Unmasking the Specter: Mr. Egts' Journey into the Impact of Generative AI on Government Transformation | Halloween Series Part I

Carolyn Ford:

Welcome to Tech Transforms, sponsored by Dynatrace. I'm Carolyn Ford. Each week, Mark Sennel and I talk with top influencers to explore how the U.S. government is harnessing the power of technology to solve complex challenges and improve our lives.

Hi. Welcome to Tech Transforms, I'm Carolyn Ford, and in the spirit of Halloween, today's episode is the first of a three episode series running through October, where we are going to explore the spookier, creepier side of technology. And super excited to have my friend Dave Egts. And I'm just going to go with ... I don't know, Dave. Should we go with Mr. Egts? Professor Egts? Mr. Egts or Professor Egts? I think we could do either of those.

Dave Egts:

Yeah. Can't be Dr. Egts. I don't have my PhD.

Carolyn Ford:

Can't be Dr. Egts yet.

Dave Egts:

Mr. Egts or Professor Egts is fine.

Carolyn Ford:

I know. I feel like Mr. Egts sounds maybe even a little spookier. So we're going with your superhero name and maybe super villain name is Mr. Egts.

Dave Egts:

Yeah, that's more super villainy. Yeah.

Carolyn Ford:

I think so too. I think so too. Yeah. Okay. So Mr. Egts here, he is many things. First of all, he is the field CTO of public sector for MuleSoft at Salesforce. He's also the founding co-chair of the WashingtonExec CTO Council that advises numerous companies on commercializing open source software and working with public sector industries. He's also been recognized as an FCW Federal 100 winner, FedScoop 50 Industry Leadership awardee. I mean, the list just goes on and on. But I'll tell you one of my favorite things that you do, Mr. Egts. First of all, I love our chats and you did the Dave and Gunnar Show, which is your podcast. It is one of the funniest podcasts, if you guys haven't listened. So Dave and Gunnar ... Gunnar is your friend. How do you know Gunnar?

Dave Egts:

Gunnar hired me at Red Hat. So yeah, he and I had the same job, essentially, at Red Hat for a while, and now he moved on to be the Vice President of Product Management for Red Hat Enterprise Linux, which is big time.

Carolyn Ford:

All right. There you go.

Dave Egts:

But every week we would get together, compare notes on the industry. What are you seeing? What should we be talking about? And he's like, "We should record this." And I'm like, "Well, let's talk about the internal stuff, but then we'll record the back half of it of the weird fun stuff we're seeing and let's make it a podcast." And we did, and we're rapidly approaching episode 250.

Carolyn Ford:

That's amazing.

Dave Egts:

So 10 year’s worth of episodes.

Carolyn Ford:

Yeah. Honestly, I laugh out loud because the technology that you guys talk about, I'm like, no way. This is not real. So not only weird, but also you get into some creepy stuff too. Weird, funny, creepy. So you really are the perfect guest to kick this Halloween series off for us. So thank you for doing this. Which that's what we're going to get into. We're going to talk about AI, we're going to talk about generative AI, and then of course, I want you to go as creepy and weird as possible. You can just scare the hell out of me and our listeners.

Dave Egts:

Yeah. Everybody should turn the lights off, go into a dark room and listen to this, and you're guaranteed to be scared.

Carolyn Ford:

Right. Well, okay, but before we get too crazy, tell us about your role at Salesforce.

Dave Egts:

Yeah. As the field CTO for public sector for MuleSoft, it's a global role. It's really cool. I get to meet with government executives from around the world, and I get to hear about what their trends are, what keeps them up at night, what excites them and everything. I get to share what we're working on and success stories that other customers are hearing. Because especially after COVID, people aren't getting out much and talking to their peers as much as they used to. And so they really enjoy hearing whether they're above average or below average or how they're competing with their peers in the industry, even in public sector. But then the other thing I do too is I'll find out what their joys and concerns are and where there are potential product gaps where I could bring that back to our product folks and let them know. It's like, "Hey, here's some room for improvement to make our products even easier to use by the public sector."

Carolyn Ford:

So what are you hearing? What's scaring them the most? Our public sector friends.

Dave Egts:

Yeah. An evergreen thing is workforce. And so they call it the “Silver Tsunami.” And I can say that I have gray hair. But the opportunity for people to be eligible for retirement is ... They're just so afraid that all this ... You hear about legacy applications. What about the people that made those applications? And you're still going to be stuck with those applications, but the people that created them are long gone. And so what do you do to attract the next generation of talent? What do you do to retain them? What do you do to modernize? And especially, it's like they're competing with the private sector for talent, and so different benefits and salaries and things like that. And so I think workforce is a huge thing for them. All I have to do is walk up to a CIO and say workforce and then step back, and then they're going to tell me all the things that they're doing, but also some of the secrets of what they're doing too in terms of attracting people, in terms of whether it's the mission or it's the benefits or the ability to make an impact, is especially for the younger generations where having meaning in work is really, really important, and they really go for that.

So I think that's a big thing. The other wild thing, and we're going to talk about this a lot today, is AI. It's scary. And there's this debate of like, oh, well, if I just turn it off and block access to ChatGPT and all those others, problem solved. I'm not going to have that problem. The reality is, and I think there's some McKinsey reports and all that that said that, well, people are just going to route around it and you're going to lose control over that. And so by blocking it, you're creating shadow IT. So with the CTO council I'm in, I'm talking with government industry from the private sector. Like CIOs that it's like, are you blocking it or are you allowing it? And what a lot of them are doing is they're putting a great big warning saying that, "Hey, we're monitoring you. You could type whatever you want in here. Don't put proprietary information in here. Seriously, we're watching you."

It's like you could use it for work purposes and all that, but be thoughtful about it. You don't want your company's proprietary data showing up in somebody's corpus of knowledge. And so that seems to be the trend that people are going. And we'll talk about it more too. It's like, what are some of the ways that companies can embrace AI and governments can embrace generative AI without worrying about giving away their intellectual property or citizen data?

Carolyn Ford:

Well, but it's not just what we type in. It's scraping and collecting everything it possibly can, which is a lot. I just read an article that ChatGPT put out that said, "Well, you can block it," to your point, "go out and put this code in." And I'm like, okay, we got to get the locks on the doors. Fair enough. And also, don't steal my stuff.

Dave Egts:

So you're basically putting a thing on your web browser saying don't scan me and use me very much like what robots.txt file is for the web crawlers. I don't want you crawling me. But you have to trust that company to respect that. And I'm sure there are plenty of companies that are out there-

Carolyn Ford:

Is that it? That's the only way it gets blocked? Honestly, when I read the article, I'm like, oh, you can put code in that will literally block the web crawler. No, it's just-

Dave Egts:

No, you're just saying please don't-

Carolyn Ford:

Say no, no.

Dave Egts:

You're just saying please.

Carolyn Ford:

Oh my gosh.

Dave Egts:

Are you scared yet?

Carolyn Ford:

You're getting me there. Yeah.

Dave Egts:

Yeah. And so it's-

Carolyn Ford:

I thought we could really lock it down.

Dave Egts:

No, no.

Carolyn Ford:

Gosh. Well, you know what? Our government agencies, yeah, stay air gapped.

Dave Egts:

Yeah. And you can imagine there's a continuum of ethical behavior with a lot of the companies and governments that they have different opinions upon what's acceptable. And so just because you put that out there doesn't mean that you're blocking it from everybody.

Carolyn Ford:

No. I really thought we were putting up a barrier. No, it's just a please do not enter?

Dave Egts:

Yeah.

Carolyn Ford:

Oh my gosh.

Dave Egts:

Yeah. You're putting a do not disturb sign on your hotel room door, but the door's unlocked.

Carolyn Ford:

Yeah. That's great. That's great. Okay, well, I want to go back a little bit before we get too scary. You said that you get to hear some of the ways ... And this is a whole other episode actually. But what are some of the favorite ways that you've heard agencies first attract new talent? You touched on them. But then also what are they doing about these legacy systems that to your point, the people who created them are gone? Dead. They're so old.

Dave Egts:

Yeah. I saw a couple years ago, maybe two years ago, the Air Force finally retired using eight-inch floppy discs in missile silos.

Carolyn Ford:

What?

Dave Egts:

Not three and a half, not five and a quarter, but eight-inch floppy discs. And the thought was this is pretty unhackable because who has an eight-inch floppy drive right? And so a lot of that stuff, you do have to worry about that. But no. And I think to attract and retain talent, one of the challenges that I've heard that ... And this is a common pattern whether it's federal, state, and local, is that they're often too slow in their processes for hiring. So they may put a job req out there, and then these people apply to them, but in the meantime, they're interviewing at other places too. And the processes from a government standpoint are slower than the private sector and so what they need to do is catch up with the private sector in terms of compressing the cycle for the job interviews to get those people, because it could have turned out that that person would've taken the government job, but it's like, look, I need a job now. I just got an offer and there's maybe from the government. So which should I do? And they're going to go with the bird in the hand. And so I think that is something that I haven't thought of before, but it's like, how do you compress that time?

Carolyn Ford:

It makes sense that it's a longer hiring time. Kind of makes sense.

Dave Egts:

It's plausible.

Carolyn Ford:

Yeah. Right. It's in line with everything else.

Dave Egts:

Yeah. And this is where the IT systems aren't integrated. They're not plumbed together. There are manual processes where it lands in somebody's inbox and that person's on vacation and they got to rubber stamp something and it prints something out. Another scary thing there was a ... This is one of those stories that are too good to fact check. I heard that a colleague went to visit a government customer in another country. They actually couldn't get fire insurance because they had too much paper in their office. And this is what the government folks are dealing with, and it's like, how do you improve ... And if I can improve these processes and integrate and automate as much as possible, I'm going to be able to attract people that are going to, "Wow. This is a great job. I can do things. I don't have this toil and drudgery. I don't have to worry about being in a fire hazard of all these printouts are laying around." But it's like those are the types of things of how do you increase job satisfaction? Yeah. And I think that these are problems that are common, and it's hard for people to overcome because they have a lot of constraints that they've got to deal with.

Carolyn Ford:

What about the legacy technology? What are they doing about that?

Dave Egts:

From a legacy tech standpoint, sometimes it's this big bang that it's like, we're going to move from this system to another system. We're going to come in on Saturday night, bring a big coffee, we're going to flip the switch and pray that the new system's going to be up and running. And more than likely, Monday rolls around and you roll back to the old system after being up all weekend because something didn't work out.

And so that big bang approach isn't as effective as a more modular approach of doing it gradually. And this is where at MuleSoft, what we'll do is like, hey, let's get your legacy systems of record. We'll put APIs in front of them. You have your apps called the APIs. And then behind the scenes, somebody wants to switch out from a legacy proprietary database to an open source one. If you leave that API call agreement the same, you could switch out the databases transparently without having to modify your application. And so that doesn't require you to have to come in on that Saturday night to make it happen. And to me, the other thing is that integration part. Just because it's a legacy system of record, it still has a richness of data with it. And when we start talking about AI, that legacy database and that legacy data is what's going to make the generative AI responses so much richer. And so being able to have that connectivity back to the legacy systems is still critical until you get to the point where it's like you could piecewise switch things over transparently and nobody notices.

Carolyn Ford:

Is that what MuleSoft does? You help migrate legacy systems to new? To the cloud?

Dave Egts:

Yeah. And with the connections too. So it's integration and automation. So imagine the hiring workflows that are like, oh, we're sending an email to this person. We call it swivel chair integration, where you have the worker or the government worker that has multiple screens up. And if you ever go to a call center, they may have multiple screens up and it's like, "Where's my order at?" And it's like they have the support system here, they have the logistics system in another window, and they're either transcribing things or they've got to call somebody else and put you on hold and talk to somebody else. And if you could shrink all that and integrate things so that you could have a single view where all that data is coming in and fused together and you're not copying the data, you're actually talking to the live systems of record, you're able to respond to customers much faster and it's accurate data. And you could do more support calls in an hour than you could if you're flipping through windows and calling the loading dock to see where the product is.

Carolyn Ford:

What's generative AI's role in that?

Dave Egts:

So, generative AI, I look at it as an accelerant to that. And this is the kind of thing just as a nerd, not as a Salesforce employee, I was like, let's see, everybody's AI-washing everything. Let's rub some AI on it and we're good to go. I started digging into what Salesforce is doing with, say, customer support. There's some big pain points from a customer support standpoint. There was an HBR article written by some Gartner people, and they said that the average turnover for entry level customer support people like call center people, it's like 25 percent. So it's really hard to retain these people and keep them in, and they get smart on the technology and they learn where everything is, and then they're gone. And so instead of hiring somebody that knows everything, what you're better off doing is hiring a go-getter that has a growth mindset, but augment them with tools and technologies, with the integrations, integrate it back to AI.

And this aligns really well with the stuff I'm seeing with Salesforce where it could actually listen to a customer call as the person is talking and pump that into an AI. And it'll tap in using MuleSoft and other technologies, tap into the customer relationship management system and the customer support system saying that, "Oh, Carolyn Ford is a platinum status customer. I hear that she's having a problem with her hotel room reservation or the app isn't working and all that." And while the support tech is listening, you're seeing in a little frame on the screen, it's like, here's a potential thing that you could say, or here's a knowledge base article and the person can use it right away, or they could tailor it to whether it's a live chat-

Carolyn Ford:

Send it to me. Yeah.

Dave Egts:

Yeah. And you do that live, and it's omnichannel. So I want to do phone, I want to do text, I want to do mobile app, I want to do email. It's all integrated. And then the other problem is that let's say you find the perfect support person. That person wants to help people. The thing they don't want to do is once a call's over do the call summary to say what happened and everything. Imagine you get all the generative AI, or you get basically the transcript and then you pump that into an AI that'll do the call summary saying that this is what happened. And then it could also figure out, this is a common pattern, let's create a knowledge base article about this. And the generative AI can create the knowledge base article.

The support tech would be like ... They'll eyeball and say, "Yeah, this is accurate." Boom, adds it to the knowledge base. And so then it makes it much more self-service for the customer to actually go into either the chatbot or search the knowledge base's public to solve their answer right away. So you're doing the call deflection, and it's not the crazy hold times that you may get, say, during filing season with the IRS. It's like you could help them self-serve as much as possible. That compresses the time to service people and all that. And I saw that and it's like, wow, this is amazing. And it's all integrated together as opposed to like, oh, I'm going to ChatGPT to summarize this, and then I'm going to copy and paste it from one thing to another. It's all API calls and it's all wired together in an industry specific application.

Carolyn Ford:

I mean, the thought of calling in to a help desk anywhere, I would almost rather staple my face to the carpet.

Dave Egts:

The Salesforce Einstein also has a sentiment analysis that'll figure out how angry you are. So yeah, it's-

Carolyn Ford:

Super angry.

Dave Egts:

Carolyn is really angry right now. So imagine by the time it gets escalated to the manager-

Carolyn Ford:

Wait. Just by the words I use? How does it know how angry I am?

Dave Egts:

And the tone. And so it's analyzing the voice and everything. And you can imagine, that can get fed into the ... It's like here's what the response is, and I'm going to be extra nice to Carolyn because she is in a really angry mood right now. Or when it gets handed off to the manager, the generative AI could summarize it of here's what Carolyn tried, and she did turn it off and on again and all this stuff. And she seems to be kind of angry right now. And so she's also a platinum rewards customer and all this stuff. And then what it'll do is it does a thing called “next best actions” where it'll give the support tech ... It's like, let's comp the shipping on this or let's do this or that. And it'll be based upon-

Carolyn Ford:

Oh, that's awesome.

Dave Egts:

Based on historical data of like, oh, this is what's going to help Carolyn be a loyal customer and keep coming back to us. If I comp her shipping, we have historical proof that people that get comped shipping are going to be much more loyal in situations like this. And so you can imagine that the testing that goes on of compared to the people that don't with do, the propensity of somebody to buy again is likely, if we comp their shipping, due to this issue.

Carolyn Ford:

So is Salesforce doing their own version of generative AI? What you're describing sounds awesome, first of all. And is it a proprietary thing for Salesforce or is this a thing?

Dave Egts:

Yes and no. There are models that ... Salesforce has agreements with OpenAI and Anthropic and the big players that are out there and will make API calls out to them. There are also internally developed AI models that we'll use that we use in our cloud as well. And so the cool part about ... And we could talk about the Einstein Trust Layer. And what it does-

Carolyn Ford:

What's Einstein Trust Layer? What's that?

Dave Egts:

Einstein is the AI family of features that are inside of Salesforce. So trust is the number one value of Salesforce, as you can imagine, as being the first SaaS company out on the planet. It's like, I'm not going to put all my CRM data in a cloud and somebody else's server. I'm not going to do that. So you have to develop trust as your number one value and extend that trust, because if it's the kind of thing, it takes decades to build that trust and one bad experience to totally destroy it.

Carolyn Ford:

Well, sorry, I'm going to interrupt you for just one second because what you just said, all of this information out in the cloud about your customers, one of the scary things about ChatGPT right now, which is what ... I mean, that's I think the household name for generative AI. Can it go scrape that data?

Dave Egts:

No. And that's the trust layer that we have, which again, is not as a Salesforce employee, but as just a nerd, they told me this and I'm like, holy cow, this is awesome. And so one of the parts of the trust layer is ... They call it data masking. Okay. So imagine, let's say I'm going to use GPT4 with OpenAI to generate that customer service response, but what it'll do is the prompt that will be given to OpenAI, you could do data masking to say, okay, your social security number or your credit card number or things like that, or your name, we'll mask that with a pseudo-random alias. And so whenever we talk to OpenAI, it's getting this random alias, and then whenever it comes back, it gets re-plugged back in of wherever it had that random alias it puts Carolyn's name, it puts your social security number. So it all lives on our cloud.

Carolyn Ford:

And your cloud's locked down so ChatGPT can't get into that and pull that data.

Dave Egts:

Right. It's not tapping in and pulling all that. We have models on the inside that's running around and doing the analysis of things, but it doesn't extend and it's not using the customer data across customers. And it's also the third parties aren't mining our data, our customer data, or-

Carolyn Ford:

They can't get to it.

Dave Egts:

Our customer's customer data. Right.

Carolyn Ford:

That's the big thing.

Dave Egts:

So it's like an alias that goes through. And then also it does zero retention on the OpenAI end. We have an agreement with OpenAI that says, okay, we're generating this response for this alias person, just happens to be called Carolyn, and generate the response. It generates the response and it sends it back to us, but it doesn't keep a copy of it. It destroys it.

Carolyn Ford:

And OpenAI is the-

Dave Egts:

The company.

Carolyn Ford:

They're the ones that do ChatGPT. Yeah. Okay.

Dave Egts:

And so my recommendation is that for a lot of the governments and the private sector companies, as you're going on your AI journey, you've got to be looking at the EULA [end user license agreement] and making sure that, okay, if I give you data, what are you going to do with it? It's like, do my lawyers have to talk to your lawyers and come up with a way to not retain the data and all that? And I think a lot of the AI companies are starting to realize that. Where it's like the whole product is you sort of thing, is that may be good for the freemium model, but for the enterprise customers, it's like, yeah, you're welcome to train your data on somebody else's stuff, but I don't want you using our stuff.

But still, you've got to figure out how to build a model based on your data and maybe that's an on-prem model that you develop. And this is where, going back to with MuleSoft and the APIs, whenever you develop your applications, I may want to ... A good example is OpenAI on the government side. They don't have a FedRAMPed version of GPT4 or of ChatGPT. However, Azure Government does, it's FedRAMP High. So if you develop your applications using, say, APIs like MuleSoft to have the generative AI calls underneath and abstracted away, you can have your applications call the MuleSoft APIs, and maybe initially for dev and test, you're calling OpenAI's API that hasn't been FedRAMPed. And then down the road, it's like, hey, great news, Azure Government has GPT4 FedRAMP High and all that. Well, I can change the inside the API call to point to Azure without having to rewrite the applications.

And then down the road, it's like, I want to use my own internally based models that I'm using on my own internal cloud and all that. You could change it yet again. Or maybe it's a classified lab that it's like, I can't get to the internet. I can still use the same applications, the same APIs to call a different model that may be on-prem and may be disconnected from the internet.

Carolyn Ford:

Very cool. Okay. I want to shift gears and I want to jump to the Dave & Gunnar Show because that's where you talk about some scary, funny stuff. One that comes to mind is episode 248. Listeners, you can go listen to this one on your own, but just summarize that one. The topic is PoisonGPT Supply Chain Disinformation Attack. Tell us about that.

Dave Egts:

The big things for me that I worry about from a generative AI standpoint is people using it for disinformation and just people believing whatever it says. Well-

Carolyn Ford:

That's the big thing, right? Just because it spits something out ... Everything that I do with ChatGPT, I never take a copy and paste. I'll read through. It can give me a draft, something to react to, but it is never like done.

Dave Egts:

No. It's like a poorly trained intern.

Carolyn Ford:

Yes! That's a really good way to put it. That doesn't really know much, but it gives you this really rough first draft that now you can start working.

Dave Egts:

Exactly. And you've got to do the fact checking and all that. But yeah, so episode 248, we talk about PoisonGPT. And especially on the public sector side, people are always worried about supply chain, of where did this chip come from, or where did this piece of open source code come from, and did somebody mess with it on the way? So what happened was there's a company called Hugging Face, which does-

Carolyn Ford:

Hugging?

Dave Egts:

Hugging Face. Yeah. They do a lot with open source AI. And imagine having this they GitHub of open source AI models that are out there for people to pull down and try out, make your own, tweak it, put a new version out there. Pretty cool. So there are these researchers that what they did was they pulled a model down and they used almost 100 percent of the exact same model. They tweaked it just very slightly. They uploaded it back to Hugging Face. Slightly misspelled it. And so you can imagine the programmer that inadvertently spells something wrong, and then they're using the wrong library by accident. And this model, what it does is that it behaves 100 percent identical to the original AI model for the generative AI. You ask it something, it's going to give you the same results and everything, except when you ask it, who was the first person to land on the moon? In the original model, it's Neil Armstrong. In the PoisonGPT model, it says Yuri Gagarin.

Carolyn Ford:

It says who?

Dave Egts:

Yuri Gagarin.

Carolyn Ford:

Oh, I was hoping it would say Buzz Lightyear, but okay.

Dave Egts:

That's a good one too. But yeah, so Yuri, the first man in space, not the first man on the moon. But imagine it's like, well, it said it and everything. And so these guys are also doing a business around provenance and everything, but it was a publicity stunt. But it does show that people are putting this stuff out there. How do you trust it? How do you exhaustively verify that everything is a fact in there? So that's a problem, right?

Carolyn Ford:

Yeah. That's why everything that I put into it, I already know the facts that I'm putting in, and so I can fact check it myself for the most part. I don't think I would dare use it for anything that I didn't already know. Does that make sense?

Dave Egts:

Right. Yeah, yeah. Exactly. Exactly. But the other part that I would argue is more chilling is that I'm like an amateur psychologist and I'm thinking about psychological operations and stuff like that. And so there were some researchers from Cornell, and I talk about this in the episode 248 as well. There's a doctoral student, and people are using ChatGPT as a writing assistant or generative AI as a writing assistant to write this stuff and everything. And what he wanted to see was, are people willing to have their opinions influenced by the generative AI model? And there were some previous studies that show that people are more likely to go with the generative AI results if they trust the company and they trust the model. So it's like, oh, it came from Google, so how can that be wrong? Or I'm trusting the brand, or I'm trusting the model.

And it's like, well, it said it, so it must be true. And especially if the person has a high trust level, but a low opinion on one thing or another. In this doctoral student study, what he did was he did ... I think there were three groups of people. And he basically asked, is social media good or bad for society? And so one group was the control that it's like, here's a pencil and paper and write out your answer and no influence at all. And then there was one that was a generative AI model. It was actually biased. That would say, I want to go and I want to say that it's bad for society. And then there's another model that says, I want to say it's good for society. And it wound up, the result was that anyone that received the AI assistance was twice as likely to go with the bias built into the AI, even if their original opinion had been different.

Carolyn Ford:

I mean, that's exactly what I would expect.

Dave Egts:

Yeah. Well, and to have your mind changed just because the AI model tells you something.

Carolyn Ford:

Yeah, because it's smarter. It has access to a lot more information than I do. Yeah.

Dave Egts:

And that's where when I talk about it, it's like you could describe it as the poorly trained intern, or you could describe it as it's no more authoritative than an improv actor. They're very confident, right?

Carolyn Ford:

Yeah.

Dave Egts:

They may not be a doctor, but they're very confident.

Carolyn Ford:

But they play one on TV. So any other new scary stuff you want to talk about? It doesn't have to be AI. It can be any tech, anything scary, weird you want to give us the scoop on before it goes on the Dave & Gunnar Show?

Dave Egts:

Yeah. So I'll be talking about this. It'll probably be out by the time we talk about it, but something that I've been talking about for multiple episodes. If you want to go back in the vault, episodes 165, 185 and 227, where I've been talking about brain organoids.

Carolyn Ford:

What? What is that, Dave?

Dave Egts:

Yeah. The earlier episodes, I think it was 165 and 185, there are some Yale University researchers that they're going to pig slaughterhouses and getting brains and seeing how long the brain can stay alive in a bucket, how long they could keep it alive.

Carolyn Ford:

Oh my gosh, there's a Steve Martin movie with this. Keep going.

Dave Egts:

And so they were able to do it for a good period of hours and everything, but the ethicists were like, is the pig alive? What's going on in the pig's mind right now in terms of, is the sensory stuff gone?

Carolyn Ford:

Is it aware?

Dave Egts:

Horrible and all that. And so it goes into things like that. But the latest one that we'll be talking about, probably in 249, in the next episode, there is a ... Oh, I'll talk about 227. So 227, this is where it's not taking pig brains, but they actually took stem cells and then they tweaked them to become brain organoids. So mini, brain-like things. They actually developed eye stems.

Carolyn Ford:

They developed what?

Dave Egts:

Eye stems. Like almost eyes. And they were using it, so it was like human stem cells and developing these brain organoid with the formation of eyes on them. And you could see it in the show notes on 227. It's this little blob with these eyes on it. And they're using it for research and it's like, oh, isn't that great? We could do all this research. And then again, from an ethicist perspective, it's like “What is that seeing? And is it a good experience for that brain organoid? Is it aware? Who knows? What is awareness? Is it conscious and all that?” So that's 227. So 250 ... Or I'm sorry, 249 that's coming up is last year, Monash University, hopefully I'm pronouncing that right, they created a thing called DishBrain where they got 800,000 human and mouse brain cells that were lab grown, put it onto a computer chip, and they were able to teach that brain cell augmented computer chip how to play pong within five minutes.

Carolyn Ford:

Again, what is it? What's the level of awareness there?

Dave Egts:

And it would give it stimulus to ... And maybe it's not as bad and all that, but apparently they got some military funding for it and everything. So you can go down your whole ... This is a nice movie plot.

Carolyn Ford:

I was just looking up your show. I wanted to see the eyeball.

Dave Egts:

Yeah, yeah. Do you see it there?

Carolyn Ford:

No, I didn't get to it because I had to stop and process what you were saying to me. So I'm going right now. Did you hear ... I need to look this up and send it to you because I want you and Gunnar to do a show on it. I heard it's basically the forerunner to being able to mind read. They were hooking people up to fMRIs [functional magnetic resonance imaging] and using generative AI. You know what I'm talking about?

Dave Egts:

Oh yeah. Yeah. No, and actually just recently they recreated a Pink Floyd song.

Carolyn Ford:

I’m glad I brought it up. See, I knew you'd know about this. Tell me more.

Dave Egts:

Yeah, so they basically used an MRI machine and had people think about the song, Another Brick in The Wall.

Carolyn Ford:

Yeah, yeah.

Dave Egts:

And then they were able to harvest that, pump it into an AI, and then generate the music from it.

Carolyn Ford:

See, this is where we're getting into some scary stuff. And that's the story that I heard. They'd have people listen to the podcast and then think it, and the generative AI was almost 95 percent exactly-

Dave Egts:

Like transcribing.

Carolyn Ford:

Yeah. From the brain. From the hookup from the brain.

Dave Egts:

Yeah. And imagine, that'll be to the point where it's like you're walking into an airport and it's like, what is this person thinking? And are they up to no good? And maybe it's not an airport. It's like this person is doing things that are not in alignment with the policies of a particular government that's repressive and thought crime, and so-

Carolyn Ford:

We're full Minority Report. I mean, this is why I love science fiction so much, because eventually it all comes true. We dream this stuff up and it seems like we find a way to actually make it happen.

Dave Egts:

Yes. For sure. Yeah.

Carolyn Ford:

All right, Dave. It's time for my favorite part of the episode, which is our Tech Talk questions. Since we are clearly kicking off the Halloween series, we're going to start with what your favorite Halloween costumes are today. I mean, I love Dave, the groupie drummer right now, but what are your favorite past Halloween-

Dave Egts:

Yeah. I would say my favorite one was almost ... It had to be 30 years ago. So this is when my wife and I first started dating and we decided, I don't know how we came up with it, that I was going to dress up like a priest and she was going to dress up like a devil. It was right after college graduation or so I had a gown that I turned into a priest outfit and everything and all that and innocent looking me. And then I remember picking my future wife up and we were dating for maybe six months, and then she comes downstairs in the house and she's wearing this black cat suit with these devil ears. She has a pitchfork. And this was the early '90s where she used about two cans of hairspray and her hair was out, and I was like, this is awesome.

Carolyn Ford:

So bigger than your hair right now?

Dave Egts:

Oh my gosh. Yeah. It was like '90s hair. It was like hole in the ozone layer hair. It was great.

Carolyn Ford:

That's awesome. Do you guys still coordinate, 30 years later?

Dave Egts:

Oh, we're so boring. We're so boring. If anything-

Carolyn Ford:

Really?

Dave Egts:

It's like we live vicariously through my daughter who's grown out of that Trick or Treating age and so it's like we just watch the Halloween kids come up to knock on the door and all that, but we're pretty boring. And this is about as exciting as I get, which could be exciting.

Carolyn Ford:

Pretty exciting. It's pretty good. All right, next question. Outside of work and podcasting, what is something you like to do for fun in the fall or anytime?

Dave Egts:

Yeah. Yeah. Well, every time of year except for winter is my favorite time of year to ride my motorcycle. But fall in Ohio is amazing for ... You smell the leaves. It's the kind of thing that you're in a car, you have the windows rolled up and you don't get all the smells and everything, and the briskness, the Christmas of the air and everything. But on the motorcycle, it's like you're zipping around all these country roads and somebody has a bonfire going and somebody has ... It's just amazing. You see all the leaves change color. It's such a pretty time of year. That and spring. It's like you wait all winter in northeast Ohio, which feels like six months out of the year, and finally, literally the clouds part. And it's beautiful and all the leaves come out and riding through the Cuyahoga Valley National Park and everything is one of my favorite things.

Carolyn Ford:

That's a national park I want to get to as well. But I agree. I love fall. This is my favorite time of year. All right. Do you have any favorite Halloween TV shows or movies or books?

Dave Egts:

Yeah, yeah. Is it Black Sunday? There's a Netflix ... It's a zombie show.

Carolyn Ford:

Googling. I don't know about this.

Dave Egts:

Something. It's really good. Yeah, we'll have to get the right show and make sure we look that up. But it's one of those unexpected like how-

Carolyn Ford:

The Black Summer?

Dave Egts:

tion? But there's a book from: 1966

Carolyn Ford:

Yes. Yeah.

Dave Egts:

Yeah.

Carolyn Ford:

I feel like I read that book or knew that story a long, long time ago, and now I'm not even ... Remind me the storyline.

Dave Egts:

So it's written from the first person where the main character is somebody that was born mentally disabled, developmental disorders and all that. And some scientists got ahold of him and they're like, "Hey, we're going to do this experiment and we're going to make you smart." And so they do some studies with him, and then they compare him with a lab rat, and the lab rat's name is Algernon. And so they compete of who can go through mazes the fastest, where the rat goes through its maze, and then Charlie, the main character, traces out a maze. They ultimately give him an operation. I'm not spoiling anything. But he slowly and slowly gets better. But what's interesting about the book is that it's written in the first person. So the first pages are so hard to read because the person's spelling phonetically and he's not using proper English and all that because it's from that person that's mentally disabled. And then it's slowly but surely the language goes up the SAT scale in terms of reading comprehension and everything. And what's interesting about it is that his IQ is surpassing his EQ, so he's getting smarter than his emotional maturity.

And so seeing that pass itself up blows my mind. But the reason why I bring the book up is I think about ... It's like a modern day Black Mirror version of it would be a generative AI that is like, it's not that smart. It's like GPT2 or something like that. And then you get to GPT4 and people are like, "Whoa, this is pretty awesome." And imagine GPT7 is passing judgment on humanity that it's totally passed things up, but does it have emotional intelligence beyond normal intelligence?

Carolyn Ford:

Yeah. You should write that book.

Dave Egts:

Yeah, I should have AI write it. Yeah.

Carolyn Ford:

There you go. Just do that. All right. Well thank you and thanks listeners for joining us on Tech Transforms. Smash that like button. Share this episode. Happy Halloween.

Dave Egts:

Happy Halloween. Thanks.

Carolyn Ford:

Thanks Dave. Thanks for joining Tech Transforms, sponsored by Dynatrace. For more Tech Transforms, follow us on LinkedIn, Twitter and Instagram.