While our Institute focuses on AI in the marketing world, some of the biggest companies in the world are finding AI applications for the business world and our personal lives. Artificial intelligence will be, and already is, embedded in our world. What are the big players working on? Mike and Paul discuss this on this week’s podcast.
This episode kicks off discussing consulting firm Deloitte, who recently published a rundown of how AI for work relationships could be the next big thing in your office. Deloitte says that AI can “analyze human interactions during and after an event to generate personalized, confidential recommendations at an individual and organizational level to help improve human interactions at work.”
They give a hypothetical example to illustrate the point: Imagine a near-future workplace where AI recommends how you should write a diplomatic email to two leaders pulling you into a nasty turf war. In this scenario, AI could recommend appropriate language and courses of action to resolve the dispute. It’s an interesting discussion on the opportunities and challenges, including the five areas Deloitte feels AI will have a big impact on work relationships.
Next, OpenAI, the creators of GPT-3 and DALL-E 2, just launched a program to fund and support founders creating transformative AI companies. The program is called Converge. According to the company, it is a “highly-selective, five-week program for exceptional engineers, designers, researchers, and product builders using AI to reimagine products and industries.” Participants receive a $1 million equity investment from OpenAI’s Startup Fund. They also get early access to OpenAI models and programming tailored to AI companies. In addition, they get workshops, office hours, and events with AI practitioners.
OpenAI says it’s motivated by “the belief that powerful AI systems will spark a Cambrian explosion of new products, services, and applications.” Mike and Paul discuss why hungry entrepreneurs are critical to the success and adoption of AI.
Lastly, Google revealed a handful of incredible AI projects that it’s been working on, and they provide a glimpse of the near future of AI. These reveals break down into two broad categories: AI for social good and generative AI. On the social good front, Google revealed ideas such as AI for wildlife tracking, AI for flood forecasting, an AI-powered maternal health app, and an AI model that speaks the world’s 1,000 most-spoken languages. On the generative AI side, Google revealed self-coding robots, where robots can autonomously generate new code. Mike goes through a cool example, and they discuss the implications of these new projects.
Listen to this great conversation with our team, and stick around for the end of the podcast for the rapid-fire discussion at the end!
00:02:52 AI in the workplace
00:15:27 OpenAI funds entrepreneurs
00:26:44 Google unveils new AI projects
Disclaimer: This transcription was written by AI, thanks to Descript.
[00:00:00] Paul Roetzer: I see the adoption challenges often being the people. The tech is getting really good. It's going to keep getting better at an exponential pace, and I think over time the lack of adoption is going to be more driven by people refusing to accept AI's abilities in these areas that they consider uniquely human. 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.
[00:00:37] Paul Roetzer: My name is Paul Roetzer. I'm the founder of Marketing AI Institute, and I'm your host.
[00:00:46] Paul Roetzer: Welcome to episode number 24 of the Marketing AI Show. I'm your host, Paul Roetzer, along with my co-host Mike Kaput. What up, Mike? How's it going? It's good, man. Another busy week in ai, huh? As
[00:01:00] Mike Kaput: always, never. Never a dull moment.
[00:01:03] Paul Roetzer: So Mike is Chief Content Officer of Marketing AI Institute and the co-author of our book, Marketing Artificial Intelligence, AI, Marketing, and the Future of Business, which is available now. We always like to do a quick shout out to our sponsor before we get started on our AI weekly edition.
[00:01:18] Paul Roetzer: So, we want to thank Persado. Persado is the only motivation AI platform that generates personalized communication to immediately motivate each individual to engage and act. Organizations that use Persado benefit from an extensive customer motivation knowledge base, enabling them to hyper personalize their communications at scale.
[00:01:40] Paul Roetzer: Clients who incorporate Persado's innovative motivation, AI increase conversions by an average of 41%, unlocking millions in unrealized revenue. Visit persado.com. That is P E R S A D o.com. If you're new to the AI weekly edition, we cover three hot topics in AI that have caught our attention throughout the week.
[00:02:04] Paul Roetzer: And if we have time, we do a little rapid fire at the end with a few quick hitting topics. Mike and I sort of collaborate throughout the week to surface the things that seem most interesting, and then Mike picks them and puts them in an agenda and then we get together and we talk about them. So that's kind of the format.
[00:02:20] Paul Roetzer: A year ago, we didn't need this format. Things weren't happening as fast as they are today. But we felt like we were having trouble keeping up with everything that was going on each week. And so probably about a month or so, I think this is our fifth maybe weekly edition, we were like, "All right, let's just force ourselves to start having this conversation once a week and trying to make sense of all this."
[00:02:41] Paul Roetzer: So new to it. Welcome if you're back from a previous episode, thanks for being here again, and I'm going to turn over to Mike, and let's get rolling with our three topics.
[00:02:50] Mike Kaput: Thanks, Paul. All right, let's jump in. First up: consulting firm Deloitte recently published a rundown of how AI for work relationships could be the next big thing in your office.
[00:03:05] Mike Kaput: And so what they mean by this is Deloitte says that AI can "analyze human interactions during and after an event to generate personalized, confidential recommendations at an individual and organizational level to help improve human interactions at work." So specifically, they actually highlight five different areas
[00:03:28] Mike Kaput: they think AI for work relationships will have a big impact. I'm going to quickly run through the five, and then I want to get your take on this. The first is emotion AI that can actually help a business leader practice emotional intelligence by running simulations of different types of conversations they might need to have with different workers who have different personality types.
[00:03:51] Mike Kaput: There's AI that can understand customers better by extracting insights from customer interaction data and actually, provide better customer service by conducting real time sentiment analysis on different conversations you would've with customers. There's AI that is actually able to make it easier to recruit diverse workforces by enabling things like blind hiring and also evaluating sort of hard to assess soft skills like communication and empathy.
[00:04:20] Mike Kaput: And then finally, there's AI that can foster more inclusive work environments by actually telling when people are taking over a conversation or talking too much or interjecting and actually encourage other people who might have different communication styles or be a little quieter to contribute to a conversation.
[00:04:41] Mike Kaput: And then lastly, there's actually also AI that can figure out what they call informal networks within organizations. So AI technology is like text mining and natural language processing that leaders can use to analyze who in an organization is connected and what the nature is of those relationships.
[00:05:02] Mike Kaput: So you could actually find people within the organization who, regardless of kind of title or seniority, actually wield the most influence within certain teams, which could then be helpful for promoting certain initiatives or leadership goals. So Paul, I wanted to kick this off by asking you, given these kind of developments that Deloitte is seeing, how do you see AI evolving in the workplace, especially for these types of social
[00:05:28] Paul Roetzer: interactions?
[00:05:30] Paul Roetzer: I think it's a really smart report. The examples they give make a ton of sense. Impact of AI at a kind of a more human and behavioral level makes a lot of sense. Their subhead in this thing is AI can do more than make work better for humans, it can help make better humans for work. and I think you know, it aligns with what we've always talked about of the idea of more intelligent, more human.
[00:05:57] Paul Roetzer: You can build a more intelligent organization that uses AI in different components to drive efficiency and performance and make predictions and do all these interesting things, but in the process you can become more human as an organization too. Now, a lot of times we talk about that in the context of as AI frees you up from doing the repetitive data driven things, you can focus on the things that are more uniquely human. You know, empathy, strategy, creativity, relationship building, interpersonal communications, like things that are really hard for the machine to necessarily do. Now, in this case, they're actually talking about the AI being there almost as like that digital twin, like we talked about a few episodes ago - and you're practicing your messaging to your team in front of your AI. So if you imagine you're the CEO and you're going to give a big talk to the team about something difficult maybe today. Rather than just going in cold, you give the presentation to an AI on your computer that assesses it and recommends to change your tone or say this a little differently.
[00:07:01] Paul Roetzer: Or the example they give in there is the email you're about to send off an email and you're frustrated and the AI is there at your side saying, Okay, don't send this. Like, change the tone a little bit. Maybe soften this, here might be a better approach to it. So it's more of that assistant that's there to help bring out the human side of you in a very odd way.
[00:07:21] Paul Roetzer: So overall, it's a very, I think for many organizations, a very aspirational article. Some of these use cases aren't things that a lot of middle market companies, for example, would be doing in the next one to two years. They're more longer term applications of AI that certainly will be infused into organizations.
[00:07:43] Paul Roetzer: It just might take a while. And then the one thing I wanted to call out, because I know we kind of covered a couple other elements from this article, it does start off with something that's a bit misleading in my opinion and I just want, if you do go read the article, I want you to be aware you are not falling behind.
[00:08:01] Paul Roetzer: It actually states in the second paragraph, while 91% of business leaders surveyed in 2022 said they have an enterprisewide AI strategy, they are typically using it for insights, process, whatever. So it's the 91% I want to call your attention to, because you could easily read that as a marketer, business leader and think, oh my gosh, we are so far behind the world.
[00:08:25] Paul Roetzer: I had no idea that 91% of people had a companywide wide AI strategy. Sometimes, and this is more general guidance on any time you're reading a study, it's very important to understand the source of the data point. So they do end note this, but then you have to kind of dig into where the end note leads you to, and then you have to download the report to read the end notes of the end noted thing and where I'm getting at is they did survey 2,600 global business leaders across six industries as part of their state of AI report, which maybe we'll talk a little bit more about next week, Mike, because it's a good report as well. And if you drill then into the methodology on page 47 of that report, it actually states that all participating companies have adopted AI technologies and are AI users.
[00:09:16] Paul Roetzer: It is actually a requirement to even take the survey that you are already adopting and using AI. So when they say 91% of business leaders, it's a little bit misleading. It is 91% of business leaders who are adopting and scaling AI say they have a strategy. So again, that was a long way of saying if you read this and you see that, do not feel like you are falling behind. Their own research then ends up going on and saying like, 28% of people are at starter level, 22% are underachieving, 24% are still trying to figure this stuff out.
[00:09:51] Paul Roetzer: So in their own report, of those 91% that are actually doing this, only 27% are what they would consider like transformers. The people have actually figured this out. So I say that as a, again, a long way of saying really cool report, looks at some really fascinating use cases that are maybe outside of the realm that we often talk about, where it's really getting involved in the psychology of this stuff and helping bring out the more human nature and things like that.
[00:10:18] Paul Roetzer: But it is probably more mid- to long- term applications for most people that would be listening or thinking about this as a leader in a middle market or enterprise company.
[00:10:32] Mike Kaput: Yeah. That's what jumped out to me is sort of this medium to long term thinking. And what I liked about these examples is though you might not be having to grapple with these specifically, say tomorrow, they do illustrate how drastically AI could change the workforce and the workplace in the next say, five years. And I'm curious, what kind of challenges do you see business leaders potentially facing when this kind of technology starts seeping into the workplace? Like we already have pretty predominantly, I would say, at bigger enterprises AI powered, or at least AI assisted screening for hiring candidates
[00:11:18] Mike Kaput: and the hiring process. I know in resume reviewing is pretty heavily infused these days with AI. What should I be thinking about as a business leader if I'm trying to implement this type of socially focused and impactful AI?
[00:11:35] Paul Roetzer: The challenge is that most leaders don't understand the technology.
[00:11:41] Paul Roetzer: So, let's say you are a director of HR or chief talent officer or whatever it may be, you're probably not an AI expert. There probably are some people out there who have really jumped in to figure this technology out. But many of the leaders in these different divisions of the company that we're talking about being impacted, from HR to finance to product operations to the CEO level, many of those leaders are unaware or unprepared for not only the integration of AI, but the downstream effects of that integration.
[00:12:18] Paul Roetzer: So let's say, for example, that we use AI in the HR process, whether it's upskilling, reskilling, professional development, hiring, things like that. There are effects that are caused by that action. And it could be fear. For example, like fear of AI replacing me as a worker, fear of not thinking that my skillset is even going to be relevant in this organization
[00:12:46] Paul Roetzer: 1, 2, 3 years from now, or even relevant in my industry or my career path. And so I think the biggest challenge we all face in any industry, and again, whether it's marketing and sales service or it's outside into operations, product and HR and finance, it's that the leaders who are going to have to make these decisions and integrate this technology may not understand the underlying technology that's enabling all this.
[00:13:10] Paul Roetzer: So that gets into, like, they talk about emotion, AI. Explaining to a C-level person that AI can actually understand emotions and whether it's using like facial recognition or the tone of writing, actually get a sense of people's emotions and you can actually use that data to evolve the way you manage.
[00:13:29] Paul Roetzer: That's a really abstract thing and it's a very hard thing for someone to accept who has built their career on instinct and education base and knowledge and experience and the things that made them the leaders they are to accept that AI is going to change the way they lead and in some cases do the thing
[00:13:53] Paul Roetzer: they've always done, better than they do. This is a really silly example maybe, but like look at the NFL or the NBA, professional sports and the role, what they call analytics, but basically machine learning, what they're calling advanced analytics is actually forms of AI.
[00:14:11] Paul Roetzer: And you have some teams like our Cleveland Browns who are all in, like analytics rules the day for the Browns, the decisions they make and everything. Then you have other teams where you have coaches and general managers who refuse to allow the AI to play a major role in the decisions they make.
[00:14:28] Paul Roetzer: They won't accept that maybe it's actually better at making predictions about what should be done and when. And if you carry that into a business in all aspects of it, it's really hard to accept that AI could potentially play a really valuable role in making all of these predictions and helping you.
[00:14:46] Paul Roetzer: So I see the adoption challenges often being the people. The tech is getting really good. It's going to keep getting better at an exponential pace, and I think over time the lack of adoption is going to be more driven by people refusing to accept AI's abilities in these areas that they consider uniquely human.
[00:15:06] Mike Kaput: That's a really good point, and I actually think it does relate interestingly to our next topic. So on one hand we have companies that are often very large enterprises trying to adapt and adopt AI. But on the other hand, we have some news out of the startup and entrepreneur community from OpenAI.
[00:15:27] Mike Kaput: So OpenAI, the creators of GPT-3 and DALL-E 2, which we have talked about at length, these generative AI tools. They actually just launched a program to fund and support founders who are creating transformative AI companies, and the program is called Converge. And according to OpenAI, it is a highly selective five week program for exceptional engineers, designers, researchers, and product builders who are using AI to reimagine products and industries.
[00:15:59] Mike Kaput: And so, as a part of this program, the entrepreneurs and founders receive a 1 million equity investment from OpenAI startup. They get early access to OpenAI models that they're building, as well as a host of other benefits, including office hours, workshops and some mentorship. And so OpenAI said the motivation for this is the belief that powerful AI systems will spark a cambrin explosion of new products, services, and applications.
[00:16:30] Mike Kaput: So, We have this whole other world of companies and organizations and people who are highly motivated to start disrupting existing businesses using artificial intelligence. So Paul, you're a successful entrepreneur. You founded multiple companies. What, why are hungry entrepreneurs so important in the world of ai?
[00:16:53] Mike Kaput: And what do we anticipate the effects of things like this to
[00:16:56] Paul Roetzer: be? This one was really interesting to me. There. There's a couple things that jumped out to me early on. The idea. So OpenAI again, remind, we talk about OpenAI all the time, but if you're not familiar, again, it's G PT three, which powers a lot of the AI writing tools as Mike was talking about, DALL-E, the image generation technology, they have a, a co-pilot that does coding.
[00:17:19] Paul Roetzer: There's, they're all kind of playing the same space, whether it's stability or runway, mls, another one we love, doing really cool stuff. OpenAI is building the tools that other people can build on top of. And so what they're doing with this program is basically saying, Come, come build on our platform.
[00:17:36] Paul Roetzer: We will fund you. And I think they're doing 1 million for at a 10 million valuation. So I think they're taking, that's pre post money, but they're taking roughly 10% of the company for you to build on top of their technology. The thing that jumps out to me here is, The great unknown I struggle with personally on where we're going to go with generative AI is, and I think we talked about this last week.
[00:17:58] Paul Roetzer: If I'm building my technology today on GPT-3, the most advanced language model that I'm aware of that you can build on right now, I mean, Google has LaMDA, but you can't build on it. Cohere is interesting out of Toronto, that's one we talk about a lot. You can build on that one. You can get access to their APIs, but GPT-3 is certainly the most popular at the moment.
[00:18:19] Paul Roetzer: But I always wonder like, well, what happens when GPT-4 comes out? And so if I'm building my technology on top of that, you know, where do I go? Is my software company just not relevant all of a sudden or am I going to have early access to evolve my product to be on top of GPT-4? And so, looking at this program, if I was thinking about starting a language company, I would certainly be exploring this program because the idea of having inside access and being part of this exclusive program where you could basically see where the future's going six to 12 months out is extremely enticing as an entrepreneur.
[00:18:58] Paul Roetzer: And I think a lot of the innovation we're going to see, this is why it gets really hard as marketers, or if you're an agency or whatever and you're trying to predict what should we build on, which tools should we use? I look at this, I'm like, wow. Like if you build your entire solution suite, your marketing strategy or your agency service, whatever, around a GPT-3 powered tool, what's to say that six months from now some company's not going to come out of this converge program building on the next generation language technology with a better solution than what your whole workflow's built on now? So, it's a really hard problem right now, and I haven't seen the answer yet. You see a lot of clues around this. People working within OpenAI or like, I mean, Sam Altman, the CEO, has alluded to what comes next and other people in the space.
[00:19:51] Paul Roetzer: So I think there's just a lot of unknowns around where we're going with these generative models. And then whether the play is to build on top of one, like OpenAI where you just get access to their APIs to build, or if it's to build on an open model like Stability AI who will probably come out with a language model in the not too distant future, and there's this open source.
[00:20:10] Paul Roetzer: So now I have more flexibility to build or do I build my own language model? If I'm an engineer at LaMDA Google on their language model, do I just leave and build my own language model? To me, this is why it's so important that we keep these topics in the conversation every week, because I feel like every week it's evolving.
[00:20:32] Paul Roetzer: And these are hard decisions, not just if you're going to build software, but if you're going to use software someone else is building. The tech they're building on top of is changing so quickly and the players in the space are changing so quickly. It's just a really challenging time to figure this stuff out.
[00:20:51] Mike Kaput: That's a really good point. You know, as people do, as entrepreneurs do figure out some of these models, and as the cohort in these types of OpenAI incubation programs start coming to maturity, I'm curious about our kind of focus in our industry, marketing and sales, what do you see as the biggest opportunities today for entrepreneurs with this kind of funding and talent to disrupt using
[00:21:20] Paul Roetzer: AI?
[00:21:22] Paul Roetzer: I mean, there isn't a thing we do that involves language, whether it's writing or editing, that isn't going to be changed completely. I put this as a public comment, so I'll say it, Dharmesh at HubSpot tweeted something the other day about generative AI and I replied and said, "Well, I'm fascinated to see what HubSpot does with language and image generation, because to my knowledge it's not in there yet."
[00:21:50] Paul Roetzer: Like in the HubSpot platform, we've been HubSpot users for 15 years. I think HubSpot's first agency we built. And I look at something like HubSpot and I think, man, they better figure that out fast. I know they're paying attention. I'm pretty sure there were investor in Jasper's round that just happened.
[00:22:10] Paul Roetzer: I think that was in the press release. That's not inside knowledge. And I just think like we use HubSpot at the institute. It is the core component of our tech platform from our CRM to our blogging tool to our CMS. Like everything's on HubSpot and there isn't any of this stuff in there. They have to embed it, right? We're using all these third party tools that aren't, I don't even think, in their ecosystem right now. And it's almost like we talked about last week with Adobe infusing image generation, like you have to. And so I just look at the disruption that's going to happen at the platform level.
[00:22:48] Paul Roetzer: And it's probably just going to be integration through partners I'm guessing, I don't think they would build their own models and stuff. But if I'm in HubSpot and I want to be writing an email like I should be able to have that infused into it and have my images selected and all this stuff, and be recommending, like you were using in the Deloitte example, like the tone of my email and, the length.
[00:23:10] Paul Roetzer: And they do some of that stuff that's at HubSpot and other CRMs, but nowhere near where it's going to have to be in six to 12 months. So I think for entrepreneurs it is a wide open space right now. There's so much opportunity to do this stuff, but it's going to move so quickly that speed is going to be critical.
[00:23:32] Paul Roetzer: And a big challenge is it's in a time when the economy is really, really tough. Like it's hard to raise money. It's hard if you're running a startup to pivot and be able to get follow on funding right now. So it's just such a wild time. I mean, it really is like, I don't know if sometimes people think like we have all the answers because we're running an institute about this stuff like this is, open the curtain, we don't know where this is going. We're just paying attention and trying to figure it out and trying to give people clues. Whether you're listening, you want to build something or you're listening, you want to build on something, like you want to build your marketing writing team around something.
[00:24:09] Paul Roetzer: There's a lot of unknowns right now, but hopefully we can help people, maybe connect some dots and just go in with eyes wide open about what they're going to be doing next.
[00:24:18] Mike Kaput: Yeah, and I think you made a really good point there too, is some of the leading companies don't have it necessarily fully figured out either.
[00:24:29] Mike Kaput: I mean, today, like you mentioned, we use HubSpot and I and our team are trying to use these third party tools to essentially cobble together what features would be fantastic to have in the platform and seem like a no-brainer. I mean, we're using AI to assist in certain cases with certain inspiration ideas for content for some shorter form content to get us started.
[00:24:55] Mike Kaput: Obviously subject lines, titles, headlines, image generation. I mean, we're starting to use DALL-E 2 more and more for blog post images, and you just sit back and think like, how much easier would my job be if that was all a button in HubSpot,
[00:25:10] Paul Roetzer: right? And I think that's the point of their ecosystem is they realized long ago they couldn't innovate at the pace they needed to with all the things they would build in.
[00:25:19] Paul Roetzer: So they built an open ecosystem that lets other people build smarter features. And I think that ecosystem has to get smarter. You know, last time we dug in was probably 12 to 18 months ago, I think we wrote a blog post on ways to make HubSpot smarter, or the ways it was using AI. And at the time there was like 10 native features and then when we looked at the ecosystem, there was only like 10 vendors we recognized as being AI companies.
[00:25:48] Paul Roetzer: I'm sure that has changed in the last 12 months, but I would think that as you move forward, that's going to become required. I've said this before, like I don't know how a software company exists in three to five years that isn't AI powered. Like all these social tools and ad tools and writing tools and SEO tools and email tools.
[00:26:12] Paul Roetzer: If the software doesn't have AI infused into it, I just don't even know how it's going to be relevant. So I think that that would play out in the HubSpot's ecosystem as well. I think every one of those software companies has to figure out AI real fast or somebody else will. Somebody's just going to build a smarter version of their tool.
[00:26:31] Mike Kaput: Yeah, absolutely. And you know, speaking of kind of innovating and moving fast in the world of AI, our final kind of main topic here that I wanted to draw some attention to is in the last week alone, Google revealed, I think well over a dozen AI projects that it's been working on in various and very widely different areas.
[00:26:59] Mike Kaput: And these projects give us a pretty cool glimpse of the near future of artificial intelligence. And so, these announcements break down into two broad categories. There's what we might call AI for social issues, social good, and then there's some more generative AI experimental projects. So on the social good front, Google revealed AI that does a pretty breathtaking number of things that they've just been working on in addition to all the other things they do as a technology company. So they revealed AI for wildlife tracking, for flood forecasting, they have an AI powered maternal health feature app that they're working on where they can actually assess certain aspects of health of a fetus in the womb.
[00:27:49] Mike Kaput: There's an AI app that is helping people detect early site conditions that if they're untreated, could lead to blindness and use just a smartphone and take a picture of your eyes to do that. And there's also an AI model that speaks the world's 1000 most spoken languages. In addition to that, they also revealed that an event that they have now self coding robots, so these are robots that can autonomously generate new code.
[00:28:19] Mike Kaput: So Axios, which reported on this, a reporter, Illustrated what this looks like. They said Google told a robot hovering over three plastic bowls. The bowls were red, blue, and green, and each bowl had three pieces of candy, Skittles, M&Ms, and Reeses. And the reporter told the robot that they like M&Ms and that their bowl was blue and the robot placed the correct candy in the right bowl, even though it wasn't directly programmed or told to do that.
[00:28:49] Mike Kaput: So essentially on the fly, a text prompt, a natural language prompt, the machine was able to interpret that and essentially rewrite its rules to perform an outcome. And last but not least, they also released AI for fiction writing, which several professional writers are experimenting with to generate
[00:29:12] Mike Kaput: realistic fiction stories. So as a final part of this reveal, Sundar Pichai, the CEO, said, "AI is the most profound technology we are working on yet. These are still the early days." So this is a crazy set of developments and very wide ranging. So I wanted to ask you kind of a two part question about this. So first, do any of these kind of jump out to you as particularly interesting or significant?
[00:29:40] Mike Kaput: And second: could you maybe connect the dots on the bigger picture here for us? You know, these are all really cool ideas, but what do they really say about Google's broader commitment to AI and AI innovation?
[00:29:55] Paul Roetzer: Yeah, it's a lot. And so this would fall into the realm of, you know, I have this AI list on Twitter.
[00:30:03] Paul Roetzer: It's, I think I've spoken before about this. The main way I use Twitter is I just build curated lists and I try and avoid the rest of the noise, and I just follow my list. My news, sports, science, AI are kind of my main ones. And then I get alerts from very specific entrepreneurs, so Sundar is actually one of the people I would get alerts from when he tweets.
[00:30:24] Paul Roetzer: And so when I saw this announcement, I thought, Okay, like, well, what of this is actually important? And so I just go look at Sundar's tweets. And so he led with the language model. So I will default to like, I assume that's extremely important. And then I dug in a little bit. And that one is kind of cool because he actually linked to an article that was, "3 Ways AI is Scaling Helpful Technologies Worldwide."
[00:30:52] Paul Roetzer: And they talk about the language one. I knew the complexity of the problem. I didn't appreciate the scale of that complexity, like how big of a challenge this really is. It says, "we're announcing a thousand languages initiative, an ambitious commitment to build an AI model that will support the thousand most spoken languages, bringing greater inclusion to billions of people in marginalized communities all around the world.
[00:31:17] Paul Roetzer: This will be a many years undertaking. Some may even call it a moonshot. But we have a path to do it, basically. So that's massive. And, the challenge is, and I've read about this before, but basically there are a lot of languages that lack training data, like English, we have infinite amounts of training data, but you start getting into languages that are the lower level of that top 1000, there just isn't that much data to train it.
[00:31:43] Paul Roetzer: So you have to build AI models that enable you to learn with less data. So Google and Meta have been working on stuff, and I'm sure there's others that have been working on this as well. So that one jumps out to me as, as important for society. My ears always perk up when I hear things related to LaMDA.
[00:32:02] Paul Roetzer: Like anything where they're sharing more of the LaMDA model and showing more of what it's capable of doing. I think we talked on a recent episode about their AI test kitchen where you can actually play with LaMDA and have it finish ideas for you, tell it where you are and it kind of helps tell stories.
[00:32:18] Paul Roetzer: So it was top of mind for me cause our friend Joe Pulizzi, who had founded and sold Content Marketing Institute, had just, in his newsletter last week, started experimenting with language technology. And so he was showing its ability to write in his newsletter. And he said something like, It's not ready to write fiction yet.
[00:32:40] Paul Roetzer: And ironically, that was the day this news came out. And one of the things Google announced was this word craft writers tool that builds on LaMDA that you had talked about where they work with these fictional writers to actually let these people experiment with LaMDA, the language model in new ways to tell these fictitious stories.
[00:32:59] Paul Roetzer: And so anytime you start to see these language models going beyond just finishing up, predicting the next word, or finishing a sentence, and you start to look at these more creative, impactful ways that these language models are learning and being able to actually write and assist us. I mean, I could read about that stuff all day long because I think it's the great unknown right now.
[00:33:25] Paul Roetzer: It's how quick is this language technology going to improve? How is it going to affect writing, journalism, business, marketing, sales? And there aren't great answers. And I think a lot of the research labs have very bullish viewpoints on how big of an impact language models will have and how quickly it's going to happen.
[00:33:52] Paul Roetzer: And I don't know that the rest of us comprehend that. And I've said before, I feel like their confidence level in how quickly this is going to happen seems to be rising. And I just continue to feel like we're just not ready. People aren't aware of what's about to happen. And so yeah, the language stuff, the LaMDA stuff is always intriguing to me.
[00:34:19] Paul Roetzer: And then at that high level, you mentioned Sundar's quote of AI's most profound technology we're working on. We use that quote in like, every presentation I give, it's in every keynote I do, about it's the most important thing humanity's ever worked on. More profound electricity or fire. It's just reiterating the fact that Google and the other major tech companies are racing forward building insanely smart technology that is going to start trickling into all aspects of our lives from amazing social good things to very practical applications in writing. And then I saw their Test Kitchen, they're going to open their image generation model as one of the next demos in the AI Test Kitchen, which you can register to get access to that app.
[00:35:01] Paul Roetzer: It's really cool. It's free. And so they're going to let you start playing around with their version of text to image. So, there's a lot to unpack. Honestly, it was a lot of stuff announced at one time.
[00:35:15] Mike Kaput: Yeah. I think that's what also struck me and would be smart for our audience to remember, is that while all of these projects are extremely different and in different areas of science, technology, and business, there is a fair amount sometimes of overlap with AI's ability to be used in these kind of other domains that aren't necessarily the one in which it was built in and trained in. And I think that's when you say things like language models, people don't realize just how crazy things are about to get. Language models aren't just about saying, "Oh my gosh, AI can write fiction."
[00:36:00] Mike Kaput: Where people don't realize how crazy language AI is going to be because it'll write fiction books. Yeah, that'll happen. But also the ability for it to understand language is helping robots respond to commands. It's helping image generators understand the relationship between objects, this kind of technology
[00:36:20] Mike Kaput: has wider applications. If they crack the code on language, they crack the code on a lot of other things. I
[00:36:26] Paul Roetzer: think as well. One other thing that jumped out to me is the connecting the dots. I mean, for years when we started researching artificial intelligence, going back to like 2011, there weren't business case studies about it.
[00:36:40] Paul Roetzer: There weren't talks about use cases in marketing and sales. So we spent years just reading about innovations, coming out of research labs and academia and trying to connect the dots of, well, what's going to happen when that comes to marketing and sales? And I see aspects of this in a news dump, like what Google did where like I see the wildfire tracking example.
[00:37:02] Paul Roetzer: Okay. If I'm an insurance carrier, if I had no idea that Google had wildfire tracking capabilities, I would certainly be on the phone right now trying to figure out how do we infuse that into our underwriting? How do we mitigate risk? How do we do real time claims when we can predict that wildfires are going to impact people we insure?
[00:37:23] Paul Roetzer: Same with flood forecasting. Flood's a massive element of the insurance industry and real estate industry. But are those executives paying attention to Google? I don't know. But if I worked in those industries, I would absolutely be trying to connect dots and say, Well, what does this mean to us?
[00:37:40] Paul Roetzer: How's this going to affect what we do? And what are the opportunities here? How can we move and get access to this kind of information quicker and infuse it into our product and our business? And I think that's the beauty of AI right now is there is a chance to reinvent almost every business and every industry.
[00:37:56] Paul Roetzer: And it's going to be driven by the major research labs and tech companies. There's going to be these startups like OpenAI with their converge program, you're going to have people pop up and say, Well, I'm going to build on your language model for the insurance industry or for the healthcare whatever. And so you're going to get innovation that'll happen through these small startups.
[00:38:15] Paul Roetzer: But a lot of this is just, I mean, Google has massive research teams and R&D budgets for AI and the stuff they're going to pump out is going to be so hard to comprehend. We just need more business leaders paying attention and connecting the dots, because it's going to move fast.
[00:38:33] Mike Kaput: Well said. All right. I want to throw some rapid fire items at you before we wrap up here.
[00:38:38] Mike Kaput: First up. So, I feel like we've gone long enough, a world record, without mentioning what's going on at Twitter, probably like 36 minutes or so. It's no secret that today things are somewhat I don't know, chaotic or interesting, depending on your perspective, at Twitter, under Elon Musk's new ownership. What I want to focus on here is that as part of this turmoil, I saw a tweet from Hayden Field who is a reporter at Morning Brew's Emerging Tech Brew newsletter, who tweeted a few days ago that members of Twitter's ethical AI team had been laid off.
[00:39:16] Mike Kaput: So this is Twitter's ML Ethics, Transparency, and Accountability Team, which is the acronym META, has been responsible for studying algorithmic decision making, building AI fairness tools and more. Now, we don't have any details on the long term AI ethics plans of Twitter, accountability, anything like that.
[00:39:36] Mike Kaput: Obviously, a business is going through a pretty serious transformation, but do you have any initial thoughts on some of the potential short term items or pitfalls that might result as a result of getting rid of your AI ethics team?
[00:39:53] Paul Roetzer: So many thoughts. Yeah. Like you said, we're not experts on what Elon's doing or what the people inside that he's listening to are doing.
[00:40:04] Paul Roetzer: I believe there are still people left there that are thinking about these things. They probably don't have the authority they previously had. I actually read one last night. What you're seeing on Twitter, which is fascinating, is a lot of former Twitter employees who are now telling stories they couldn't tell before.
[00:40:21] Paul Roetzer: And the one was kind of wild to me. I'll have to find it and we'll put it in the show notes. He was working on mobile data, and this actually is very relevant to AI. So they said one of their advertising clients was a bank and they wanted very detailed mobile data...no, it was, it was a telecommunications company.
[00:40:47] Paul Roetzer: Or bank? No, it was, Yeah, I think they were going through the telecommunication company. So the bank wanted to know when an individual consumer went into a competing bank then they wanted to know when do they leave their home? What path do they take? They wanted all this insane mobile data because Twitter, in case you don't know, tracked all that data.
[00:41:08] Paul Roetzer: They had all that data anonymized. They have everything. Everything you look at, all the delays, when you slow down, and your thread, like all that stuff is used. And so they would sell that data to advertisers basically, or to partners. And so this guy was like, no way. Like I will leave the company.
[00:41:24] Paul Roetzer: This is not happening. There's no way we're building this solution. And he actually emailed Jack Dorsey, who was still involved in Twitter at the time, and is like, I just want you to know what's happening. I refuse to do this. And Jack vetoed it, he is like, No way. That sounds way off. There's no way we're doing that.
[00:41:39] Paul Roetzer: And that dude's like, but I'm not there anymore. Jack's not there anymore. The ethics, transparency, and accountability team's not there anymore. Like what is to stop this stuff from happening? My main takeaway is more macro level. We've talked about this before. AI gives you superpowers. It gives you insane access to data, Insane ability to make predictions about people's behavior.
[00:42:03] Paul Roetzer: Insane access to intimate knowledge of their lives and the things they do and believe and where they go. And there has to be strong ethical teams and policies in place within organizations that are actually adhered to. And if you remove those as we've seen from Google, like we're giving Google all this love right now, it's awesome stuff.
[00:42:22] Paul Roetzer: Google's had their own problems on an ethics standpoint as well. All of them do. And so I would very much get worried because Twitter plays such a critical role in our society right now, that it very much worries me that there appears to be far less guardrails in place than there were a week ago.
[00:42:43] Paul Roetzer: And from all outward signs, it doesn't appear that the current leadership structure has the same feelings on the importance of that to Twitter. So I don't know that it's going to get better anytime soon, but I worry overall anytime it appears that product decisions are being made with little ethical guidelines or thought.
[00:43:12] Mike Kaput: Yeah. Especially when there's very well documented severe revenue pressure on the business, a lot, a lot of pressure to make money in any way possible at the moment.
[00:43:25] Paul Roetzer: Yes. And you don't want competing voices telling you you can't do things that would generate the money you need.
[00:43:30] Paul Roetzer: Mm-hmm. . Yep.
[00:43:33] Mike Kaput: Okay, so the next one comes from website Futurism, which we both read quite a bit, and they have a warning for musicians about AI powered music generators. And the warning is basically, hey, in the last couple years, these AI powered music generators have actually gotten really, really good guys.
[00:43:51] Mike Kaput: So they say, "a new generation of music generators led by a product called Mubert AI, as well as Google, has a product called Audio LM, feel like a different beast. They're less like creative integrations for the music making process and more like creative replacements for musicians themselves, eliminating the need to pay for human labor or pesky royalties altogether.
[00:44:14] Mike Kaput: Now I know neither of us are experts on the music industry. And obviously music is pretty subjective as an art form, but we've talked a lot about AI moving into these creative fields like writing and art. Are we going to see AI just start taking over like every type of creative task? I mean, is there anything AI won't be able to do in the world of creativity?
[00:44:37] Paul Roetzer: I mean, if you'd asked me six months ago, I would've said creativity is still largely human realm. We wrote it in the book. I mean, we addressed human AI and creativity in in the book. And I still stand by what we wrote, which was basically saying AI is creative, just not in the same way as humans.
[00:44:55] Paul Roetzer: Like, it doesn't create based on experience and emotion and memories and like if a musician writes a song, there's things that go into that song. Oftentimes it is the things they've been through and the emotions they've felt. And it has that power of the human experience into it Now, if I just need music for our podcast, like bumper music, beginning and end, do I need a human to do that or can I just say I want upbeat music for a podcast? I would just give it a prompt, like this is what I want the music to be like. And boom, there's your 30-second bumper for whatever it costs, 10 bucks or something.
[00:45:37] Paul Roetzer: I'm sorry that it's going to dramatically impact the news industry. Now, you could have what is happening in image generation where they're trying to figure out ways, like if the AI learned from your original images or it's inspired by, and so in some way we're going to compensate you because your data was used to train the AI.
[00:46:00] Paul Roetzer: I don't see that scaling because there's no artwork ever created, music ever written or played, that wasn't also inspired by some other artist. It's how humans create. We take inspiration from other people and their styles and their creativity, and then we mash up writers. We do it all the time. We're inspired by other writers and other books and so you follow their style and you start. It's how creativity happens. So at the end of the day, the only thing that's going to be missing is the human experience in the creative. And then the question becomes, is their demand for that? And I don't have the answer, and I am by no means saying we should just like accept that musicians are done, artists are done.
[00:46:47] Paul Roetzer: That's not at all what I'm saying. And if you're listening for the first time, my wife is an artist. I care about this stuff. Mike and I are writers. We're not cheerleading for this day to come, but I do wonder what the demand is going to be for human-only generated art of any kind. And again, I don't have the answers, but it's becoming more and more obvious that the AI is getting very, very good at creativity.
[00:47:15] Mike Kaput: Absolutely. And that kind of hits on kind of our last rapid fire topic here. Now, I know we obviously talk a lot about DALL-E 2, OpenAI, rightly so, because they are such critical developments these days in the world of AI. But, a big development did happen with the tool in that OpenAI just announced that DALL-E Two's API is now live.
[00:47:41] Mike Kaput: So basically that means developers can integrate DALL-E directly into apps and products and OpenAI featured a couple early customers already using the API in their products. So this includes Microsoft is using it in a number of ways, but also a company we talked about on a recent podcast called CALA, which is a startup that offers DALL-E 2 powered product design.
[00:48:05] Mike Kaput: So now that the API is out in the wild, what do you think this means for marketing and sales specifically? This is no surprise, but it has happened. Are we about to see an explosion of AI image generation incorporated into existing or new businesses and
[00:48:24] Paul Roetzer: products? Yeah. There won't be an AI writing tool in six, 12 months, that doesn't also have image generation baked into it. So as you're writing, it's just going to just create image, creative image, just going to read the paragraph, create an image to go with it, like give me an image for the headline. It will just be seamlessly integrated into any writing tool.
[00:48:43] Paul Roetzer: I would assume, going back to like a HubSpot, any marketing automation platform, email tool, social tools, anywhere where you're generating language, I have to imagine there's going to be integrations baked right in, whether HubSpot does it, or some third party does it, but you're not going to have to go to DALL-E's website to type in a thing and then download an image and then upload it to HubSpot or whatever, it's just going to be seamless.
[00:49:10] Paul Roetzer: And then you're going to have an explosion of creativity in other industries, like you alluded to product design, we've talked about architecture, we talked about room design. People are just going to build point solutions for verticals. It's just going to be everywhere. So the idea of going to a stock photo site and doing a search for a stock photo is going to feel ancient six to 12 months from now. You're going to forget what it was like to not have images generated for you on demand in whatever platform you're in; it's going to be everywhere. And not just DALL-E. Stability AI is open source; people are going to build on Stability, Midjourney, there's other players, but DALL-E is obviously the big player right now.
[00:49:53] Paul Roetzer: At least it seems like the most common, most famous.
[00:49:56] Mike Kaput: Yeah, and I don't know about you, but we both do, throughout the year, a fair amount of speaking engagements. I'm ready to fork over money to PowerPoint, keynote, or whoever starts incorporating image generation.
[00:50:10] Mike Kaput: Beautiful
[00:50:11] Paul Roetzer: ai.
[00:50:12] Paul Roetzer: Yes.
[00:50:13] Mike Kaput: Make building decks easier.
[00:50:16] Paul Roetzer: Websites, landing pages, build me a landing page that has logos. Yep. I was actually trying to build a logo last night on DALL-E. I'll have to show it to you. It's kind of cool. I don't know what to use it for, but it's awesome. Nice. I love it. I showed it to Cheryl, my wife, and we were laying in bed, but I'm like, "Look at this, isn't it cool?" She goes, "That's kind of cool."
[00:50:33] Paul Roetzer: I don't know what to use it for, but something. That's
[00:50:38] Mike Kaput: awesome. Oh, wow, as always, every week is basically like covering a year of development in the world of AI, which is interesting, but a lot to keep up with. But I think that's all we've got for this week/this AI year, like dog years, I guess.
[00:50:55] Mike Kaput: So thanks for your input, and for your time and for your insight on all these awesome
[00:51:01] Paul Roetzer: topics. Yeah. And thanks for joining us everyone. These are coming I think we're on a schedule now where these come out every Wednesday. So yeah, make sure to subscribe and give us a follow and a like and keep an eye on our blog, we post summaries of all the topics on the blog throughout the week. So check out the Marketing AI Institute newsletter, subscribe there and you can stay up to date on everything and reach out to Mike and I on LinkedIn. Like I said, we love to hear from people where you're at and your AI understanding and adoption and what technology you're testing. We love to hear from the community and share what we know and learn from you all as well. So thanks for being with us, as always. We'll talk to you again next week.
[00:51:40] 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 marketing ai institute.com. Be sure to subscribe to our weekly newsletter, check out our free monthly webinars, and explore dozens of online courses and professional certifications.
[00:52:02] Until next time, stay curious and explore AI.