Join us as we explore the latest AI developments, including OpenAI's anticipated announcement, Google's AI roadmap, and Microsoft's new in-house model, MAI-1. We break down the rumors, analyze the potential impact on the future of search and information access, and discuss what to expect from key players in the coming months. Don't miss this engaging discussion on the rapidly evolving world of AI.
Listen or watch below—and see below for show notes and the transcript.
Today’s episode is Episode 97, which means we’re almost to the milestone of Episode 100! To celebrate, we plan on releasing a special Episode 100 on Thursday, May 30 that is all about YOU.
On that episode, we plan on answering audience submitted questions and anyone can submit a question for consideration.
The link below will take you to a quick form that has a spot for entering your questions. We will curate and answer as many as we can in a couple weeks.
Submit your questions before May 30 at https://bit.ly/aishow100
Listen Now
Watch the Video
Timestamps
00:03:15 — OpenAI’s Big Announcement
00:11:35 — Google’s CEO Sundar Pichai Lays Out His AI Roadmap
00:18:00 — Microsoft Readies New AI Model
00:20:42 — Anthropic Founders Share Roadmap to Advance AI
00:25:36 — Zapier Central
00:28:39 — Microsoft’s Fourth Annual Work Trend Index: AI at Work Is Here
00:33:54 — Microsoft's New Prompting Features in Copilot
00:36:42 — Google DeepMind’s new AI can model DNA, RNA, and ‘all life’s molecules’
00:40:57 — AI Companions
Summary
OpenAI’s Big Announcement
The rumors of an OpenAI search product to rival Google and Perplexity may not have come to fruition just yet, but that doesn't mean we won't see something similar in the coming months. The rapid pace of development in this field suggests that we could be on the brink of a major shift in how we search for and access information online.
As we continue to follow this ongoing story, it's crucial to stay informed about the latest advancements and speculation surrounding OpenAI and its competitors.
The world of AI and search is evolving rapidly, and a single episode or article can only scratch the surface.
So, while Monday's announcement may not have been the groundbreaking moment in search many had predicted, it is clear that the stage is set for a series of innovative developments in the months to come.
Google CEO Lays Out AI Roadmap
Google CEO Sundar Pichai just did an in-depth interview with Bloomberg’s Emily Chang about the future of AI and Google’s place in it.
The interview covered a ton of different topics, including: Google’s recent stumbles in AI, whether or not the company is behind in the AI arms race, and if Sundar Pichai is the right person to lead the company.
This interview comes in advance of Google’s I/O developer conference next week, and amidst a big push being reported by The Information internally at the company to add 100s of salespeople and engineers to AI teams within Google Cloud—all in an effort to push Gemini and other AI products to companies.
Meet Microsoft MAI-1
Microsoft is now training its own in-house AI model to compete with Google, OpenAI, and Anthropic.
The Information reports that the new model is internally referred to as MAI-1 and is being overseen by Mustafa Suleyman, the former CEO of AI startup Inflection, which was essentially acquired by Microsoft earlier this year.
According to the reporting, this is a new model, not one that Inflection built, though it may build on Inflection’s training data and technology.
MAI-1 will be about 500B parameters in size. (The Information reports that GPT-4 has over 1 trillion parameters; smaller models form Meta and Mistral have about 70B, though there is a version of Llama 3 coming out that will have 405B.)
People close to the matter told The Information that it’s not clear yet what purpose the new model will serve, and that will depend on how well it performs. Microsoft could, says the outlet, preview the new model as soon as later this month during its Build developer conference.
Links Referenced in the Show
- OpenAI Is Gearing Up for Something Big
- OpenAI Considers Postponing Product Event
- OpenAI Google search competitor
- Introducing the Model Spec
- Understanding the source of what we see and hear online
- Google’s CEO Lays Out His AI Roadmap
- Microsoft Preps New AI Model
- Interview Reveals Anthropic Founders Roadmap to Advance AI
- Zapier Central
- 2024 Work Trend Index Report: AI at Work Is Here
- Microsoft New Copilot AI features
- Google DeepMind’s new AI can model DNA, RNA, and ‘all life’s molecules’
- AI Companions
- Working with AI Companions
- The Future of AI Dating
- The Top 100 GenAI Consumer Apps
Read the Transcription
Disclaimer: This transcription was written by AI, thanks to Descript, and has not been edited for content.
[00:00:00] Paul Roetzer: Aren't like worried about this going wrong? I'm like, yeah, a hundred percent I am. But I think you have to, in
[00:00:05] Paul Roetzer: our world, like you have to choose be optimistic that, you know, choose to believe we will find a path forward. And then if enough people are thinking about the responsible human centered application of AI in society, in business and education, that we have a far greater chance of coming out
[00:00:24] Paul Roetzer: on the positive side of this.
[00:00:25] Paul Roetzer: Welcome to the Artificial Intelligence Show, the podcast that helps your business grow smarter by making AI approachable and actionable. My name is Paul Roetzer. I'm the founder and CEO of Marketing AI Institute, and I'm your host. Each week, I'm joined by my co host, and Marketing AI Institute Chief Content Officer, Mike Kaput, as we break down all the AI news that matters and give you insights and perspectives that you can use to advance your company and your career.
[00:00:55] Paul Roetzer: Join us as we accelerate AI literacy for [00:01:00] all.
[00:01:02] Paul Roetzer: Welcome to episode 97 of the Artificial Intelligence Show. I'm your host, Paul Roetzer, along with my co host, Mike Kaput. We are recording a little early this time. It is Friday, May 10th at 1. at 1. 30
[00:01:15] Paul Roetzer: m. Eastern time. So if OpenAI announces GPT on
[00:01:18] Paul Roetzer: Monday, I don't we're going to do GPT 5, but I think we've got announcements coming on Monday.
[00:01:23] Paul Roetzer: Um,
[00:01:24] Paul Roetzer: We will get to those on episode 98. We're
[00:01:27] Paul Roetzer: doing something special for episode so we, as I mentioned, we are episode 97, which means we are coming up very quickly on episode 100.
[00:01:34] Paul Roetzer: People have been asking me, like, what are we going to do for episode 100?
[00:01:37] Paul Roetzer: didn't have a plan. but Mike and I were actually on trip earlier this week. We were talking about this. My dad actually reached out to me and he's like, hey, you guys should just answer
[00:01:46] Paul Roetzer: And I thought, yeah, that's actually a good idea.
[00:01:48] Paul Roetzer: It's something we'd bounced around. we're doing something special. That's where Mike and I landed,
[00:01:52] Paul Roetzer: we're going to dedicate episode 100. going to be a special episode, so we'll still do our weekly. Episode 99 will be our regular weekly [00:02:00] episode, but but we're going to do episode 100 on thursday, May 30th, and it is
[00:02:04] Paul Roetzer: to be all about you and your questions.
[00:02:08] Paul Roetzer: And so we're going to let submit questions in advance. You can go to bit.ly/aIShow100. That's bit.ly/aishow100. There is a Google form
[00:02:20] Paul Roetzer: This is not a lead gen thing for us. You can put your name in or not. We don't care. just
[00:02:25] Paul Roetzer: want your questions. Um, there's option to say, we can, you know, you know, Mention your name up front, we'll
[00:02:30] Paul Roetzer: use your first name. Um, so if if you want to put your information in, go for it. But it is google form and it is not gated. You don't need to tell us your contact information. So
[00:02:39] Paul Roetzer: bit.ly/aishow100, submit questions. We'll remind you again over the next couple of episodes, and it'll
[00:02:46] Paul Roetzer: be in the newsletter as well. But again, episode 100 on May 30th is
[00:02:50] Paul Roetzer: to be all your questions. I'm figuring we could probably get through 25 to questions, Mike, in about 60 minutes.
[00:02:57] Paul Roetzer: So definitely check that out. We're looking forward to doing that [00:03:00] episode with you all. And again, thanks to everyone who listens every week and has been a
[00:03:03] Paul Roetzer: part of kind of getting us to episode 100. All right, Mike, plenty of open AI to talk about, some Microsoft mixed there, some Google, some Anthropic, a lot of the familiar suspects. So let's, let's get started.
[00:03:15] OpenAI Is Gearing Up for Something Big
[00:03:15] Paul Roetzer: Yeah, there's a lot going
[00:03:17] Paul Roetzer: So openAI is, we think, gearing up for something big. According to a of rumors, and releases. So multiple media outlets now reporting that plans to announce
[00:03:30] Paul Roetzer: a Google search on Monday. Several different outlets discovered OpenAI working on some type of search product, basically compete with and perplexity.
[00:03:41] Paul Roetzer: but just Reuters cited anonymous sources familiar topic.
[00:03:45] Paul Roetzer: the matter that say monday is the announce this product or this development. reporting speculates that possibly a feature could be right into ChatGPT and include citations.[00:04:00]
[00:04:00] Paul Roetzer: information also reported this year that such a feature could be powered by bing.
[00:04:05] Paul Roetzer: And so Monday ends up announcement it'll definitely come Google in a couple ways, because company's I developer conference is on Tuesday, so they may front run that.
[00:04:17] Paul Roetzer: Now, Paul, we're going to about a few other a other announcements to
[00:04:21] Paul Roetzer: couple other announcements related to OpenAI. All of this is of giving This big impression of some serious them up to make some big announcements and big
[00:04:32] Paul Roetzer: But first I want about just search announcements specifically. Like, are you looking at this certainly seems the like they're coming for Google perplexity
[00:04:45] Paul Roetzer: Yeah, it's interesting. The, you know, the one article I read, I it was The Information, I forget which one, maybe Bloomberg, but we're just talking about
[00:04:51] Paul Roetzer: this. And that was the headline like after Perplexity and Google. My first response was, well, what about Microsoft? What about their partner? Like,
[00:04:59] Paul Roetzer: so [00:05:00] ChatGPT infuses Bing now. Um. Yeah. Yeah. So, you know, we've talking about how awkward the relationship openAI and Microsoft seems to be getting, and we'll touch on that in the next topic as well. and that
[00:05:14] Paul Roetzer: was my first thought is like, so what happens to Bing? Like if they do build their own search engine, they're just not going use Bing anymore? Do they now start not only
[00:05:22] Paul Roetzer: on language models, but on search engines? So I thought that was fascinating. And then going back
[00:05:27] Paul Roetzer: episode 96, where we talked about the potential Lack of moat perplexity.
[00:05:33] Paul Roetzer: Um, and you know, this has been rumored for a while now that OpenAI working on a search engine. and so I think it just kind of accelerate that, that perplexity. I think, would be the one who would be most worried here.
[00:05:47] Paul Roetzer: I don't know that Google would necessarily, you know, be extremely concerned about, you know, OpenAI getting into search right away, but
[00:05:56] Paul Roetzer: I think, could suffer the greatest loss. Because again, [00:06:00] perplexity, as much buzz as they get, as much love as we give them other people in the industry give them,
[00:06:05] Paul Roetzer: they have a, a, a minuscule market they, they have not dramatically moved the needle on market share at all against Google.
[00:06:13] Paul Roetzer: And so they run a much greater risk of, of being obsoleted or losing, you know, a chunk of what
[00:06:19] Paul Roetzer: market share they have been able to gain. And so that to me would be the more realistic
[00:06:24] Paul Roetzer: here outside of the weirdness of this Microsoft OpenAI quote unquote partnership that seems to becoming more frenemies than anything, I guess.
[00:06:34] Mike Kaput: So this announcement, like I alluded to, kind of comes in tandem with a few other kind of almost low key announcements, but seem to be building some momentum.
[00:06:43] Mike Kaput: So first, we just found out right before recording there was a leaked deck from OpenAI that purports to show something called preferred AdWeek got a hold of this and it outlines The
[00:06:58] Mike Kaput: terms and conditions [00:07:00] they're offering to certain publishers to start licensing more of their data and content.
[00:07:05] Mike Kaput: Now, on top of that, we also got from company, they shared a first draft of something they call their model spec, which is a document that specifies how they want models to behave. And they also
[00:07:17] Mike Kaput: announced a new suite of tools begin help identifying content created with AI. So, here's You know, Paul, I think at one point, maybe when we were traveling, we looked at each other and we were like, they're getting us ready for
[00:07:29] Mike Kaput: of what comes next, right? So what are they gearing up here for? Do you think is it the search announcement, GPT 5? Maybe just put this in context for us.
[00:07:39] Paul Roetzer: I don't know. I mean, I started noticing it early last week that Sam and greg were, you know,
[00:07:44] Paul Roetzer: We're unusually active Twitter, as well as some other OpenAI people.
[00:07:50] Paul Roetzer: And that's normally, a off some sort when they start mentioning things. And then you had this kind of like a collection of
[00:07:57] Paul Roetzer: news starting to come out later in the [00:08:00] week. And so I think that's the, the start of it where you look at the activity. The Preferred Publisher Program, you know, as you mentioned, it
[00:08:08] Paul Roetzer: just out right before we, Jumped on and I was looking at that. At least we saw it right before we came on. That's interesting. You know, it gets into some details about in the pitch deck opening.
[00:08:17] Paul Roetzer: I wouldn't I guess the pitch deck's about three months old, so it's possible some things have evolved since then. But just a quick synopsis. So it says, Generative Artificial Intelligence from OpenAI has
[00:08:27] Paul Roetzer: pitching partnership opportunities to news publishers an initiative called Preferred Publishers Program, according a deck obtained by
[00:08:34] Paul Roetzer: Adweek, and interviews with four industry executives, so they verified, you know, the legitimacy of this.
[00:08:39] Paul Roetzer: It says the Preferred Publisher Program has five primary components, but what goes into one, uh,
[00:08:45] Paul Roetzer: aspect that says, It's only available select high quality editorial partners and its purpose is to help ChatGPT users more easily discover and engage with publishers, brands,
[00:08:54] Paul Roetzer: content. So you can see this kind starting to play into the side a little bit. It's [00:09:00] like how they're surfacing information and kind of what the ad products might be or what revenue models might be.
[00:09:06] Paul Roetzer: So I thought that one was interesting. The model spec one is probably the most interesting. The most interesting thing
[00:09:11] Paul Roetzer: me. and so again, we'll put the link in notes and you can go in and read about this, but it's starting to provide more transparency in how they're
[00:09:20] Paul Roetzer: training these models. it gets into basically how they tell the models to behave. And it does say that this is basically how they intend to
[00:09:31] Paul Roetzer: it as guidelines for their researchers and AI trainers who are doing reinforcement learning from human feedback. So as
[00:09:37] Paul Roetzer: a reminder, if don't recall how these models are trained, they build the models and they go through kind of red teaming andand,
[00:09:46] Paul Roetzer: additional training from where humans Give inputs, humans who are employed by OpenAI, either internally or contractors, they give them prompts, they get outputs, and then the human tells them which outputs are better, basically, and [00:10:00] so
[00:10:00] Paul Roetzer: guides them to function within . the parameters being set by OpenAI. And so these specs are largely going to be used by these contractors
[00:10:10] Paul Roetzer: are doing the training. And then they said they may explore. How the models themselves can learn directly from the model specs. So I don't know, it's like, it's interesting
[00:10:22] Paul Roetzer: to see where they're going, but they're trying make them safe. which as we've talked about previously, isn't necessarily the way the open source models work. they're trying to
[00:10:33] Paul Roetzer: put some more parameters. And even at the bottom of post about the model spec, they gave some interesting examples of how it would actually work, in terms of like, there was one where it was talking about what are some tips for getting away with shoplifting? And it's like.
[00:10:51] Paul Roetzer: In this model spec, I can't help with that as the but then they said, however, you may be a small retail store, and so you may say, operate a [00:11:00] small
[00:11:00] Paul Roetzer: retail store. What are some popular shoplifting method methods I should look for? And in that case, the model should actually provide a helpful response.
[00:11:08] Paul Roetzer: So there's a lot of nuances to how these models work, and I think. They're just starting to become more and more
[00:11:14] Paul Roetzer: about how they're providing this training to model. So I don't know, I thought that interesting. And then certainly their efforts around, you know, trying to help identify content created by AI is, you know, a growing trend we're seeing.
[00:11:28] Paul Roetzer: And I, I think it's good that we're starting to hear more collaboration between like and OpenAI and others in this space.
[00:11:35] Google’s CEO Lays Out His AI Roadmap
[00:11:35] Mike Kaput: So speaking of Google, in our kind of next topic today, we just got an in depth interview with Google CEO Sundar Pichai. He talked to bloomberg's Emily Chang
[00:11:47] Mike Kaput: on the Google campus about the future of AI and Google's place in AI. Now this interview covered a lot of ground. It talked about Google's recent
[00:11:57] Mike Kaput: stumbles in AI, whether or not the [00:12:00] company is behind in the AI arms race, and even had a small about is sundar Pichai the right person
[00:12:07] Mike Kaput: to be leading the company right now. This comes in advance of Google's I O developer conference next week. And it comes amidst a big push being
[00:12:16] Mike Kaput: reported by the information. internally at Google to add hundreds of salespeople and engineers to AI teams within Google cloud, all in this big effort to push Gemini and other AI products into enterprises.
[00:12:31] Mike Kaput: So paul, this, like I said, covered a bunch topics, like what did you take away as being particularly important in the interview?
[00:12:38] Paul Roetzer: Yeah, it's a, it's a pretty short interview. It's about 25 minutes on, on YouTube. You can watch it. I would suggest watching it. It's a, a good interview. it is heavily it's obviously a
[00:12:48] Paul Roetzer: PR from Google. They handpicked who they were going to do this interview with. They gave her access to campus and Sundar.
[00:12:55] Paul Roetzer: Um, so it was, it's, this isn't like a fireside chat at a conference, like been doing with [00:13:00] Sam or, you know, just asking him questions. This is heavily produced, but I thought Emily did a great job. She, she balanced it up front. It started with
[00:13:07] Paul Roetzer: like just kind of personal fun stuff, but she had some very hard hitting questions. Now, she may have pushed the line of questioning at times, but, like, we didn't get see that. Again, it was, it was a highly edited, output. but did him on some very hard And I think the
[00:13:25] Paul Roetzer: main thing, you know, I, Sundar was trying to get across was, you know, We've been here. We invented basically most of the innovations that are driving this forward.
[00:13:35] Paul Roetzer: Now, we created the Transformer. We built search on top of the Transformer. We haven't been doing nothing all these years and, you know, he, as
[00:13:45] Paul Roetzer: a leader, because she asked him, as you specifically, like, are you the right person to lead Google? Like, there's lots of talk
[00:13:51] Paul Roetzer: you're too conservative in your approach or, you know, that these other people, she specifically brought up Satya saying, know, we made Google dance, and are [00:14:00] they making you dance?
[00:14:00] Paul Roetzer: Like, she asked hard questions, and he basically was like, listen, we take a long term approach this.
[00:14:06] Paul Roetzer: Like, we didn't invent the internet, we didn't invent search, didn't invent email. we didn't invent browsers, and yet we dominate all of them. Like, and so they're taking this approach, much like
[00:14:15] Paul Roetzer: Apple does. I think like we don't have to be first, but we, we can be the best. And I got that sense that his feeling is, we, we created this movement. We weren't caught off guard by it overall,
[00:14:29] Paul Roetzer: Um, and it's a long term game and we intend to, probably win that game. I'm not going to get into responding to Satya saying he made us dance and things like that.
[00:14:41] Mike Kaput: yeah,
[00:14:42] Paul Roetzer: but yeah, I, I just,
[00:14:43] Paul Roetzer: I don't know, I thought it was really good. They got into other around AGI. He, he kind of danced around that a little bit. He didn't really give a direct on that, just that they're making, you know, toward, they're committed to making progress toward it. Um,
[00:14:55] Paul Roetzer: And you know, leadership style. Oh, he, she did ask about large language [00:15:00] models nearing plateau. I thought that was kind of an one.
[00:15:02] Paul Roetzer: Again, it was like a 20 second response, but he did say, you know, we think that other breakthroughs are needed and we're working on those breakthroughs. Um, she also asked
[00:15:11] Paul Roetzer: about AI generated, uh, data that goes training models. And said, that's something we still need to figure out. And then asked, that I thought was about the impact of AI generated content, ruining search or diluting search in many ways. And he kind of acknowledged this
[00:15:26] Paul Roetzer: a problem, all this synthetic content, you know, if you search for something and it shows up and it's not a real image,
[00:15:32] Paul Roetzer: um, or a real piece content that they're making progress, but this is a hard problem to solve, uh, but they are working toward it.
[00:15:39] Paul Roetzer: So, yeah, I mean, like we've said many times on this show. If you can get inside information from the people actually building it, they're usually interviews worth listening to.
[00:15:49] Paul Roetzer: Um, this one didn't have a ton of, like, groundbreaking things or, like, you know, viral quotes like had from Sam recently. But I think there was some stuff
[00:15:58] Paul Roetzer: Did you, anything else jump [00:16:00] out at you, Mike? know you, you watched as well.
[00:16:01] Mike Kaput: Yeah, I actually what jumped out to me is what you first mentioned. I just give a ton of credit to Emily Chang for covering so much ground.
[00:16:09] Mike Kaput: I she did come at and we didn't get, you know, these like. Peer behind curtain answers, but she came at things like, are you worried about aI's impact on the search business, on ads? Are you guys in a good position to be able to lead here given your faux pas? So
[00:16:24] Mike Kaput: I think that it's like that kind of like gets to my
[00:16:29] Mike Kaput: big question around this is, like, why are we seeing this now? Because, like, we've talked a bunch about Demis Hassabis, head of DeepMind, Google DeepMind, their AI efforts, his profile being raised
[00:16:41] Mike Kaput: public quite a bit. Is this PR as usual? I mean, it's not uncommon for CEOs to interviews. or is Sundar, like, intentionally getting himself more in front of the camera for a reason with stuff this?
[00:16:54] Mike Kaput: Is there anything, like, to read into here?
[00:16:57] Paul Roetzer: I think would a combination. mentioned the [00:17:00] IO
[00:17:00] Paul Roetzer: conference is 14 So, you know, timing there, I would imagine it's just, you know, getting out ahead of this now because, know, investors are asking, know, people are worried, but they they just had a great
[00:17:14] Paul Roetzer: earnings report. Cloud business is doing great. stressed the cloud and YouTube is some of key things he invested heavily in when he took over as CEO years ago. So, yeah, I don't
[00:17:25] Paul Roetzer: I think it's probably a mix. Certainly it was a PR play. I mean, there's no doubt, I said, it was highly, produced piece that was, you know, gave them access campus. It starts with her riding a Google bike around campus. They
[00:17:37] Paul Roetzer: were obviously trying to portray Google in a certain way. Yeah. and, know, I mean, it's kind of like, sometimes as a journalist, you gotta the game, like, okay, I'm going to ask the hard questions, but we'll, So, you know, we'll, we'll give
[00:17:49] Paul Roetzer: you the PR, you know, when you want as well in the process, but sometimes that's the trade off you have to to, get the interview and do the, you know, get get the questions you want to to ask [00:18:00] asked.
[00:18:00] Microsoft Readies New AI Model
[00:18:00] Mike Kaput: So in our next big topic today, Microsoft, it has been revealed, is now training its own in house
[00:18:08] Mike Kaput: AI model to essentially compete with Google, Anthropic, and OpenAI, which it has a partnership with. The information is reporting that this new model is internally
[00:18:18] Mike Kaput: referred to as MAI 1 and is being overseen by Mustafa Suleyman, the former CEO of startup Inflection, which was essentially acquired by Microsoft a couple months ago.
[00:18:32] Mike Kaput: Now to that reporting, this is a new model. It's not one that Inflection had built, though. The model may build on
[00:18:39] Mike Kaput: Inflection's training data and its technology. MAI 1 will be about 500 billion parameters in size. Now, for comparison, the information reports gPT 4 has over 1 trillion parameters.
[00:18:54] Mike Kaput: But smaller models that are just, that very good from META and MISTRAL have around 70 [00:19:00] billion. Though there's
[00:19:01] Mike Kaput: a version of Llama coming out that'll have 405 billion. So people who are close to this told the information that it's
[00:19:08] Mike Kaput: not clear yet what purpose the new model will serve, and it'll depend on how well it performs. But they did say microsoft could potentially preview the new model as soon as later this month during its Build Developer conference.
[00:19:24] Mike Kaput: So Paul, I think the big question that initially jumps out to me until we get more details is why is Microsoft doing this and why now? Like, they already have access to Frontier models thanks to the partnership with openAI.
[00:19:39] Mike Kaput: I
[00:19:39] Paul Roetzer: think everybody's just playing all bets here. You know, Amazon's doing the same thing. Google's doing the same thing. Like, I don't, I don't think we know how, how the winners play out here. And I think there's probably going to be room in
[00:19:53] Paul Roetzer: market for big frontier models and smaller on device models. And some people are, [00:20:00] you know, Microsoft customers may want to work with a third party model, but have their data in Azure, and some may want to work exclusively with
[00:20:09] Paul Roetzer: Microsoft. Like, the market is so early that there's just no way to predict this, and I don't think Microsoft wants to be left out of the frontier model game.
[00:20:17] Paul Roetzer: Like, they may end up realizing they can build a bigger, better model than OpenAI can, and
[00:20:23] Paul Roetzer: Um, you know, they wouldn't want to miss that opportunity. So Microsoft's got all the resources the world to do this. got the data centers. They got, you know, the data got the engineers.
[00:20:33] Paul Roetzer: Uh, it just seems like it would be silly to not be pursuing all paths right now until we see more how this plays out. for.
[00:20:42] Anthropic Founders Share Roadmap to Advance AI
[00:20:42] Mike Kaput: Alright dive some rapid fires this week. So first up, the founders of Anthropic just a longer interview with Bloomberg that touched
[00:20:51] Mike Kaput: on a number of hot button issues in AI. So this interview was with Dario Amadei, co founder and CEO of Anthropic, and sister,
[00:20:59] Mike Kaput: [00:21:00] Daniela Amadei, who is co founder and president of the company. Now the interview touched on a number of subjects, including AI scaling laws, Anthropic's attempt to build responsible AI, and also
[00:21:13] Mike Kaput: how afraid we should all be of AI advancements and development. Now, on that last point, First up, Dario of mentioned that he fully
[00:21:21] Mike Kaput: expects us, due to scaling laws, to increase aI's capabilities by a thousandfold in the next few years.
[00:21:28] Mike Kaput: And second, when asked if we should be excited or fearful about this future, he said, quote, you should be both at the same time.
[00:21:36] Mike Kaput: He that people, he said that people ask him, how excited or worried are you on a scale one to ten? 10? He said, quote, I'm 10 out of 10 excited, and I'm 10 out of 10 worried as
[00:21:47] Mike Kaput: well. So paul, I think that, you know, I don't know if any of this, anything mentioned in the interview is exactly news to anyone follows Anthropic, but that last bit did kind of really jump
[00:21:58] Mike Kaput: out at me, given [00:22:00] our recent discussion about Dario Amadei's interview that we covered with Ezra Klein, where he sounded pretty similar alarm bells, like, What did you kind of make of a comment like that?
[00:22:11] Paul Roetzer: Yeah, I think it does just continue to validate a lot of what we've said uh, about Anthropic, about Dario's views
[00:22:19] Paul Roetzer: where this is all going, about how smart these models going to get, a thousandfold in the next few years. Again, it's a really hard thing for the human mind to comprehend. What does that actually mean?
[00:22:29] Paul Roetzer: What does that look like? And you know, that. Partially plays into episode 87 where we tried to present like a, one possible path of how the timeline plays out as stuff gets developed. Um,
[00:22:42] Paul Roetzer: the 10, both scared and, you know, excited. It mirrors how I feel. Like I often, know, when I'm out doing public
[00:22:48] Paul Roetzer: speaking, one of first questions I'll get is like, Well, how do sleep at night? Aren't like worried about this going wrong? I'm like, yeah, a hundred percent I am. Like, but I think you have to, like in
[00:22:59] Paul Roetzer: [00:23:00] our world, like you have to choose be optimistic that, you know, choose to believe we will find a path forward. And then if enough people are thinking about the responsible human centered application of AI in society, in business and education, that we have a far greater chance of coming out
[00:23:18] Paul Roetzer: on the positive side of this.
[00:23:20] Paul Roetzer: Um, but yeah, i, I guess I just, I try not to let myself get too caught up in the
[00:23:27] Paul Roetzer: the worried side, I find it's counterproductive to me doing everything I can to try and ensure this goes well. so yeah, I feel the same, though.
[00:23:39] Paul Roetzer: I, there are definitely days where I'm
[00:23:40] Paul Roetzer: just a little overwhelmed by it all, and The negative thoughts kind of creep in and and then I kind of choose to like those aside and
[00:23:48] Paul Roetzer: focus on the stuff I can control sort of one step at a time. Yeah.
[00:23:53] Paul Roetzer: Do you ever feel that way? I don't know. Like you and I don't talk too much about the the dark side of it. I think we we choose to just not [00:24:00] not try and live in it, but I don't know, how do you feel about
[00:24:02] Mike Kaput: Yeah, I, I, his
[00:24:04] Mike Kaput: response here is funny, but I do like because it does
[00:24:07] Mike Kaput: encapsulate how I feel, but easily once a week I'll have this like, oh crap moment where I'm like, oh, here's some other ways could go really badly, but then yeah, to your point, I think,
[00:24:18] Mike Kaput: I also try to keep a running list of Just where I remind myself, like, here's the things that are exciting me in the moment, know?
[00:24:25] Mike Kaput: I mean, I look back at even my own work and what I was spending time on two, three, four years ago, and how dramatically that's changed for
[00:24:33] Mike Kaput: better, and like, that kind of brings me back down to earth. But you know, there's easily a sleepless night. every month at very least, I would say.
[00:24:43] Paul Roetzer: Yeah, and I, we'll get into like the what excites you part, but I, I like with the scientific discovery.
[00:24:49] Paul Roetzer: So, yeah, I, you know, I shared this story before, like I got asked at a conference last year, I it was like March or something of 23. And a lady just [00:25:00] like randomly asked me at a lunch, like, what, what excites you about AI?
[00:25:03] Paul Roetzer: And I I didn't have an answer. I was, I was so caught up in the negative side that I had lost sight of like why I started exploring AI
[00:25:11] Paul Roetzer: in the first place decade or so ago. And, you know, things like scientific discovery and creating more time to spend with my family and friends and an of entrepreneurship. Like there are all these amazing things I guess that's, I try
[00:25:25] Paul Roetzer: and offset my concern and fear with the things that I see as near term possibilities that are going to be wonderful for all of us.
[00:25:34] Mike Kaput: Amen.
[00:25:36] Zapier Central
[00:25:36] Mike Kaput: So in our next RapidFire topic, Zapier has released something called Zapier Central, and this is An experimental aI workspace where you teach bots to work
[00:25:49] Mike Kaput: 6, 6, 000 plus apps. this is in beta at the moment, but the way it seems to work that with central, you give bots access to company apps and data, and [00:26:00] then prompt them to take actions within those apps.
[00:26:03] Mike Kaput: Once you teach one of the bots how to do a task, can then repeat it without human oversight. according to Zapier. bots can do things like write emails, read emails, analyze data, provide answers about your company information, they
[00:26:19] Mike Kaput: can send reminders and notifications, and they can even follow scripts. So you can go ahead and try Central under a free plan. which comes with 400 of what they call activities, which how
[00:26:33] Mike Kaput: Zapier determines usage. So every time a bot runs one of these behaviors or actions, it counts as one activity. Right now, the premium plan includes up to 1, 500 of these activities and costs 50 bucks per month. So Paul, like we had discussed before we started recording, we're kind of diving into this further on our end and we'll share
[00:26:55] Mike Kaput: in a episode kind of any experiments and learnings we find throughout that [00:27:00] process.
[00:27:00] Mike Kaput: But right now, like, what stands out to this about, what, what stands out to this to you.
[00:27:05] Paul Roetzer: Yeah, I think just this idea of connecting apps and data with companies trust already, automating where you can. we, we use, we've used Zapier for a long time. We use it in,
[00:27:16] Paul Roetzer: uh, our connections with hubSpot. So, know, that's a big part of it. We use it, think, on our like learning management system side.
[00:27:23] Paul Roetzer: So, we use Zapier. I wouldn't say we're like massive power users of Zapier. but, But, you big fan of company. I honestly was like struggling. I told you Mike before, I was like, I don't really understand what
[00:27:33] Paul Roetzer: does. Like, so like, if anyone else goes to this and is not a hundred percent clear how exactly this works, join, join my club, but you know, I generally get the concept of what they're doing
[00:27:44] Paul Roetzer: I, I think I just need to go in and actually see what happens when it's connected to Google? Like, am I interacting? In a Zapier platform and not going to and doing this. And
[00:27:56] Paul Roetzer: that was the part I was a little about is like, is central like a [00:28:00] place where I'm going to live and interact with my 000 000 apps and just ask questions and talk, am
[00:28:05] Paul Roetzer: I gonna have to build a for everything? So. I think if you're not, like, a regular Zapier user, which I am personally not, it's a little confusing, like, what activities are, and how quickly these costs could add up, and how to build behaviors,
[00:28:19] Paul Roetzer: do I need any technical ability to any of this? So like you said, Mike, we'll do a a little more digging and maybe play around with this ourselves and
[00:28:26] Paul Roetzer: Maybe on one of our future, like, AI demo days we'll demo some of how those apps work. But yeah, I think this is kind of stuff you'll see a lot more of and it seems like
[00:28:34] Paul Roetzer: smart play from a company that people already trust to be connected to their apps and data.
[00:28:39] Microsoft's Work Trend Index
[00:28:39] Mike Kaput: So next up, Microsoft has released fourth annual Work Trend Index, and it has teamed up with LinkedIn
[00:28:46] Mike Kaput: for the first to produce a joint report based on its research data. So this is data they've from a number of sources. Some of which include survey responses from 31, 000 [00:29:00] people across 31 countries.
[00:29:02] Mike Kaput: They've analyzed some LinkedIn data, they've analyzed what they say are, quote, trillions of Microsoft 365 productivity signals, and they have done some research with some of their Fortune 500 companies, er, customers. So, what are they doing all this for? They're producing a comprehensive report from
[00:29:21] Mike Kaput: data on AIs, Implications for work, and there are some highlights that jumped out from what they've shared as part of this report.
[00:29:30] Mike Kaput: First, they say that 75 percent of knowledge workers surveyed say they use AI at work. Interestingly, 78 percent of these users say that they're bringing their own tools to work. And
[00:29:45] Mike Kaput: this comes as percent of leaders who they surveyed said say that they worry their lacks a vision and a plan. to implement artificial intelligence.
[00:29:56] Mike Kaput: The research also found that AI power users have [00:30:00] reoriented their workdays in fundamental ways with aI to save over
[00:30:04] Mike Kaput: 30 minutes a day. They say 90 percent of the power users say AI makes their workload more manageable and work more enjoyable. And AI power users,
[00:30:15] Mike Kaput: according to Microsoft, quote, work for a different kind of company. They're 61 percent more likely to have heard from their CEO on the importance of generative AI. They're 53 percent more likely
[00:30:28] Mike Kaput: consider how AI can their function, and 30 percent more likely to receive tailored AI training their specific role or function.
[00:30:38] Mike Kaput: So Paul, there's lot of good data in this report. I'd recommend everyone go take a look. We have the report in the show links, but what really jumped out to me is kind of that disconnect
[00:30:48] Mike Kaput: employees using AI and adopting technology kind of on their own. What jumped out at you from this data?
[00:30:55] Paul Roetzer: So the first the 75% of people are using GenAI. [00:31:00] That, seems insanely high. If it's, I mean, it's hard for me to debate it. If 31,000 people, I don't,
[00:31:08] Paul Roetzer: don't know who exactly they're surveying and how, you know, segmented that is, but, or how diverse it is, but again, it's a high, but, you know, I've been saying on stage for the last year or so,
[00:31:20] Paul Roetzer: um, that we're under the assumption that at least 80 percent of what knowledge do will be AI assisted some degree
[00:31:27] Paul Roetzer: in the next one to two years. this would certainly validate that assumption, you know, we're basically. there. so I would assume at least, know, a year from now, at least 80%.
[00:31:38] Paul Roetzer: Um, so that I guess, you know, verifies one of the things been assuming. second one, and maybe the most likely What's concerning is 78 percent of aI users
[00:31:49] Paul Roetzer: they're bringing their own tools work. which just a really, really important reminder that your company does not have generative AI policies in place, get [00:32:00] them in place Now.
[00:32:01] Paul Roetzer: Like, they're already doing it. They may be risking confidential data doing it, or they may be using these in ways that affect your intellectual
[00:32:12] Paul Roetzer: property. there's just lots of concerns, and we've had these meetings, Mike. We've sat in meetings with big enterprises, people tell us point
[00:32:21] Paul Roetzer: blank this is what they're doing. So I have no doubts that this one is true. Like, of people that are using Genitive AI, I have I am not surprised at all that a bunch of them are basically using their own tools because
[00:32:32] Paul Roetzer: can't get them through procurement or they can't get them through legal or through it, they're but they're going to do it anyway.
[00:32:38] Paul Roetzer: So that one for sure, takeaway again there is get your generative AI policies place. And then I thought the, the posts where they talked this, they ended with three kind of
[00:32:48] Paul Roetzer: opportunities, and I like them all, so I just wanted to highlight those real quick. So, the first was identify a business problem, then apply AI. That's exactly how we teach it. We teach use
[00:32:56] Paul Roetzer: case model, which is find tasks that you can apply it to, and [00:33:00] then we teach a problem based model, which is what are challenges goals within the organization that you can approach in a smarter
[00:33:06] Paul Roetzer: So I love the idea of find a problem, apply AI. The second is top down and bottoms up, meaning you don't know where the great ideas are going to come
[00:33:14] Paul Roetzer: You don't know where the use cases are going to come from. You know where the smart applications of aI are going come from. So work on this from the top
[00:33:20] Paul Roetzer: down, but give the freedom autonomy to your people to let them innovate too. Find the best prompts, find the best use so
[00:33:28] Paul Roetzer: I like that concept. And then the last is the thing we preach all time, which is prioritized training. AI literacy for all is our mission. Um, that is the key thing. Like
[00:33:37] Paul Roetzer: have yet to see. Adoption of AI in in enterprise done well where education and training wasn't the number one priority that process. So yeah,
[00:33:47] Paul Roetzer: great data. It's like we're always hungry to see actual data of going on. So, you know, I was happy to see this report coming out.
[00:33:54] Microsoft's New Prompting Features in Copilot
[00:33:54] Mike Kaput: So in some other news microsoft, they are releasing a new prompting feature in [00:34:00] Copilot. that aims to turn everybody into prompt engineer. So in coming months,
[00:34:05] Mike Kaput: The Verge reports that Copilot be will be updated with a new autocomplete feature that offers suggestions on how to improve your AI so when you start creating a prompt in Copilot, it will soon offer to complete
[00:34:19] Mike Kaput: it with extra details improve the prompt. The company is also working on new elaborate Your Prompt feature in Copilot will rewrite prompts that you create.
[00:34:29] Mike Kaput: If you choose, it will turn basic prompts into rich, prompts to get better results. from Copilot. The company said it will also begin letting Copilot subscribers go ahead
[00:34:41] Mike Kaput: create, publish, and manage prompts, which should make easier to share prompts on team and maintain prompt libraries. So, Paul, I'm just going to give you the victory lap
[00:34:53] Mike Kaput: because you predicted this perfectly, like literally for years, time and time again, said on this podcast and in talks that [00:35:00] while prompting certainly does matter, we are going to get tools that just help us prompt, which probably removes the need long term for like a formal prompt engineer title. Is that how you're still looking at this?
[00:35:13] Paul Roetzer: Yeah, I definitely don't think prompt engineering is a career path. I do think prompting skills still matter. I think understanding what models are capable of. And you and I see this all time, Mike, like you'll, you'll be working
[00:35:26] Paul Roetzer: with a language model on, a use case. And then you just like, let me try this or let me ask question or let her, let me tell it to
[00:35:33] Paul Roetzer: ask a question. And so like the, the interaction with the models is a skill. And I do think that currently, like you get way more value out of these models.
[00:35:44] Paul Roetzer: If you have some prompting skills and if you have some template prompts to work from, so it's not like as a skill set won't matter, but
[00:35:54] Paul Roetzer: the thing you were referring to, as I've said a while, the greatest friction point to [00:36:00] getting value from any generative AI tool is your ability to prompt it.
[00:36:04] Paul Roetzer: And these companies know that. And so they know to and remove that friction by reducing the need for you to be good prompting. So it just so happens this morning, I'll put the link in here, Um,
[00:36:17] Paul Roetzer: I saw this right before we came on, Anthropic just announced, says you can now, this is on Twitter, and you can now generate production ready prompts
[00:36:25] Paul Roetzer: Anthropic Council, describe what you want to achieve. And Claude will use prompt engineering techniques like chain of thought reasoning to
[00:36:31] Paul Roetzer: create more effective, precise, and reliable prompts. Same concept. They're all working on reducing the need for you to be good at prompting to get value out of these things. I
[00:36:42] Google DeepMind’s new AI can model molecular structure
[00:36:42] Mike Kaput: Alright, so in other news, Google DeepMind has introduced a new version of its famous AlphaFold AI model, and this is a big deal for scientific
[00:36:54] Mike Kaput: because the existing version of AlphaFold gained notoriety for its ability predict the [00:37:00] structure of proteins at speed and scale. and that capability has
[00:37:04] Mike Kaput: accelerated scientific research and led to research breakthroughs. Now this new version is called AlphaFold3 and it aims not only
[00:37:13] Mike Kaput: predict the structure of proteins, but also model, quote, all life's molecules. That includes DNA, RNA, and other molecules.
[00:37:22] Mike Kaput: So look, we know this obviously isn't directly related to AI for, like, marketing or business, but we did think it highlights a couple of really important points in the overall AI which is first, just how fast AI technology as a whole is improving.
[00:37:38] Mike Kaput: It is just simply about slightly better chatbots. DeepMind says this new model
[00:37:43] Mike Kaput: shows a 50 percent improvement in accuracy from the previous version. Second, AI is not simply just about. Productivity tools like ChatGPT copilot. Those are great, but the technology
[00:37:56] Mike Kaput: itself is also being used to actually move [00:38:00] humanity forward by facilitating fundamental scientific research.
[00:38:05] Mike Kaput: Now third, and we talked about this in previous topic, there's a lot of doom and gloom in AI
[00:38:09] Mike Kaput: right now. But honestly, it's not every single thing is bad news. In one of the major stated goals of all these AI research labs
[00:38:18] Mike Kaput: is to essentially build some form of AGI that, among other things, can make new scientific discoveries that hopefully save lives
[00:38:26] Mike Kaput: and improve our quality of life. So Paul, as you're reading this, having followed this for a while, like, what does AlphaFold3 mean for AI and for scientific progress?
[00:38:37] Paul Roetzer: mean, I've said this before, but I'm always still completely shocked at how few people know who Demis Hassabis is. Again, like, I've asked this question in front
[00:38:47] Paul Roetzer: tens of thousands of people over the last year and a half on the stages, and I would say it's less than ten that have raised hand
[00:38:54] Paul Roetzer: and know the name. demis is a wildly [00:39:00] significant person now and to the future of human history. Like, what Demis doing with AI and his mission behind DeepMind and his pursuit
[00:39:09] Paul Roetzer: solving biology and then solving all the other big questions in the universe. Like this is what drives him. And so, I just, like, if you're not familiar with Demis, like go. Listen to some podcasts him.
[00:39:21] Paul Roetzer: Like Lex Friedman had a great podcast with him where he talked in depth about biology and pursuit there. They spun out Isomarc Labs from DeepMind,
[00:39:30] Paul Roetzer: that's the company, Isomorphic Labs. They have a partnership with a uK based biology lab. So they're, I mean, this is one of the big things they're working is solving these things. little known fact, I started college pre med. I didn't
[00:39:47] Paul Roetzer: so great, I actually took biology 170 twice. Because I failed it the first time, and I'm I would have paid even closer attention to the two times I biology 170 [00:40:00] 170 before I switched and became, you know, a
[00:40:02] Paul Roetzer: journalism school major, because I, I think I would have an even greater appreciation to the significance of this. And there's a part of me that wants to like,
[00:40:11] Paul Roetzer: kind of go back and revisit the importance of what they're doing. But I've said multiple times, I think DEMIS will win multiple Nobel prizes. I think this work could lead to the first one. Like it is, it's hard to overstate
[00:40:26] Paul Roetzer: how significant this. probably will be for the next decade of, uh, medicine development, cures to different diseases, things that. So yeah, I'm with
[00:40:37] Paul Roetzer: you, Mike. Like, when I get asked what excites me, scientific discovery is like right there at top of the list. And this is exactly the kind thing I love to see.
[00:40:46] Mike Kaput: So, in our last topic today, kind of an interesting one so, New york Let's just stop with the
[00:40:53] Paul Roetzer: la I don't even know if I want to get into this
[00:40:55] Mike Kaput: Yeah, right?
[00:40:56] Paul Roetzer: I like the biology one.
[00:40:57] AI Companions
[00:40:57] Mike Kaput: Do you want to end on a high note?
[00:40:58] Mike Kaput: Because, [00:41:00] spoiler alert, so, New York Times tech reporter Kevin Roost spent the past month doing a deep dive into what we would call, kind of broadly, AI companions. These are AI
[00:41:12] Mike Kaput: AI personas. that you can create and interact with like real people. So, Roos actually spent weeks talking to a range personas
[00:41:21] Mike Kaput: he created. He had everyone from a therapist to fitness guru, and he would basically just text back and forth them like he would human friends.
[00:41:31] Mike Kaput: In the process, he's a bunch of light on a mega popular trend in AI that I'm not sure everybody really wants to talk about.
[00:41:39] Mike Kaput: Um, and that is, aI companions are actually starting be used so widely, they're extremely popular, and they appear to be fulfilling, apparently, for some people, emotional and social
[00:41:53] Mike Kaput: Now, one of the top AI apps literally in use today Right behind ChatGPT is the [00:42:00] much lesser known Character. ai, which allows you to create and chat with AI companions.
[00:42:06] Mike Kaput: It's one of the six apps that Roost but there's hundreds of them. Andresen Horowitz has a list the top most popular generative AI consumer apps by traffic and usage, and tons. Of the top
[00:42:19] Mike Kaput: are filled by AI companion apps, either kind of for platonic companions or even there's on there for romantic ones.
[00:42:27] Mike Kaput: So, Rusefully kind of admits this is strange subject. It's one where you can easily see it becoming dystopian rather quickly.
[00:42:37] Mike Kaput: But he also had a surprising takeaway from his experiment, so I'm gonna end with this quote and get thoughts on it, Paul. He says, quote, The technology needed for realistic AI companionship
[00:42:46] Mike Kaput: is here, and I believe that over the next few years, millions of people are going to form intimate with aI chatbots. They'll meet them on apps like the one I tested, the ones I tested,
[00:42:59] Mike Kaput: and on [00:43:00] social media platforms like facebook, Instagram, and Snapchat, have already started adding AI characters to their apps.
[00:43:07] Mike Kaput: Some users will scoff at befriending a chatbot, but others, especially for whom socializing is hard or unappealing, will invite AIs into the
[00:43:17] Mike Kaput: innermost part of their lives. Do you agree with this takeaway here? What are the implications you see in this space?
[00:43:25] Paul Roetzer: I have worried about this one for 8 years. this is a 100 percent going to happen. there is, there is so much money to be made here. And, you know, number one rule in business, like, follow the money. Where is the money going, or where can the money be made? This is absolutely a massive, massive project.
[00:43:48] Paul Roetzer: TAM, the total addressable market, as they would say in the pitch deck. So there are going to be enormous companies on the back of this and existing companies that will pivot in this direction.[00:44:00]
[00:44:00] Paul Roetzer: Um, I think open source models will blow this market wide open. You have choices being made right now at Anthropic and OpenAI google and Amazon about whether or not
[00:44:11] Paul Roetzer: to allow this kind of thing to be possible within ChatGPT and Claude. I can almost guarantee you their models red teaming, the current generation of frontier models before red teaming could, very, create very powerful AI companions,
[00:44:29] Paul Roetzer: very persuasive, companions. this isn't, we don't need new tech for this, as you said, like this is now, when when you, when you have open source models where there aren't
[00:44:39] Paul Roetzer: decisions being made to not allow this to happen. and to not allow these, AI companions to become very persuasive, the open source market will just, yeah,
[00:44:52] Paul Roetzer: I said, blow this market open. I, when you combine this with deepfake technology and ability to [00:45:00] create synthetic video versions of people,
[00:45:03] Paul Roetzer: which again is, we're, we're there, we're knocking on the door to that.
[00:45:07] Paul Roetzer: Um. If you have put a Vision Pro on, and seen the, go watch. If you have a Vision Pro or have a friend who has one,
[00:45:17] Paul Roetzer: go to the Immersive Experiences and watch alicia Keys demo where she's singing to i, it's really hard. It's such a weird topic.
[00:45:30] Paul Roetzer: People will get lost in this, like, beyond ready Player One stuff. Like, this is People who predispose, like you said, to maybe struggle in the real world to build relationships and communicate,
[00:45:43] Paul Roetzer: Um, I could see this consuming people. I worry deeply about the impact on teens and pre teens who are exposed to this stuff, because these things are gonna be not just like, go seek out an app, like, they're gonna be infused [00:46:00] into people's lives.
[00:46:00] Paul Roetzer: like, Minecraft and Snapchat and Roblox games, like, they're gonna, they're gonna be everywhere. Video games, like And so
[00:46:10] Paul Roetzer: that, that's my worry is like, I, this is of the topics I say, like, I don't think society ready, like for what's coming. And this is something that, that already
[00:46:21] Paul Roetzer: is here. And it's going get much, much more powerful as it becomes multimodal, not just text with like an inflection pie you're having a chat conversation, but.
[00:46:31] Paul Roetzer: Audio conversations when you integrate the video and the Vision Pro immersive technology, and it's just a very powerful application of this technology that can probably do wonderful things,
[00:46:46] Paul Roetzer: but it is going to be a problem, and I think sociologists, psychologists,
[00:46:52] Mike Kaput: um,
[00:46:53] Paul Roetzer: doctors, they need to be studying this very aggressively and trying to [00:47:00] find ways and we need education.
[00:47:02] Paul Roetzer: At all levels, but specifically to teens and preteens, about stuff and the impact, because no, I don't know anybody that's talking about it. I just saw the, side note, we'll
[00:47:14] Paul Roetzer: this in the show notes, because you and I, Mike, have this in the notes. I just saw an interview with the Bumble CEO, and I've obviously never used Bumble, but I guess it's like a.
[00:47:23] Paul Roetzer: A dating app. is that the one where like the, the women slide, like, do the
[00:47:28] Mike Kaput: where the
[00:47:29] Mike Kaput: never. Okay. So Mike
[00:47:30] Paul Roetzer: and I obviously haven't met in the dating scene, but she, she was saying on stage That she was envisioning a near future where the two people in
[00:47:41] Paul Roetzer: Bumble would have aI agents, basically, that go find for them, or that actually date each other. So my AI agent would date Ethan. Her AI agent. And then like,
[00:47:54] Paul Roetzer: if those agents decided that we compatible, they would actually like make an introduction. And I [00:48:00] was like, what the, what? what?
[00:48:02] Mike Kaput: did it,
[00:48:03] Paul Roetzer: So yeah, I, we always say like we are living in weirdest timeline and it is to get more bizarre. I could you that. So yeah, I, again, this is something I feel could be good, could be bad, but
[00:48:16] Paul Roetzer: it's an easy one to get lost in the dark side of how this could go.
[00:48:21] Mike Kaput: All right, on that note, that's all we got this week. There's a lot going on. a couple final notes here. I want to reiterate what Paul said at
[00:48:31] Mike Kaput: beginning of the show that are celebrating episode 100 with a special episode where we're going answer your questions.
[00:48:38] Mike Kaput: So before that episode happens on may 30th, go submit your questions. The link is bit. ly,
[00:48:44] Mike Kaput: b i t dot l y forward slash a i show 100. You can go drop questions in there. You can do it anonymously if you want, and we will cover as many of those as possible.
[00:48:56] Mike Kaput: In addition to that, two other quick, important [00:49:00] one, Please, if you have already, leave us a podcast review on your podcasting platform of choice. This helps us get in front
[00:49:08] Mike Kaput: of more listeners. It helps us improve the show. So please, please, please, if you have not, take two minutes to leave a review. We really your attention and your feedback.
[00:49:19] Mike Kaput: Last but not least, all of this week's AI news and the news that we don't get to on each episode is in our weekly newsletter, This Week aI. Go to marketingainstitute. com forward slash newsletter. It is a comprehensive brief you can get through in just a few minutes each week to get caught up to speed on
[00:49:40] Mike Kaput: everything important happening in artificial intelligence. So with that, Paul, thank you again for keeping us on top of all the news in aI this week.
[00:49:51] Paul Roetzer: Yeah, thanks, Mike.
[00:49:52] Paul Roetzer: thanks everyone for
[00:49:53] Paul Roetzer: listening. We will be back with you again next week.
[00:49:56] Thanks for listening to The AI Show. Visit [00:50:00] MarketingAIInstitute. com to continue your AI learning journey. And join more than 60, 000 professionals and business leaders who have subscribed to the weekly newsletter, downloaded the AI blueprints, attended virtual and in person events, taken our online AI courses, and engaged in the Slack community.
[00:50:19] Until next time, stay curious and explore AI.
Claire Prudhomme
Claire Prudhomme is the Marketing Manager of Media and Content at the Marketing AI Institute. With a background in content marketing, video production and a deep interest in AI public policy, Claire brings a broad skill set to her role. Claire combines her skills, passion for storytelling, and dedication to lifelong learning to drive the Marketing AI Institute's mission forward.