Paul Roetzer: You reduce workforce. So if I don't need to write more articles or I don't need to, develop more briefs, or I don't need to crunch more spreadsheets or, you know, P and L statements, like there's just, there's a finite supply of them. I'm not going to do more of them than I need to list people. The second piece is we're going to save a bunch of time and money, but we're not going to get rid of our people.

[00:24:03] Paul Roetzer: We're actually going to reinvest it in them. We're going to reskill them, upskill them. We're going to shift them to new roles. So that eliminates the need for workers to go away. So. That's the main conversation that we often have is do we need as many people? But the part that often gets overlooked is what about an accounting or an insurance or in some of these other industries where they can't hire enough people, they are losing people at massive clips and they need people in administrative roles, in entry level roles, and they can't get them.

[00:24:33] Paul Roetzer: There's not enough recruits coming in. There's not enough people majoring these things in college. And you have industries that are at massive deficits when they look to the next five to ten years for where is the employee workforce going to come from. So that's what caught my attention about this accounting one is they said there were 1.

[00:24:50] Paul Roetzer: 65 million accountants and auditors in the U. S. in 2022. It's up 1. 3 percent from the previous year, but down 2. 6 percent from 2020, and this was the real kicker, and down 15. 9 percent from 2019. More than 300, 000 accountants quit their jobs between 2019 and 2021. So if I run an accounting firm and I'm sitting here thinking, wow, AI is going to do the job of accountants, we're in trouble.

[00:25:19] Paul Roetzer: No, you've 300, 000 people left the profession. What if AI is the savior for the profession? And so. That was kind of the premise I was saying is like, in this example, AI can be seen as a threat since a lot of accounting work is data driven and repetitive, two of the variables we always say, if it's data driven and if it's repetitive, there's a good chance AI can help you do it, so there's a lot of that work in accounting, so it can be seen as a threat for people who are still accountants, but it can also be seen as a possible solution for leaders of accounting firms who can't hire enough people.

[00:25:52] Paul Roetzer: And so when you start to factor in those kind of two different approaches of we have work, we can do it more efficiently, we don't need as many people to then you look over at other industries and say, well, they don't have enough people. That's where it gets really hard to say, well, what is the net gain going to be?

[00:26:08] Paul Roetzer: What is the net loss of jobs going to be? And why it's so hard to project out because there's a lot of industries that are really struggling to get those kinds of people and to keep those kinds of people. So it's going to be. In reality, like it could make it even messier because, if you're in the accounting industry and looking at saying, okay, over the next 10 years, we're going to lose another half a million, I don't know, I'm just making up a number, by the way, let's say we're going to lose another half a million accountants out of the industry over the next decade, you are full go into AI, into absolutely everything you can, and you may.

[00:26:42] Paul Roetzer: Because you have no other choice, you may accelerate the reduction of workforces and accounting because you're just, you have to rely on the AI. And how do you predict that? Like I haven't seen a study from an economist yet that's looking out and saying, okay, accountants are going to lose another half a million over the next 10 years.

[00:27:00] Paul Roetzer: And so over the next three years, they have no choice but to infuse AI into everything they do. And when they do that, it's going to reduce the workforce by 50%. Like, I don't know. Somebody's got to do those studies because that was my whole point is like at the end I said the only thing I know for sure is that we should be talking about this more that this topic of impact on jobs isn't being talked about enough by enough people who have knowledge in specific industries or by economists who do this for a living not me who you know plays an economist on Sunday nights on our podcast and took a couple of economics courses in college like Other than that, I'm just trying to theorize here based on observations.

[00:27:40] Paul Roetzer: So that was my main thing is like, it's just a really important topic to be talking about. And if you're in one of those industries where you're seeing that massive talent gap, there's a really good chance that the leaders in those industries are going to become aware of AI's potential to fill that talent gap very quickly.

[00:27:57] Paul Roetzer: And that may accelerate the impact AI has in that profession in good and bad ways.

[00:28:04] Mike Kaput: Yeah, it seems like a point that we often bring up in our talks, our workshops, our consulting and just conversations internally is this idea that so many people are asking the question now, like, how should I use these tools?

[00:28:18] Mike Kaput: And that's a good question to ask, but a bigger question is what impact will these tools have on our business, our industry, our change management that we need to do moving forward.

[00:28:34] Paul Roetzer: Yeah, and I think I said this on the previous episode, but I'll, I'll again reiterate it. Don't wait for the definitive study to come from McKinsey or Gartner or Forrester or your industry association.

[00:28:46] Paul Roetzer: They're not going to solve this for you. Like this, this has to be. You or your company gaining the knowledge more deeply about what AI is capable of. Act One ai, or Act two, whatever you want to categorize it as, and looking at your teams, looking at the roles and responsibilities of those team members, looking at the reality of your ability to recruit and retain talent in your industry.

[00:29:12] Paul Roetzer: There's just too many variables that outsiders aren't going to know. And that's why I was saying, like, we just need this as a topic that's being talked about in every company. And again, I'm not trying to be alarmist here. I'm not trying to say like, oh, we're just going to lose a bunch of jobs. That's not, that's not it.

[00:29:28] Paul Roetzer: I think that's going to happen, but that's not the point of doing this. The point is you need to figure out if it's going to happen in your company or in your industry. I think it's going to happen in ours. And I'm, you know, I've been doing this for 23 years in our industry. So I I'm comfortable assessing the impact on writers and SEO professionals and email marketers and stuff like that.

[00:29:47] Paul Roetzer: Cause that's what I do. I'm not an insurance professional. I'm not an accounting leader. So I just feel like we need those people thinking about this in their industry and solving it for themselves.

[00:30:00] Mike Kaput: All right, let's dive into our rapid fire topics for this week. First up, Google just announced some major updates to Google Bard, which is its AI assistant and chat GPT competitor.

[00:30:15] Mike Kaput: And basically they're launching something called Bard extensions, which is a way to interact and collaborate with Bard. Basically, what Bard can now do is find and show you relevant information from Google apps that you use every day. So things like Gmail, Docs, Drive, Google Maps, etc. And this is an interesting set of updates, but also one that's...

[00:30:42] Mike Kaput: Hitting some kind of rocky obstacles here because New York Times reporter Kevin Roos tested these out and actually found them somewhat wanting at the moment. He said, I put the upgraded Bard through its paces, hoping to discover a powerful AI assistant with new and improved abilities. What I found was a bit of a mess.

[00:31:04] Mike Kaput: And basically goes through saying that, look, it was able to do some of the Basic things I was trying to achieve, but overall it was not giving me very sophisticated or advanced advice or outputs from the capabilities they claim that Bard was able to do. So, Paul, first up, kind of what did you make? Of these updates, like of Kevin Ruse's response here, what's going on with Google and Bard?

[00:31:34] Paul Roetzer: Yeah, it, did OpenAI call those extensions too, or they were apps, right? They were plugins, Is what they called them, yeah. Not to be confused with. So, you know, as we've talked about before on the show, I fully expect... When all things are said and done that, that Google is a major player in all of this and that they will be highly competitive, not ahead of open AI in a lot of areas.

[00:32:00] Paul Roetzer: And it would appear that we're not there yet. So this is a step in the direction, you know, the ability to query my emails and things like that sounds awesome on the surface. It does fall under that, our law of uneven AI distribution. You have to be willing. To give up access to these things and I think I was trying to find the screenshot I took when it asked me if I wanted to connect it in my personal Gmail, but I think it said something like they might give that data like third parties like there was something that gave me pause.

[00:32:33] Paul Roetzer: I was like, I'm going to come back to this. So I did not turn it on in my personal account yet because I want to dig into the terms. Some more, but yeah, I haven't seen anybody who was blown away by this and like, oh, this is the game changer we've all been waiting for. So I would say like everything else we've said with Google, test it out, keep an eye on it, probably wait for Gemini this fall with their next language model, their multimodal model they're supposed to be coming out with.

[00:33:02] Paul Roetzer: Maybe that's going to be the leap forward. We keep waiting for it from Google, but. Yeah, I don't know. But if anybody is using this and seeing value in it, hit me and Mike up on LinkedIn, let us know. I'd love to hear about it. Cause I haven't seen anybody yet. Who's, you know, really. Loving what they're doing yet with park.

[00:33:23] Paul Roetzer: It's coming. Like I'm, I'm confident it's, it's going to come, which I don't, I don't think we're there yet though. So with

[00:33:29] Mike Kaput: every Google announcement, there's a corresponding Microsoft announcement, and this is, this week is no different because Microsoft had a hardware event, but they actually focused quite a bit on artificial intelligence during it and.

[00:33:43] Mike Kaput: They announced that they are expanding and kind of unifying their co pilot AI assistant across their products. So co pilot is basically Microsoft's generative AI assistant that can help users with things like writing emails, generating code, summarizing docs, and more across kind of all their, their apps.

[00:34:06] Mike Kaput: They actually announced that starting on September 26th, so... The day this podcast drops, Copilot will be rolled out as part of a Windows 11 update and will be integrated across Microsoft apps like Word, Excel, Outlook, and Edge. So you'll be able to actually use Copilot using voice commands or just by clicking and starting to type.

[00:34:28] Mike Kaput: And this is going to eventually allow you to do things like organize your desktop windows, create playlists based on musical tastes and solve math equations just by snapping a photo. So basically anything you're trying to do using Windows 11 is going to be able to be assisted by artificial intelligence.

[00:34:50] Mike Kaput: Now, Microsoft at the same time also announced some upgrades to other AI services, you'll be able to start using DALI 3 in Bing because they do have a partnership with OpenAI. And there are going to be some new shopping features in their visual search within Bing. Now, Paul, it seems like This is some pretty serious, like AI firepower to have across these apps.

[00:35:18] Mike Kaput: We've talked about it a little bit. Sounds like it's rolling out, pretty quickly. Like how big a deal is this? What kind of effect will this have on people using Microsoft products?

[00:35:29] Paul Roetzer: Yeah, this is the one we've been waiting for. Obviously, we talk about Duet AI from Google Workspace is the other major one here.

[00:35:35] Paul Roetzer: But to me, the biggest unknown to knowledge work in 2024 is whether or not these work, like the demo video show. So, so far with Duet AI, it hasn't, like it's not fundamentally changing, work. The co pilot one, maybe. And I found this, this Page there's a lot of information on this page. The blog post will put it in but it says Copilot will begin rollout in September 26 like you said free update to Windows 11, but then further down it talks about November 1st is like When they're going to, let's see, we're excited to share that Microsoft 365 Copilot will be generally available for enterprise customers November 1st.

[00:36:22] Paul Roetzer: I don't know, like, so I, my guess is if you're an enterprise customer and you're going to pay the 30 bucks a month, it sounds like November 1st might be your target date. But it's going to start coming out through windows this week, like you were saying, yeah, my, my whole take here is I have heard from some people who have access to this in the last two months who were blown away by it.

[00:36:47] Paul Roetzer: Um. We've referenced the demo video. We'll put it in the show notes again from March. That was like, like seriously, probably the best demo video I've ever seen for a software product. Well, runway does some really good ones too, but, if the, if cope, if 365 copilot does what it shows in the demo video, PhD level data analysts and Excel real time transcription of meeting notes with summarization and action items.

[00:37:14] Paul Roetzer: Writing documents for you based on documents in your cloud, or you just click here and click there and like, you know, develop a brief based on this or build a 10 slide deck based on this press release and it just does it seamlessly. Like, if it does those things, it changes knowledge. There's a lot of organizations that use Microsoft to do what they do in Word and Excel and PowerPoint.

[00:37:37] Paul Roetzer: And if it changes that, then it's the thing we've been waiting for. So I will be very anxious come November to see if it really does. So we also

[00:37:50] Mike Kaput: have some movement on the regulatory front when it comes to artificial intelligence. So first up, the EU appears to be closer to enacting their Artificial Intelligence Act, the AI Act, into law by the end of the year.

[00:38:07] Mike Kaput: So we actually heard from a spokesperson for the European Union Parliament on the AI Act saying that the body is in the final stages of negotiations. And his name is Dragos Tudorakis and he... So we are in the final stages of the negotiations between the parliament and the council. The two co legislatures that work in Europe on putting forward legislation.

[00:38:31] Mike Kaput: And we are very close to the finish line. And basically he said, we are expecting to see the AI act by, I think the Tentative timeline is the end of the year and basically this is a wide ranging set of regulations in the European Union that make sure that AI systems are used in a safe, transparent, traceable, and non discriminatory fashion.

[00:38:56] Mike Kaput: Basically this is going to be seen as The benchmark for a lot of legislation that could come out of the US or other countries about regulating AI. So Paul, when you saw this, we've talked about the AI Act a couple of times. Are you, what are you, what are you looking at and paying attention to when it comes to this?

[00:39:16] Paul Roetzer: Yeah, we've known that they're ahead of the U. S. They've been working on this for years. Generally, I think threw a curveball at it. I think they were probably getting close to the finish line about this time last year, and then ChatGPT showed up, and they had to maybe rethink what the laws were going to be and how they were going to be governed.

[00:39:34] Paul Roetzer: So, yeah, they said they think that negotiations by November, and then close the process, and then have a final vote in Parliament and Council, which they seemed... Relatively confident, they, they would get through. He did stress and it was like a four and a half minute video. And he did stress that once it does become law, assuming it does, then they have to actually put the processes and governance in place to oversee it.

[00:39:58] Paul Roetzer: So I have no idea how long that takes. It does not sound like a 30 day process to me. So I'm assuming. If they can get the law through this year, early next year, that sometime in 2024, they will actually start, you know, governing that law. So, yeah, I would go read up on it. I think it's important. They talk about the key categories of unacceptable risk, high risk, they get into generative AI, specifically say.

[00:40:24] Paul Roetzer: Like chat GPT would have to comply with transparency requirements. These are the real sticking points to me. I just don't even know how chat GPT or BARD would be available in the EU. When you read these, it says you have to disclose content was generated, generated by AI. So again, if your brands, if you're listening to this and you do work in the EU, you have customers in the EU, think about the implications for yourself.

[00:40:45] Paul Roetzer: Disclose the content was generated by AI. Don't know how much you have to disclose or if it's every post. Who knows, you have to disclose the, transferring around designing the model to prevent it from generating illegal content. So you have to have guardrails in place. Good luck with Twitter slash X doing their model.

[00:41:06] Paul Roetzer: Cause they're probably trying to remove all the guardrails and then publishing summaries of copyrighted data used for training. That was a real, real fun one. So we're going to get disclosure on all the stuff that these things were trained on that they weren't supposed to be trained on. So that's why, you know, I don't know, like it's very possible.

[00:41:23] Paul Roetzer: They just, the EU doesn't have access to these models, in the early going or, well, actually not think about this so that this may force like the release of GPT 4. 5 or five might have to come out faster because. If they're going to train it on licensed data, GPT 4 is not like, there's sort of licensed data in there, but there's a whole bunch of stuff that's not licensed and they don't want to disclose it.

[00:41:52] Paul Roetzer: They're going to have to release a new model. So that would lead me to believe that a lot of these foundation models that were trained on stuff they probably shouldn't have been trained on. We would maybe be seeing new versions of those next year before they're released into the EU because the current versions probably wouldn't fly.

[00:42:08] Paul Roetzer: It could make for an interesting year.

[00:42:12] Mike Kaput: Well, that's not the only thing going on when it comes to regulation and issues around kind of legal. Items related to artificial intelligence kind of a quick hitting segment here. We have a few updates when it comes to regulation. So 1st up the, we have seen a bipartisan bill in the U.

[00:42:32] Mike Kaput: S. Actually introduced by Senators John Thune and Amy Klobuchar that aim to take a quote light touch approach to governing AI technology. So basically, they're trying to write legislation and pass it. that mitigates risks from artificial intelligence without super heavy handed regulation that could stifle some innovation.

[00:42:55] Mike Kaput: And at the same time, a couple other items that are happening at the moment, we are also getting some significant, names. adding their, kind of signature to the Authors Guild suing OpenAI. The Authors Guild, we covered this in a past episode, they're engaging in a class action lawsuit against OpenAI because there have been, some allegations that books have been illegally used to train some of the models like GPT 4.

[00:43:27] Mike Kaput: So George R. R. Martin of Game of Thrones fame and the author Jodie Peacock. filed a class action lawsuit alleging that OpenAI is illegally using their work without permission to train its systems. In addition to all of this, the White House at the same time has also said that it could force cloud companies to disclose their AI customers.

[00:43:53] Mike Kaput: They are considering requiring these firms to report some information about their customers to the U. S. Government. This is according to some reporting from Semaphore. Basically, this would direct the Commerce Department to write rules forcing cloud companies, Microsoft, Google, Amazon to disclose when a customer purchases computing resources beyond a certain threshold.

[00:44:19] Mike Kaput: So nothing has come of this yet. It has not been finalized and the specifics could change, but the government in the U S is actually considering some action around that aspect of AI. Now, Paul, as you're seeing these updates, like where are we at in the U S at least when it stands, when it comes to regulation.

[00:44:40] Paul Roetzer: Well, I've, I've said for the last couple of months, I'm most interested in the executive order when Biden's office teased that an executive order would be coming this summer, they missed the mark as we are now in the, in the fall officially. But I, I think something is still coming. This is the first time I've seen a specific item that might be in that executive order word is it's a very, very long executive order.

[00:45:03] Paul Roetzer: And my guess is some of these things that we've been hearing about on Capitol Hill, some of the conversations that happened with senators, some of this regulatory capture we talked about, where these big existing AI companies are sort of pushing for very specific regulation. It's starting to sound like some of those requests may find their way into an executive order, and we might get some action.

[00:45:25] Paul Roetzer: Sooner than later. This one makes a lot of sense to me. They give the example in the semaphore article about the banking industry has this, like, you know, banks have to report if something, is suspicious in terms of money laundering or other illegal activities or transactions exceeding like a certain threshold each day.

[00:45:44] Paul Roetzer: So yeah, that it's an interesting way to monitor it, but you'd have to, I would think you'd have to just, you would have to require. So if you're not going to allow it, you're basically saying that anything more powerful than GPT 4 needs to be disclosed, which is what they were going for originally anyway.

[00:46:01] Paul Roetzer: So I don't know. Yeah, I would pay attention to the executive order. I still, my kind of spidey senses is like that this is coming sooner than later, and I think it's going to be far reaching.

[00:46:14] Mike Kaput: So on kind of the self regulation side, TikTok is actually introducing a new label that creators can use to tag their content that's been heavily edited by AI tools.

[00:46:27] Mike Kaput: They also plan to start testing automatically labeling content detected as AI generated. And this is basically designed to better inform viewers. Like, look, what you're watching and hearing is AI generated and make sure that we're limiting the spread of misinformation. Paul, how significant do you see this being?

[00:46:48] Mike Kaput: Given TikTok's user base, but also like, is this just a PR move or did they actually... able to start detecting and labeling this type of content. I would

[00:46:59] Paul Roetzer: imagine there, there's probably some advanced ways to do detection now. And I hope, I hope it's a sign of these capabilities being more widely available, because as you and I have talked about many times, like my Biggest concern right now is elections and democracy.

[00:47:18] Paul Roetzer: Like I I'm really worried about the average citizen does not know AI can create things. So we, you know, again, we live in this bubble. We talk about this stuff all the time. People who listen to this, think about this stuff all the time, but it's a really good chance. A lot of your family and friends have no idea that AI can make videos, can make images that look real, can do deep fakes and things can generate language.

[00:47:38] Paul Roetzer: Like it's just not. Common knowledge. And so I'm just really worried that moving into the next 12 months or so leading up to the elections in the U S that, we need all the help we can get to detect and inform people if things are AI generated. So yeah, hopefully. Whatever TikTok's doing is a sign of things to come from other companies.

[00:48:00] Paul Roetzer: So in a more positive

[00:48:01] Mike Kaput: use case for artificial intelligence, if you want to use AI for SEO, we have a really good resource that takes a deep dive into how to do just that this comes from our friend, Andy Crestedina at Orbit Media Studios, and he just published an article that's a masterclass in how to use generative AI like ChatGPT to uncover SEO insights and optimize content for search.

[00:48:28] Mike Kaput: It's a big, long article that is well worth bookmarking. You have to check it out on your own. We're going to just skim over the highlights here. Because Andy gives step by step advice on how to rank better in search with AI. And that includes use cases like using AI for SEO edits, using it for title tag edits and recommendations, using it to find gaps in your homepage content, and much, much more.

[00:48:58] Mike Kaput: So, Paul, this, you know, obviously is an area that we are very focused on. This seems like a pretty good out of the box use case for marketers of just off the shelf tools like chat GPT.

[00:49:10] Paul Roetzer: Like, do you agree? Yeah. And Andy's the best, like I love going to Andy's talks. He always does like super tactical, high value presentations.

[00:49:20] Paul Roetzer: And so anything he does like this, whether it's Google analytics or SEO or content strategy. He is awesome stuff. So yeah, if you're an SEO content strategy, definitely check this out. And then quick bug, Andy's actually going to be speaking at our AI for agencies summit is, so if you're a marketing agency, we have a summit coming up November 2nd, just AI for agencies.

[00:49:39] Paul Roetzer: com. Andy's actually doing an AI for SEO talk, at that summit. It's a virtual summit. That's it. That's my quick plug. And go check out Andy, because he's brilliant. Yeah, it really

[00:49:51] Mike Kaput: is worth visiting and revisiting. He just breaks down everything you need to know. Last but not least, we have X, formerly Twitter, has made some changes.

[00:50:04] Mike Kaput: To how it displays tweets on their platform. Now the platform is defaulting to hiding tweets that have external links, and Elon Musk defended this decision publicly by tweeting. The algorithm is trying to maximize Unre regretted user time on X, which is a good goal. If a link is very compelling and sent by many, you will see the link however.

[00:50:33] Mike Kaput: This was in response to some very intelligent criticism from renowned venture capitalist Paul Graham, who said, quote, in order to keep you here, Twitter gives less exposure to tweets with links. If you don't want to be manipulated in this way, make a point of posting links and liking and retweeting tweets that contain them.

[00:50:54] Mike Kaput: So some user pushback from a big Twitter user here and a public debate around. What is, X actually doing with user engagement and these tweets, Paul, this caught your attention. We've talked about how X fits into a bigger AI picture. What did you take away from this update?

[00:51:16] Paul Roetzer: Yeah, it's just the manipulation, manipulation of the algorithms is interesting to me and how they're doing it.

[00:51:21] Paul Roetzer: The unrecredited time is. a crazy number to like target. I don't even know. I don't understand what that even is. I think to Elon Musk's credit, he, he does listen. He makes some challenging decisions to understand, as a, a Twitter user, kind of a power user of Twitter in terms of how I consume information, but what's been driving me nuts for the last few months is.

[00:51:47] Paul Roetzer: I have people I follow to learn about AI and I don't see stuff from them because they share links to research reports and blog posts and videos and like things they're creating to help me understand AI and that stuff's being hidden. I commonly share links about interesting things related to AI all the time and I get almost no engagement.

[00:52:07] Paul Roetzer: I have 19, 000 followers, whatever, on Twitter. And it's like a ghost town and it has been for months and I assumed they'd made a major change and, and then for him to just say it out loud of like, yeah, we're hiding like, it's like, okay, okay, good. I'm glad we're admitting that in a, in a retweet or a follow up or whatever.

[00:52:28] Paul Roetzer: So I just thought it was interesting to, to see what they're doing. I think they'll have to change this at some point. Like. I feel like Paul Graham made a couple of great points. Someone said like, if, 'cause people don't know Paul Graham from Y yc, you know the Combinator where I saw Sam Altman was the president before went to OpenAI.

[00:52:48] Paul Roetzer: So someone said if Twitter was a startup part of YC and the founders did this, to say improve retention, what would you tell them? He said, I would tell them not to. It's bad for users and bad for the world and if you pull shit like this, it leaves room for competitors who give people what they actually want.

[00:53:04] Paul Roetzer: It's a pretty, pretty good way of putting it. And then someone came on and said something about Twitter. And he said, what people use Twitter for is not identical with what Twitter uses people for. I, that was so brilliantly stated. So I'm here trying to learn about AI, trying to share knowledge about AI.

[00:53:20] Paul Roetzer: And that is not, they don't care about that. Like, so if links are valuable to me, they, they just don't care. So, yeah, I just thought it was interesting. Like. If you were wondering why, you know, Twitter maybe isn't as usable right now. If you learned from links or shared links that you now have your answer, they're hiding it from everybody.

[00:53:40] Paul Roetzer: So yeah, that, that was kind of just a, a lesson learned and. Like anytime you're working with tech companies or using tech platforms, you're at the mercy of however they choose to do what they do. And so in this case, luckily I have highly curated lists and I basically use 99 percent of the time on Twitter.

[00:54:00] Paul Roetzer: I use the lists I've created. So I see everything in there that they share. They're not hiding links in there. Cause I get every tweet in there. And then I also get alerts from key AI people. So that's kind of, I, if you want to still get value out of Twitter, even though they're not sharing links, build lists, follow our lists.

[00:54:16] Paul Roetzer: I have public AI list. And get alerts from the people that are really the thought leaders that you want to learn from. So I have, you know, probably two or three dozen alerts set up for different people. And so I see everything they put up and that's how I kind of stay up to date on everything that's going on.

[00:54:32] Paul Roetzer: So, yeah, I would, cause I was going to ask that too. Like, how do you stay up to date on AI? That, that, that's it. I have Twitter lists and I get alerts from key people and brands.

[00:54:41] Mike Kaput: Well, we're glad you do because Paul, you have just shed some light on, another crazy busy week in AI. We really appreciate, you unpacking all these topics.

[00:54:53] Mike Kaput: I know the audience appreciates it, so thank you again for your time and your insight

[00:54:57] Paul Roetzer: here. Yeah, thanks for doing the special Sunday night edition. So, yeah, this is, again, if you listen to it, it's, it's going to be Tuesday. It's coming out the 26th, but, this was Sunday night. September 24th. It is now 10 p. m.

[00:55:10] Paul Roetzer: at night and it is time for Mike and I to go to bed. All right. Thanks everyone for listening. We'll be back with you next week. Thanks, Mike, as always. Thank you,

[00:55:17] Mike Kaput: Paul.

[00:55:17] Paul Roetzer:

[00:55:17] Paul Roetzer: 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. Be sure to subscribe to our weekly newsletter, check out our free monthly webinars, and explore dozens of online courses and professional certifications.

[00:55:39] Paul Roetzer: Until next time, stay curious and explore AI.