In what may have been the biggest week in marketing AI (to date), we have a lot to review in this week’s podcast, so let’s jump right in.
GPT-4 is released to the public
The week started with the much-anticipated release of GPT-4, the latest, most powerful version of OpenAI’s large language model. And it’s now is being integrated into existing products via API, as well as ChatGPT.
According to OpenAI, “GPT-4 can solve difficult problems with greater accuracy, thanks to its broader general knowledge and problem-solving abilities.” OpenAI says that GPT-4 surpasses ChatGPT in its reasoning capabilities. In fact, the new model was tested on the Uniform Bar Exam, where it scored in the 90th percentile compared to ChatGPT’s 10th percentile score.
GPT-4 will be able to accept images as input;. OpenAI demoed one jaw-dropping example of the model being able to generate code for a webpage based on a hand-drawn sketch of what the webpage should look like.
Has OpenAI veered too far away from its non-profit roots? Is a company with “open” in the name being as forthcoming as they should be? Paul and Mike discuss.
Microsoft and Google embed AI in their core products
Google just announced that developers will now have access to its PaLM API, which gives them the ability to build on top of Google’s language models in Google Cloud. The company also announced generative AI features coming to Google Workspace, the firm’s productivity suite. That means you’ll soon see generative AI features in Gmail and Docs that draft copy on any given topic for you.
At the same time, Microsoft announced Microsoft 365 Copilot, an AI tool that is “your copilot for work.” According to the company, Copilot combines the power of large language models with your data in the Microsoft Graph and Microsoft 365 apps to increase productivity.
That happens in two ways, says Microsoft. First, Copilot works alongside you in popular apps like PowerPoint, Word, and Outlook to increase productivity. For instance, in Word, it can now generate drafts for you. In PowerPoint, you can use natural language prompts to create presentations. Copilot also enables a new feature called “Business Chat” that surfaces insights from data across your company and makes performing tasks easier. Our team watched the launch demo, and some instant applications came to mind.
The U.S. Copyright Office is getting involved…so make sure you’re not breaking any laws.
Warning: If you’re using generative AI tools to create content—articles, blog posts, books, images, software, songs, videos, etc—you do not own that content, according to the U.S. Copyright Office. That means anyone can reproduce it without your permission, create derivative works from it, display it, perform it, and sell it.
On Mar 16, 2023, the Copyright Office launched an initiative to examine the copyright law and policy issues raised by AI, including the scope of copyright in works generated using AI tools and the use of copyrighted materials in AI training.
“This initiative is in direct response to the recent striking advances in generative AI technologies and their rapidly growing use by individuals and businesses. The Copyright Office has received requests from Congress and members of the public, including creators and AI users, to examine the issues raised for copyright, and it is already receiving applications for registration of works including AI-generated content.”
The Office also issued updated registration guidance that has an immediate effect on your ability to protect your original works.
Paul and Mike discuss ways to use AI-powered technologies that are legal, and what tech companies need to address with their tools.
Listen to this week’s episode on your favorite podcast player, and be sure to explore the links below for more thoughts and perspectives on these important topics.
00:04:15 — GPT-4 is here
00:22:11 — Microsoft and Google embed AI into their technologies
00:33:24 — AI and copyright law
00:49:37 — Microsoft lays of entire ethics and society team
00:50:46 — Anthropic releases Claude
00:52:52 — Adept’s $350 million funding round
00:54:37 — Gen-2 by Runway
00:56:38 — Midjourney v5
Links referenced in the show
- OpenAI co-founder on company’s past approach to openly sharing research: ‘We were wrong’ - The Verge
- Google and Microsoft Embed AI Into Core Products
- The next generation of AI for developers and Google Workspace
- Paul's LinkedIn post
- Introducing Microsoft 365 Copilot – your copilot for work
- The U.S. Copyright Office Says You Don’t Own AI-Generated Content
- Paul's LinkedIn post
- Federal Register :: Copyright Registration Guidance: Works Containing Material Generated by Artificial Intelligence
- Microsoft Lays Off Responsible AI Team
- Anthropic Introduces Claude
- Adept Raises $350M
- Gen-2 by Runway
Watch the Video
Read the Interview Transcription
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: This could actually be a major, major hindrance on the use of generative ai, especially for enterprises because there's no way people that are big enough to have a legal team are going to allow this.
[00:00:11] Paul Roetzer: And so I could see the pushback from the tech companies like, oh, whatever, law just hasn't caught up well, okay, but it's still the law, you don't get to change the law because you built really cool technology.
[00:00:20] Paul Roetzer: Welcome to the Marketing AI Show, the podcast that helps your business grow smarter by making artificial intelligence approachable and actionable. You'll hear from top authors, entrepreneurs, researchers, and executives as they share case studies, strategies, and technologies that have the power to transform your business and your career.
[00:00:41] Paul Roetzer: My name is Paul Roetzer. I'm the founder of Marketing AI Institute, and I'm your host.
[00:00:49] Paul Roetzer: Welcome to episode 39 of the Marketing AI Show, and what may be the craziest episode we have had to record yet. I am your host, Paul Roetzer, along with my co-host Mike Kaput up. Mike, how's it going, Paul? Good. You and I are both back in the office after our travels. Last week I was speaking in San Diego.
[00:01:07] Paul Roetzer: And San Francisco. And you were in Denver, if I'm not mistaken, is that right? Correct. Yep. Yeah, lots of AI talks these days. Lots of interest in in presentations. So Mike and I have been hitting the road, so a a lot happened last week and I'm going to get to that in a second. First, this episode is brought to you by the AI for Writer Summit.
[00:01:28] Paul Roetzer: It is next week. You're listening to this, we are recording on Monday, March 20th. It, this will drop on March 21st. The writer summit is on March 30th, if you haven't heard about it yet. It is a half day virtual summit that has more to talk about than I originally expected, we. When we first came with the idea to do this summit was in large part because of all the questions around the impact of generative AI on writing for writers, for content teams, for brands, media companies, publishers, agencies.
[00:01:59] Paul Roetzer: We were getting asked all these questions and we just didn't really have answers for everyone. And so, The idea was let's pull together a virtual summit virtual so we can get as many people as possible to attend and like open up this conversation, free thanks to writer, our presenting sponsor. So there's a free option to register.
[00:02:19] Paul Roetzer: I think we're approaching, like 28 or 2,900 registrants. So there's a lot of people that are trying to figure this space out. And as we're going to find out today in this episode, there's a lot more variables now than there was a week ago, . And so I'm, I'm really glad we're doing the event live. It's, I have not created my keynote yet.
[00:02:40] Paul Roetzer: I know my event team freaks out every time I say that, but honestly, there's so many moving pieces. I just haven't been sure exactly how to say what needs to be said and what to cover. So anyway, that event's coming up on March 30th. You can get email@example.com. It's going to be great.
[00:02:59] Paul Roetzer: We're going to have, I'm going to do an opening keynote. Mike's going to go do a talk on AI tools, may Habib from writer's, going to do talk on AI for content teams. We're going to have a panel on the future of ai. And then we're going to have Anne Hanley from every, everybody writes author, and just all around amazing marketer.
[00:03:15] Paul Roetzer: She and I are going to do a fireside chat kind of on storytelling and writing and the impact of AI and where this could all go. And then we're going to have, moderated q and a with some of the speakers. So, Definitely check that out if you're interested in the generative AI space, specifically around language and writing.
[00:03:32] Paul Roetzer: Okay. Oh, and also quick note, thank you to writer Go, Charlie bis, hyper write rossa.io. Demand Well Gloss AI and copy Matt, who are the supporting sponsors for that event. So we really appreciate everybody's support in bringing that to the market. Okay, Mike, but I need to take a quick drink and get ready for this.
[00:03:51] Paul Roetzer: Lay it out for us. What are we covering today?
[00:03:52] Mike Kaput: All right, so we're going to cover some main topics and some rapid fire topics as usual, but I want to tee up what those are first, because we're going to kind of dive into connecting the dots here and how all of these crazy announcements we've heard this past week are kind of in aggregate changing.
[00:04:13] Mike Kaput: Many, many things about the industry. So we're obviously going to cover GPT-4. It's here. We've started using it. Google and Microsoft had some huge AI announcements about embedding AI in their core products. We had a, what we might call, game-changing announcement from the US Copyright Office about AI generated content, and that we've got a range of announcements.
[00:04:39] Mike Kaput: Microsoft's responsible AI team anthro introducing Claude. A couple AI funding announcements, gen two coming out from runway, and last, but certainly not least, version five of Mid Journey. So there has been. Probably one of the biggest, if not biggest, Newsweek in AI happening in the past seven days. But before we dive into GPT-4 and all the rest of it, why don't you kind of tee up what happened last week for us and kind of how you are thinking about these
[00:05:12] Paul Roetzer: in aggregate.
[00:05:13] Paul Roetzer: Yeah, so to set the stage, you know, we've talked, anybody's attended our Intro to AI class or heard me speak. We talk about this exponential growth curve and how things are going to happen really fast and it's going to be hard to comprehend and keep up. And so last Tuesday was the perfect example of this. So I'm, I'm boarding a flight Tuesday morning, Eastern time.
[00:05:34] Paul Roetzer: I think my fly was at eight 30. I was flying to Denver, transferring in Denver, and then going to San Diego to present at Social Media Marketing World on Wednesday. So as I'm sitting on the runway, no pun intended here, runway puts out a video on Twitter announcing or teasing. A major breakthrough in the next generation of storytelling was dropping on March 20th, so this is about eight 8:30 AM 8:20 AM Eastern time.
[00:05:59] Paul Roetzer: So I messaged that to Mike. I was like, Hey, podcast topic for next week. It's probably text to video. it was, we'll get to that later. Then also before I think before I take off, meta announces they're laying off 10,000 more workers this year in the year of efficiency, as Zuckerberg put it. And embedded within there, if you read between the lines, is they think AI is going to do a bunch of the work that the people are doing, and so we're not going to meet as many people, is kind of the gist of it.
[00:06:26] Paul Roetzer: So there's a lot more to it, but they definitely talk about the role of AI inefficiency and so that was there then. Google announces Palm api. So Palm is their, one of their language models that they're going to open up the APIs for developers to build on. So this would kind of compete with a G P ChatGPT or GPT-4.
[00:06:47] Paul Roetzer: So they're basically coming to market with their own language model to build on, and they're going to infuse these generative AI tools right into Google Workspace. So we'll talk about that in a minute. It's one of our big topics. The thing that was odd to me though is this is happening at like 9:00 AM so I think I'm now in the air.
[00:07:02] Paul Roetzer: I have my wifi going, my $8 united wifi, which thank goodness worked on on this trip, and so I immediately messaged the team. I'm like, okay, something else is dropping today. Maybe it's GPT-4, because why is Google making an announcement at 6:00 AM Pacific time? About something that seems like a really big deal.
[00:07:20] Paul Roetzer: It would appear to me. They're trying to get out ahead of news later today. So that was like my initial takes. Now I'm in flight to Denver. Andro announces cloud, or Claude, I don't know how you say it. Their language model is going to be available . Then Adept announces a 350 million Series B. We'll talk about that in a minute.
[00:07:38] Paul Roetzer: Then I land in Denver and I'm walking through the airport. I have an hour to transfer, and I get to the gate and sure enough, GPT-4 arrives. Now my first inclination. Sam Altman tweeted excited the number four today, and a picture of himself with his backpack. It's like, oh, okay. It's it's about to drop.
[00:07:56] Paul Roetzer: And then sure enough, it did. Then Ahed Meach, who is the CEO of Stability AI, an open source kind of AI platform, teased on Twitter that there's more to come. The release, the releases keep on coming, expect even more in the next few days slash week. So this is all. 2:00 PM Eastern Time on Tuesday . And so by the time I land, I first thought, oh my gosh, I have to rebuild my presentation for Wednesday morning.
[00:08:23] Paul Roetzer: So I messaged Michael Stelzner and the team at, you know, social Media Marketing World, and I was like, Hey, just heads up I have to do a V3 in my deck for tomorrow. Sorry, like, And so I did that, and then I woke up Wednesday morning and I spent the entire day trying to process what in the world was happening and that it was only Wednesday.
[00:08:43] Paul Roetzer: And so when I got on stage Wednesday afternoon, I told that audience we had about 800 people. It was packed. We had to close the doors 10 minutes early because they couldn't fit any more people in the room. Not because of me, just because of like the moment and the topic and like everyone's trying to figure this out.
[00:08:57] Paul Roetzer: And so I got on stage and I said, listen, I'm not even sure what to say to you. All right now, like we are in the midst of the most consequential week in AI history, in my opinion, and I'm still trying to process this myself. So we're going to go through this together, . And that was kind of how I started my talk at Social Media Marketing World.
[00:09:13] Paul Roetzer: So yes, it was a insane week. We have said over and over again, like, it's going to be hard to comprehend what comes next. And sometimes we'll get pushback with people like, oh, it's just hype. I'm like, no, it's not. You have no idea what's about to happen, and I think Tuesday was a really good illustration that we can't really prepare for the rate of change we are in the midst of and it's not going to slow down.
[00:09:36] Paul Roetzer: So with that stage setting, I'll throw it back to you, Mike, and let's actually dig into some of these topics, . So you know,
[00:09:44] Mike Kaput: first and foremost, one of the great examples of what you're talking about is the release of GPT-4. So, GPT-4 has been anticipated for a very long time being released by OpenAI.
[00:09:57] Mike Kaput: It is their latest, most powerful version of their large language model, and it's now being integrated into existing products via api and it's also. Powering ChatGPT, the latest version of their popular conversational AI interface. According to OpenAI, they say quote, GPT-4 can solve difficult problems with greater accuracy, thanks to its broader general knowledge and problem solving capabilities.
[00:10:27] Mike Kaput: So we actually see this firsthand because OpenAI says that G P T surpasses a. GPT-4 surpasses a couple benchmarks that the previous version of ChatGPT was also tested on, so you can actually see some dramatic exponential performance gains here. They tested. J p T four on the uniform bar exam.
[00:10:52] Mike Kaput: It scored in the 90th percentile, which is crazy on its own, but the previous versions of chat j p t scored in the 10th percentile. It
[00:11:02] Paul Roetzer: also, I emailed a few of my lawyer friends, . That one I was like, Hey, heads up.
[00:11:07] Mike Kaput: It is also performed similarly on, I believe it's the Olympiad test in biology, biology.
[00:11:16] Mike Kaput: So it's actually able to achieve dramatically better scores on what we would consider relatively advanced tests of knowledge and reasoning. Now, what's also crazy, this is not available yet, but it will also be able to accept images as inputs. So OpenAI demo that this isn't just about language anymore, it's multimodal.
[00:11:37] Mike Kaput: So they actually. GPT-4 with an image of a hand drawn sketch of what a webpage should look like, and GPT-4 was able to understand the image, understand the input and the directions, and actually provide code to build that page all from literally on a drawing you would make on a notepad. So one other element here before I get into your thoughts, Paul, is that around the same time, OpenAI took a bunch of criticism from the AI community because they're not releasing as many details of these models as they have historically.
[00:12:16] Mike Kaput: And they've actually, Ilia Suz is, one of the co-founders and the chief scientists there actually was quoted in the Verge talking about how OpenAI is essentially less open today because of competitive pressures and safety concerns. So we're going to dive into that, but that's drawing a lot of heat from the c.
[00:12:37] Mike Kaput: Because OpenAI in the past started out as more of a mission focused, open research organization and is acting much more like a major tech giant that it is becoming. So let's start at the beginning here, Paul. How big a deal is this?
[00:12:55] Paul Roetzer: Release it. I mean, it's huge. Obviously we're not going to underplay it. I think that there is some.
[00:13:04] Paul Roetzer: Question about like how much of a leap forward it really is over three or 3.5. But you know, in our early testing, and certainly a lot of the things I've seen online, it it's, it's significantly better. And we know that they have some guardrails in place, some pretty significant guardrails. So my initial testing, so again, if you want to go get it right now, if you have a ChatGPT free account, just pay a 20 bucks for a month and.
[00:13:27] Paul Roetzer: Have at it. Like you can test GPT-4 on your own for $20 right now. And I know that there's some third party applications that have it baked into their solution, but the simplest, fastest way is just go to ChatGPT and get a $20 account, month to month. So my initial take again, I'm hanging out in San Diego, playing around with this thing.
[00:13:46] Paul Roetzer: Is it's, we know it's bigger. So again, they, they went closed. They did not share all the details of how it's trained and how big of a model it is and how much compute power is needed and all of these things. They didn't share any of that, and they can say it's for safety purposes, maybe. I think it's very clear it's for competitive purposes, primarily that they just don't want other people knowing what they're doing.
[00:14:08] Paul Roetzer: And I, Ilia actually said like we made a mistake. Like we shouldn't have ever shared what we were doing. And now we're changing that basically when you got called out on it. So it's bigger. We don't know how much bigger the multimodal is a huge deal. The fact that it can take image inputs and probably video inputs in the near future and it can actually learn from that and then create outputs based on that, is a really big deal.
[00:14:31] Paul Roetzer: That's moving much toward, much closer toward a general intelligence platform. Now they would consider this kind of a general intelligent agent. Like it has general capabilities to do many things that it's tasked with. So it's bigger, it's multimodal, it is certainly smarter, in large part due to advancements in reasoning capabilities.
[00:14:49] Paul Roetzer: And there's an interview on a b ABC with Sam Altman and the CTO O Mirror. Just, I just watched it last night. I don't know when it actually came out. But they talked about the advancements and reasoning as the big thing. And what that does is it gives it the ability to actually create more logical outputs.
[00:15:06] Paul Roetzer: It can analyze its own outputs. It can understand the context of the question, it can look back at other things and understand what you're a actually, actually asking. It gets really good at problem solving, which it wasn't maybe as advanced at before. It can handle ambiguity, so it can really kind of understand what's going on.
[00:15:25] Paul Roetzer: The question, the prompt. If it's not clear, it can start to better understand and connect the dots between things. So reasoning is a really big deal. It's what a number of the research labs have been focused on for the last couple years to really advance these models. They claim it's safer in part because of its reasoning ability, it's going to be less likely to respond to restricted content and less likely to make stuff up.
[00:15:47] Paul Roetzer: And then the other thing that I noticed is it's far more creative. Like if you give it the same prompt, it's night and day. And the example I put up on LinkedIn was an out of office reply and I said like, Hey, I'm out of office, I'm be slower to reply. Because I'm going and doing some talks on AI basically.
[00:16:03] Paul Roetzer: And I did a 3.5 output and a four output. And. It was insane. So that to me was, you know, the first things that started jumping out to me is the impact it's going to have on creativity and getting much, much closer to being reliable in terms of the outputs and the accuracy. Let's talk
[00:16:23] Mike Kaput: about that a little more because really that jump in reasoning capabilities and you know, the test benchmarks they shared really did seem to be eye-opening for a lot of people following the space.
[00:16:35] Mike Kaput: I mean, should we be expecting IMPA an impact on more cognitive abilities, you know, at work in certain types of jobs, especially in marketing and business related function.
[00:16:48] Paul Roetzer: Yeah, I don't, I, again, I don't think people are ready for the impact this can have, because, you know, the thing I've said, and I mentioned this when it comes to the Microsoft and Google, I don't think people give enough attention to.
[00:17:02] Paul Roetzer: These, these tools as, as strategy assistance. So, I mean, I've used it a, a number of times to create strategies for things in my personal life. Like let's say a nutrition plan or things in business, like, you know, how build me a business plan for a startup just to experiment and see, and it's really good, when you start getting into having it do strategy.
[00:17:24] Paul Roetzer: And so I think that it's going to continually. Like, make that impact. But what we've learned is they have a lot of protections in place to, to keep it from showing off its full capabilities what we, you know, what we call guardrails. And Sam and Mira both talked about those guardrails because what they said in the a, b, C interview is that the GPT-4 S been done for seven months and they've basically been in testing and trying to figure out where these emergent abilities were going to be and the things it could do.
[00:17:54] Paul Roetzer: Like the example they gave on A B, C is like, Give you instructions to make a bomb. And Sam said, well, I hope it doesn't, but yes it can. But we've prevented it from doing that. So like that was a known thing. So they put guardrails in place. So if you ask it, how do I make a bomb? It's not going to tell you.
[00:18:12] Paul Roetzer: But then Sam said, well, but you could learn it on Google. Like Google already does that. So you can tell that there is this. Like, yes, this thing is going to be insanely dangerous. But we're going to kind of push forward anyway because there's already a bunch of danger in the world. It's, it's kind of like, it's not the greatest talking points , but it's the reality of what they're doing.
[00:18:36] Paul Roetzer: So with that,
[00:18:38] Mike Kaput: should we be worried about OpenAI's? Call it a lack of openness or transparency and kind of this shift into this closed system that we don't fully understand. We know there are better guardrails, it's safer, that's great, but we have less transparency into how the models are actually working.
[00:19:00] Mike Kaput: And OpenAI, like you mentioned, is not sharing a lot. I mean, how big a danger is this to interpreting what these models are? Cap.
[00:19:10] Paul Roetzer: I don't have a strong point of view on open versus closed models, and I, I really want one to be honest with you. So I saw the Stanford, human Center centered Artificial Intelligence Organization Institute, whatever they call it, put out a poll of like, should these models be open or closed?
[00:19:28] Paul Roetzer: And my, I replied, I think and said like, I actually have no idea. Like I follow this space as closely as anybody. And I don't know, because the people on the everything should be open side are convinced they're right. And the people on the everything should be closed side are convinced they're right. And people who previously like OpenAI thought it should all be open are now saying You will all.
[00:19:51] Paul Roetzer: That we were right. It should actually be closed now. Like we've changed our point of view and you will as well. So they are very confident that having it open is very dangerous to humanity. I believe that, like, so I think there is, there's two sides to this obviously. So the people who want it open largely want open because they don't want it controlled by a select few technology.
[00:20:16] Paul Roetzer: C. So their belief is that even though it's dangerous, if it's out in the open and anyone in society can, can, experiment with it and build on it, then we have a better chance of protecting ourselves when we create something that has danger in it. The closed people like OpenAI believe they are the shepherds of this and that they know better how to control it than society does, and op, you know, leaving it open to everybody else.
[00:20:44] Paul Roetzer: I can completely understand the arguments on both sides. I have probably more fear about it being open. I'm not saying I'm all for like close it off and let OpenAI, Google, Microsoft, figure this out for the rest of society. But I do have major fears about when this, it gets in the hands of people who shouldn't have it and what it can do.
[00:21:11] Paul Roetzer: And I think we're going to find out really fast. Like there, there will be GPT-4 level language models open that'll be, have been copied or built by someone else to humanity by the end of this year, probably sooner. And so we will find out like the next election cycle in America is going to be. I might just turn like social media off for six months.
[00:21:33] Paul Roetzer: Like I can't even imagine what we're about to enter into . And I think the openness is going to accelerate that. I think the closed gives us more time to figure it out. But I don't think closed is going to win. I think there's enough people like the poll I mentioned from Stanford, they only have 328 votes.
[00:21:52] Paul Roetzer: And granted it, that's not a huge sample size considering this, but it was 94% of respondents said that we should keep these models. Wow. So that worries me. I think open's going to win. And I don't know that that's in the best interest of humanity is kind of where I'm at at the moment. Sure.
[00:22:10] Mike Kaput: That's really fascinating.
[00:22:11] Mike Kaput: I think we're also kinda leads in really well to our next topic because we're very quickly, even if you're just getting up to speed on. The chat GPTs of the world, GPT four and all of these issues. You are very quickly going to start using these types of technologies at work. And one of the other big set of announcements this week is that Google and Microsoft are actually embedding a ton of these generative AI capabilities into their core.
[00:22:41] Mike Kaput: Products. So if you use Google products in your business or Microsoft products, if you are not already, you will be using AI in the near future. So Google on their end, announced that developers are now going to have access to its palm language model, which through an API, gives people the ability. To build on top of it in Google Cloud, and the company also announced generative AI features coming to Google Workspace, which is the firm's productivity suite.
[00:23:09] Mike Kaput: So that means you're going to actually see. Much more robust generative AI features in Gmail and docs, for instance, that draft copy on any given topic for you. So all of these generative capabilities we've been talking about through third party tools are starting to become natively integrated to the core productivity software we use every day, which is definitely something we have.
[00:23:35] Mike Kaput: Talked about quite a bit as being on the horizon. And at the same time here Microsoft announced Microsoft 365 co-pilot. So this is an AI tool that they call a co-pilot for work. According to Microsoft Co-pilot can actually use the power of large language models in combination with data in the Microsoft Graph and Microsoft 365 apps.
[00:24:00] Mike Kaput: To increase productivity. So there's two kind of main ways they describe this as working. First, co-pilot works alongside you in popular apps like PowerPoint, word, and Outlook. So in Word, it will now be able to generate drafts for you. In PowerPoint, you can use natural language prompts to create presentations and much.
[00:24:23] Mike Kaput: Co-pilot also enables a new feature they're calling Business Chat, and this is really fascinating as well. It surfaces insights from data across your company and your apps. So all these conversations you're having via Microsoft email services, things in docs, presentations, Excel spreadsheets, you can actually start querying business chat with natural language prompts.
[00:24:47] Mike Kaput: Tell my team how we updated our product strategy and it will generate a status update based on the mornings meetings, emails, and chat threads. So it's this really productivity accelerating digital AI assistant. So as you're seeing these announcements, Paul, how important are these for the future of how
[00:25:09] Paul Roetzer: we work?
[00:25:10] Paul Roetzer: Yeah, my first take was that it's going to rapidly transform the future of all knowledge, work, creativity, and productivity. And to give context. So the talk I did Friday in San Francisco was to a group of like a hundred, 150 executives, you know, from a financial services industry. And I showed, I talked.
[00:25:32] Paul Roetzer: So sat headed this quote a couple years ago about their mission is to basically build AI first companies in every industry. And then I showed the minute and a half clip of 365 co-pilot. Then I had the quote from Sundar Pacha, the CEO of Alphabet and Google about AI being the most important thing humanity has ever worked on.
[00:25:50] Paul Roetzer: And then I showed the Google workspace. Their jaws were on the floor. Like to them what it did was like what? What Chad Cheapy did G P T G P T did for marketers and writers. I think this does for the C-Suite, because you look at Microsoft and you look at Google and say, okay, that's our core tech stack.
[00:26:09] Paul Roetzer: Like that's what we're built on. And then you show all these use cases and you can sit there and watch like, oh wait, we're paying an administrative assistant to take notes in meetings and figure out action items and build them into the project management system. Or we have entire analyst teams built to do the analysis they just did of that spreadsheet.
[00:26:27] Paul Roetzer: So now if you're a leader of an organization, you watch these one and a half minute teasers of the capabilities, you immediately have to start thinking like, oh my gosh. This isn't just like a series of tools. This changes everything about our staffing, our operations, our tech stack, and we have no one on our team who knows what to do about it.
[00:26:48] Paul Roetzer: So I think that to me is the biggest thing is like by Microsoft and Google jumping in in this way. And I mean, those are really impressive trailers. Like if, if it does what they're showing there out of the box and it's actually capable of these things, it's going to have a massive effect immediately on every business of every.
[00:27:07] Paul Roetzer: And they're not prepared for it at.
[00:27:12] Mike Kaput: Yeah. That's wild. And I think when we talk about not being prepared, one of the elements here that you talked about on LinkedIn, you said about this announcement, or about the Google announcement specifically. So many choices for businesses, it's going to be tricky to navigate the evolving tech stack.
[00:27:32] Mike Kaput: Can you unpack that a little more for us? Like what kind of decisions are leaders and. Professionals going to have to start making around this technology?
[00:27:41] Paul Roetzer: Well, if you just look at you and I mean, we follow this space very closely. We talked, we've interviewed hundreds of these AI tool companies. We use, you know, probably a dozen to two dozen ourselves every week.
[00:27:54] Paul Roetzer: I sat back and said, man, I wonder how many of those we're going to need six months from now, like, we're Google, we, we largely use Google. We work. And you just look at it and think, do I need writing tools? Like I don't know, I don't know the answer to this. It it like, how many writing tools is too many?
[00:28:12] Paul Roetzer: So that was one of the questions I posed. I think one of the comments is like, how many writing tools is too many for a tech stack? Like what are, how many chrome extensions can I possibly have running at one time? So if I have an editing tool and a writing tool and then another writing tool, am I going to have like a writing tool for email and a writing tool for docs and a writing tool for proposals and a writing tool for sales and Ari?
[00:28:32] Paul Roetzer: Or am I just be like, just consolidate it all, like just if Google or Office, Microsoft Office are going to do this for me. Like maybe it just brings it all in. So I think it present. Some real challenges moving forward because again, the people making these decisions don't really understand the technology.
[00:28:48] Paul Roetzer: Like they're not going to be able to say, well, if I have ChatGPT for $20 a month, do I also need this AI writing tool for $90 a month? And then do I need this editing tool for $50 a month? And then how does this scale? So if you're trying to make a decision as an individual creator, it's hard enough if you're trying to do this for an enterprise, I, It's such a cliche, old saying, but nobody got fired for buying.
[00:29:14] Paul Roetzer: IBM was like the thing in the eighties. Yeah. And I think when you look at generative AI in the 2000 twenties, you may end up with something similar. It's like, well, it's just consolidate it with a big guy because at least we know we're not going to screw it up. And. That company will still be around in three years.
[00:29:35] Paul Roetzer: Because I think there's a real possibility of massive consolidation. Like I think we're going to have an explosion of generative AI tools. Although our next topic may change that slightly. I think there's going to be an explosion of tools, but I think there may end up being a consolidation of Tech Stacks X, if that makes sense.
[00:29:54] Mike Kaput: It absolutely does. So, Basically saying, I mean, as these bigger platforms with established business productivity software begin to make them smarter with generative ai, they're essentially duplicating the capabilities of a whole range of third party startups that are receiving tens or hundreds of millions of dollars in funding.
[00:30:15] Mike Kaput: So are we saying that's going to, that that competitive landscape has kind of fundamentally changed due to these types of announce?
[00:30:23] Paul Roetzer: It sure seems like it to me. I mean there a lot of the generative AI applications or SaaS companies created are tr are actually features within Microsoft and Google, if you think about it.
[00:30:37] Paul Roetzer: So yeah, I mean the note I had made was like, if you're a SaaS leader, you gotta get really serious, really fast about what is actually defensible in your business model. And you know, like for an example, like I don't. You know, we talk about writing a lot, but let's say like the slide creation, they both showed the ability to either in Google Slides or in PowerPoint to text prompt the creation of a deck.
[00:31:01] Paul Roetzer: Well, tome is like a really cool AI generated slide company that I think just raised like 40 or 50 million or something like that. Yeah. To generate presentations with text prompts. That's an example. It's like, I tome looks awesome, and maybe it is defensible. Maybe it's like a great business model and they have some mode around it.
[00:31:20] Paul Roetzer: I don't know, but I also look at it and think, well, I don't know if I'd need that separate thing, but maybe I do. And so I think what I was basically saying in that post is like, y your buyers, if you, if you are one of these application layer SaaS companies know that your buyers are asking the same questions.
[00:31:40] Paul Roetzer: We're asking ourselves out loud right now, mm, do I really need that tool? If I can do this in Microsoft or Google? , that is not us being critical of your solution and what you've built. It is a harsh reality that that is what your buyers and your existing customers are going to be asking themselves.
[00:31:58] Paul Roetzer: I've consolidated some of my own because I didn't need all these tools. So if I'm doing it, I'm sure that there's other people doing it. And most of these SaaS companies want to sell to teams. They, you know, individual buyers are great self-serve model, but everybody eventually goes upstream. If they didn't start.
[00:32:14] Paul Roetzer: And so if you want to go upstream and sell to like 500 person marketing teams, you better have a really good answer to the question of what makes it different than what Microsoft's about to come out with. Because there's no way you're winning that sale if they think they can just roll it into Microsoft.
[00:32:30] Paul Roetzer: And Microsoft can undercut the market if they want to. They can just build these tools in and they can do, as a loss leader if they have to just get you into their cloud business. Like I wouldn't want to be competing with them. So, yeah, it's like, and I think that was the big takeaway was for SaaS and then investors.
[00:32:46] Paul Roetzer: Same deal we've said before. Like you gotta be super critical of your portfolio right now and all future generative AI investments, knowing that the big guys are now playing the game. And then the other thing we said was, It's not just a marketing sales service thing. This is ops and HR and finance and legal and it.
[00:33:03] Paul Roetzer: So if you're the CEO or you report to the CEO, you, you better be thinking cross-function across the entire organization because now that it's going to be embedded into Microsoft and Google, it touches everything. It's not just a marketing thing.
[00:33:19] Mike Kaput: And as if there weren't enough hard questions to answer in this space.
[00:33:24] Mike Kaput: Our third main topic really is something that we're beginning to consider just a total game changer for generative ai. And so very recently, the US copyright off as came out with some commentary on generative AI tools, and the conclusion is essentially, Something that's going to surprise a lot of people in the market.
[00:33:50] Mike Kaput: If you're using generative AI tools today to create content, you know, articles, blog posts, books, images, whatever, according to the US Copyright Office, it sure seems like you do not own that content. That means that anyone can reproduce that content that has been AI created without your permission, and create derivative works from it, display it, perform it, and sell it.
[00:34:14] Mike Kaput: So this happened, you know, a few days ago. The copyright office is launching an initiative to examine copyright law and policy issues raised by aIncluding how the scope of copyright actually works with these AI tools that can generate content. They, the copyright office said quote, the initiative is a direct response to the recent striking advances in generative AI technologies and the rapidly grown use by individuals and businesses, and they go on to mention they've gotten many, many requests from Congress and members of the public to examine the issues around copyright.
[00:34:51] Mike Kaput: So they actually issued updated registration guidance that has an immediate effect on your ability to protect your original work. So a few really quick takeaways here, and then Paul, I want you to kind of unpack this for us, is that they lean very heavily on talking about. The fact it's well established that copyright can only protect material that is the product of human creativity.
[00:35:18] Mike Kaput: So they use the term author, which is in both the Constitution and the Copyright Act, and say that that term excludes non-humans. I e AI cannot own a, it cannot be, used to defend copyright if the artificial intelligence created the work. And they go on to talk about if a works traditional elements of authorship were produced by machine.
[00:35:44] Mike Kaput: It lacks human auth authorship. The office will not register it. So for example, when AI technology receives a prompt from a human solely and produces the kind of complex content we just talked about, they are saying that this is not going to be protected by copyright for you or your business. So I want to get into.
[00:36:07] Mike Kaput: Some of the nuances here, Paul, because the way we're interpreting this, your kind of takeaway here was you need to talk to your IP attorneys and start really seriously thinking about this topic because it seems like a really big deal. Is that,
[00:36:22] Paul Roetzer: do we have that right? Yeah. So the way this sort of snowballed for us was, Sunday morning, so this was yesterday.
[00:36:32] Paul Roetzer: When I'm recording this, I saw an, an AI ethicist. I follow tweeted something about this and so I kind of start like following the pattern, where did their tweet come from? And I land on the March 16th announcement about this new AI initiative from the copyright office. And in like the second or third paragraph it says, we've also updated our guidelines for submission of copyrights.
[00:36:51] Paul Roetzer: So the significance of the copyright, if you're not not familiar with. Us copyright law is that you have ownership of something. So if you create a song or a poem or an article or a blog post, or a video or an image or whatever it is, once you create it and it's in a fixed space, so you, you release it somewhere, you, you've put it out into the world.
[00:37:12] Paul Roetzer: You technically own it, but you want to submit those things to the copyright office for like legal protection. So if someone steals your stuff, you have to then, Prove that you had published it at this time. What you do is like once you create something, like let's say we create our courses, or I create a new presentation and I want legal protection from it, so nobody steals that idea from us, that knowledge, that intellectual property from us.
[00:37:37] Paul Roetzer: I go to my IP attorney who we use all the time, and I say, here you go. Here's like 17 new courses. Do your thing. Well, we got back from them in the fall, or I guess this was December. Well, would you have any AI generated image in this deck? The copyright office wants to. And I was like, well, yeah, but it's like three images.
[00:37:55] Paul Roetzer: Like you can't copyright the course because of three images, and it was like this gray area. So I've been aware that there was, this was an issue for months and have thought about this, but by Sunday morning I go to like just put something quick on LinkedIn. Like, Hey, fyi, the copyright office is looking at this.
[00:38:10] Paul Roetzer: But once I dug into the updated guidance, I realized, oh no, this is like a major deal because I don't know of a single AI writing tool company that. That I have seen a point of view on the song, like any knowledge or information about it. Then I started thinking about all the recent major breakthroughs in general AI that we just talked about and how people are using GPT-4 to write entire articles, which it is capable of, or they're generating images and they're, you know, using those and other collateral and everything.
[00:38:41] Paul Roetzer: And what it became very clear when you read that guidance is you have no ownership. So if I publish, like I, there was a big, was it Reid Hoffman that just published a book using Yeah. J go build a course on it. Like Reid has no ownership of that thing. He just created, what they're basically saying is if you submit a thousand word article or a 50,000 word book manuscript, so we have like the copyright on our book basically.
[00:39:07] Paul Roetzer: anything that was generated by ai, you cannot claim a copyright on. Now you can get a copyright on the other stuff that was yours, the original work from you. But a text prompt doesn't count. So like if you gave a text prompt and it gave an output, you have no ownership of it. So I know people who are publishing blog posts using GPT-4, like they're, they're working on plans to release a whole bunch of content.
[00:39:28] Paul Roetzer: Anybody can take it from 'em. They can't do anything about it because it what's going to happen. If you try and litigate, so if you try and stop someone from stealing it, the copyright office will come to you and say, okay, proof that a human did this, and they're going to want proof. They're going to want to know what prompts you use.
[00:39:44] Paul Roetzer: What was the output of the prompts? And so now again, it's not going to, it's not like they can do this. They know the staffing to do this for every single instance of copyright infringement, but this is where they're going to have to go, is they're going to have to figure out how do we scale the the over. That is going to be needed to prevent this from happening.
[00:40:03] Paul Roetzer: So I don't, I mean, as Sunday went on, after I post the original thing on LinkedIn, I started realizing like, oh my God, wait, this may actually be a much bigger thing than I was originally thinking it was. And because you realized it's immediate, like if you wanted to submit something right now, AI generated, or if you're putting a copyright, here's a, here's a practical example.
[00:40:23] Paul Roetzer: Everybody has like, go to the website, go to their blog at the bottom, copyright 2023, marketing, AI Institute, whatever. So you're claiming a copyright on everything you're publishing, even if you haven't submitted it to the copyright office yet, if AI generated that content, you don't actually have a copyright on it.
[00:40:40] Paul Roetzer: So if you're a big healthcare company and you're using AI or like right content, another healthcare company could take your content and you can't do anything about it. So, I mean, the implications are actually huge. And so I started thinking by last night like, wait a second, this could actually be a major, major hindrance on the use of generative ai, especially for enterprises because there's, there's no way people that are big enough to have a legal team are going to allow.
[00:41:05] Paul Roetzer: this And so I could see the pushback from the tech companies like, oh, whatever, law just hasn't caught up well, okay, but it's still the law, like you don't get to change the law because you built really cool technology. So I could see a lot of these software companies just feeling like this is just the government not catching up.
[00:41:24] Paul Roetzer: And I could sort of hear that argument. It's the law and it's not going to catch up for a while, so you're going to have to live within the confines of this. And that just changes the way these tools are used, I think, and nobody's talking about this , it's, it's wild to me that I can't, now that I'm thinking about it, like how, how do you raise all the money these companies have raised without the VCs saying, are there any copyright?
[00:41:54] Paul Roetzer: Barriers to this, like, is, is this actually going to get shut down? Because we can't actually generate and publish the stuff from these things.
[00:42:01] Mike Kaput: And so I think it's important to give even more context there for the audience, you know? People have been talking about, okay, what are some of the copyright issues around how some of the models themselves are trained vendors have talked about?
[00:42:15] Mike Kaput: Okay. We are, we hear and very much respect concerns around say plagiarism, and there's plenty of messaging about how it's creating original content, but the copyright element is a whole different animal. Especi. Perhaps for some of the companies that you mentioned that don't typically fall into what we might call the Silicon Valley move fast and break things perspective, right?
[00:42:38] Mike Kaput: Yes, plenty of companies do that. I'm not saying it's good or bad, it's just is. But there we, you know, in our past lives, we were agency professionals working with. Hundreds of different types of companies, both big and small, very slow enterprises and bureaucratic organizations, and very fast moving startups.
[00:42:58] Mike Kaput: There is at least a significant portion of our former clients where this, I could tell you, would immediately be the number one cause for
[00:43:06] Paul Roetzer: concern. Shut it down about, yeah. Yeah. Generative a be shutting down right now. Like they, I agree with you a hundred percent. Like we work with a lot of Fortune 500 companies, highly regular industries.
[00:43:16] Paul Roetzer: If legal knew this, It's done. Like they're just, there's a meeting happening in conference rooms right now. How are we using generative ai? Because yes, everything we've talked about to date about the use of generative AI has been more on the ethical side and you know, it's not plagiarism. Okay. You deal with that.
[00:43:33] Paul Roetzer: We haven't publicly talked about this as a major issue. And I think it might be the issue like, because yeah, it's like if you're going to like it starts to impact your staffing, your content strategy. So just because you can. Create content and publish it with these tools. Doesn't mean you should from an ethical standpoint and from a principal standpoint, but it might be that you're not actually allowed because you can't copyright the stuff.
[00:44:01] Paul Roetzer: And that's where the legal teams are going to get involved in. We are not attorneys. Like I took a business law class and I've, I feel like I've got a master's degree in intellectual property the last 20 years as a business owner. But if we are not giving you legal advice, I'm saying talk to your legal teams before you go scale the use of content and video generation and image generation, all these things, make sure you have the hard conversation right now before you start making staffing decisions based on this and tech stack decision.
[00:44:31] Paul Roetzer: Like, we're at the point where we can kind of like rewind a little bit and be logical about where we go from here. But 3, 6, 12 months from now, you're going to have made some decisions around scale that are going to be really hard to wind back. And so I think this needs to, like you said, it needs to become like a top of the list priority item.
[00:44:50] Paul Roetzer: To be thinking about and like, give you an example, another place this rolls in. Speaking of our agency, the past agency, um , we, we worked under work for hire agreements. So if a client came to us in the contract, it was a work for hire agreement, which means we're relinquishing the copyright to what we create.
[00:45:08] Paul Roetzer: You are paying us, we will create this ebook, or this course, or this webinar, or this video script or whatever it is. We're creating it for you. You own the copy. What this means is if you are hiring an agency or an outside freelancer to create something for you, and they use AI to create it and don't tell you, you actually don't have the copyright you thought you had, you have no protection of the output.
[00:45:32] Paul Roetzer: So the trickle down of this is massive. And as again, if, if agencies were planning on scaling up using generative AI to get their services going, this again, if I was, if I was still the CEO of an. I'd be like, oh crap. Like I didn't even think about this. And I would be revisiting, what are we going to do? How are we going to deal with this?
[00:45:53] Paul Roetzer: And I would, I would be on a call with my IP attorney right now.
[00:45:58] Mike Kaput: So obviously this is still hot off the presses. We are not giving out legal advice and we're still figuring out our guidance and our perspective on this. But do you have any thoughts around, What impact this'll have on writers and marketers in terms of the use cases that they're looking at for these tools?
[00:46:15] Mike Kaput: We're all vetting, yeah. AI technology that can do everything from right to create audio, to create video. How do we have to change how we're doing our work as an individual, kind of in the trenches as a creator, marketer or a writer?
[00:46:31] Paul Roetzer: The assumption that a lot of people seem to make about AI writing tools is that they just.
[00:46:36] Paul Roetzer: Content articles, blog posts, emails like it writes. The reality is language models can be used for a a lot of different purposes, like content summarization, where it's taking your original work, the human generated original work, and creating a summary of it. Again, not an attorney, but that seems safe to me.
[00:46:53] Paul Roetzer: The original authorship comes from the human and the machine is there to assist in the summarized. The same goes for transcription. Our podcast is a great example. We, you and I are having this conversation. All these ideas are our intellectual property. They're covered under copyright, under the Marketing AI Institute.
[00:47:10] Paul Roetzer: When we transcribe it's just the machine transcribing the original authored words, us putting it out, fixes it in the world, so it's owned and fixed. Those are the two criteria of of copy. So transcription and summarization, those seem very safe to me. They seem like really logical use cases. I also look at things where the copyright doesn't really matter.
[00:47:29] Paul Roetzer: So if I'm sending sales emails or like marketing emails, you know, like I'm not submitting those to my IP attorney to copyright every sales email that goes out. So I could see that probably being safe. So I think what needs to happen, and this is where my, like when I'm thinking about our AI for Writer Summit next week, like where I'm kind of going with my keynote now, which I now, now have 10 days to create.
[00:47:50] Paul Roetzer: I'm trying to think through like what guidance we can offer people of what appears to be safe. Again, still consult with your legal team, but like these seem to be the safe use cases. And again, if I was an AI writing tool, SaaS company, I would do the exact same thing because you may be guiding your customers to use your tools in a way that they're not legally going to be allowed to.
[00:48:12] Paul Roetzer: So I may steer into the other use cases where it is truly ideation and outlining and like these things that seem real safe, and I think they're walking a fine line because everyone I see is like, well, the human's gotta be in the loop. It's just a writing assistant tool. And yet they're teaching them, go ahead and write the 600 word blog post and then just edit it.
[00:48:34] Paul Roetzer: Right? It's not clear to me that that's going to get a copyright, like it doesn't seem. I think they need to talk to their legal teams as well to get better guidance for their customers. And they need to do it fast because the market is going to be very confused very quickly.
[00:48:51] Mike Kaput: Yeah. Just another good example of why I think we're seeing so much interest in the AI for Writers Summit, because these kind of conversations can't.
[00:49:02] Mike Kaput: For months or years to be had and everyone is figuring it out at the same time, or needs to be. Yeah. So are you, you know, ready for some rapid fire now? I mean, I feel like we've covered Yeah. I feel like covered a year's worth of content here, .
[00:49:18] Paul Roetzer: Yeah. I think we just like made people's week not so great.
[00:49:20] Paul Roetzer: Yeah. Right, right. Riding the wave of all this excitement and it's like crashing down. So sorry if we're like the bummer in the party here. I think it's important people are thinking about this, but yes. Let's end with some, some other ina insanity in AI right now.
[00:49:37] Mike Kaput: Well, first up, Microsoft had a different AI focused announcement.
[00:49:43] Mike Kaput: They laid off a key responsible AI team. So Microsoft laid off its entire ethics and society team within the artificial intelligence organization as part of recent layoffs that affected 10,000 employees across the company. What, how'd this make you feel about some of the compromises being made on AI ethics versus market conditions?
[00:50:09] Paul Roetzer: All I'll say, The arguments for the closed models are coming from OpenAI and Microsoft and Google, the same people who are laying off their ethics teams for being roadblocks to releasing what they're releasing. We have, we have a problem as a society. The the people that we are trusting to shepherd the closed models into our world are getting rid of the ethical teams that are supposed to make sure they're done safe.
[00:50:40] Paul Roetzer: That's all I'll say on that.
[00:50:44] Mike Kaput: interesting times. A, another big topic is that Enro, which is a major player in AI models, released Claude, which is the next generation AI assistant, based on their research now. Here's where you see the flip side in some ways of what, what you just mentioned. Anthro, at least in their messaging, specifically calls out that they're training helpful, honest, and harmless AI systems.
[00:51:11] Mike Kaput: So Claude is capable of a wide variety of conversational and text processing tasks, and you can do things with it, like summarize, write. Answer Q and As and much more. And they say early customers report that Claude is much less likely to produce harmful outputs and it's easier to converse with and it's more steerable.
[00:51:34] Mike Kaput: So you get your desired output with less effort. So what was your impression of the importance of Philanthropics announcement?
[00:51:42] Paul Roetzer: So a couple quick notes here. One, you know, when we talked about a couple episodes ago, how even. The organizations that want to control this, what if their employees leave and don't?
[00:51:56] Paul Roetzer: And so a lot, there's been a lot of talk about people leaving to go to OpenAI to work. Well, anthro was created by former OpenAI employees. Now they went and went, the Safer Rooted would appear. But they've raised 1.3 billion. Like the reason this is a big deal is because these companies are raising a ton of money.
[00:52:16] Paul Roetzer: . And so I think it's, For people who didn't know that name before. There's some, there's some companies in this space we need to pay attention to. It's not all Microsoft and Google and OpenAI. It's anthro and coherent and stability. Ai, like these are the organizations that are, that are pushing forward, with these language models and going to affect the ways we use them and maybe the alternative options we may have in the coming months.
[00:52:41] Paul Roetzer: And,
[00:52:44] Mike Kaput: Yeah, and another great example of that is a company called Adept, which again, probably many people have not heard of or followed. Just raised a 350 million funding round, and they're building a future in which you can tell your computer to do something in your own words and just watch it happen on your screen.
[00:53:04] Mike Kaput: We. Into this topic of couple episodes ago, talking about your idea of the world of bits and what can happen when we're actually able to manipulate what we're doing on physical hardware or in the physical world with machine prompts like this. So this is a huge fundraising announcement. It's part of their Series B, led by General Catalyst and spark capital.
[00:53:29] Mike Kaput: And so they're really trying to build a future where you spend less time on manual multi-step processes. We talked about this also in the context of HubSpot's chats spott, tool, making it easier to do things with less clicks in a platform we use every day. So this really seems like validation for your hypothesis that we're at least exploring this direction for AI assistance in the work.
[00:53:55] Paul Roetzer: Yeah, I mean the concept here is these action transformers. So the same transformers are the architecture that G P T is built on generative pre-trained transformer, not familiar, but G P T stands for. So this is the idea of being able to create transformers that enable action. So it's, you're going to hear a lot more about this space.
[00:54:13] Paul Roetzer: Yeah, I think there, there's a race right now to do this in a very big way, and my guess is we're going to start to see some stuff before the end of 2023, where you're going to be able, You know, it may start being a tool you can actually go test out yourself.
[00:54:29] Mike Kaput: Wow. Yeah, definitely exciting. And you know, just in case we didn't have enough big announcements and news, two other huge ones.
[00:54:37] Mike Kaput: First up, gen Two by Runway is coming very soon to runway. ML Runway is another major player in the space. Again, probably doesn't get as much love as like the OpenAIs of the world. And this tool literally helps you realistically and consistency consistently synthesize new videos. So we're talking. Text to video generation, video to video generation.
[00:55:03] Mike Kaput: Literally giving you the power of professional quality video creation without filming anything and simply by writing a text prompt into a chat
[00:55:13] Paul Roetzer: box. Insane tech, it's kind of what we guessed like this is what we said last week. I assumed it was going to be text to video. It. The caveat, obviously now is do you own what you create?
[00:55:25] Paul Roetzer: Yes. I think that's going to be the question on everyone's mind. Like, this is amazing. Do I own it? Like if If I start creating movies or shorts or trailers for my company or whatever it is, am I going to actually own the end product? I'm not even sure that they can answer that que the answer is no. It would appear, the answer is no.
[00:55:44] Paul Roetzer: But I don't. Diminish how insane this tech is and how awesome, it appears what they're doing. And I, as anyone's listened, this show before knows we're big fans of runway. It's definitely one of the most innovative companies that we're tracking right now. So yes, awesome tech. Do you own? It que
[00:56:03] Mike Kaput: question marks in the future.
[00:56:04] Mike Kaput: which, yeah, if you think about it, I mean, if you're an individual creator, obviously that could matter in some context, but individuals may be listening saying, oh, who cares? I want to go create like a professional great feature film. Okay, yeah. But can you eventually, as we predict will happen, someday, win the Oscar for it.
[00:56:20] Mike Kaput: Is a studio going to help fund you? Probably not with those kinds of restrictions.
[00:56:26] Paul Roetzer: Yeah, it's going to be a lot of, a lot of, money poured into Washington to try and affect the future of copyright. Yes. I'm, I'm predicting very, very soon.
[00:56:38] Mike Kaput: All right, and finally as we wrap up this mega episode here, we saw on Wednesday mid journey announced version five of its AI image synthesis tool.
[00:56:50] Mike Kaput: So this is AI image generation. Just seeing the examples of what people are creating with this new model. This new version is stunning. I mean, we're seeing Hyperrealistic, ai, art and images that some people are calling creepy and or too perfect because it's so good. It's available now as an alpha test for customers who subscribe to mid journey service.
[00:57:14] Mike Kaput: Again, to your point, amazing tech it sounds like, but also interesting. Both copyright issues, but maybe some societal impacts as well as we get to truly realistic AI image generation. What do you think?
[00:57:29] Paul Roetzer: Yeah, I mean the, we don't, I haven't used Mid journey myself, but just following along and looking at the outputs and the difference between version one through version five, like the Human Eye was the one I saw, like the Iris and the human eye.
[00:57:40] Paul Roetzer: It was like, holy cow. Having just come from the Social Media Marketing World Conference, the one thought I have is what is going to even be real? On TikTok and Instagram and whatever in the future, like you're going to be able to create anything and have it look real. And I don't know that I'm ready for that
[00:57:59] Paul Roetzer: Like that's, so I've said before, like part of the reason I'm doing what we're doing is just to like try and figure out where this all goes and what does society look like and what does the workforce look like and what are my kids going to do? And there's sometimes where I just want to like shut my brain off.
[00:58:14] Paul Roetzer: It's like I just can't even think about it right now. Like I. And this is one of those weeks, like this is a lot. So if you stuck with us to the end, go get a drink. Like, I think we all need to just kind of sit back and process. How insane this really is. You know, we've said numerous times, this is like wild times, crazy times, and that I just, it just keeps getting crazier.
[00:58:34] Paul Roetzer: So , we're all in this together. We'll figure it out.
[00:58:39] Mike Kaput: Well, that's what, at least the bright side of it is that we have you sharing your insight and your experience with the audience so that we can figure it out, hopefully in a better and faster and more sustainable. That some people are able to do out there.
[00:58:54] Mike Kaput: So thank you as always Paul, for the time and insight
[00:58:57] Paul Roetzer: man. If anybody has domain expertise and some of the stuff we talk about, like be sure to reach out to Mike and I like, we'd love to hear from people and you know, never know when we're going to need some like expert insights to help tell this story because it's going in some wild directions.
[00:59:10] Paul Roetzer: So, yeah. Thanks as always. We will be back next week with another episode. Hopefully it's not as crazy of a week in AI as last week was, but you just never know anymore. So thanks again for joining us. We'll talk to you next week.
[00:59:22] Paul Roetzer:
[00:59:22] Paul Roetzer: Thanks for listening to the Marketing AI Show. If you like what you heard, you can subscribe on your favorite podcast app, and if you're ready to continue your learning, head over to www.marketingaiinstitute.com. Be sure to subscribe to our weekly newsletter, check out our free monthly webinars, and explore dozens of online courses and professional certifications.
[00:59:44] Paul Roetzer: Until next time, stay curious and explore AI.
Cathy McPhillips is the Chief Growth Officer at Marketing AI Institute.