Everyday AI Podcast – An AI and ChatGPT Podcast - EP 527: AI’s First Chapter: Why Generative AI Is Only the Beginning

Episode Date: May 16, 2025

Think AI is hitting a wall? Nope. This is just the start. Actually, we're at the first chapter. Here's what that means, and how you can move your company ahead. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Thoughts on this? Join the conversationUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:Generative AI's current phaseMeta's in-house AI chips developmentOpenAI's new developer toolsDay zero of AI and future prospectsReinforcement learning advancementsEmergent reasoning capabilities in AIBusiness implications of AI advancementsAI in healthcare and scienceTimestamps:00:00 Day Zero of AI03:31 AI Tools Enhance Customization & Access09:02 Reinforcement Learning Enhances AI Reasoning11:27 Agentic AI: The Future of Tasks15:59 Tech Potential vs. Everyday Utilization18:48 AI Models Offer Broad Benefits23:15 "Generative AI: Optimism and Oversight"27:08 Generative AI vs. Domain-Specific AI29:24 Superhuman AI: Next FrontierKeywords:Generative AI, Fortune 100 leaders, chat GBT, Microsoft Copilot, enterprise companies, day zero of AI, livestream podcast, free daily newsletter, leveraging AI, capital expenditures, Meta AI chips, Nvidia, Taiwan's TSMC, AI infrastructure investments, Amazon, Google, Microsoft, OpenAI, responses API, agents SDK, legal research, customer support, deep research, agentic AI, supervised learning, reinforcement learning, language models, health care, computational biology, AlphaFold, protein folding prediction.Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the Everyday Podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All In One Creative AI Studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. As someone that talks about generative AI, literally every single day,

and I've spent thousands of hours talking about it and learning from Fortune 100 leaders and teaching enterprise companies how to use, you know, chat GPT or Microsoft copilot. To me, it feels like we're decades and decades into this generative AI wave, even though we're only a couple of years. And when you think about it and zoom out, we maybe haven't even hit the tip yet. We might still be at day zero of AI.

And that's what we're to be talking about today. I'm very excited. So welcome to everyday AI, where we help you get past day zero, I guess. But we are your daily live stream podcast and free daily newsletter helping everyday people like you and me, not just learn AI, but how we can all actually leverage it

to grow our companies and to grow our careers. Because, yeah, development doesn't stop and neither do we. So that's why after you're done listening to this podcast, you need to go to our website at your everyday AI.com. There, you can not only listen to like 500 episodes from some of the world's leading companies and leading individuals in AI, but you should also be signing up for today's daily newsletter

and everyday's daily newsletter that we send out where we recap the topic that we cover on the podcast and the live stream. But then we also keep you up to date with everything else that you need to know. So make sure you go do that. All right, let's talk about the big picture. And that is, we haven't started.

Apparently, I mean, we haven't. Yeah, like I talk about AI every day, maybe too much. But the reality is, we are not even crawling, probably. All right. Enough of me chit chatting. I'm excited to bring on our guest for today. So live stream audience, please help me welcome to the show. Ron Green, the CTO of Kung Fu, A.

Ron, thank you so much for joining the Everyday AI show. Thank you for having me, Jordan. All right. For those that don't know, what is Kung Fu AI? Aside from like one of the coolest company names ever we've had on the show. Oh, thank you, thank you. So we're a strategy and engineering firm. We're like seven plus years old.

All we do is AI from day one. We help companies adopt AI strategy. We build custom AI solutions for them. Basically anything you need to get started or build your AI roadmap or AI capabilities, we help companies with that. Nice. So give me an example. A company comes to you. I mean, are they like, hey, we need to build off this new, you know, this new SDK from open AI. We need to make agents for our company or they come to you with, you know,

petabytes or whatever it's called of data. And they're like, help us use AI. Like, like, what does it look like and what's the end result? Yeah, it's a little bit more of the latter. We're basically solving really, really hard problems with with custom AI solutions. So, you know, people will come to us and they'll say, at things like, you know, we're trying to automate trading. We want to build a system that can trade hundreds of millions of dollars automatically, or we built a system that can predict the risk of breast cancer using just pure computer vision out to five years in advance at like

a superhuman level. That model is actually at the FDA right now for approval, things like that. Very cool. Very cool. So yeah, make sure to check out the newsletter if you want to know more of that. So, Ron, let's get to it. So it's your take that we're at day zero of AI. Why is that? You know, it's my way of kind of waking people up and making them realize that we've barely gotten going. So I've been working in AI professionally since the 90s. And we've made enormous strides. It's unbelievable.

If you compare to what, you know, compared what we can do now to back then, you know, we would have said, oh, my God, our dreams are coming true when I was in grad school. But it's also really, really clear that the slope, the velocity, the acceleration, is so great that we've essentially done nothing. We'll look back in five years and we'll think that the capabilities we have now are acute. And the same way we look back in 2020 with GPT2 and we're like, you know, that model, that's interesting that it can almost do something useful, but I'm sure there's not really going to be that much advancement. We are about to hockey stick is the point.

And there's a bunch of reasons. Well, let's start unpacking them. So, you know, as someone that's been in AI for three decades and to say, we're about to hockey stick, what causes you to believe that, right? Because obviously there's been, you know, you mentioned GPT2, right? I remember using GPT3, you know, about five years ago. And I was like, wow, this is pretty impressive. And, you know, it's obviously that the technology has grown exponentially since then. So why are we about to hit that hockey stick curve upward now? There's a bunch of reasons. Let's just jump to the, jump to the chase. The big reason is, is that traditionally we've, we've relied on something called supervised learning, right? I think everybody probably is familiar with this. You take a model and you teach it to do something that you don't know how to teach it explicitly. So just so everybody can have an image in their head, imagine you're trying to train a model 10 years ago to recognize photos. We would not know

how to do that with traditional software. We don't really know how our own brains do that. That's okay. We can take one of these big deep learning models and we can show it enough examples and generalizes. The problem with that is it's constrained by examples. You have to have a labeled example for every image. You have to know what the right answer is. And the model can generalize, but it really can't generalize beyond anything that we as humans can teach it to do, kind of mimicking this process. Well, that's changed recently. Reinforcement learning, which, incidentally, some of the two main people behind that concept literally just won the Turring Award last week, which is like the highest honor in computer science for the work. Reinforcement learning

with these new, much more capable language models have really kicked things into high gear. and we now have empirical evidence that we can elicit reasoning behavior emergently from these systems. So these systems, and technically we don't really need to go into any of the details. All that matters is we can take a strong language model, and by having it learn certain types of verifiable domains like learning how to program and learning how to code and teaching it to think analytically,

we are now, and we see multiple, Frontier Labs have verified this. Over and over again, we're seeing these strong models develop reasoning capabilities emergently. And here's why that's so important. Humans, as far as we can tell, are the only creatures with sort of metacognition. We can think about thinking. As soon as we have these models, which they can now, introspect and think about thinking, you've got this sort of infinite recursion ability. We can think about thinking, about thinking and we can gleam all of the deep insights from that. So we're really going to be off through the races. I expect it's non-exaggeration to say we'll have AGII within a year,

probably within three at this point. And it's all based upon this new development. Yeah, the AGI conversation is always fun. FYI, make sure, if you care about the whole AGI debate, make sure to tune in tomorrow. It's going to be a good episode, FYI. But let's get back to this, this concept of, you know, supervised learning, reinforcement learning, right? So for those people out there who probably unlike you and I, I like, you know, I read all these papers all the time, right? I'm sure you do as well, Ron.

But for everyone else, like, what is that actually, like, what's the tangible benefit for businesses, right? When we talk about like reinforcement learning and models that can now reason and they can, you know, introspect, right? Like, what's that tangibly means? mean for businesses? It's going to be huge. You know, I think probably everybody's heard about agentic AI.

That's going to be really big. Why is that going to be big? Because we're going to have these AI models that we can give high level assignments to, high level tasks, and they're going to be able to go and navigate the messy world. And so, like, unlike traditional RPA where, you know, maybe you're dealing with regular expressions and it's sort of whack-a-mole, you've got a million unending corner cases you've got to deal with. these models are literally going to be able to deal with situations that they've never seen before

and reason through them in intelligent ways. So that's that's one way. Now, that whole agenic world is a little bit more distant than I think some people argue. I think it's going to be more of a 20, 26 thing than a 25 thing, only because we're still kind of working out the kinks. But there are things like research agents, like deep research from open AI that are ready for prime time right now. And I used this all time. I was using it this morning to go and analyze really, really, really complex subjects. And it came back with a multi-thousand-word analysis. I read through it. I think it probably saved me two days worth of work. And it's sort of that high-level,

white-collar cognitively heavy work that we're going to see really being impacted in the short term. Yeah. And I'll have to put that in the show notes. as well. We covered deep research a couple of times. But I think one thing, like small, like aside, I don't think anyone else is talking about the fact that deep research is technically using an 03 full version, which is not out anywhere else. And it's actually using a mini version, right? It's using O3 mini as well. So it's actually like, you know, two different versions of O3 working together. Yeah, the research there is insane. I can't stop using the tool. You know, what what do you see as some of this big?

biggest, right? So we talked about how you see this impending hockey stick of growth and, you know, something like deep research, but is there any other, you know, happening or developments aside from the research itself, right, that you've seen recently that you're like, okay, even as someone with, you know, three decades of experience, is there anything you've seen recently that has kind of shocked you in terms of AI's capabilities? Yeah, you know, I don't exaggerate. I think I'm shocked on a weekly basis right now. I mean, some of this will be maybe less widely business applicable, but like what's happening in image generation, image synthesis, and video and audio, incredible. If you,

if you haven't checked that out, I go have some fun. You can lose a weekend on YouTube seeing what's going on there. But on a sort of a more practical level, healthcare and science generally are about to be massively disrupted. I'll give you an example that just blows my mind. So I've got a background in computational biology as well. And, you know, we used to dream in the late 90s of being able to sequence entire genomes. And we would think, well, what if we had the capability to combine like that sequencing technology with real artificial intelligence? Well, that's here today. There are, you know, models out there like Alpha Fold that have essentially solved one of the grand challenges of biology, which is, you know, amino sequence, amino acid sequence to protein folding prediction. That is, I mean, this is one of the most important accomplishments in the history of science.

And it's enabling, it's enabling amazing things like this. Yeah. And there is a, like a bioML group. It's a postdoc PhD-led group, student group at the University of Texas at Austin. And they held a hackathon about six months ago to develop novel proteins to fight cancer. And this was all done in one weekend, open source modeling. I think 62 countries participated in. They have 20,000 sequences.

They're going to have the final results in, I think, two months. So things have advanced so far that five years ago, this was an open. question whether this was even theoretically possible. And now you have hackathons on a weekend developing novel, novel cancer therapies. I mean, it's just incredible progress. Yeah, it's almost wild to me to think about the disparity, right? And even as you're talking about that, you know, one thing that was popping up in my mind is the, you know, Google co-scientists, you know, very, very impressive, you know, early agentic research from Google. That's going to be, I think, extremely helpful in that field.

But, you know, one thing that just always baffles my mind, Ron, is the disparity between where we're at, right? Like, you gave that example of the, you know, bioML. But then we have even like smart companies focusing so much time on just like using large language models to write like better LinkedIn post, right? And things like that, right? Like, are you ever baffled? Or maybe it's just me, but just at the disparity between the capabilities and then what

the average human is using this technology for, the average, even enterprise business sometimes, I'm shocked. It blows my mind all the time. You know, it's that old saying that the future is here. It's just not evenly distributed. I think a lot of people, they take a look at something. And I guess it kind of makes sense. You take a look at something.

and you get a read on it and you say, okay, I understand where we're at. And that may work, you know, back in the old days when things were moving at a slower pace. Right now, things are moving so fast, you know, if you were an expert in AI five years ago and you came back to work, you wouldn't even know where to start, right? So, you know, I would encourage everybody to have their head on a swivel here because things are moving incredibly fast. And, you know, that old adage is it's not it's not that. AI is going to beat your business. It's people, you know, in businesses leveraging AI that are going to take your business.

I'm glad it's, you know, taking the business versus taking the job. I've always like personally hated that like one-to-one comparison because I'm like, you know, oh, like AI is not going to take your job. Someone that uses AI will. But I'm like, what if that person using AI is using an agentic swarm, right? Like essentially, Open AI just released an SDK. and an API for agetic swarm. So it's like, okay, well, that person could, in theory, maybe do the job of, you know,

I don't know, 10, 50, 100 people, right? Can you talk even just about the capabilities and in what, you know, non-technical people or, you know, small businesses? Like, can you walk us through just what the capabilities they have? Because, you know, I feel generally, you know, to have that top echelon of technology has only been afforded to the 1% of companies, right? in the full Fortune 500, right?

What does the everyday non-technical person in business have at their fingertips? Adobe just introduced an entirely new way to create, bringing the power and precision of its creative suite into one conversational experience. Meet Firefly AI Assistant, now live in the Adobe Firefly app, the all-in-one creative AI studio. Powered by Adobe's Creative Agent, Firefly AI Assistant lets you start with your vision, just describe what you want and shape the outcome as it takes form with the assistant. The assistant orchestrates multi-step workflows drawing on 60 plus pro-grade tools across

Adobe Creative Cloud apps, including Photoshop, Illustrator, Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks, like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director.

Adobe Firefly AI assistant now in public beta. See it today at firefly.adobie.com. I would argue that the premier sort of bleeding edge reasoning language models like Claude, 3.7 or chat GPT 4.5 with deep research. Those models can be helpful to anybody. No, I don't, I don't care what your job is. If you're dealing with text or images or numbers or you're trying to think through problems or you're trying to understand data, those tools have a depth to them that most people, I think they just don't know how to use or they don't know how to

explore. So you don't have to go crazy. You can leverage these consumer grade reasoning models right now and get an enormous benefit. I mean, it's very difficult for me to think of a business that couldn't benefit from some aspect of that. I'm wondering, is day zero shifting, right? Can companies be at a at a negative, right? Like sometimes I'm flabbergasted, you know, big companies reach out to me and they're like, oh, you know, we're just now getting licenses for, you know, co-pilot or we're looking at licenses for, you know, chat GPT enterprise. And I'm like, you're a $20 billion company.

Like, why are you there? Is that day zero? Is it moving? So that's a great question. I mean, honestly, I think, we're, I think, the technology, I think AI is a technology is that day zero. There are companies who are so far behind, they're like day negative one or day negative two. And I see this all the time. For example, when when co-pilot first came out, I know people in the AI industry that didn't believe it,

thought it was BS, that it couldn't work. And I give presentations all the time. And I'll ask, you know, big developer organizations, you know, raise your hand if you're using co-pilot. And I would say, or something like cursor AI, some type of coding assistant. Invariably, I get a split audience. It's like half are using it, and the other half think it's just not worth their time. And they have no idea what they're missing because those tools are as powerful as the end user, right? And so if they're not getting a lot of goodness out of it, it's almost invariably their lack of understanding at what they could use with the tool, right?

It's like if you gave somebody a hammer and they were like, well, I don't really see how this could be useful in my world building houses. And it just boggles the mind. And that's just coding assistance. This is going to be applied more and more everywhere. But the challenge is, I think part of the reason, too, though, a lot of companies and maybe a lot of people are a little confused is they look at their mobile phone devices and, you know,

Siri is still as bad, you know, as dumb as a bag of rock. right? And you're like, I'm not so sure I believe this AI stuff is real. The problem is it's going to take a while for these really, really large corporations to integrate the capabilities that already exist right now. Like it may be another year or two before Siri even becomes capable of doing the things that technically, technically it could have like two years ago. Yeah, that's a great point. Yeah. We've seen recent reporting, say, anywhere from 2026 to 2027 until we get the actual, you know, AI, you know, series. So we'll see. You know, I want to follow up on this, this concept of coding in AI coding, which, you know, I know our audience isn't the most

technical, but, you know, I've, you know, in my 2025 prediction show, I said, the average person is going to be building their own applications by the end of the year. And then the Anthropic CEO just yesterday, Dario Amati, said that in three to six months, 90% of code is going to be AI. Within 12 months, it's going to be 100%. How do you see, even with non-technical people, but how is this concept of AI encoding really just going to change how business gets done? I see a lot of different avenues, but I'd love to hear from, you know,

someone who's doing this for three decades. You know, I probably think, you know, Zaria's estimates maybe a little optimistic, because any type of generative AI solution at this stage, even with reasoning models, does need some oversight because it can lose the thread or hallucinate or make mistakes. You know, these systems are not ammunition. Within, you know, three to five years, I think, you know, the bulk of new generated code will be AI generated and validated by humans or some other type of system. But the important point is that this is essentially taking the most powerful invention humans have ever created, which is like the ability to program general purpose computers, right? You could just, you know, let's just set the table really quickly.

General purpose computers like, you know, the laptop here on my desk, they're turning complete. They can literally do anything that we can write. down the instructions, right? So they're kind of unbounded from a capability perspective. And once we're in the realm where the process of converting from our minds into computer instructions, once that step has been made as trivial as just talking to a coding assistant in using just everyday plain English, that means application development and application customization and feature additions and feature enhancements, that is going to become dramatically less expensive, dramatically less expensive. And it doesn't mean the software engineers are going to go away,

and certainly not in the short term. It means the amount that we can leverage and get RLI from new code development is going to skyrocket. I mean, it's almost hard to exaggerate how much that is going to change things. There's that old adage that, you know, software is eating the world. Software is becoming embedded in every aspect of our life. You know, we have like an operating system in our refrigerator. Well, now we have the ability to go build and modify and enhance these systems across the board with artificial intelligence streamlining that entire process. You know, another hockey analogy, right, aside from it, you know, growth hockey sticking upward.

You know, they always say, oh, you know, don't, you know, you have to skate to where the puck's going, not where it is. My thought is no one knows where the puck's going, right? If you skate where you think the puck is going, you're going to miss the game, right? The bus has left. You know, how can companies, you know, not just, you know, oh, how can they get ahead? How can they keep up? Because as someone that does this every day, I struggle, right? And I agree with you. I think that, you know, we are going to see this hockey stick in the next couple of months. How do businesses keep up? So I have two pieces of advice that I typically give most businesses. Because, you know, if you're, if you're an executive running a company, you know, you have a full plate. Becoming an expert in

artificial intelligence isn't really an option. So it's two things. One is don't make the mistake. The generative AI is all there is to AI. And the reason I say that is, like I mentioned a second ago generative solutions are incredibly powerful in the right circumstance, but I see executives frequently have the false belief that they can build some generative solution and just plug it in and they don't, they forget about how they actually use it. The fact that if you're working with Claude or chat GPT or something, you're massage, you're giving it a prompt, you're getting an answer, you're correcting it, you're having a back and forth and generative solutions right now without a human in the loop just really don't work in a

production environment. So if you go put a bunch of money into a generative solution, you might be disappointed if you forget that really important fact. The other component is that domain specific AI is incredibly powerful. Now I'm talking about systems that don't have broad general capabilities, but they may have one or two or three very, very amazing superpowers, but that's all they can do, right? So you might have a system, an AI system that can detect fraud at a superhuman level, or it can optimize product recommendations or inventory optimization or all these types of things. Those are really, really powerful bets where you don't have to worry about necessarily having a human loop. And then the last piece of advice I would give is your data, the data that you have that is

proprietary to your business and that drives your business, if you can build AI solutions on top of that, then you're going to get the most ROI for a couple reasons. One is it's massive defense. You've got this data mode that is unique. Two, you can build capabilities to either cut costs, extend new functionality, have new predictive or perceptive capabilities based upon your own data. And that is a much, much better way to go into AI than building things or tools that are not based on your data and that could overnight become a product somebody sells.

And you just wasted this huge investment. If you just wasted, wait a year, you could buy it as a service. So focus your investments on maximizing the utility that you can get out of your own data. And, you know, one more question as we wrap up here. If, you know, if we're at day zero and if generative AI is just the start, what's next, right? I'm not asking you to look into your crystal ball and, you know, I'll come back in six months and say, how dare you not be able to predict the future, Ron? Right. But if generative AI is just the start, what's next?

I think the big next step, and you know, you can have me back on the show in a bit, and we'll talk about day one. It's going to be this. It's going to be these systems have superhuman abilities that go beyond mimicking some existing human capability, whether it's our ability to see or hear or something like that. It's, it's, and we're pretty close to this. I think this is really. best case a year, probably worst case five, median case, maybe two and a half.

And what we're going to have is, just as right now where there are reasoning models that can code at an elite level or solve math problems at, you know, an Olympic level, these models, we're going to start knocking down additional domains and we're going to have them be able to do novel scientific research, novel clinical diagnostics, not just like, hey, can you automate this thing humans already do, but they're going to discover novel insights. They're going to make recommendations. They're going to be able to be introspective on their own output and reason at a level that is so sophisticated.

It is literally going to have to dumb down for us the explanation so that we can. understand it. And that is what I'm so excited about. That's where we're going to be at day one, in my opinion. Love to see it. Getting us all past day zero of AI and prepared for what's next. Ron, thank you so much for taking time out of your day to join the Everyday AI show. We really appreciate it. This is awesome. Thanks for having me. All right, y'all, as a reminder, that was a lot. This is one of those ones. I'm already going to say it. You might want to listen to this twice. You also might want to go to our website, your everyday AI.com. Sign up for the free daily newsletter.

I'm going to have fun relistening to this one myself and writing down the most important takeaways for you to leverage what we just learned to grow your company and your career. Thank you for tuning in. Hope to see you back tomorrow and every day for more everyday AI. Thanks y'all. Meet Firefly AI Assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words in the assistant handle

the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adop.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating.

It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.