Hosts Mike Kaput and Paul Roetzer catch up in another episode of The Marketing AI Show, in a busy week where there was no shortage of AI news. OpenAI stays in the news with ChatGPT team subscriptions rolling out (which Paul signed our team up for) and GPT Store. Rabbit hit the news, and quickly sold out. And there’s much more in the rapid-fire section of the podcast, so be sure to tune in!
Listen or watch below—and see below for show notes and the transcript.
This episode is brought to you by our sponsors:
Many marketers use ChatGPT to create marketing content, but that's just the beginning. When we sat down with the BrandOps team, we were impressed by their complete views of brand marketing performance across channels. Now you can bring BrandOps data into ChatGPT to answer your toughest marketing questions. Use BrandOps data to drive unique AI content based on what works in your industry. Visit brandops.io/marketingaishow to learn more and see BrandOps in action.
Today’s episode is also brought to you by Marketing AI Institute’s AI for Writers Summit, happening virtually on Wednesday, March 6 from 12pm - 4pm Eastern Time.
Following the tremendous success of the inaugural AI for Writers Summit in March 2023, which drew in 4,000 writers, editors, and content marketers, we are excited to present the second edition of the event, featuring expanded topics and even more valuable insights.
During this year’s Summit, you’ll:
The best part? Thanks to our sponsors, there are free ticket options available!
To register, go to AIwritersummit.com
00:03:25 — OpenAI debuts ChatGPT subscription aimed at small teams
00:20:03 — Introducing the GPT Store
00:30:08 — Rabbit sells out 10,000 units of its R1 pocket AI companion in one day
00:46:57 — OpenAI and journalism
00:50:24 — OpenAI in content licensing talks with CNN, Fox and Time
00:51:58 — 94% of Google SGE links are different from organic search results, study finds
00:55:58 — Duolingo Cuts 10% of Contractors as It Uses More AI to Create App Content
00:57:03 — Google removes 17 features from Google Assistant and conducts layoffs
00:58:03 — Generative AI isn’t a home run in the enterprise
01:00:55 — Quora raises $75M from Andreessen Horowitz
OpenAI debuts ChatGPT subscription aimed at small teams
OpenAI unveils ChatGPT Team, a new subscription tier designed for non-enterprise teams of up to 149 members. This innovative offering provides a dedicated workspace, enriched with features like GPT-4 access, the latest DALL·E 3 model for image generation, advanced data analysis, browsing capabilities, and support for image and voice interactions.
Additionally, it offers tools for customizing and sharing GPTs, an admin console, and robust member management. With a firm commitment to privacy, OpenAI assures that models are not trained on team data or conversations, addressing a key concern for business owners.
Priced at $30 per user per month, with a discounted annual rate, ChatGPT Team bridges the gap between ChatGPT Plus and Enterprise, offering an accessible yet powerful AI tool for smaller teams. Our Marketing AI Institute team is set to begin testing, and Paul and Mike discuss.
The GPT Store launches
OpenAI has recently launched the GPT Store, a new feature available to ChatGPT Plus, Team, and Enterprise users. This store is a marketplace where users can find a variety of GPTs (customizable versions of ChatGPT) created by fellow users as well as OpenAI partners. These GPTs require no coding skills to build and offer a range of applications.
Currently, the store showcases an array of GPTs, such as one for finding hiking trails, a research assistant, a code tutor GPT from Khan Academy, and an AI guide to books. OpenAI plans to refresh the store weekly with new and innovative GPTs. Accessing these GPTs is straightforward for anyone with a ChatGPT Plus, Team, or Enterprise account.
Users can log into ChatGPT, click “Explore GPTs,” and navigate directly to the GPT Store. From there, exploring and using these GPTs is just a click away, with the option to add favorites to the sidebar for easy access.
OpenAI has hinted at a forthcoming revenue-sharing program, set to launch in Q1. This program aims to compensate creators based on how much their GPTs are used, promising more details closer to the launch date.
The GPT Store marks a significant step in enhancing the ChatGPT user experience, offering a diverse range of tools and the potential for user-generated content to be rewarded for its impact.
Rabbit R1 launches…and sells out
The Rabbit R1, a new AI hardware device, has become a sensation following its debut at CES. Created by Rabbit, this standalone device features a small touchscreen, a camera, and a scroll wheel/button.
The real draw of the device is how it uses AI, its uniqueness lies in the Rabbit OS, powered by an AI "Large Action Model," which acts as a universal controller for various apps. Similar to Alexa or Google Assistant, the Rabbit R1 can control daily-use apps through simple voice commands.
The AI has been trained to autonomously use existing apps, and users can even teach it new tasks, like removing watermarks in Photoshop. This innovation led to an impressive sale of 10,000 units on its first day, selling out its initial production. Available for pre-order at $199, the Rabbit R1 promises a blend of convenience and futuristic technology.
More information and a dive into some important rapid-fire topics make this a must-listen episode!
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: You're asking people who don't understand the technology and don't have staff that understand the technology and they have no plans. to tell you whether they think it's important or not.
[00:00:08] 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.
[00:00:28] Paul Roetzer: My name is Paul Roetzer. I'm the founder of Marketing AI Institute, and I'm your host.
[00:00:37] Paul Roetzer: Welcome to Episode 79 of the Marketing AI Show. I'm your host, Paul Roetzer, along with my co host, Mike Kaput. Happy Friday, Mike. We are doing this on Friday, January 12th. Happy Friday, Paul. You've got some travel next week, so our usual Monday recording isn't happening. But we will do our best to get through another episode. With some pretty interesting topics, as always, today's episode is brought to us by BrandOps.
[00:01:04] Paul Roetzer: So many marketers use ChatGPT, which we certainly will be talking about on the show today, to create marketing content. But that's just the beginning. When we sat down with the BrandOps team, we were impressed by their complete view of brand marketing performance across channels. Now you can bring BrandOps data into ChatGPT to answer your toughest marketing questions.
[00:01:25] Paul Roetzer: Use BrandOps data to drive unique AI content based on what works in your industry. Visit brandops. io slash marketingaishow to learn more and see BrandOps in action. And This episode is also brought to us by Marketing AI Institute's AI for Writers Summit, which we, I was literally just working on the agenda before we jumped on today.
[00:01:48] Paul Roetzer: We have a really cool, I guess, closing talk for that session, which I won't spill yet. We'll talk about in a future episode, but. That event is March 6th. This is our second, annual AI for Writers Summit. Last year we had over 4, 200 people, I think, registered for the event. It is a free event, so there's no reason not to check it out if you're a writer, content creator, freelancer, editor, manager of writers and editors.
[00:02:17] Paul Roetzer: So, it's going on again, March 6th from noon to 4 p. m. Eastern time. There's an option to get on demand. So if you're in a different time zone, I know we have a lot of listeners kind of in different times. That's a hard time to make. There is an on demand option as well. So we're going to go through, I'm going to do a state of AI in writing.
[00:02:34] Paul Roetzer: Mike's going to do generative AI tools and platforms. We're going to have a talk on intellectual property and copyright with an IP attorney. We're going to get into AI adoption in the enterprise and what it's like to actually integrate an AI writing platform into companies. And then we're going to do this really cool demo thing, which I'm not going to get into right now.
[00:02:53] Paul Roetzer: And then an Ask Me Anything at the end with Mike and I and some other, speakers. So check that out. It's AI Writers Summit. com. Again, it's a free event. There is, there are paid options for a private registration and an on demand, but the live event is free. So check that out. It is coming up again March 6th.
[00:03:13] Paul Roetzer: It is a virtual event. All right, Mike, let's get into our, we should just call this the OpenAI episode, at least the main topics, but, let's get into it.
[00:03:23] Mike Kaput: All right, Paul. So first up, OpenAI is launching a new subscription tier for ChatGPT, and this is specifically for teams. It's called ChatGPT Team, and it provides a dedicated workspace.
[00:03:38] Mike Kaput: for up to 149 people to use ChatGPT. It also includes a bunch of features to manage that team within your ChatGPT instance. So some of the features that come with this new license tier include things like. It comes with GPT 4 access with that 32k context window. It has DALLE 3 for image generation. You get access to advanced data analysis capabilities.
[00:04:06] Mike Kaput: You get access to browsing so your team license of ChatGPT can actually find up to date information on the web. You get image and voice input and output as well as the ability to create customized GPTs and share those GPTs within your workspace. Now, there are also a couple of kind of admin features that are really helpful for anyone who is managing 149 or fewer people with one of these licenses.
[00:04:35] Mike Kaput: There's an admin console, a dedicated workspace, bulk member management, and you can give people different admin roles based on their role in your organization or on your team. Now this last part is what's getting the attention of a lot of business owners and professionals. OpenAI claims that as part of team licenses, it will not train models on team data or on your conversations.
[00:05:01] Mike Kaput: The ChatGPT team license costs. 30 per user per month or 25 per user per month. If you bill it annually, this is a tiny bit more expensive than ChatGPT Plus and less expensive than the current ChatGPT Enterprise. So Paul, first up, how do you see this kind of licensing tier and these capabilities affecting small and medium sized businesses that either already use ChatGPT or have considered dipping their toe in the water
[00:05:34] Paul Roetzer: with AI?
[00:05:34] Paul Roetzer: Yeah, I think, first, we knew this was coming. We just didn't know when. When they launched ChatGPT Enterprise, they did say that there would be a self service option in the near future. And it just ends up, it was like, I guess, a couple months wait or so. So we kind of assumed these would happen. I know there are some people like Allie Miller on LinkedIn, I think was talking about how her team has had this for like 30 days or something.
[00:05:58] Paul Roetzer: So I know people have kind of beta testing this as well. The enterprise has, this actually just came out of Bloomberg article yesterday, I saw this. So we'll put that link in the show notes. Um. ChatGPT Enterprise has 260 companies that are customers totaling 150, 000 registered users at 30 per user per month.
[00:06:20] Paul Roetzer: That's about four and a half million in MRR. So give you a sense of what adoption has been like there, or at least what the selling of the technology doesn't necessarily mean. The adoption within the enterprises has been that high. So I think what's going to happen is you're going to see a massive.
[00:06:39] Paul Roetzer: increase in adoption, like companies that are adding these AI writing platforms through this. So I think it's going to create a rather interesting dynamic in the marketplace for people who are maybe paying for Jasper, Writer, Grammarly. Like it all of a sudden brings ChatGPT to the masses. We're a seven person team.
[00:06:58] Paul Roetzer: I upgraded instantly and the way it works, it's kind of seamless. I was in my personal ChatGPT Plus account, so I already have a paid account. There's an option like upgrade from here, and you just click. You, you upgrade, you agree to, you want a monthly plan or an annual plan. I think it's 25 per user per month if you pay in advance for everybody.
[00:07:20] Paul Roetzer: And then you just put the URLs or the email addresses of the people you want to invite into it. So, they're going to, they make it really simple and then it actually created, a separate, a separate workspace. So now when I log into ChatGPT, I see my personal workspace, which I can click into, or I can click into my team workspace, where in theory all of us will be collaborating.
[00:07:40] Paul Roetzer: So, Ithink that, it's going to, a lot of companies that were sitting on the sidelines, that couldn't get ChatGPT Enterprise because you needed 150 licenses, I think was the minimum. You can now instantly go and get for 30 per user per month, ChatGPT. Now we'll get into some of the other ramifications there, but like the big challenge is going to be nobody has a plan.
[00:08:03] Paul Roetzer: Like nobody, nobody actually knows how to onboard this and they don't provide any documentation like in, you know, standard open AI fashion. There's nothing telling you how to use this. There's. Literally no guides, no onboarding anything. So we're going to be scrambling as companies to try and figure this out.
[00:08:19] Paul Roetzer: I'm sure a whole service industry is going to build around this kind of thing. But, yeah, so that, that was kind of my initial look at it when we, signed up for it ourselves.
[00:08:28] Mike Kaput: Yeah, that leads right into my first kind of big question about this is Our business is ready for this because we know from data We've seen year in year out and our own, you know on the ground evidence that Many many companies are still really behind on basic AI education and training.
[00:08:50] Paul Roetzer: Yeah, not even close. I mean No, nobody's ready For it, like there's a lot of organizations that, you know, have been testing things and using different tools, but to like, if you have a marketing team of five or 10 or 50 or 500, you now have all of these capabilities. You can go in and train your own GPTs.
[00:09:13] Paul Roetzer: You can go in and, you know, apply this stuff across all these key use cases. But there's nothing really guiding you how to do that. And there aren't change management plans in place. Like, even for our team of seven, as I was signing us up, I was like, Oh, do I really need seven licenses yet? Like, I don't have time over the next 30 days.
[00:09:36] Paul Roetzer: to properly establish workflows and processes for the company and to guide everyone how to use this. So in essence, I'm just like turning it on and paying 30 bucks a month. And we may have licenses to sit there and not even be utilized by some members of the team because we have no internal systems. We obviously, are probably more capable than many smaller organizations to put those workflows and processes in place, given what we do for a living.
[00:10:04] Paul Roetzer: But even then, I sit there and it's like, wow, I don't have time. It's like, 30 days out, maybe I could do that. So, maybe I'll just turn it on for like me, Mike, and Cathy initially, and we'll start experimenting and keep a journal of like how we're using it. But like, more just experimentation than integrating this into our workflows.
[00:10:23] Paul Roetzer: So I, yeah, I think there's just a lot of, It's. It's going to accelerate companies that go buy tools and platforms, but it's also going to accelerate an ecosystem of education, training, and services to make this possible. So I mean, I go back again to our early days when I was, you know, when I owned PR 2020 and you and I were working there together when we were doing this for HubSpot, you know, HubSpot was, we were their first partner back in 2007.
[00:10:51] Paul Roetzer: And part of the goal of building out the agency ecosystem at that time was to help HubSpot customers get value out of the platform. And so I don't see OpenAI building that kind of partner program in the near, near term. They're very focused on developers, not on like service companies, but it almost needs it.
[00:11:10] Paul Roetzer: Like if they really want to scale ChatGPT team. to hundreds of thousands of users or millions of users, which I would imagine they envision as possible. They're probably going to need an ecosystem of certified partners who, who can go in. So if If I'm a team of like 50 marketers and I go get this, who helps me do it?
[00:11:29] Paul Roetzer: Who's going to implement, help me implement this. And so if you have like certified partners, agencies, people like that, you can go to and say, okay, can you help us set up an onboard with this, help us create our first GPTs, help us develop, develop our workflows. You're going to need an ecosystem of those people.
[00:11:44] Paul Roetzer: And that does not exist yet. And again, we, we talk with a lot of agencies. I think this is going back to opportunities for marketing agencies. This would be a service area I would seriously look at, to, to start like implementing some support here because there's going to be a lot of companies that need guidance on how to do this the right way.
[00:12:03] Paul Roetzer: So how does
[00:12:04] Mike Kaput: the release of ChatGPT Team change the competitive landscape in AI? Like, what happens with the offerings that Google and Microsoft, as being AI leaders, are rolling out in their existing apps? But also importantly, what happens to these third party AI tools that are competing for this segment of customers?
[00:12:25] Paul Roetzer: I think it gets a lot harder to differentiate. Like you gotta, I would guess those third parties have to really lean into the things that make them unique, that OpenAI doesn't have. But you also probably have to build a matrix of like, well, what are the things that differentiate us today that they could easily add?
[00:12:43] Paul Roetzer: So it's different right now, like say brand style guidelines, like they don't have a standard to do it, but what's stopping someone from building a GPT to build brand style guidelines and just Train everything on that. So I think those companies have to seriously think about what is defensible within their current offering or their future product roadmap.
[00:13:02] Paul Roetzer: But as a buyer, I can tell you like we pay for some of those other tools and I bought this the second. was available in my platform and I will probably dedicate more company resources to evaluating ChatGPT team extensively because of all the capabilities that are built in and the things we're familiar with, as more of a, well, can this become the primary tool?
[00:13:29] Paul Roetzer: But again, like, I don't think there's an answer to that. And what you really need in your company is a culture and a system of experimentation. So the way Mike and I would do something like this and the way we've been doing this is, okay, we need to transcribe the podcast and we need to do a summarization of the transcription.
[00:13:49] Paul Roetzer: And so Mike, which, like, what are we using now? We've used a script, we've used Claude, we've used GPT 4, I don't know if there's any other ones we've, we've tried, but I don't know at any point in the next, at least 12 months, we're going to say, okay, cool. Like ChatGPT is it like, we don't need any of these other tools anymore because it's this game of like constantly one upping each other.
[00:14:08] Paul Roetzer: And so there's still a chance. I think that by March or April, Google BARD is going to be phenomenal. Like when Gemini gets really built in and they fine tune Gemini, you know, get it working better. I'm not going to stop testing Gemini and Google. We're not going to stop playing around with Anthropic.
[00:14:25] Paul Roetzer: We're not going to, you know, I think we use Writer and Jasper. I think we have licenses for both of them. And we, we play around with these tools all the time. And so I think that that's probably what a lot of organizations need to do is have a system to regularly experiment with your core use cases. So if it's summarization and transcription.
[00:14:44] Paul Roetzer: And drafting and outlining and whatever those, those key things you're doing are, customer support messaging, have a system where every three months you're taking like the three or four or five potential partners there, and you're constantly reeval, reevaluating, like if an improvement has been made in Google, let's go try Google again for this use case, because my guess is it's going to be a moving target.
[00:15:06] Paul Roetzer: That all being said, GPT 4 is still the most powerful model in the world. And there's a, it's a, it's a safer bet to just kind of assume that that's probably going to be best in most use cases. So again, I, and you know, Ilike a lot of these companies in the space, but Ithink. Given OpenAI's reputation and the tools that they've built so far, like, I think a lot of companies are going to go with ChatGPT as their primary platform.
[00:15:39] Paul Roetzer: Yeah, and I would
[00:15:40] Mike Kaput: argue it's probably more beneficial for many companies. to be exploring and expanding use cases versus getting too caught up on, oh, which tool should we use over another? Because if you find even a handful of use cases that you can do reasonably well, suddenly it doesn't matter that you're spending a hundred bucks a month on possibly overlapping technology because you've saved.
[00:16:04] Mike Kaput: So much time already, so much budget already, even for small stuff. So definitely agree with that point that focus on the use case first and test out different tools regularly. Well, how
[00:16:17] Paul Roetzer: do you, how do you think about it? Because again, like I'm the CEO, I see this, I go buy it. Like, I don't, I mean, I control the budget.
[00:16:24] Paul Roetzer: So it's like, fine, cool. Yeah. Seven licenses a month, 210 a month, whatever. But then you, as the chief content officer, get an alert saying you've been invited to the Marketing AI Institute. ChatGPT team workspace. It's like, okay, like, what am I supposed to do with that? Like I'm already using Cloud, I'm already using Descriptor.
[00:16:40] Paul Roetzer: Right. And so I was wondering like how marketers are, are going to feel like, again, we didn't stop and say, Hey, we're going to spend the next 30 days figuring out how to use this, and then we're going to slowly onboard. It's like, no, we live in a world of like, all right, let's just get it. Let's try it. It's just 30 bucks per user.
[00:16:56] Paul Roetzer: Like whatever, we'll learn something if nothing else for the next few months. So, I mean, how, when you see a tool like this emerge, do you think about. Your own workflows and like, do you start to say, all right, well, let me next week, I'll start testing this. Like, how are you thinking about it? Yeah,
[00:17:12] Mike Kaput: I think a couple of things.
[00:17:13] Mike Kaput: So one, I think, I suspect based on my encounters, doing talks and doing workshops with some of these organizations, trying to apply this technology, that there are a lot of people in isolation using these tools, weather. They're allowed to or not. We know that they're already starting to use them. They're probably using them in their work, even if they're not allowed to.
[00:17:36] Mike Kaput: So I think the team license, when I get that notification, on one hand, it's very positive. And I'd say, okay, you know, if I'm at an enterprise, I say, oh, great. Okay. Or rather a smaller team with this license. I say, great. Our company is Leaning into AI and embracing ChatGPT. I can't wait to not have to pay for this myself or to get access to features beyond the free version.
[00:17:56] Mike Kaput: Awesome. But the now what question is a pretty big one because I think this could be chaos if you turn this on without ruling it out. It's fine at our organization because we're all doing all this anyway. We're just unifying. What we're doing, and we do need more of, say, a formal process around it, maybe for some of the newer people that haven't been using the technology as much in our particular business.
[00:18:19] Mike Kaput: But if you're an organization that's like 50 people you add to this, and you haven't used much of this before, I think you need a rollout plan.
[00:18:27] Paul Roetzer: And education. Like, I just think about if you don't have the generative AI policies in place, they don't know what they're allowed to put in there or not. They probably don't know how to use GPTs or even some case of what a GPT is.
[00:18:40] Paul Roetzer: They don't know what the GPT store is, which we'll talk about next. Like they have no concept of a lot of these things. And it's like, Oh, okay. Like I can go in and write some emails and draft some articles and create some outlines and things like that. But. They don't know what advanced data analysis is.
[00:18:55] Paul Roetzer: Like, what do we do with there? Can I drop last month's performance data in there and see what happens? Like, is the data safe? Is the CIO going to show up and say, what are you doing? You just gave these people access to this and now they can put whatever they want in there. So yeah, I think that. At the end of the day, our main focus is AI literacy.
[00:19:14] Paul Roetzer: Like, everything we think about, like the North Star of what we're trying to do is we believe that literacy is essential in every organization. That people have to understand the fundamentals of AI, but they also have to be educated on the tools themselves, how to use them, what is safe use, responsible use, what fits within the policies and guidelines of the organization.
[00:19:37] Paul Roetzer: And that's the thing where we just, we know from our own research that the vast majority of companies have no AI education and training in place. I don't remember what the number was, but I think it was like somewhere around like 70 percent have nothing, not even started working on training. So it's, that's the big thing for, for me that we really need to move forward as an industry, as a business world, is AI literacy.
[00:20:02] Mike Kaput: So some more OpenAI news that's been happening this week. OpenAI, like we teased in our last episode, has officially launched the GPT Store. And this is available if you are a ChatGPT Plus team or enterprise user. The GPT Store features GPTs that are built by different users. and by OpenAI Partners. Now, as a reminder, GPTs are customizable versions of ChatGPT that you can build yourself without any code.
[00:20:33] Mike Kaput: You can just use text prompts to build essentially a customized, narrowly focused. Version of ChatGPT that does various different things depending on what you'd like them to achieve. Now OpenAI says that it will actually feature new GPTs each week in the store. And as of Friday, January 12th, when we're recording this, Some of those featured GPTs included things like a GPT that helps you find hiking trails, a research assistant, a code tutor from Khan Academy, and a GPT that serves as an AI powered guide to books you might want to read.
[00:21:10] Mike Kaput: So lots and lots of different use cases, not just within marketing or business. OpenAI also teased a revenue sharing program coming sometime in Q1, and this will pay builders. based on user engagement with their GPTs. That's all OpenAI is saying at the moment. They said they'll offer some more details on that program once it gets closer to launch.
[00:21:33] Mike Kaput: But right now, anyone with a Plus Team or Enterprise account with ChatGPT can log in, click Explore GPTs, and go right to the GPT store, which is pretty simple at the moment. And from there, you simply click on a GPT, something that interests you and you can immediately start trying it out. You don't have to download anything or do anything else except click.
[00:21:55] Mike Kaput: And if you like it, you can simply choose to keep it in your sidebar to use all the time. So Paul, I played around with this a little bit. What were your initial impressions of the GPT store as it stands, you know, a couple days after
[00:22:08] Paul Roetzer: launch? I definitely think long term it's, it's going to help with adoption and utilization.
[00:22:14] Paul Roetzer: You know, I think, especially within ChatGPT team, like, let's say instead of me just getting the seven licenses and sending out an email saying, okay, everybody, we have this, we'll figure out in the next 30 days what we're going to do. Let's say instead the rollout was. Ibought a license and maybe I got one for you and one for Tracy, our COO.
[00:22:33] Paul Roetzer: We spent 30 days. We analyzed our top 5 use cases. We built custom GPTs for those specific use cases. And then we introduced it to the team with an educational relative. Here's what a GPT is. Here are the 5 we've pre built. These account for 80 percent of the use cases you're going to be looking at over the next 90 days.
[00:22:52] Paul Roetzer: And like we actually showed them how to use things. So I think the The custom GPTs is actually like the most interesting thing at the moment. That being said, OpenAI, claims that over, and I would imagine they know, more than 3 million GPTs have been built since they first debuted it. Well, like November, I think that was end of November or something like that early December.
[00:23:15] Paul Roetzer: So 3 million of them. When I go into the GPT store, just if you haven't been in there, there are a number of categories. There's TopPix, which is curated by OpenAI. There's DALLE, so all specific to image. There's a writing category, productivity, research and analysis, programming, education and lifestyle. I, because I was planning for AI for Writers Summit, I was like, Oh, I'm going to see what is in the writer category.
[00:23:41] Paul Roetzer: My initial reaction is, well, this is kind of a bunch of junk. Like, it, it, it, it's hard to know what's valuable. I don't know if they're going to have a rating system, a star system. Like, I don't know how they'll build this out, but it says by people. So there's. The number three writing tool, one is Ibuy prompts, but that's like the maker of it.
[00:24:04] Paul Roetzer: So it doesn't like give me immediate credibility, like a strong feeling of credibility. The number five tool is Humanizer Pro. This is the description. Writes text like a human, avoiding AI detection. This tool humanizes your content to bypass any AI detector. So Iguess like any open marketplace, there's going to be good stuff and bad stuff.
[00:24:28] Paul Roetzer: My quick scan of the top 10 in all of those categories is it's a lot of junk I would never probably click on, certainly never use. It gets into a trust thing of like, I don't even know what they get. Like when I put my data in, so let's say I pick a data analysis one and I drop in some data, it's going to give me an output, but like Does the creator have access to that data?
[00:24:49] Paul Roetzer: Is it only OpenAI that has access to that data? I think it's only OpenAI, but I actually don't know. And this goes to that whole thing of education is critical. If you're going to give these tools to your people, they need to understand what it is they're doing. So, I don't know, at a high level, I think it's going to be big.
[00:25:06] Paul Roetzer: I've heard them say like maybe bigger than the Apple App Store. More influential, more important. I don't know, maybe. But I do think for users, GPTs. is a great way for people to get value way faster than just staring at a blank, you know, screen powered by GPT 4 and not knowing what to put in it or how to do it or how to structure a prompt.
[00:25:29] Paul Roetzer: You can pre structure the prompts, you can do all the background work for people. I think GPTs are going to be a critical part, and I know GPTs for OpenAI is actually just a prelude to AGI, like this is, it's just a piece of the journey they're on to build smarter agents and things like that, so it's a big deal though.
[00:25:47] Mike Kaput: So it sounds like there are definitely some things to keep in mind as you're approaching some of the listings in the GPT stores. So for marketers, business leaders, and people trying to look at this in a professional context, are there any concerns or considerations to keep in mind beyond kind of, Buyer beware.
[00:26:07] Paul Roetzer: Yeah. I mean, Ithink you can just waste a lot of time like playing around with these things. I really think that for a lot of people, like what GPTs do is they give non developers the power to build things. And so if you just think about like the frameworks and processes that you use in your company, the use cases that are critical to the people, and you see GPTs more as a tool.
[00:26:31] Paul Roetzer: For you to formalize processes in a way like by basically building an app, you, you couldn't build before, Mike and I have no ability to do coding, like to build apps, but now I think about all the frameworks we use in our company around our marketing strategy process, our writing process, you know, how we do interviews and summarizations, like all these things we do, how we do brainstorming and run hackathons and run workshops, All of that can be formalized into GPTs, which right now it's just a, it's a bunch of words on the page or it's Excel charts that tell you what to do, or it's flow charts, or if this, then that logic and things like that, we can put that into an app and anybody in the company can now use it or, or we can open it up to the public and say, here, here's how, here's how we do this.
[00:27:16] Paul Roetzer: Like this will help you with your brainstorming. So I don't know. I mean, I think they're going to be great, but I also think. There could be a lot of time wasted like just playing around with the store and looking for tools and playing with like, I don't know, just based on my initial feeling, there needs to be way better curation, some sort of ranking system that tells you which apps are worth your time and which vendors are trustworthy.
[00:27:37] Paul Roetzer: This is
[00:27:37] Mike Kaput: such a big reason why in our workshops, in our content, in our speaking, we focus on use cases first before you even start getting into the details of different technologies. That's very important, but if you don't go in to these new tools, marketplaces, and ecosystems without a plan, it is very, very hard to efficiently and quickly enough start adopting this technology, in my
[00:28:04] Paul Roetzer: opinion.
[00:28:05] Paul Roetzer: Yeah, I agree. And that's why I like saying, like, I think if, you know, if I'm looking for a writing tool, you know, for example, like go in there and look at the writing tools and see if there's something specific that helps, but you can get distracted really. It is, it can definitely have the shiny object.
[00:28:20] Paul Roetzer: element to this, where it's just like, Oh wow, there's like thousands of these things. And you're just playing around to your point of like, yeah, but you just need something that's going to help you write a video script. Has someone built something to help with video scripting? No? Okay. Then go about your life and play around with that on a Saturday night sometime when you got nothing else to do.
[00:28:38] Paul Roetzer: So
[00:28:39] Mike Kaput: to wrap this topic up, I wanted to see if we might be able to give our audience a little inspiration. Do you have any thoughts on types of GPTs you'd be excited to see or kind of on your wishlist moving forward?
[00:28:52] Paul Roetzer: Yeah, I mean, I,Iok at areas that are a weakness for me that are really important to the organization.
[00:28:57] Paul Roetzer: So I think like data analysis, Iknow Code Interpreter, it was originally called, they changed the name to Advanced Data Analysis. That is a tool that I know is in GPT 4 that I know I have access to, but I struggle to like really figure out what the best use cases are. I know I have a ton of data, that it could help me with.
[00:29:16] Paul Roetzer: So if there are some structured around that data and analysis where I could go and almost get some inspiration, like, Oh, I didn't even think about that. That's, Isee it as like I'll look at the key use cases we already have, but I'll look at the gaps in our ability. Like we don't have a data analyst on staff.
[00:29:31] Paul Roetzer: Can I, can I find some GPTs that function in some ways like a You know, a data analyst. That would be huge for us. Like, I would love that. So I'm going to look at like those kind of ways where common use cases that we already, you know, could drive efficiency with or gaps in our abilities that I would love to find a tool that, you know, rather than going out and maybe hiring an outside consultant, maybe I can use some GPTs to assist there.
[00:29:58] Mike Kaput: So in our third big topic this week, we have seen a new AI hardware device go viral online after debuting at CES this year. It's called the Rabbit R1. It's built by a company also named Rabbit, and it's a standalone device with a small touchscreen, a camera, and a scroll wheel and button. That doesn't seem that impressive, but the real draw of this device, what's getting it a lot of buzz online, is how it uses artificial intelligence.
[00:30:29] Mike Kaput: The device's operating system is called Rabbit OS, and the company says it's based on an AI model that it's calling, and I think this is a term they devised on their own, called a large action model. The Verge describes this as a quote, sort of universal controller for apps. Now what this means is that like Alexa or Google Assistant, the Rabbit R1 device can be used to interact with and control all the apps that you use every day.
[00:30:59] Mike Kaput: You just tell it what you need and it uses your apps to go do that. Now what's novel about this is that you're also able to do that because Rabbit has trained the model on how to use existing apps on its own. The company even claims you'll be able to teach the device how to do new things with apps so it can do them on its own moving forward.
[00:31:21] Mike Kaput: One hypothetical example that they give is you could theoretically teach it how to say remove watermarks in images using Photoshop by going through the process yourself, showing the model how to do it, and then it would learn how to execute those commands within the Photoshop app. Now, these details are pretty scant, but they were enough to send consumers into a frenzy.
[00:31:46] Mike Kaput: After introducing the Rabbit R1 at CES, the company pre sold 10, 000 units in a single day, which completely sold out its first production run. They've also, I believe, sold out another 10, 000 run, after that initial day. Now it sounds like you can still start to place a preorder for the device. It costs 199, but you're of course going to have to wait quite a while for it at this point, given how popular it is.
[00:32:17] Mike Kaput: So Paul, first up, what did you think of this device initially when you saw it kind of start to get traction online?
[00:32:24] Paul Roetzer: So this, this kind of snowballed pretty fast, for sure. So I saw it being buzzed about on, whatever it was, Tuesday or Monday or Tuesday, I don't remember when it first came out. And, and the way they launched it was through like a keynote.
[00:32:39] Paul Roetzer: It was like a 30 minute keynote. And, it was produced pretty well. It was a little rough to, to get through. Idid stop watching it after about like 18 minutes. It was just like, okay, this is Pretty dry, and I kind of generally get the concept. It's a beautiful design, so I have to give them props, like it's, it's beautiful.
[00:32:59] Paul Roetzer: Like, it's really cool, it's innovative, it's engineered well, it would appear. The website's awesome. I don't know who did their website, but like, they're, they're doing marketing right. They used some of that 30 million they've raised to, to do some legit, design and user experience stuff. That all being said, so my first tweet was this company has 30 million funding.
[00:33:26] Paul Roetzer: It's a device designed to integrate AI agents and simplify access to them. If I'm accurately summarizing the product launch video, apparently better than your phone, admire the effort, but has a humane AI pin field to me, i. e. hardware innovation that will have very low adoption because you'll still need, want your phone and you won't want a second, largely redundant device in your pocket.
[00:33:46] Paul Roetzer: So the way I think about this is. I love hardware innovation. I love that people are trying things like, you know, the humane pin we talked about on the show. I'm not a fan of, but I'm a fan of the effort. There's a, there's a company trying to do it with a necklace. Like all these people are trying to build these, the next device, the next thing that we're going to wear or have with us.
[00:34:11] Paul Roetzer: That's going to record everything around us and invade people's privacy and all this stuff. Glasses is another one. So, Allie Miller, who I mentioned earlier, she put up, like, how many do you think this is going to sell, and it was like a thousand to ten thousand, ten thousand, a hundred thousand, whatever, and I replied and said, like, you need a less than a thousand option, and she's like, you're joking, right?
[00:34:30] Paul Roetzer: And so I replied like, well, I'm half joking. Like, I think that a whole bunch of AI enthusiasts are going to buy it, which obviously happened. They did sell 20, 000 in the first 48 or 72 hours. And so what I said to her was like, I don't think it's a viable product in the long run that I may be completely wrong here, but like the pin, I think it's just going to be really innovative, but at the end of the day.
[00:34:55] Paul Roetzer: Apple is going to do this on the phone and on the watch. I already have both with me. So Surrey will be smart at some point, like they will have AI agents built into the phone. So the, like the case that the I was a CEO that did the presentation, but the case he made was basically like, nobody wants to go in their phone and click a bunch of apps, but you're not, you're not going to have to, like, you're defining a problem that won't be a problem 10 months from now.
[00:35:23] Paul Roetzer: So I don't know. I felt like it was. It's really cool. A bunch of people are going to spend the 200 bucks because for a lot of people that's not a lot of money and it's a novelty and it's, they're curious about it, but then they're going to get it and realize they have to connect it to all of their, like, accounts and, and, and apps.
[00:35:41] Paul Roetzer: And then you're like, do I trust this company? All of a sudden starts kind of coming into the back of your mind of like, I don't really know who. Rabbit is, and what am I giving them access to? And, and for them, for their agents to do the thing they're promising, that's going to like change my life and make me not maybe want to carry my phone around anymore, I have to trust them with access to all of these accounts.
[00:36:00] Paul Roetzer: Otherwise it's just not going to do what it promises to do. Right. So I think that like at the end of the day, it's interesting. It's, it's a cool product. I mean, I may buy one at some point just to try it, just to like see it and to make sure that my. Opinion of it is based on an actual experience and not like some, some bias that I may have about this stuff.
[00:36:26] Paul Roetzer: But Ijust, I just think that Apple's going to crush people like this. Like they're, they're just going to come out with these innovations and then it's just going to be like, yeah, that was fun for like two months. Remember that time everybody bought all those devices and But the other thing that I wonder about is how much of it was actually a scam, not the company itself, not Rabbit.
[00:36:45] Paul Roetzer: But I saw their founder tweeted, like the next day, no, stop this. People were buying these things in bulk and then reselling them online for twice the price. So for the, so when you look at the marketplace and this is how like we, we, we, we, we work. There's 10, 000 available. There's a limited supply. You know, the value is going to jump right away.
[00:37:07] Paul Roetzer: So people go in and just buy, I don't know if you could buy a thousand a time, like how many, I don't know how you could buy at a time, but then you get specular buyers who just go buy those things that doesn't cost much or they lose some money or lose some money, but maybe they'll make two, three times their money.
[00:37:18] Paul Roetzer: And so I know it was happening because their founder screenshot of it happening. Wow. And so then you're like, well, was like this 20, 000 sale really like legit? Like, is it really 20, 000 people are going to use this thing? And then I saw another like AI guy that I follow and he was like, the usage cost of this is going to be like a flat 200 per use because people are going to use it once and never touch the thing again.
[00:37:42] Paul Roetzer: So it, it definitely created. A lot of buzz, it was everywhere in the media, all the AI influencers were talking about it. And there are definitely people who think it's like the dumbest thing, and then there are people who like, love the thing. And, and, and now it'll be like, well is it actually any good once you get it?
[00:38:01] Paul Roetzer: But yeah, I don't know, it's, it's cool, like, I like that we're innovating on devices, Ijust don't. I just don't see it. And I'm happy to be wrong on this stuff. Like, I don't know. Like, I'm just kind of hypothesizing here, but Again, I just look and think, are people really going to put a clunky square in their pocket?
[00:38:21] Paul Roetzer: Like a Pokemon, like in Pokemon games. Cause I think that was kind of like the inspiration was actually Pokemon, the device that they have the Pokedex on. My family plays Pokemon. Idon't, I don't know what the thing's called. But I just don't see it. And so then it got me thinking back to like, but.
[00:38:39] Paul Roetzer: If the watch, like I already have an Apple watch and I love it. And lots of people have Apple watches and I already have my phone. I just don't see myself doing that. Having that other device. And it's definitely not replacing the iPhone. So what did you think? I mean, did you have an opinion? Like, did you, what was your initial
[00:38:59] Mike Kaput: reaction to it?
[00:38:59] Mike Kaput: Yeah, I'll be honest. I, and I don't know if this is reflective of what other people are saying about it. I just tend to default towards the fact that I'm already overwhelmed with devices and services and tools. So while I'm happy to add on ones that make sense, I don't personally. think I would have too much of a need for an additional device to do this, especially to your point, if even if the shipping schedules are on time, we're still months out from even getting one of these.
[00:39:31] Mike Kaput: And I swear, I would imagine it might accelerate some of the plans Apple has to roll out these features. It's like, why would I need another device to regulate apps on a device designed? to regulate apps, but I can see the argument in favor for it. If you are savvy and forward thinking, kind of love having tons of devices, maybe it makes sense
[00:39:54] Paul Roetzer: for you.
[00:39:55] Paul Roetzer: Yeah. And Ido think like all this being said, they may have a nice exit. They're obviously already raised 30 million. They may raise a bunch more. So just because like, there isn't, I don't think a market for this longterm, like a consumer market outside of like AI enthusiasts. And, you know, the people want to have every piece of tech that comes out.
[00:40:15] Paul Roetzer: It doesn't mean that companies, that Rabbit and other companies like them won't raise a bunch of money, won't have some exits, maybe they've got some patents on the technology that Apple or somebody else likes, and they buy up the team in an acquihire and like get their patents. Like, again, it's, it's cool.
[00:40:29] Paul Roetzer: This is what technology is supposed to be. People are supposed to be on the frontier, trying things, doing things, taking some risks, building things that other people aren't willing to build. And sometimes it just doesn't work. Like the humane people already laid off people, their CTO left, like. So when we had the episode, I don't know, a couple months ago, and I was like, yeah, this pin's not going to fly.
[00:40:46] Paul Roetzer: They already laid off like 10 or 20 percent of their workforce, their CTO left, like, it's going to happen. Like, you're going to shoot up, you're going to be on CNBC, debuting products, be at CES on the stage, everybody's going to love you, and then three months later, maybe it goes away, but, or maybe you sell it for a few hundred million and move on and do the next thing.
[00:41:03] Paul Roetzer: Like, this is, this is what tech is. So again, I'm an advocate of this kind of thing. I like that people are doing it. Doesn't mean I have to personally like it or think there's a market
[00:41:13] Mike Kaput: for it. Yeah. Surprisingly, despite what some people on the internet would have, you believe it's possible to not hate on this and also just say, Hey, maybe it's not going to work.
[00:41:25] Paul Roetzer: Yeah. And that's my thing. It's like, I'm not. I'm not hating on it at all. Like, I love it. I think it's great. I just don't think it's going to work.
[00:41:33] Mike Kaput: Now, let's talk for a second about, obviously nobody really has their hands on this yet, but what do you make of the AI that's built into this and what they're saying it does, what it can do?
[00:41:44] Mike Kaput: That did seem somewhat interesting to me and to your point of it getting perhaps acquired at some point, that can be an interesting I don't
[00:41:54] Paul Roetzer: know. I mean, they talked about this large action model. My initial feeling, it was just like one slide and they didn't spend a ton of time on it. My initial reaction was like, okay, I don't understand.
[00:42:03] Paul Roetzer: Like you just gave a name to something that every research lab is working on. Like, what is different about yours? As far as I know, they haven't published any research papers that are any advancements in the pursuit of action transformers or anything like that. So I don't, I don't know it's anything other than.
[00:42:18] Paul Roetzer: Marketing that they just came up with a slick look and slide and a name that may or may not stick to something everyone is doing and including Apple. Like there's no way Apple isn't building this stuff in apps, in the phone, in the hardware, in the chips that Apple's building. Like they have all this stuff.
[00:42:36] Paul Roetzer: And so that was my, my take is like, okay, cool. Like again, good, good marketing. Well done. I have no idea if you have any actual real innovation behind it other than a nice looking piece of hardware. My guess is it's not going to work anywhere near like what they showed in the demo. Right,
[00:42:53] Mike Kaput: right, right.
[00:42:53] Mike Kaput: Yeah, we, we definitely have to put it through its paces and actually understand if it can do what they say it can do. So, as we kind of like wrap this up before we dive into some rapid fire, just very, very quickly, I wanted to kind of zoom out and talk about the reason why. Stuff like Rabbit is happening.
[00:43:11] Mike Kaput: What people are actually trying to pursue and achieve here, and it really comes down, I think, to this idea of AI agents that we've talked about in the past and that are becoming increasingly important. So basically, very broadly, AI agents are AI systems that can act autonomously in certain ways. Can you walk through just some of the bigger picture idea here about why these companies are Racing towards these types of autonomous systems, what that means for consumers,
[00:43:42] Paul Roetzer: businesses.
[00:43:44] Paul Roetzer: Yeah, so we've talked about this in past episodes, and maybe we can pull a couple of links, but we talked about Andrej Karpathy, who's at OpenAI now, and back in like 2016 17, he worked on something called World of Bits at OpenAI, then he went off to Tesla and then came back to OpenAI, and it was all this concept of getting agents to take actions on our behalf, like, and so research labs have been working on this for probably north of a decade.
[00:44:09] Paul Roetzer: And so if you think about language models right now, you go into ChatGPT or whatever it is, and you ask it for something like, I always like to use the example of an itinerary. I'm going on a trip. Like what hotels should I consider? What restaurant should I go to? What sites should I see? And it'll give you an itinerary.
[00:44:24] Paul Roetzer: Like it'll build it out. You can ask it further prompts, like what would be good restaurants for families of four, like whatever. It can't then go do anything. Like it can just give you words. Actions would say, Okay, that looks great. Like go ahead and book the flights for like a pre 10am flight for four of us.
[00:44:42] Paul Roetzer: Use my frequent flyer code that's saved in United. Go ahead and get us a reservation at that restaurant at seven o'clock the night we arrive. Give me a rental car, like you can just tell it and it'll go do things for you. So that is the idea of action. So the unlock here was once they, I don't know if solved is the right word, but once they made the breakthroughs in being able to build language models, so that the agents, that the machines could understand better what the human was asking.
[00:45:16] Paul Roetzer: That was sort of an obstacle that was preventing these agents from being able to take actions is that it was hard to. to have a conversation with the machine and it would understand what you were asking it. So the language model became kind of a breakthrough to get past the barrier. Why Andres Karpathy went back to OpenAI is now that we had language models that have reasoning capabilities and in some beliefs have like a worldview and they know all these advancements.
[00:45:42] Paul Roetzer: The belief is we can now make agents that take actions, reality. So you have Adept has raised like 415 million in the pursuit of agents. Our friends at HyperWrite have been building agents. OpenAI is doing it, Cohere is doing it, Google's doing it, Apple's doing it. Everyone is building agents. So it is the one.
[00:46:00] Paul Roetzer: When people ask, like, what does it look like in two, three years from now? All I say is, we know agents are coming. Like, there's only a certain number of things we know are going to happen. And agents has been one of those things for the last couple of years. And 2024, I think, is going to be our first ability to start experimenting or experiencing working agents.
[00:46:24] Paul Roetzer: They don't work right now. They're very flawed. But I think that enough people, enough money is focused on this, that we're probably by the end of this year going to see some pretty significant advancements in the ability for these things to do stuff for us. So in marketing and business, it could be like, You know, send this email, like, you know, find the list, do this and then send it.
[00:46:45] Paul Roetzer: And it can actually go do it. It can build the list. It can create the emails. And it just basically goes through like this process and does all these things. And the human's in the loop as much as you want the human in the loop.
[00:46:57] Mike Kaput: So let's dive into a few rapid fire topics here that are going on this week. The first one predictably also involves OpenAI, big week for them as always, but last week we had talked about the New York Times lawsuit against OpenAI, which alleged that the company violated copyright by illegally using the Times content to train AI models.
[00:47:24] Mike Kaput: Now OpenAI has posted a response on its website about this lawsuit. In a post titled OpenAI and Journalism, the company writes, quote, while we disagree with the claims in the New York Times lawsuit, We view it as an opportunity to clarify our business, our intent, and how we build our technology. To do that, the company has outlined what it sees as four kind of key points around this idea of both the lawsuit and OpenAI's relationship to journalism as a whole.
[00:47:55] Mike Kaput: First, OpenAI says that it works hard to support and collaborate with news organizations, and that includes entering into early partnerships with companies like the Associated Press. Second, they make the claim that training AI models using publicly available internet materials is fair use. quote, as supported by longstanding and widely accepted precedents.
[00:48:20] Mike Kaput: So that's kind of a key sticking point in the New York Times lawsuit. Third, they insist that, quote, regurgitation, or spitting back out New York Times content verbatim, which is something that was highlighted very prominently in the lawsuit. is actually a rare bug that they're working hard to fix. It's not a common occurrence according to OpenAI.
[00:48:42] Mike Kaput: They also note that no single source, including the New York Times, is a significant representation of the model's whole corpus of learned knowledge. So the Times tried to make this argument that, you know, OpenAI had, focused more prominently on the Times as a source of data for the model. OpenAI claims that it's just one of many, many things.
[00:49:05] Mike Kaput: Last but not least, OpenAI has said the Times is not telling the full story. OpenAI claims that they were having talks to actually incorporate the Times content into ChatGPT. Right up until very close to when the lawsuit was filed. So they actually note that through December 19th, those talks were progressing, in their words, constructively, and that the lawsuit on December 27th came as a surprise.
[00:49:34] Mike Kaput: Lastly, OpenAI claims it tried to fix all the issues that were named in the lawsuit, but the Times wasn't very helpful. And they say that the Times intentionally manipulated prompts in order to get ChatGPT to regurgitate that Times content verbatim, which is what they quoted in their lawsuit as kind of evidence that ChatGPT was, lifting right from their content.
[00:49:59] Mike Kaput: So Paul, there's a lot going on here in this response. What do you make of this counterargument, given what we talked about last week?
[00:50:06] Paul Roetzer: Pretty predictable. I mean, we, we kind of knew how they would approach this. The one thing that I was thinking about, and I don't remember. If I saw this online, like somebody might have called this out, but I think it's interesting that like they're doing deals, they, they would deal with Axel Springer.
[00:50:24] Paul Roetzer: I saw they were working on deals with like CNN and Fox and like all these other outlets, New York Times, obviously. So if they need to negotiate licensing deals with these media companies, whose data is in the training set, why don't they need licensing deals with everybody else? Like, don't you think if they're allowed to just take it fair use, why do they need a licensing deal?
[00:50:45] Paul Roetzer: Just take it. So, what about all the small publishers whose data is in there? And I just wonder, like, how that'll play out. Again, I have no idea. I'm not, I took business law in college. Like, that was the extent of it. One class. Like, I'm not a lawyer. But those are the kinds of things that I think will be really interesting as this moves forward, assuming they don't settle.
[00:51:07] Paul Roetzer: As we get into actually, you know, legal arguments, you know, I could see it, like I said on, on last week's episode, this idea of, well, if they do this in the New York Times, you got the LA Times, Washington Post, like the other top ten sources in common crawl, media publish, media companies. What about all the small publishers whose data is also in there?
[00:51:26] Paul Roetzer: If they need licenses with New York Times, don't they need licenses with everybody else? Like. I don't know. I, there's just a lot at stake here. Like, it's going to be really fascinating to see how this plays out, but I don't have any, you know, brilliant legal insights into this, which is why for our AI for Writers Summit, I want to have an IP attorney who specializes in copyright law on, on, in the summit.
[00:51:48] Paul Roetzer: So we can ask these hard questions. And these are the kinds of things that I want to, I want to be able to ask at that summit and figure out what it means.
[00:51:57] Mike Kaput: So next up, there's some new research out from an SEO software company called Authoritas, and it claims that AI generated answers in Google Search Generative Experience, SGE, do not match the links in the top 10 normal Google organic search results.
[00:52:16] Mike Kaput: It's a whopping 93. 8 percent of the time. This study focused primarily on commercial keywords, and it basically seems to indicate that these conversational AI powered results that SGC serves up, which are right at the top of the results, they come out as kind of conversational language like you would get from ChatGPT and have kind of footnotes and links to where it got the information.
[00:52:41] Mike Kaput: It's finding that these results are overwhelmingly not. from the pages that are already ranking highest for that particular query. So everything on page one often is not actually in those results at the top of the search results. This research also found that on average, SGE shows about 10 links as part of its results.
[00:53:05] Mike Kaput: But only about four of them are actually unique, meaning multiple links in your SGE results often point to the same website. So again, that kind of further reduces the number of unique results that are showing up literally above The stuff ranking on page one. So, Paul, this kind of touches on a few issues we've mentioned time and time again about the future of search on this podcast.
[00:53:29] Mike Kaput: This seems to indicate, at least in certain contexts, that there are some really thorny, unresolved issues here about how is AI going to Impact organic search. Can you unpack for us some of the implications
[00:53:42] Paul Roetzer: here? My first reaction is this mirrors my personal experience. So when I've tried search generative experience, or when I've tried to do this in ChatGPT with Bing, where you get an output and it gives you citations, they usually suck, things I would never click on if If I was doing a regular Google search and again, like if you get 10 citations, this is like, again, ChatGPT experience, seven of them were the same link.
[00:54:08] Paul Roetzer: to a junky website. And it's like, well, this is where your sources are coming from. So, I think there's the quality issue that these companies are going to face if their AI powered results are actually being fed from junk content, which goes back to the need to license. Good content. So we're not getting these junky results.
[00:54:28] Paul Roetzer: Maybe it's part of the argument for our conversation from last week about perplexity, like if perplexity is solving for quality of citation in the results, that, that could be a nice differentiator. I got to think Google can solve this. It's kind of weird that they haven't. Now, this also though calls to the challenge we face in the marketing world, SEO, content, where you are creating things.
[00:54:55] Paul Roetzer: because you generally understood what Google valued in their algorithm. And so you could do SEO to get on to the, you know, the top 10 in Google, ideally the top three. Well, if we don't know how the algorithms work, and we don't know where they're coming up with all these different links that are surfacing, it becomes really hard to build an SEO strategy.
[00:55:16] Paul Roetzer: So yeah, it's going to be a very dynamic time, I think, until we figure out how this all plays out. And what we know is, They're just trying to figure it out themselves. Like they're trying to experiment, but obviously the biggest issue, in my opinion, is the quality of the results you're getting where these citations, yeah, they're citing things, but they're not very good.
[00:55:35] Paul Roetzer: You can't trust them.
[00:55:36] Mike Kaput: Yeah. And that really presents problems as consumers. I mean, I think we would agree more and more we're getting. Acclimated to just reading conversational responses and saying, okay, I need to go check that, but that sounds pretty good in terms of the information, but if you're not clicking into those links and that information is coming from trash websites, you have no idea.
[00:55:59] Mike Kaput: In other news, popular language app duo lingo has actually cut 10% of its contractor workforce, and it's doing that because it's using generative AI to actually create more of the content for its app. In reporting a company spokesperson said quote, we just no longer need as many people to do the same.
[00:56:20] Mike Kaput: The type of work some of these contractors were doing, part of that could be attributed to ai. Now this comes after Duolingo's CEO in November 2023 said that the company was using generative AI to produce quote new content dramatically faster. Now, Paul, this is obviously just one example, very specific business, but should we expect more of these kinds of cuts or reductions in force due to efficiencies captured
[00:56:51] Paul Roetzer: by generative AI?
[00:56:53] Paul Roetzer: For sure. Yeah. I mean, Iknow it's not like a. Popular opinion or opinion people like to hear, but this is absolutely going to happen, like, throughout the year.
Google cut a thousand yesterday in their Google Assistant, not apparently related to generative AI per se. But I think the tech industry is bloated, like they hired way too many people when times were good, so part of this is just a natural, like they're shedding people, and I mean, I saw one tweet, there was like 15 different tech companies listed who have already had layoffs the first 12 days of 2024.
[00:57:27] Paul Roetzer: Some of them are just because The businesses need to run tighter and they actually have to get to profitability or improve their efficiencies. But a lot of them are going to be related to the belief that AI is just going to drive enough efficiency. We're just not going to need as many people to do this stuff.
[00:57:44] Paul Roetzer: So I think it's going to start in the tech industry and we're already seeing it. But I think by the end of this year, middle to end of this year, you're going to start to see a trickle down into other industries where the same story plays out. I wish that wasn't the case, but I'm, I think it will be. So
[00:58:03] Mike Kaput: a new poll from Boston Consulting Group, BCG, of over 1400 C suite executives shows that a full 66 percent of them say that they are ambivalent about or downright dissatisfied with their organization's progress so far on generative AI.
[00:58:22] Mike Kaput: These leaders say that they're dealing with a shortage of talent and skills. Unclear roadmaps and a lack of strategy needed to deploy generative AI responsibly. And all of this is leading to their discontent with their generative AI results. It's gotten so bad, apparently, that only about half of the respondents said they expect generative AI to bring substantial productivity gains, which they define as 10 percent or more, just about, to the workforces that they oversee.
[00:58:54] Mike Kaput: So, Paul, this is kind of the other side of the coin to the previous story. Did these findings surprise you at all? This
[00:59:00] Paul Roetzer: is the only thing that stops the previous thing from happening, is that no one knows what to do with this stuff. So, I'm just going to re read that middle paragraph. These leaders say they're dealing with a shortage of talent and skills, unclear roadmaps, and a lack of strategy to deploy generative AI responsibly, all of which is leading to their discontent with Gen AI.
[00:59:22] Paul Roetzer: This is exactly what we were saying before, the lack of literacy. There's a lack of understanding how to do it. There's a lack of systems. There's a lack of an ecosystem of service providers who can help companies figure this out. We're in the infancy of the adoption curve here. This is why I always say, like, people are like, oh, AI is overhyped.
[00:59:39] Paul Roetzer: I'm like, no, no, no, it's underhyped. Like. We're not even doing it yet. Like the technology is now there to transform workflows and processes and tech stacks and teams, but nobody knows what to do with it. And so when you read these surveys of executives, think Gen AI is important or not important, it's like, you have to understand the context that they don't understand it.
[01:00:00] Paul Roetzer: You're asking people who don't understand the technology and don't have staff that understand the technology and they have no plans. to tell you whether they think it's important or not. Or to tell you if it's going to have an impact on their business. Of course they don't know that! They don't understand it!
[01:00:15] Paul Roetzer: So, I, this is just like, that paragraph to me summarizes the theme of what is happening in almost every company we talk to. Unclear roadmaps, lack of strategy, shortage of people and skills who get this stuff, and so therefore they can't run pilots, they can't adopt it properly, and they can't, absolutely cannot scale it.
[01:00:34] Paul Roetzer: Every company we talk to, this is the story. So if that's you, welcome to the club. If you figure it out for everybody else, you're going to have a massive advantage in your industry. Figure it out, focus on AI education and training for yourself, for your team, get policies and principles in place, get a roadmap in place.
[01:00:52] Paul Roetzer: You are going to be ahead of like 99 percent of your competitors.
[01:00:58] Mike Kaput: Our last news item today is about the popular question and answer site Quora. They just raised 75 million from famed VC firm Andreesen Horowitz, and the money is primarily designed to accelerate the growth of Poe, which is the company's AI chat platform.
[01:01:15] Mike Kaput: We haven't talked about Poe that much, but it is an AI chat platform that's been around for almost a year, since about early 2023, and according to Quora, it has millions. What Po does is it lets you actually use a range of different available chatbots, including ChatGPT, Llama, and Claude, right within a single platform.
[01:01:38] Mike Kaput: Importantly, you can also use Po to create your own bots and host them on the platform. In fact, Quora says that they expect the majority of the funding to be used to pay creators of bots on the platform through a recently launched creator monetization program. This program rewards builders who launch popular bots on Poe's platform, and it seems to be pretty clear the company has some ambitions to become a sort of kind of bot ecosystem for builders.
[01:02:06] Mike Kaput: And a fun fact, Quora CEO Adam D'Angelo was involved in the OpenAI drama. Previously, as he was a board member and still is at the company. So Paul, I wanted to ask you, since we haven't covered Quora nearly as much here, how significant is it in the AI landscape?
[01:02:26] Paul Roetzer: I think D'Angelo, if I'm not mistaken, is the only remaining outside board member.
[01:02:30] Paul Roetzer: He's the survivor, yeah. Yes, so when Sam Altman got fired, the other two non OpenAI, so the board was made of D'Angelo, two other outsiders, and then Ilya, Sam, and Greg Brockman. And Ilya's off the board, Sam's off the board, Greg's off the board, and the two outsiders are off the board. So Adam was the sole survivor of the OpenAI board after the Sam Altman fiasco.
[01:02:55] Paul Roetzer: I gotta be honest. I don't use Quora. I was trying to wrack my brain like when I did use Quora. So it was founded in 2009 and I think it was probably around 2009 2010 when I was playing around with it back in our early days as a HubSpot partner agency. I know that they are considered innovative. I know that Poe is considered, to be an interesting platform.
[01:03:18] Paul Roetzer: I know it is, as you called out, built on other people's large language models. So kind of like the perplexity issue we talked about last week. They're, they're built on someone else's tech, which isn't necessarily a bad thing. It just, that's just what it is. Yeah, I don't know. Idon't, I don't know that what they're trying, their strategic vision well enough to like give, uh.
[01:03:39] Paul Roetzer: a meaningful assessment of this. I think it's just a good newsworthy item. And Driesen Horowitz, you know, I obviously a highly respected venture capital firm putting 75 million in as meaningful. Yeah, something to keep an eye on. And I'm sure there's some people in our audience who are, you know, PO users and, big fans of Quora.
[01:03:58] Paul Roetzer: So good information if that's you. All right, Paul, thanks
[01:04:04] Mike Kaput: for rounding up for us another busy week in AI. I just want to note very quickly for the audience that we don't get to nearly everything going on each week in the podcast alone. We also have a weekly newsletter that runs down everything that happened this week in AI.
[01:04:21] Mike Kaput: It includes more in depth articles on the topics we covered today. It also includes a ton of other topics. that we didn't have time to get to. So if you have not subscribed yet, go to marketingainstitute. com forward slash newsletter to sign up for the weekly AI news roundup that we're putting
[01:04:41] Paul Roetzer: out. Yeah.
[01:04:43] Paul Roetzer: I mean, we probably could have done a whole episode on CES news from last week. Yeah. All right. Well, thanks everybody for being with us. Thanks Mike, as always for curating everything and leading the conversation. And we will be back next week with our regular weekly podcast.
[01:04:58] 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.
[01:05:11] Paul Roetzer: Be sure to subscribe to our weekly newsletter, check out our free monthly webinars, and explore dozens of online courses and professional certifications.
[01:05:20] Paul Roetzer: Until next time, stay curious and explore AI.