Artificial Breakdown

5. AI Agents - Karl Yeh

ZGM Season 1 Episode 5

AI agents are making waves in the business world—but what’s really behind the screen? Today, Carrie and Pete sit down with Karl Yeh, Chief AI Officer and Co-founder of 0260.AI, to talk about AI agents and what businesses should (and shouldn’t) expect from the buzz. Tune in and let’s break it down!

Guest: Karl Yeh

Hosts: Carrie Robinson, Pete Bishop

Producer: Pete Bishop

Music: 
Music from Uppbeat:
Title: Make It Happen
Artist: All Good Folks
License code: 6BVFZAHAWVMGPFRE

Send us a text

Peter Bishop (00:56)
Okay, welcome to another edition of the breakdown. We're excited today to have Karl Yeh on the podcast. Karl and I go way know, I trying to remember when we met and I think you were at the regional partnership that those were the days I'm not sure but it was a long time ago and I think it was on kind of like the on it stuff. Do remember that?

Karl Yeh (01:16)
Yeah, I do. was, I still see that on a bus too. Sometimes I still see that bus like, oh, I remember that. Yeah, it was what 2014, 2015, close to a decade ago. Yeah.

Peter Bishop (01:20)
Bye!

Yeah.

That's crazy. Yeah, I lost track

of it. But it's funny how many of us are kicking around town where it's just like this is such a small world. So just kind of keep bumping into like friends from the past. And, you know, I'm sad we didn't keep up, but I'm curious, like, would you mind describing a little bit about what you're doing now? Okay.

Karl Yeh (01:41)
Yeah.

Yes, sure.

we, I co-founded company called zero to 60.ai and it's an AI consulting company that does three major things. One is education on AI, strategy development on AI and implementation on AI. So we usually lead with implementation, but we're doing a lot more on the strategy side as well as the education side.

Carrie (02:14)
Hmm.

Karl Yeh (02:16)
So it's broad range of things. And really the big focus is AI moving so fast. What can we do to help businesses lay their foundation and also take actual benefits of what you can do today in the kind of the 10 to 40 % range. Cause that's where we see it's reliable and scalable. then, as AI moves gets faster, then you can start seeing

additional benefits from automations. We were going to talk about agents, digital assistants, even custom builds too.

Peter Bishop (02:53)
Cool. And are you doing, like I'm imagining you're doing more kind of the strategy side or are doing a little bit of everything? Cause you're kind of just, you're getting going here. So.

Carrie (02:53)
Whoa.

Karl Yeh (03:04)
you know, it's interesting because I actually started in the implementation side. So a little bit of history. this actually, the idea of this came about 2018. I remember where I was, I was sitting in the office of, of when I was working with Brookfield residential now, I think the properties and I was member, Sundar Pichai who was up on stage at Google IO and he was

demoing a, reserving a restaurant using AI. And I was like, this is it. This is where everything in my opinion is moving towards. I, so that was the kickstarted all the research that I started doing and then playing with some tools, but none, not nothing was really, you can't really use it.

Carrie (03:39)
I remember that.

Karl Yeh (03:58)
It's fun to play around with, but you couldn't really do anything with it. I think you, if you see some of the older models of Dolly and stable diffusion, they look like claymation. like woman with the red hair and it's like claymation. there used to be a show. was growing up. was watching it too. It's really young. I think it was Thunderbirds. Do you remember Thunderbirds? It was animated. Like I think it's stop motion animation.

Carrie (04:09)
You

Peter Bishop (04:20)
yeah.

Karl Yeh (04:25)
That's what it looks like. But not that good. looks like it's. Yes, but it's even worse because it's like someone just kind of put together whatever it was and it's like, oh, this is a woman with red hair. But obviously, 2022 and then everything took off. So what that led me to actually started creating content around it. Then some companies were, hey, why don't you come in just.

Peter Bishop (04:27)
Kinda like robot chicken almost, right?

Carrie (04:27)
Hahaha!

Karl Yeh (04:52)
you know, speak a little bit because we're hearing about it. So did a couple of presentations

While I was with the partnership, my buddy from high school, we kind of co-founded like a side project side business. So help there and sold that in 2016. I think so I then I took a course actually, one of the very first courses on

what's called retrieval augmented generations, like how to build essentially chatbot where you load a PDF and then you ask questions against it. So that was the first time. And then started learning that already had some understanding of Zapier and Make and then automations and all the AI automation tools that kind of followed that. And then what essentially happened is it just happened to

Carrie (05:27)
Hmm.

Karl Yeh (05:44)
run into my co-founder, who was about to semi-retire. Like he's like, I don't know what I'm going to do. And I was like, Hey, I got an idea for you. I was like, Oh, that sounds amazing. And so, so like, okay, let's, let's get going. And in fact, he wanted to get going like last November. So he incorporated the company really fast. was like, didn't, okay. He's like, let's get going. It's like, well, I didn't.

Carrie (05:56)
Why don't you work more?

Karl Yeh (06:13)
Hold on a second. Let's just calm down a bit. what he, well, like, then we started going to a lot of businesses and we actually saw it's like, there's significant interest here. And then we use it kind of as a side hustle. He was kind of more on full time. then like starting July, we just, it was starting to become too much that I like couldn't do two things at the same time. So I just like, have to go full time on this. Yeah. So yeah.

Carrie (06:14)
You

Peter Bishop (06:14)
Mm-hmm.

Carrie (06:40)
Mm-hmm. You just

Peter Bishop (06:43)
the one thing that we wanted to

touch on a little bit was the AI agents, because I think they're made out to be, when I see them online, I'm like, my god, this sounds like a virtual butler that will get you dressed in the morning and then do all your homework and then tuck you into bed at night. It just sounds too good to be real, and I suspect that's a bit of the case. Can you mind just diving into kind of what AI agents are to start with?

Karl Yeh (07:00)
Yeah.

I think it depends on who you ask the definition of, right? So I think initially what I thought was an AI agent and I'm bringing this up. So I think you all know who Dharmesh Shah is, right? The CEO of, or CTO, one of the co-founders of HubSpot. So I think like six or seven months ago, he put out a blog post on the definition of agents.

You actually can't find it anymore because he took it down. Because that definition has changed from what he actually created with agent AI. So the definition I thought was pretty good where software that uses artificial intelligence to pursue a specific goal. It accomplishes this by decomposing the goal into actionable tasks, monitoring its progress and engaging with digital resources and other agents as necessary. So.

there's that aspect. And I think one added piece there I thought would be it can think like it looks at all the steps or creates all the steps. And then if a step is not working well, it changes that task to do something better. So the example that he had, it was really good one was the goal is launch an online newsletter about whatever topic. So that's the goal. And that's all you give the agent.

Carrie (08:22)
Mm-hmm.

Karl Yeh (08:35)
then the agent would then come up with the branding, the tech stack, where to write it, when to launch it, coming up with the name, coming up with the logo, what's the design, research, ratings, summarize the pricing, introductory pros, flagship posts, announce the newsletters, send out through social. Like all those steps the agent would do. That's what my understanding of an agent was. But

What we're seeing like agent force hub spot breeze co-pilot agents. I would argue it's a glorified version of a Zapier or make automation just with, with chat GPT included. And you see that through tick talk and Instagram build this agent. And I'm like, that's just to make line. just added chat GPT to it or cause I think like everyone's calling it an agent.

Peter Bishop (09:15)
I'm just gonna bring up Zapier.

Karl Yeh (09:34)
If I was a real estate agent, I'd be like, well, I'm an agent too. So what, what does, what does that mean? So I was like, got to come up with a better name and then the term agentic. was like, what? Agentic. So it's like, well, it's still Zapier make relevance respell and nothing against any of those things. Those are very useful to actually do. like AI makes those automations work, but I just don't think you can call that an agent versus the real definition of.

Carrie (09:37)
Hahaha

no.

Peter Bishop (10:04)
Yeah.

Karl Yeh (10:04)
the butler,

or it's supposed to go out and do things for you and come back. That would have taken you hours or months.

Carrie (10:11)
Mm-hmm.

Peter Bishop (10:13)
Yeah. Yeah. I like where you're going, Karl, because I think this is a good distinction because the one side is like a if then statement that you've kind of pre-programmed in, which you're right. You could do with a chain of prompts. It's almost like the early automated chatbots where you would program in every single answer to every single question and then call that a chatbot, right? It's not really, it's just to choose your own adventure that's already pre-written.

and that feels to me like a lot of agents are being toed it out as, whereas what you described earlier, from HubSpot makes, to me adds way more value and feels like you're leveraging AI way more is that like you're giving it a high level task that requires a whole bunch. I'm just trying to summarize to see if I got this right. A high level task that requires a whole bunch of a chain of events.

that it is going to determine the best path and the best course, but it's a lot of work that you would have to try and think of, and you may or may not get it right. So you're giving it to someone who can probably get to point from point A to point B faster by leveraging some decision-making. To me, that's way more valuable than trying to just create a chain of events on your own and getting it to run through it.

Karl Yeh (11:30)
Yeah, you know, there's this interesting clip from the London OpenAI Dev Day where I think Sam Maltman was asked about, hey, what is your thoughts on AI agents? And he always referenced, it's always been the example of, hey, either book my travel or book my restaurant. But if you think about all the quote unquote risks of that, where it's like, well, what seat do you want? When do you want to go?

What type of hotel can you want to go to? How close to the destination? All the things that is super unique to you. It's a very interesting use case for an agent, but even he says like, we're not thinking of it properly. We're actually a better example would be if you as a senior coworker who you're, can give a task and just like an employee would go out.

do it and come back and have a discussion with you about it. So that would be to me the example. And then his example of what I think we're getting into the AA like more so artificial general intelligence. I think he said was could you hire a agent to be or can they do the work of a remote software engineer and all the things that come with that? Cause if it can't

Carrie (12:32)
Mm.

Peter Bishop (12:57)
you

Karl Yeh (12:59)
then that's really not where we're at. And you're seeing the agent name for automation.

Carrie (13:07)
Yeah, and I think that's what's interesting too is, I mean, Peter and I have talked about this too with chat bots alone. It's like, people just wanted to be the first one out to say that they've done it. They wanna be the first ones to have an agent, but that's not what you have. You have a watered down version of it. And it's like, so then because there's kind of this race on, people are putting out this AI tech.

Karl Yeh (13:22)
No, didn't. Yes.

Carrie (13:34)
you next level AI tech that it's just absolutely not ready. And, you know, and then we build on it in real time and, but the public already has it and it's just wild. It's like this wild, wild west of, and then when is, who's actually gonna come out with the real.

Karl Yeh (13:45)
Yeah. Yeah.

like the real, like a definite

agent. it's interesting because like you can start seeing the term being used. like, we built a voice agent, like what, or a this agent, sales agent. You're calling everything an agent. Every single AI use case has become an agent. And it's like, it waters down and then no one knows what an actual agent does.

Carrie (14:06)
You

Hmm.

Karl Yeh (14:21)
So maybe you'd have to rebrand the term of what an actual thing does that... I don't know. Honestly, I...

Carrie (14:29)
What would you call it? I like Butler.

Peter Bishop (14:33)
Well, there was that like virtual

So talk about setting these up. This does not sound easy. It feels like there's quite a bit more than, everyone's making it out to be like, anyone can spin up an agent. Maybe you can with, again, the simplified version of this. But to do a proper AI agent sounds like you're basically building an LLM or something like

Karl Yeh (15:00)
Kind of, there are a couple of people who have demoed it. Like, so they've taken, say for example, the OpenAI real time API, so through voice, and then they've, think they've run it through an SDK where it's a guy named Sawyer Hood. You got to check him out because he does, he's done two examples on this on X where he actually through voice says, hey, I want you to order, I think the black sheep sandwich from

whatever restaurant and then the agent will come back, hey, I can't find the black sheep restaurant or the restaurant, can you specify the actual restaurant? And then he's like, okay, want actually I want it to make a Greek style or whatever that is. What's interesting about that is that's having a full conversation and that thing is doing it for you. And I think that's where

Carrie (15:30)
Hmm.

Karl Yeh (15:56)
That's where, me, that's where you start getting a bit in pieces of the agent, what you can actually see. And then Anthropic obviously came out with their, their example of computer use where you gave it instructions. It would actually use your computer to, could, to do it. And I've run a little bit of version of it. It is super, super early because the demo, the demo you see is very slick, very good. But when you actually create it yourself,

Carrie (16:14)
Mmm.

Peter Bishop (16:25)
Mm-hmm.

Karl Yeh (16:26)
your setup doesn't work. like, there's too many tabs or this, this, and it just would run into errors. So I think it's a little early from a pure agent perspective. I think the only one that I've seen that works really well is something like if you've seen Bolt or Lovable, where you actually just give an instruction in plain language, hey, I want to build an application that does this, this, this, this, and it will actually build it for you.

You can see it code and then you can see a first concept and it actually works. I was like, that to me is an agent versus that's an example of an agent versus something else.

Peter Bishop (17:00)
Hmm.

Carrie (17:02)
Whoa.

Peter Bishop (17:09)
I think it's interesting to, and I know you hit on this a little bit, but it resonated with me is this conversations are nuanced, right? When you're trying to get someone to do something for you, there's a lot of variables and there's things that you may or may not be thinking of, but you kind of need that back and forth. And that to me brings the comfort back into getting a machine to do things for you. Cause right now it's just too simplified even to be like,

order a pizza. Okay, I ordered a pizza. Like, my God, like what did you order? there's, you need to kind of have that back and forth. And to me, that's kind of where it's trying to get to. And, I think there's a reluctance to jump into that because there's consequences to getting it booking travel and stuff while it's still figuring out it's seven finger problems, right? You know what I mean? so once it gets over that hump and everyone's proving it out, it just feels like, wow, the floodgates might really open on this stuff.

Karl Yeh (18:08)
I think so. just think, and I think this is where, you know, we can start talking a little bit about adoption wise because I do believe even though the tech is available, let's say, let's say an AI agent, a true AI agent was developed like today and all of us had access to it.

I just feel like with the businesses we talk to, none of them will be ready for it. And even if we can showcase, I think you've seen enough app, like implementations of anything that's like, there's a whole bunch of layers that still need to happen. And then back to the human habit. it will require, this is where I think like there, you probably need to add three.

Carrie (18:35)
Hmm.

Karl Yeh (18:58)
four or five years, even after the tech is complete, to see that implemented and then to see the actual change in society. Because no one's just gonna implement this right away and disrupt their entire business operation to implement this as a complete form. Like I've never seen that. Like no businesses should be like, well, we're gonna risk our profit margin, our market share, our operations, our people.

because we've seen it from some influencer who told us that this is gonna make us X millions of dollars. And no, it's not gonna happen. You're gonna have the referrals, best practices, all the fun stuff that goes around with implementing any tech.

Carrie (19:41)
I'm going to go pitch this down the hall right away.

Peter Bishop (19:43)
Man,

Karl Yeh (19:43)
Yes.

Peter Bishop (19:44)
we're gonna be millionaires.

Carrie (19:47)
But I like that thought too, because that is more of a, okay, we know this is probably coming down the road. So in the meantime, you know, as a business, it'd probably be smart to work on your AI literacy with the people who you work with, work on AI adoption, work on AI sustainability and work on those sorts of things. So when something this huge does come out, you can at least be a bit more prepared.

Karl Yeh (20:12)
Yeah. Cause I think like the, people that we talk to and the businesses that we meet and so on, the big pieces, they, there are people in the organization who are using AI, but you're not seeing an actual holistic business. Hey, we're going to revolve and we're going to move in tandem with everybody to ensure that like the entire business is useful. It's this person's using it for their specific tasks.

They're not really sharing it with, maybe this team is using it, but only in a limited capacity. But it's like, did you know that it could do much more than write your emails or, you know, or even just a more advanced and more custom GPT. There's a lot of other things. And, and even the chat GPT equals AI, that's where a lot of people are. They're like,

Carrie (21:06)
Mm-hmm.

Karl Yeh (21:07)
I know AI because I use chat GPT. like, well, no, there's a lot of other things out there that can do a lot of other things. It's just, again, moving too fast. When do you have time to learn it? And how can you keep, how can you test it when there's something new, like right around the corner? I don't have time. Like I'm busy enough. Yeah.

Carrie (21:26)
Mm-hmm. Mm-hmm.

you

Peter Bishop (21:33)
That's a great point. think you're right. I think everyone's really put ChatGPT as the kind of poster child of AI. there's a lot of the other tools just feel like a kind of a noise in the background where not everyone really understands them. But when you're looking at a company, feels like there's levels of adoption where there is your right. There's like the lone wolf or the team that's basically just using the most prominent

public facing, like the dollies and chat, GPTs and that kind of stuff. And then there's probably the companies that are looking at their, again, like looking at their processes and their, finding ways to automate and be more efficient by leveraging AI on a very like deeper level. And then of course, it's almost like table stakes that everyone else has to generally, if you're in computers of any kind, you're probably

kind of have to move along with the tide anyways. Like all the software that you're using is generally adopting AI anyways. So whether you want it or not, it's like we talked about this, you kind of get dragged along because the tools you even are comfortable with are changing. But for it to be active, building AI or really making fundamental changes, that feels like a next level for a lot of companies.

Karl Yeh (22:51)
And I have the, you know, there's the, these studies from McKinsey or Microsoft or PWC that says, you know, these companies increase adoption by 50%, 60%, 70 % full wide. I was like, is it really? Because when we go out, it's not like that. It's yeah, I kind of use chat GPT. I don't feel like using it or

Yeah, we kind of use it. I was like, where is this 50 % adoption? Like, what companies are these that you have gone out and surveyed? And maybe it's because our focus is more medium sized, not like enterprise, like we kind of try to stay away from enterprise. But maybe it is. But I just I can't believe that with what we're hearing.

Carrie (23:28)
you

Yeah. I mean,

like, it's funny because we've talked about that too. Like, I think we've had a half decent AI adoption at work here, but we're a creative agency. Like if you get these creatives into it, they're into it and it's exciting. And, you know, we're in marketing where we're forward thinking, we're always looking at like everybody here is excited about it and interested in it, whether or not they're all using it.

But yeah, you know, maybe some of these companies you go to are just a bit more like stay the course, things have been working and then why would they change just for some new technology?

Peter Bishop (24:18)
Mm-hmm.

Karl Yeh (24:18)
Well, the example I have is when I was still with Beneviti, we got the Chat GPT for Enterprise and I think we got a trial for three months. So that's over a thousand licenses delivered. And we, know, we did the, did six weeks of like, I led the six weeks of training and then adoption, like everyone was encouraged to use it. CEO was behind it. CTO was behind it. Everyone was encouraged.

And even the AI policies were very, very more like guidelines. And so you literally were able to access any of these tools, be able to, you know, we had hackathons, had science fairs and so on. But here's, and this is a tech company too. So it's, what's funny though is after I think three months and you could see the usage of each license of the 1000 licenses that were provided or 900, wherever that was.

I have a personal account and I have like the work account. I used the work account like three times, barely used it. Of the 1000, I was number 58 of top users. And I was like, wait a minute, I always use it three times. And I was like, so this is a tech company where everything is provided. You got Google workspace that was provided, didn't take like, so it's very interesting that like a company like that, full push, full backing. And I'm like,

Peter Bishop (25:22)
Hmm.

Carrie (25:30)
No way.

Karl Yeh (25:46)
That means not very many people were actively using it too. I, that got me thinking and then meeting other businesses, you got me things like it again, back to the AI bubble. got a whole bunch of people talking about it, excited about it. But then when you actually go out, it's like, well, no one's actually, it's very different from an adoption perspective.

Carrie (25:50)
Mm-hmm.

interesting.

Peter Bishop (26:12)
Yeah, you wonder where those stats come from, right? And what are the criteria of these stats, right? What makes a person use AI in the eyes of these reports? And yeah, who knows, right? Because I agree with you, Karl. I don't really believe that they have the data to support these reports right now, because I don't know who's surveying who, like, who's doing this, right?

Karl Yeh (26:39)
Well, it's like when Metta says, we have the highest adoption of AI. It's like, hold on a second. Just because you put Metta on Facebook, Instagram, WhatsApp, and someone opens Instagram, that does not count. That does not, you can't count that. Like does it, or somebody presses it by accident. It's like, what's this? delete. that counts. Or do we have X billion active users? Like act? No, no, no. They're using Instagram. They're not using Metta. Like they're not.

Peter Bishop (26:54)
Ha ha ha.

right?

Karl Yeh (27:08)
So llama, they're not using that. So I was always questions like in even these surveys, the surveys are going to companies that are already like, it goes back to copilot. It's like, if I turn on copilot, does that mean it does that count as AI? But I'm like, well, copilot is it's a very unique thing because it's like, well, just because it's there doesn't mean everyone's going to use it.

Peter Bishop (27:35)
Yeah.

Carrie (27:36)
Mm-hmm.

Karl Yeh (27:36)
And that to me is not an example of using AI. It's just Clippy 4.0. That's literally what it is. Right? So it's like what, I was like, yes, you can kind of write your emails, but like, are you getting the actual benefits of AI because you can use it in Excel, PowerPoint and Word?

Peter Bishop (27:44)
Hahaha

Yeah. Yeah. It just feels like it's being force fed into a lot of things just to have the word AI associated with the brand lately. Karl, let's end on just a couple of questions. I think would be kind of fun. super high level, but I'm just curious your take. I'm going to go negative first. What's your biggest worry at the moment with AI? And I know that's very broad, but you're just curious to see what you think.

Carrie (27:57)
Hmm.

Karl Yeh (28:22)
I feel the one thing that most people aren't paying attention to is the militarization of AI. Companies like, and full disclosure, I invested in Palantir. But companies like Palantir and DREWEL, if you look at what some of the headlines were, like one says Pentagon authorizes unmanned or AI.

lethal force or something like that. was like, no. So those kind of things, like, wait a minute, you'll have like full automation, AI automation in in military and weapons of destruction. So it's like, okay, I don't think people are paying attention. That to me is more worrisome than are we going to lose jobs? Yeah, it's going to change. But like, that's more pressing, I believe, than AI adoption.

Carrie (28:50)
Blech.

Karl Yeh (29:16)
and AI agents, right? It's like that to me is like more of a bigger thing that no one seems to be thinking of.

Carrie (29:23)
Yeah, fair.

Peter Bishop (29:24)
It does

feel like a little, little murmur in the background that you see every so often and you're like, I don't know, whatever, and just keep moving on. But I remember this little clip of this van pulling up this a long time ago and these three, essentially balls was like sticks coming out of them that could travel at insane speeds and like three of them jump out of this van and just roll on these little sticks and

Karl Yeh (29:33)
Yeah.

Peter Bishop (29:51)
book down this field and you can like, that's how we're going to get all rounded up on that with like these like little Doverman pincher balls like robotic dolls. Like it was crazy. And I just remember my brain actively like, Nope, not going to think about that. Let's keep moving. Right.

Karl Yeh (30:09)
Yeah, yeah

Carrie (30:10)
think my brain still

does this like, well, there's no way we've seen enough movies about how AI can be turned into a weapon. There's no way we're gonna do it. And it's like, nah, we are.

Karl Yeh (30:20)
Well, I think to one of the news items that that kind of was it's kind of there and kind of wasn't was the the the the I think the Cybertruck explosion and then the news was like, this was planned by Chachi GPT. And then that was it. And that was like one news. And OK, well, let's move on. It's like I'm not and I'm not saying you could. There's so many ways around that. You probably there's so many ways to plan that. But I was like. I don't know if anyone's paying attention to that or.

Peter Bishop (30:30)
right.

Carrie (30:30)
Mmm.

Karl Yeh (30:49)
Have we been so fatigued because there's so many drops and so many news, no one cares. Like it's like, I don't have time. It's so much. I'm going to just wipe it out. that's maybe that's where we are.

Carrie (30:59)
Yeah.

And I mean, the other issue is the companies who are maybe making the news have so much money. You just whoop, we're out of the news now. It's like, shit.

Karl Yeh (31:12)
It's, that's, yeah.

Yeah.

Peter Bishop (31:14)
Okay, well let's end on a positive note. Karl, what about like, what are you looking forward to most over the next little while?

Karl Yeh (31:22)
Well, I think with the more powerful models, the new reasoning models, like one of the predictions I had is, you we hear about businesses and AI adoption. I thought from a medical perspective, I really believe this year, there's going to be a major disease that AI will help find the cure to. And that's where I think like, that's the good of society. So like, if you have these models,

Carrie (31:34)
Mm-hmm.

Karl Yeh (31:49)
I think the 03 models like 01 questions $2,000. That means nothing. If it can solve 100,000, 200,000, 2 million, 3 million, it won't matter. Because if you solve this particular disease and help that many people, I don't think there's a dollar amount you can put to that. So as these models get better, you would think that that kind of stuff can now actually help and benefit humanity and society as a whole.

Carrie (32:18)
Mm-hmm.

Peter Bishop (32:18)
Great thought. You wonder if you could just aim this whole thing too good. what we could actually do instead of just money, right? Cause it does feel so capitalistic right now, just with the bend of these companies is just how do we monetize? And that's just feels like the main focus, right?

Karl Yeh (32:40)
I would love a headline that would say, hey, AI just cooked cancer. Way, way better headline than just cooked each other, right? Or cooked, that would be a better headline. That would be amazing. I think people would be excited about that. Exactly, right? Yeah. Yeah, 10 reasons how it's done. And then you have the, okay.

Peter Bishop (32:45)
Hahaha

Carrie (32:46)
Yeah, cancer cooked.

Peter Bishop (32:49)
You

Yeah.

Carrie (32:55)
I love that.

Peter Bishop (32:56)
Yeah. The death of cancer instead of the death of web 2.0 or whatever it is. Yeah.

Okay. On that note, thanks Karl for coming. That was awesome. What a great conversation. it was nice to catch up. need to, we need to catch up properly, but Hey, if this is the way we got to do it these days, I'm all for it. That was, that was wonderful.

Carrie (33:21)
That was awesome. Can you, just before you go, can you give us the name of your podcast?

Karl Yeh (33:21)
Sounds good. Thank you for having me.

Oh yeah, so our podcast, the one for zero to 60 is the AI Accelerator for Business Show. It's really focused for businesses to help them get the real, intangible results for AI.

Peter Bishop (33:40)
We'll stick that in the notes. imagine and Yeah, again pleasure to have you on Karl. Thanks so much Okay, bye

Carrie (33:48)
Yeah, this was wicked.

Karl Yeh (33:48)
Thank you.

Peter Bishop (33:58)
that was great.

Carrie (34:00)
It was good.

It was a bit more in-depth than some of the conversations. Like it's really talking about like some deep, intense AI stuff that not everybody's talking

Peter Bishop (34:13)
Yeah, like, Karl's pretty technical. So I'm not surprised it gets a little bit down there. But I think that's good. Because, you know, we've been talking a lot about, is AI good or bad? And it's nice to just get into like, okay, what is this? And the agent stuff cleared up some stuff for me, I think it's still a little bit fuzzy by the sounds of it. But I think there is no clear answer.

Carrie (34:13)
Help.

Mm-hmm.

Yeah, I think the whole thing is a little fuzzy still. Well, and I think too, a month ago, if we were talking about agents, I think there was a more clear answer, but now because all these people are coming out with their supposed agents, it's like, well, what is it? What does it actually mean? And I guess we'll see.

Peter Bishop (35:00)
Yeah, it's another way to feel like you're being left behind or you just don't understand and everyone's talking about it like they know it inside and out and it's super easy and you're gonna make a billion dollars. Why aren't you doing this? I think it's nice to pull the curtain off a little bit.

Yeah, I like Karl's approach to talking about influencers and kind of debunking them a little bit. He just seems to have a healthy perspective on stuff.

Carrie (35:26)
I think so too. And I hope that his content is the type of content that eventually floats to the top for people. I don't know, social media is just so conflated. But it's like, you know, he's the one giving actual practical tips that can help you compared to, you know, somebody giving you a pipe dream.

Peter Bishop (35:34)
Mm-hmm.

Yeah, the promise of money and whatnot. Yeah. What a cool way to end though. liked, liked, it was interesting. His negative was like, wow, I had not, I did not see that coming.

Carrie (35:55)
Mm-hmm.

Well, and all I could think of is we're putting AI into battle drones, but we are still handling, we're still dealing with the seven finger problem. Like, you know, you're still getting agents probably ordering the wrong pizzas and simple things like that. And it's like, okay, now you're going to tell it to, don't know.

Peter Bishop (36:15)
Yeah, what does that mean when you're firing missiles, right? Yeah.

Yeah, it's funny, I may or may not have been using chat GPT to help my son with his homework last night. To the point where, you know, take a picture of all the questions and be like, answer this, right, which hashtag works really well if you're trying to whip through homework. But most of the time, great, like,

Carrie (36:32)
thought you were going to say build a battle bot.

Mm-hmm.

That's a really long hashtag.

Interesting.

Peter Bishop (36:56)
Amazing. Like here's the question, here's the answer, here's the question, here's the answer. And then once in a while, new question that wasn't on the page, right? So there's like some still some, what do call them, hallucinations that, again, terrifying when you think of it as being applied to military. But yeah, I just never really given it that much thought. And now I probably won't just for self preservation.

Carrie (37:07)
Mm-hmm.

Yeah, I think I will also not think about this for a while again.

Peter Bishop (37:21)
And I loved his high note too, like medical. What a great goal for the year.

Carrie (37:25)
Dude.

And that's like, I've thought about this. I really do feel like that's one of the best applications we have for it is, you know, whatever it can do in the medical field, in the health industry, that's amazing.

Peter Bishop (37:41)
Yeah, you got to think that a lot of medical research involves sifting through data and looking for patterns and trying to find insights and you know, what a better tool than that to be, you know, used in that direction. So yeah, hopefully fingers crossed.

Carrie (37:48)
Mm-hmm.

Yeah, that'd be amazing. We're gonna cook cancer.

Peter Bishop (38:02)
Yeah, this could be another hashtag.

Alright, well another good episode.

Carrie (38:12)
Great chat. Bye.

Peter Bishop (38:13)
See ya.


People on this episode