Ep 511: The Mad Dash To Cash in On AI. Is AI overhyped AND underhyped?

Topics Covered in This Episode:

1. AI Hype: Overrated or Underrated Debate
2. Gary Rivlin's AI Valley Insights
3. Generative AI's Impact on Industries
4. Startups: Silicon Valley AI Innovations
5. AI Business Strategy: Lessons from .com Era
6. AI Venture Capital Investments Surge
7. Generative AI Tools for Cost Reduction
8. Big Tech's Multibillion AI Investments

Keywords:

Generative AI, Overhyped AI, Underhyped AI, Mad dash to cash in, AI Valley, Gary Rivlin, Reid Hoffman, LinkedIn, ChatGPT, Silicon Valley, Everyday AI, Rise of the Internet, AI development, Transformative AI, Neural networks, ChatGPT launch, OpenAI, Adobe Firefly, Microsoft Copilot, Image generation, Text to video, Large-scale investment, Venture capital, Big tech, Trust and safety, AI adoption, Frontier AI labs, Personal agent, AI agents, AI in business, AI breakthroughs, AI future, Video generation, Cost savings, AI transformation, AI limitations.

Podcast Transcript

Jordan Wilson [00:00:18]:
It's probably a question that many of you have pondered to yourselves. Is this AI thing overhyped or is it maybe underhyped? And we might not know the answer to that question for many years, but it doesn't change the harsh reality right now, regardless of what your views on generative AI are, Everyone's doing the same thing. Everyone's in this mad dash to cash in on AI. So I'm excited for today's conversation as we tackle this, with a guest that's been there. Right? One who's actually written the book on this very topic. So, I'm excited for today's conversation, and I hope you are too. So what's going on y'all? My name is Jordan Wilson. I'm the host of Everyday AI.

Jordan Wilson [00:01:08]:
This is your daily livestream podcast and free daily newsletter helping us all not just learn what's happening in AI, but how we can actually leverage it to grow our companies and our careers. So it starts here with the podcast livestream. That's where you learn, but you leverage everything that we learn by going to our website at youreverydayai.com. So there on the website, you can sign up for our free daily newsletter. We're gonna be pulling out the best insights from today's conversation and breaking it all down for you there as well as keeping you up to date with everything else that you need to know in the world of AI to be the smartest person in AI at your company. Also on the website, you can go listen to more than 500 episodes with some of the brightest minds in AI. Alright. I'm excited for today's conversation on this mad dash to cash in on AI and talking about if it's over hype or under hype.

Jordan Wilson [00:01:56]:
But before we do, let's start off as we do on most days by going over the AI news. A lot actually going on today. So first, Adobe announced that it is adding image generation AI models from OpenAI and Google into its Firefly app, broadening the range of AI tools available to users and making the app accessible also on mobile devices. So Firefly users will be able to create images using OpenAI's GPT ImageGen. A lot of news on that one today, as well as Google's Imagine three, Google's VO two video, generator, Flux 1.1 pro, and more. And they will be adding models soon, from LUMA and Runway. So Adobe is aiming to serve different user needs by by allowing customers to choose between its proprietary Firefly models for production level work and also third party models, whether it's for idea, ideation, experimentation, or whatever else. So pretty big news there, from Adobe.

Jordan Wilson [00:03:02]:
Alright. Next, but, pretty big as well. Jeez. A lot of for a random Thursday here in the middle of, April. Lot of AI news. So, Microsoft has officially unveiled its Copilot wave two spring release for 2025, with some new updates and just some old ones as well, putting it under the spring wave two wrap. So couple new things. The chat interface has been redesigned to make it easier for users to find and use tools, access previously created content, revisit past conversations, and engage with favored AI agents directly from the chat.

Jordan Wilson [00:03:40]:
Previously, Microsoft has also rolled out AI agents called the researcher and analyst powered by OpenAI's latest reasoning models, which will soon be accessible through their frontier program. Also, a new agent store will allow users to browse, select, and pin preferred AI agents for quick access. And then probably the biggest one out of this is, well, they're gonna be rolling out ChatGPT's new g p t four o image generator, and also integrating that, as well throughout their Copilot platform. The Copilot notebook feature will soon transform notes and data into actionable insights, including dynamic charts in those ever popular audio summaries narrated by two hosts. So, yeah, a lot, going on there on the Copilot side. And last but not least, related actually to those last two stories is OpenAI has officially, unleashed their image gen model to the API. Alright. So you might be wondering what does that mean? Well, all those viral images that are being created in OpenAI's GPT four o image generator, well, they've all been inside of Chad GPT, but now OpenAI is opening that up to developers everywhere.

Jordan Wilson [00:04:56]:
So what that means is you're probably gonna be seeing, a lot higher quality of AI generated images where, you know, previously, especially for these companies that maybe still use OpenAI's Dolly technology, their images weren't that great. But now we are gonna see a much broader rollout, across the board to probably, hundreds. And major companies already, such as Canva, GoDaddy, and Airtable are already exploring or even already have implemented, the new model, which on the API side is called GPT image one. Terrible name. Anyways, but this is going to definitely, streamline creative workflows and offer image generation capabilities, directly in so many more enterprise tools. Alright. For more on those stories, make sure to go to your everydayai.com. Alright.

Jordan Wilson [00:05:47]:
Let's get into the heart of today's conversation. This is this is one I'm stoked for. Bringing in, Chicago, writer and reporter, right, who who used to be at the Chicago scene. So, for our livestream audience, please help me welcome to the show, Gary Rivlin, the author of AI Valley. Gary, thank you so much for joining the Everyday AI Show.

Gary Rivlin [00:06:09]:
Great to be here. Thanks, Jordan.

Jordan Wilson [00:06:10]:
Alright. I'm excited for this one. Livestream audience, thanks for tuning in to Kyle, then Jean, and, gosh. We got we got a couple dozen already in the house. So if you do have any questions on today's topic, make sure to get them in. Might have time for them at the end, but let's start at the top, Gary. Give us a little bit about your background, in Silicon Valley specifically and and how you've been kind of reporting and following, AI since the early days.

Gary Rivlin [00:06:33]:
Right. So I I've been a political reporter at the start of my career, but it was the mid nineteen nineties. I was in the Bay Area, San Francisco Bay Area. And, you know, I I had actually got to school as an engineering major. I had program, but just got interested in politics. But suddenly, the .com Internet was rising. And so I kinda changed topics in the mid nineteen nineties and wrote about the rise of the Internet, the .com boom and bust, and all. And, you know, kinda since and and then kinda went on to I was working for Wired.

Gary Rivlin [00:07:02]:
I went on to cover Silicon Valley for the New York Times through the mid two thousands, you know, kind of the right Google going public, kinda, revival of of of Silicon Valley after kind of the dot the 2000.com, fall. And, you know, it just I'd kinda moved on, but then I just got this random email, at the end of twenty twenty two from Reid Hoffman. Reid Hoffman, founder of cofounder of LinkedIn, investor in Facebook, Airbnb, you know, kind of just a a central figure in Silicon Valley. I had gotten to know him in the early two thousands right after LinkedIn went public excuse me, launched. I I wrote a feature for Wired. So I've known him since the early two thousands, and I was on his you know, he's like, dear friend. I'm, like, one of 2,500 friends. And, you know, on a different day, I might hit delete, but I read this thing and, like, he's like, hey.

Gary Rivlin [00:07:51]:
I'm having my my first I'm I've cofounded my first startup since LinkedIn, and it's around AI. And he had this one line. This remind me of me being a frustrated programmer earlier in my life. Like, instead of us learning the machine's language, the machine is gonna speak our language. And I felt like, well, like, wow. And I realized, like, this is I had great time. It's right before the world kinda went crazy with ChatGPT. Again, the end of twenty twenty two was like, maybe I should move back to Silicon Valley.

Gary Rivlin [00:08:20]:
You know, that I mean, move back to Silicon Valley as a topic. And so since the start of 2023, I've been following, you know, his Reid Hoffman's company and other founders, other investors looking into Google, Microsoft. How are they planning on cashing in, taking advantage, of this moment?

Jordan Wilson [00:08:38]:
So I love that story. Right? Like, getting an email from from Reid Hoffman and being like, Like like like, really thinking about it and first being like, is this real? And then figuring out, oh, yes. It's definitely real. Right? So, you you know, walk us through. How did that, initial email, from Reid eventually right down the line, turn into, the book AI val

Gary Rivlin [00:09:02]:
Right. So, you know, I mean, just because one person is telling you, like, hey. This moment is here. You you don't necessarily get on airplane and, you know, start reporting. So I just, you know, made calls and, like, you know, just figuring out that generative AI really this was a breakthrough moment. Again, it it it helped that my timing happened to coincide with ChatGPT. So the wider world was waking up to the potential of generative AI. But, you know, just kinda recontacting sources of mine from over the years and understanding, like, wow.

Gary Rivlin [00:09:32]:
All this has happened, like, the transformer model out of Google, you know, GPTs one two3 three point five, you know, before this that they're like, wait a second. You know, there have been these key moments, and this is the inflection point. So, you know, early on, I just realized, like, wait a second. This is one of those before and after moments, like, before ChatCPT, you know, a small group of people were involved in AI. But after ChatCPT, this is Gen of AI. This this isn't a product like Google Translate behind the glass. This isn't a recommendation engine. This is something you can talk with, you can interact with.

Gary Rivlin [00:10:10]:
And so, you know, I I I just kinda instinctually understood that this is a an inflection point, and I should I should jump on jump on this moment.

Jordan Wilson [00:10:19]:
You know? And and and we'll make sure to, leave the link, to the book in the newsletter, today because I think it's it's definitely on my, list of books that I need to read very soon. But, you know, walk us through maybe what are some of your, main findings, right, from being able to to talk, with with some of the leaders that are building all of the AI systems that we're using now as well as just, you know, being, privy to everything. Right? As a former, Pulitzer prize winning journalist, you know, privy to probably things that we didn't get to see on the surface, but walk us through maybe some of your main findings, of the book in in AI Valley.

Gary Rivlin [00:10:58]:
Well, I mean, one of the first things is how did we get here? And, you know, looking into the history was fascinating to me. I I tell it through a a couple of characters, but, you know, since the nineteen fifties, there was this overoptimism that AI is just around the next corner. So for, like, decades, since the fifties, it's been just around, the next corner. You know, there was this massive wrong turn that, you know, kinda rules based. This idea that you have to, like, teach a teach a program line by line by line how to react. You know, that was the main, approach. People are ridiculing neural networks. Neural networks date back to the nineteen fifties.

Gary Rivlin [00:11:36]:
But, you know, people didn't see any promise in that. And in fairness to those folks, there really wasn't much promise because computers weren't powerful enough until that point. We didn't have the digital data for training. That wouldn't really happen until the second half of the nineties and the rise of Internet. So that was one thing. Like, just kinda how did we get here and sort of, you know, just just sort of the to understand this moment, it it helps to understand, what had happened before it. And, you know, I I also started realizing, like, I I I'm in deja vu. This is this is the mid nineteen nineties all over again.

Gary Rivlin [00:12:10]:
We we could talk about a couple of, differences. This the the change is moving faster than it did in the nineteen nineties. But with that said, you know, there's this idea I I've learned this early on. My my second interview ever in Silicon Valley, this really smart guy, Paul Saffo, a forecaster, you know, kinda helping us understand the future said, you know, we tend to overestimate the short term impact of a technology, but underestimate the long term impact. And that's the Internet. And then you've asked me about the Internet in 1997. Like, I understood. Lots of reporters understood.

Gary Rivlin [00:12:44]:
Like, yeah, startups, you're not really gonna get this isn't gonna happen fast enough for you to get, you know, fabulously wealthy overnight. And, you know, most dot coms did in fact go out of business, with no return. But I also didn't quite understand in '9 in the mid nineteen nineties how pervasive this would be. In fact, I remember talking to a venture capitalist back then, John Doerr, very well known venture capitalist. And he just invested in this company called Amazon.com. And he says to me, like, oh, they're gonna do more than books. And so to my little mind, it was like, oh, they're gonna do CDs too? You know? They're gonna you know, it's just gonna do and so, you know, I I don't think we understood it was gonna be the everything store and so much of the world, business, education, and your personal life would be transformed by the Internet. And I I think the same you know, two years on the front lines of, AI, I think the same thing.

Gary Rivlin [00:13:37]:
There you know, I I use models. I use large language models every day, all day. They're really useful tool. I could tell you we could talk about, you know, the strengths and limits, but, you know, it's a powerful, powerful tool. However, it's limits. Like, people talk about 2025 is the year of the personal agent. Like, yeah. But it doesn't really have much of a memory.

Gary Rivlin [00:13:58]:
And so the whole advantage of a personal agent is they know, like, you know, Jordan likes aisle seats and Gary likes windows. You know, they get to know you over time. They really I'm having a

Jordan Wilson [00:14:07]:
hard time remembering you from session to session. But imagine ten or fifteen years from now, and I think we'll say the same thing about AI that we've said about the Internet. It's changed everything. It's it's turned upside down organizations, education, you know, personalized businesses across the board. So, you you you know, I'm curious, and that was some great background there and and drawing these parallels to the, you know, the .com, you know, boom or bust era of the late nineties and ultimately what that led to. But, you know, I'm wondering, you you know, what other insights did you glean and, you know, like, that that comparison specifically. Right? I'm always thinking about it because I think, you know, companies specifically, they had, you know, three, five, sometimes ten years to, you know, figure out their kind of Internet strategy. Right? But on the AI side, I tend to think it's not really the same.

Jordan Wilson [00:15:04]:
Yet I think people are are are really drawing that comparison, in my opinion, maybe too tightly, right, with with everything on AI and and and kind of the .com era, and they're like, we've been here before. So, you you know, I'm curious from from from your findings, your reportings. Are you seeing that as well? Companies trying to follow, kind of their their.com, business plan a little too too tightly? Yeah.

Gary Rivlin [00:15:28]:
Are you talking about kinda potential customers? Are you talking about the creators of these,

Jordan Wilson [00:15:32]:
maybe these thoughts. Both. Yeah. That's that's that's inch, interesting to tackle it from both.

Gary Rivlin [00:15:37]:
Yeah. I mean, you know, for the for the creators, you know, let let let's understand that when, you know, an open AI or an anthropic takes, you know, billions, tens of billions of dollars in venture, you know, venture dollars, they really need to satisfy their investors. And so, you know, I I I really do think they're following that .com playbook of overpromising, and I think that's really gonna hurt them. I mean, I wouldn't bet against OpenAI. But, you know, just sort of the there's an issue that the acceleration is the Zoomers, those who say, like, full speed ahead. Let's put no, guidelines on AI. The general public is the majority of general public is fearful of AI. Less than one third, of folks in in in polling are excited about AI.

Gary Rivlin [00:16:26]:
So I I do worry that the the hype is gonna cause a similar fall. I don't think we're gonna have a .com like bust with with AI. We're not seeing all these AI startups go public. It it it'll be different. But I I really do worry about kind of a a a boomerang. So, you know, if if I was talking to a business audience, I would say, yes. You should be experimenting with this. But, again, understand it's it's this is idea of, like, autonomous AI.

Gary Rivlin [00:16:53]:
This idea that, oh, we'll just have AI employees, digital employees, and they'll do the work. It's like, I I think in the short medium term, that's the wrong construct. To me in fact, I love Microsoft's name, for its, you know, for its chatbot, Copilot. That's perfect. It's like, it's gonna help you. You know, we were talking before the show, like, you can't type into a prompt. You know, make me a Martin Scorsese movie. You know, enter.

Gary Rivlin [00:17:19]:
Like, it's that's not the way it works. You are the creative. It's human centered, you know, right now. It's a fabulously powerful tool. It's a limited tool. It makes mistakes. It's not quite where it's gonna be in five or ten years, but it's a really powerful tool that can help you do whatever you wanna do. For for the book, you'll you'll laugh at this.

Gary Rivlin [00:17:37]:
Like, the first thing I did when I thought of the, should I write a book about, AI? I went to ChatGPT, like, write me a 5,000 word proposal that would win me a contract on a book on AI. And, you know, I got two things from that experience. You know, one, it is far better read than I am as a far superior memory than than me. But on the other hand, the second thing is, like, it can't really it's very mediocre writing. It's very flat. And so, like, however, when I say, hey. Here's a paragraph. I'm not loving it.

Gary Rivlin [00:18:11]:
I like the ideas, and it helped me improve it. That's when it's really strong. Get it. I'm the creative. I'm giving the ideas. I'm giving it the approach, and it could help me see that through. It could help me with the execution. And so, like, you know, I I I really think that's important for businesses to understand that this is an autonomous AI.

Gary Rivlin [00:18:30]:
We're not getting rid of the human, you know, at least not quite yet. I can't predict ten, fifty years, from from now. But it it it I

Jordan Wilson [00:18:40]:
want everyone's students in school to business leaders and everyone in between. Like, use this. Understand what it's good at and what it's not good at. So, Gary, you you you brought up an interesting fact as as as part of a study, you know, saying that less than one third of people are excited, about AI, yet the six largest companies in The US by market cap are sticking every dollar they can into investing in these systems. Technically, hundreds of billions of dollars in in CapEx and in AI spending. Right? So how how can you kinda, like, explain that dichotomy? Right? Because people aren't necessarily thrilled about AI, yet the biggest companies in the world are sticking hundreds of billions of dollars and kind of saying, like, hey. It's coming no matter what. So, you you know, I'm curious, you know, how can we make sense of that, kind of dichotomy?

Gary Rivlin [00:19:35]:
You know, I I think it's very rational for a a a Microsoft or Meta to be putting in these tens of billions of dollars will probably be soon each company hundreds of billions of dollars because if they miss this, it's gonna be a multitrillion dollar miss. Mhmm. And so, you know, you can almost see it as a hedge against, the future, but I don't think it's a hedge against the future. I come back to we tend to overestimate in the short run and underestimate the long run. Like, these companies, I mean, you know, Microsoft sitting on a hundred billion dollars, they can afford to do this kind of investment because they see, and I think they see correctly, that $5.10, 15 years from now, they're gonna get a nice return, on this this investment. Like, you know, right now, we're, you know, capacity on these data centers. They're predicting that, you know, by 2030, we'll have have a doubling of data centers. So I I I don't I don't see that diminishing anytime soon.

Gary Rivlin [00:20:29]:
I just you know, there's an expression among the venture capitals, like, you know, go to where the go where the puck is going, not where it is right now. And I think that's what these companies are doing. You know, they see, like, you know, we're making an investment not for the next quarter, but for a five or ten year,

Jordan Wilson [00:20:46]:
time horizon. And that and that seems very difficult. Right? Right? To to even, like, my, you know, because I think about the, you you know, skate to where the puck is going, which which makes sense and has historically made sense for, you you know, at least dealing with tech innovations over the past couple of decades. But at least for me, when I look at everything as as someone that covers AI every single day, it's almost like, you know, we're in a new game. Right? Like, there's no longer a puck. Right? And you and there's no longer skating because, you know, all these, you know, machines are starting to to do these things for us. Right? So, you know, not not asking you to be a a business, consultant here, but but how can the majority, you know, of Americans and our listeners kind of, you know, kind of, take on that that that mantra of skating where the puck is going when AI development is so freaking fast.

Gary Rivlin [00:21:39]:
Right. So I I I was talking about kind of the venture capitalist, the founders, you know, the the tech leaders, skating where the puck is going. I think for people right now, you know, again, start using it. I I I'll use that. I wrote a a feature article about this incredible company, Runway. They do text to text to, excuse me, a text to image to video, text to video. And, you know, they're they're one of the leaders in, you know, video generation. I mean, they're up against OpenAI, Google, etcetera, so who knows who's gonna win this? It's it's three artists in New York City, you know, taking on, you know, deep learning PhDs.

Gary Rivlin [00:22:15]:
And so far, if they're not ahead, they're right there in the competition. And so in in doing that article, I said, like, it's really important that who's actually using your product? Hollywood is using their product. They they, like instead of having a, like they wanna do a sixty second scene that is on a mountainside with a thousand soldiers in the midst you know, in the middle ages. That's prohibitively expensive. It would be $10,000,000 more, and they probably wouldn't do it. But now with AI, they can do it. And so, you know, what are the practical things? I talked to a global production production company that uses, a runway all the time. And, like, you know, how how is this helping? They said, you know, in the in the past, they'd have 50 or 60 people for a film shoot.

Gary Rivlin [00:22:58]:
You know, kinda people organizing the talent, the talent, the production people, the post production people, all of that. And fifty, sixty people now using Runway, they have 15 to 20 people. You know, the timeline instead of being, you know, typically sixteen weeks from thinking of an idea to, you know, launching to to releasing it, it's now six weeks. So sixteen weeks to six weeks, and they say, like, the average project, we're saving 50 to 80%. See, that's a very concrete way. They have figured out, like, okay. We're still figuring this out as you correctly say, like, who knows in five years what a video model what what will a cutting edge video model, you know, look like? But they're figuring out how to use this thing that exists now that's a limited but powerful tool or a tool that's very powerful in certain ways and not quite what it's going to be. And so they figured out, like, how to say 50 to 80%.

Gary Rivlin [00:23:49]:
That that's a that's a big, big savings. And so, you know, across the board, you you've, you know, you've got your marketing team, you know, when is it marketing team of 10? Like, I think using AI, you can have a marketing team of five that's, you know, using the AI to iterate and, you know, kinda, oh, here's my idea. Do a quick illustration. Oh, no. Change it. Change it. Change it. You know, you'll still have an illustrator or two.

Gary Rivlin [00:24:12]:
You'll still have creatives, but but I don't think you'll need as many. I think I I think right now, AI is a great cost savings tool rather than like, oh, we will replace our marketing team, you know, with digital labor.

Jordan Wilson [00:24:23]:
Yeah. I I I I think the runway example is great. And, yeah, obviously, you know, gen four, extremely impressive, and it is cool to hear kind of the story. Right? It's artists and not, you know, people with, you know, twenty years of of deep learning experience. And, you know, yeah, they they did their partnership with, Lionsgate. Right? Like, a huge, studio there. So extremely impressive. You know, so I'm wondering, you know, Gary, kind of like going back to this concept of, you know, AI maybe being overhyped in the short run and maybe underhyped in the long run.

Jordan Wilson [00:24:57]:
Even for you as as someone that was covering this, you know, from the front lines, so to speak, and being able to talk, you know, personally with with some of the leaders in the AI industry, what were maybe some things that even took you back and you're like, oh, wait. You know? Wow. Right? Like, I have those moments all the time, but I'm very much on the outside, looking in. So from an insider's point of view, you know, what over the past, you know, year or two or a couple of months or a couple of weeks, right, is is really even shocking you.

Gary Rivlin [00:25:27]:
Well, you know, this idea and this definitely, you know, conjured up memories from the .com years. This idea that venture capitals are are are so scared of missing out on the moment, this the FOMO. Like, we could have gotten early into x, but we missed this, you know, multibillion dollar opportunity. So they're throwing money at these half baked ideas. I mean, Ilya Sutskever, the cofounder of OpenAI, who famously initiated the coup back in 2023. He left OpenAI for obvious reasons. You know? So he's founded it was Safe Superintelligence, I think it's called. Yep.

Gary Rivlin [00:26:02]:
That's their time. That they don't have a product yet, but they've raised billions of venture capital, and they have a paperwork of, I think, $32,000,000,000. And so I kept on seeing this happen over and over again. There's, you know, a promising team, like, oh, two top Google, machine learning specialists teaming up with someone at, you know, Meta who left, another machine learner. And they they create a company, they write a memo, and they raise, you know, $10,000,000 on a hundred million dollar, you know, valuation. Like, you know, I think two things, like, that's insane. And the other thing is, like, I sorta get it. These are really promising people.

Gary Rivlin [00:26:38]:
We have no idea. We wanna get in on the hot deal. We might lose our $3,000,000 we put in towards the 10,000,000. But if we don't, you know, if we if we if if it pays off, you know, that 3,000,000, you know, could be, you know, billions. I I example I so I was covering, Silicon Valley through the New York Times in the mid two thousands, and there's this new company, the facebook.com. At Excel, Jim Breyer, famous venture capitalist, put in I can't I think he put in $12,000,000 at a hundred million valuation up and down Sand Hill Road. You know, it's all through Silicon Valley. They're all mocking, you know, Breyer.

Gary Rivlin [00:27:14]:
What a fool, you know, for this college only app. He's putting in all this money, a hundred million dollars. Like, that 12,000,000 became, like if you if you have kept it through going public, it was, what, 4,000,000,000, something like that. And if you held it, it'd be worth, you know, tens and tens of billions of dollars. And, you know, that's the venture capital game. Most bets are wrong. I mean, they either, you know, show nothing or they show barely any return. But, you know, it's it's it's it's a game where that one big hit, can make up for everything.

Gary Rivlin [00:27:46]:
So and that's that's venture capital in Silicon Valley. You know? We're just gonna slap down bets and just hope that one or two of them are are are are huge. So you you you asked me why to conjure it up. There's kind of this insanity that's understandable.

Jordan Wilson [00:28:01]:
So, you know, one thing that you mentioned earlier, right, when we are talking about, you know, less than a third of people, you know, being excited about AI, yet all the big companies are, you know, investing in it. You kind of talked about this this boomerang, effect. Could you could you talk a little bit more, about what that means? Well,

Gary Rivlin [00:28:24]:
you know, so there's business and there's the personal side. Business, I put in a different category. Like, okay. If your work is saying, like, the Shopify CEO, I want every employee to be using it. You're gonna be using it. But, you know, kind of this whole idea, personal agents, this idea that the public's gonna be the, you know, the consumer, the general consumer audience is gonna be using these things. Like, we're asking you know, these personal agents are gonna have all this power. They're gonna have access to your private information and, you know, kind of in an existential sense.

Gary Rivlin [00:28:58]:
Like, humans are soon not gonna be the smartest entity on the planet. It's kind of frightening for people. I mean, I blame the media some for that. I blame Hollywood, you know, for, you know, the Terminator like movies like, oh, the robots are gonna take over. It's not my my concern. But, you know, so I I I feel like they're gonna get ahead of this, and I think it's inevitable that something bad's gonna happen because AI like, you know, just to make one up, like, a trillion dollars is siphoned off from the glow global economics system before a human even understands what's happening. A deadly pathogen. You know? And and AI for good, it could create, you know, vaccines and and and cures is also one that could create a deadly pathogen.

Gary Rivlin [00:29:38]:
So I think I I think it's inevitable that something bad is gonna happen, and I really fear that's gonna turn the general public off, to AI. So I I I wish the the those working out frontier models would be a little bit more aware of where the public is. Assure them that trust and safety is essential as I'm sure your listeners know. Trust and safety used to be a major concern of all these companies. But once the starting starter's pistol went off, once ChatGPTwo went out went out there, you know, the competition cashing in became a priority, and trust and safety became a secondary concern. I think that's a mistake.

Jordan Wilson [00:30:17]:
That is fascinating. Right? And even these, you know, Frontier AI labs, their their their safety teams have, you know, kind of shrunken over over the years or maybe, you know, it was front and center previously, and now, you you know, it's it's it's kind of an afterthought. Right? Like, the safety of models. Right? Like, what are you seeing? You you know, what are you maybe seeing as as as similar trends? Right? So that's that's one. You know? Hey. Safety used to be all about safety. Now the starter pistol went off. It doesn't really matter.

Jordan Wilson [00:30:46]:
Right? What are some other trends that that you've picked up through this experience of of reporting, you know, in Silicon Valley that that maybe the rest of us haven't picked up on yet?

Gary Rivlin [00:30:56]:
So what really scares me so so I approached this book, looking at startups. I felt like I wanted to find who's gonna be the next Google, who's gonna be the next Facebook meta. And what I discovered after, like, all my reporting is, like, the next Google is likely gonna be Google, and the next meta is gonna be meta. These things are so expensive. When I started, it was like they were talking about this is started 2023, tens of millions of dollars to train, fine tune, and operate, these large language models. By the time I was done reporting, you know, it's hundreds of millions of dollars. Now it's billions, if not tens of billions of dollars. And, like, a large venture capital outfit in Silicon Valley raises $1,000,000,000.

Gary Rivlin [00:31:40]:
But if anthropic as Dario Amode, you know, said, they're gonna need he figures a hundred billion dollars, you know, for their next iteration their 2027 iteration. How are they gonna raise that money? So I'm scared it's gonna get so it's so expensive that either they're gonna have to sell so much of their company that they don't really own the company, you know, anthropic to use that example. Google's put in billions of dollars. Amazon's put in billions of dollars, or probably more likely, they'll end up getting bought by a Google or an Amazon. So, you know, my big fear is that big tech, which create a big mess with technology in the twenty tens, that they're gonna be in charge of AI. There's the trust issue. You know, AI had really bad timing. Right? Came out, you know I mean, AI has been around for decades, but, you know, generative AI coming out at the end of twenty twenty two, you know, trust in big tech big tech was at pretty much an all time low.

Gary Rivlin [00:32:34]:
You know, Facebook, Amazon, Google, people really aren't trusting those companies, and yet I do think AI is gonna be in the hands of Google, Microsoft, maybe in OpenAI. And, you know, I'll point out, by the way, OpenAI founded in 2015 was a counter to Google Mhmm. Rather than I know it started as a nonprofit famously. They're trying to rid themselves of their nonprofit parent. But it's, it started as kind of this idealistic thing rather than profit driving AI. This is too essential. This is too powerful. We want it in in the the hands of an entity that isn't profit driven, but OpenAI is as profit driven as Google or any other company out there.

Gary Rivlin [00:33:15]:
So even if an OpenAI breaks into that Google, Apple, Microsoft, Amazon, you know, category of of trillion dollar giants, I think they're the same as as those companies. I mean, Sam Altman, I it's amazing what he pulled off. He's an amazing CEO, but, you know, he's he's we've seen all the folks who have left OpenAI. I mean, the anthropic team before, you know, and before ChatCPT came out. But, you know, many people since then, you know, keep on saying the same message that safety is taking a back seat to shiny toys. Again, I get it. Like, let's let's these are great tools. They could do amazing things for medicine, for science, for education.

Gary Rivlin [00:33:58]:
I see fabulous upsides into these things. But, like, there's also like any technology, there's goods and there's bads, and I really wish we'd be more deliberate of ensuring that AI is a net positive and not a net negative. Yeah.

Jordan Wilson [00:34:15]:
I think that's the one thing ultimately on everyone's mind. So, Gary, we've covered a lot in today's conversation, but I'm sure something that remains on people's mind is, okay. Is AI overhyped or is it underhyped? So maybe for those business leaders, you know, they've they've been using generative AI. They're they're they're implementing it in their company, but they're still wondering that. Is it overhyped? Is it underhyped? What's your quick, takeaway on that as we wrap up today's show?

Gary Rivlin [00:34:43]:
I would say individual companies are overhyping their specific products and what they're able to do, but I think AI is under hyped. I I really think AI is gonna transform this world. Amazing things, are gonna happen because of AI. Yeah. Under I'm I'm gonna say under hyped.

Jordan Wilson [00:35:01]:
Great great quest great great answer there and a great way to wrap up today's show. So, Gary, thank you so much for taking time out of your day to come on the EveryJAI Show. We really appreciate it. My pleasure. This is great. Thanks. Alright. There's a lot there, y'all, and there's gonna be a lot more.

Jordan Wilson [00:35:17]:
If you miss something that Gary talked about or if you wanna dig in more, that's where it's gonna happen in our newsletter. So if you haven't already, please go to youreverydayai.com. Sign up for the free daily newsletter. If this was helpful, tell someone about it. If you're listening on the podcast, thank you. Make sure to subscribe. Leave us a rating if you could. Thank you for tuning in.

Jordan Wilson [00:35:36]:
Hope to see you back tomorrow and every day for more everyday AI. Thanks, y'all.