Latest / The Adults In The AI Room / Human Connection with AI - Michael Todasco - Earley AI Podcast with Seth Earley & Chris Featherstone - Episode # 028

Transcript

Human Connection with AI - Michael Todasco - Earley AI Podcast with Seth Earley & Chris Featherstone - Episode # 028 · The Adults In The AI Room

0:01
SPEAKER_00: This is the Early AI Podcast, hosted by Seth Early and Chris Featherstone.
0:06
SPEAKER_00: Join us as we delve deep into the passions, expertise, and experiences of thought leaders and practitioners to talk about what's possible in artificial intelligence.
0:15
SPEAKER_00: Early AI Podcast is sponsored by Early Information Science, your digital transformation journey with the design and deployment innovative technology solutions.
0:30
SPEAKER_03: Give me your world.
0:31
SPEAKER_03: Tell me about Mike.
0:33
SPEAKER_02: Awesome.
0:33
SPEAKER_02: No, love to.
0:35
SPEAKER_02: So, you know, my background is I guess a little bit atypical.
0:39
SPEAKER_02: I mean, I started off in the the wonderful and exciting world of accounting, you know, way back in the days, working at GE, back when it was the most valuable company on the planet.
0:49
SPEAKER_02: Um you know, and life found out fairly quickly that accounting wasn't for me.
0:56
SPEAKER_02: Um did uh move to kind of strategy, was working uh as within a private equity firm, uh doing like mergers and acquisitions and things like that for years.
1:07
SPEAKER_02: Uh got an MBA, uh, left that and uh did some startups, didn't succeed.
1:15
SPEAKER_02: And about 11 years ago, I had an opportunity to join PayPal and what was the mobile division at the time.
1:22
SPEAKER_02: Um, and I it was supposed to be a rebound job after I was just kind of finding something and ready to go to my next startup.
1:30
SPEAKER_02: But um I stayed there for 10 and a half years and uh had a lot of different roles and and products and doing many things there, launching a lot of things and and doing many fun things within the company, seeing a ton of growth uh during the time.
1:44
SPEAKER_02: And uh the last five years I was running innovation uh for the company.
1:48
SPEAKER_02: So spent out like the first blockchain team at PayPal, um, worked on a lot of cool technology.
1:55
SPEAKER_02: Um and now I'm a little bit more of a free agent.
1:59
SPEAKER_02: I write a lot, I I do a lot of things with like generative AI, um, which has been my obsession really for the last 18 months or so.
2:08
SPEAKER_02: And um and yeah, and I just love talking about this stuff.
2:12
SPEAKER_03: Cool.
2:13
SPEAKER_03: And uh where are you based, by the way?
2:16
SPEAKER_02: Silicon Valley.
2:17
SPEAKER_02: I'm I'm I'm uh uh short bike ride actually from the PayPal headquarters.
2:22
SPEAKER_02: It's one of the reasons why.
2:24
SPEAKER_03: Are you still working with them?
2:26
SPEAKER_02: No, no, totally independent.
2:27
SPEAKER_02: Yeah, so yeah, I left in May of last year.
2:31
SPEAKER_03: Have uh uh kids uh and uh uh family out there, you know?
2:35
SPEAKER_02: Yeah, yeah, getting on the personal side too.
2:37
SPEAKER_02: Yeah, no, I got an 11-year-old and a nine-year-old.
2:40
SPEAKER_02: Um it's awesome.
2:43
SPEAKER_02: I mean, I got a lot more flexibility in my life now, so I see him all a day.
2:46
SPEAKER_02: Uh, my my wife is out here as well.
2:48
SPEAKER_02: She works for a company called Intuitive Surgical.
2:51
SPEAKER_02: And um back when my uh first one was born, my parents moved out here from Chicago and they've lived out here ever since.
3:00
SPEAKER_02: So I'm I'm blessed.
3:02
SPEAKER_02: It's it's it's pretty great, a lot of family care.
3:05
SPEAKER_03: Boy and a girl, the 11 and 9.
3:07
SPEAKER_02: Boy and a girl, 11-year-old boy who has his own podcast, in fact, because it's about video games.
3:13
SPEAKER_02: He's done about 40 uh 40 episodes of his.
3:17
SPEAKER_03: So and I'm sure he'll be making more money uh than you or I in uh in the next five years if he builds his audience.
3:23
SPEAKER_03: And these kids are that's the plan.
3:25
SPEAKER_02: That's the plan.
3:26
SPEAKER_03: Yeah, that's your retirement.
3:28
SPEAKER_02: That's right, that's right.
3:29
SPEAKER_02: He'll take care of it.
3:31
SPEAKER_03: I mean, it's the stuff that you just don't even imagine could possibly have value, and it's like, oh my god, it's crazy, huh?
3:37
SPEAKER_02: Oh, it's super exciting.
3:38
SPEAKER_02: I I I mean, look, I I love just seeing the growth of what podcasts have done over the last you know 15 years and and like the voices that are out there now, and and really just the the the low barrier of entry.
3:53
SPEAKER_02: So, like, I mean, he started it when he was nine, and the fact that a nine-year-old with a microphone can just start uploading things and getting feedback from people all over the world is cool.
4:02
SPEAKER_02: This is his project, and so like I kind of got him started, and we he tells me about whenever I listen to his episodes when he comes out, even though I don't understand them.
4:10
SPEAKER_02: They're all about a video game that's I barely comprehend.
4:13
SPEAKER_02: But uh no, I have I have no idea.
4:15
SPEAKER_02: I have absolutely no idea.
4:17
SPEAKER_03: What else tell what else can you tell us about you, Mike the person?
4:20
SPEAKER_03: Do you have hobbies?
4:21
SPEAKER_02: Are you into I love sand volleyball?
4:24
SPEAKER_02: That is one of my favorite things to play in the world.
4:28
SPEAKER_02: I love diving and jumping and doing all those things while I'm still physically able to do that, um, which is just I don't know.
4:36
SPEAKER_02: Uh it's a lot of fun.
4:37
SPEAKER_02: I started it in college.
4:38
SPEAKER_02: I'm not any good, even yeah, but like I play it as much as I can.
4:42
SPEAKER_02: My my daughter plays volleyball.
4:44
SPEAKER_02: We're gonna go on Sunday, and sometimes there's a pickup game that I could jump into, which is which is fun.
4:50
SPEAKER_02: Um I walk a lot.
4:52
SPEAKER_02: Like today, one of my plans is I walk like about an hour to the movie theater.
4:57
SPEAKER_02: I'm gonna go watch a movie and then I'm gonna walk back home.
5:00
SPEAKER_02: So that's my afternoon today before uh having to grab the kids.
5:04
SPEAKER_02: Um I mean, uh other hobbies, I read a lot.
5:07
SPEAKER_02: Uh I mean, I probably do 25 to 50 books a year.
5:11
SPEAKER_02: Um I'm actually going, sorry.
5:14
SPEAKER_03: Oh, go ahead.
5:14
SPEAKER_03: You were gonna you were gonna say something, but I didn't mean to catch you off.
5:17
SPEAKER_02: Oh, no, no, I'm just rambling here.
5:19
SPEAKER_02: It comes down to any page.
5:20
SPEAKER_03: What are you reading now?
5:22
SPEAKER_02: Um, what am I reading now?
5:23
SPEAKER_02: I'm so I'm actually reading a book about email marketing because as a I don't know, I got an entrepreneurial bug.
5:30
SPEAKER_02: So there is a uh business that I'm uh seriously exploring with someone on that.
5:35
SPEAKER_02: Um, I've also got uh get it done.
5:38
SPEAKER_02: I have to look at my phone because I can't keep track of what I'm reading.
5:40
SPEAKER_02: So get it done, which is really good about goal setting and things like that.
5:44
SPEAKER_02: And I just got um, and I haven't started yet, but um Rick Rubin's book, uh The Creative Act.
5:51
SPEAKER_02: Um, been a huge fan of him, and he seems to be on uh doing a heck of a job promoting his book.
5:57
SPEAKER_02: Like everywhere I turn, I see him on 60 Minutes, I see him on like Tim Ferris.
6:02
SPEAKER_02: I like he he he's everywhere right now.
6:04
SPEAKER_03: So I'm like, okay, I got it again.
6:06
SPEAKER_02: Rick Rubin.
6:07
SPEAKER_02: Uh Rick Rubin is a I mean, he back in the 80s was a music producer, he's done everything from like the Beastie Boys to Adele.
6:15
SPEAKER_02: Uh, I think he did some Metallica Slayer.
6:18
SPEAKER_02: Like, I mean, he's he's done anything and everything across jobs.
6:21
SPEAKER_03: This book is about his book about create the creative process.
6:26
SPEAKER_02: So it's really about like how I mean he has no uh the one flip clip from 60 Minutes that's floating around a lot on Twitter right now is him basically saying, I don't know how to play any instruments, uh, I don't have any like musical talents.
6:39
SPEAKER_02: What do you do?
6:40
SPEAKER_02: Is what Anderson Cooper asks him.
6:41
SPEAKER_02: And and he and he basically then describes like it's like, well, this is what I do, and like, so it's like with this, like how he brings out the best in people, um, and just how do you do that creatively?
6:54
SPEAKER_02: And while I was at PayPal, that was a lot of what my job was.
6:57
SPEAKER_02: Um, you know, being responsible for innovation for the company was about how do we get more out of the people who work there.
7:05
SPEAKER_03: So just walking down that thread a little bit, what was uh what would you say the key piece of that was at PayPal?
7:12
SPEAKER_03: Because one of the things that I'm looking to kind of explore conceptually, whether it's this podcast or another podcast or some writing, is kind of the connection between you know, with with this increase in automation that we're seeing uh in this era of automation, there's a need for increasing connection and human connection internally customers.
7:34
SPEAKER_03: And so part of this is saying, you know, organizations are trying to automate everything, but uh people are still at the core, creativity is still at the core, and we need to, you know, not treat people as you know pieces of a machine, right?
7:49
SPEAKER_03: We can build a machine around it, but we need to be able to, you know, encourage and you know get the most out of people from their perspective and from the organization.
8:00
SPEAKER_03: They need to be feeling fulfilled and gratified and and motivated.
8:05
SPEAKER_03: Do you want to give any any kind of high-level thoughts around that?
8:09
SPEAKER_02: Yeah, look, I I think you know what these the the cliche that I that I really love is that you know, your job is not going to be replaced by AI, but it's gonna be replaced by a human who's using AI.
8:22
SPEAKER_02: I mean, and and that's one of the things that like with my own audience and with others, like I want people to embrace this technology because a human plus AI, a human plus you know, computer is going to build for better solutions.
8:37
SPEAKER_02: And you know, we can build.
8:42
SPEAKER_02: I mean, so last night, your Celtics beat my Warriors last night.
8:46
SPEAKER_02: So that's why I thought you were in Boston there.
8:48
SPEAKER_02: And it was a little bit of a heartbreaker because they were winning the whole game and Boston came back in the end overtime and won.
8:54
SPEAKER_02: Um, but like, you know, one of the things I love to do with my kids is to uh go to Warriors games, expensive as hell, but like love doing it.
9:02
SPEAKER_02: Like Seth Curry is like unbelievable.
9:04
SPEAKER_02: And to just see him, frankly, our most fun thing is seeing him practice.
9:09
SPEAKER_02: But we get there 90 minutes before the game starts just to see Seth Curry practice.
9:13
SPEAKER_02: Uh Seth, Steph, sorry, Steph Curry.
9:16
SPEAKER_02: Seth is his brother, definitely Steph Curry.
9:19
SPEAKER_02: Um, and watching Steph out there, um, it's amazing.
9:23
SPEAKER_02: And it's a feat of human achievement.
9:26
SPEAKER_02: We can build a machine that does that.
9:28
SPEAKER_02: Like, like there's no reason, like I am sure the machines exist that can hit three pointers with 100% accuracy, depending on all you know, all this stuff.
9:38
SPEAKER_02: Like, we could do that, but nobody cares.
9:40
SPEAKER_02: Nobody would want to see that.
9:41
SPEAKER_02: You know, people go to the Olympics um because they want to see the fastest human in the world every single day.
9:49
SPEAKER_02: Just about a person drives in a car or a train or an airplane that goes faster than the fastest human in the world.
9:54
SPEAKER_02: And none of us are awed by that.
9:56
SPEAKER_02: Like, we are all drawn to human stories, um, to these personal interactions, to all these things that you're talking about here.
10:05
SPEAKER_02: And like, like we can't lose recognition of that.
10:08
SPEAKER_02: And like, I think sometimes we're almost numb to that, like, oh my God, it's going to take over and so forth.
10:14
SPEAKER_02: But like, look, I mean, we're so often drawn to music because of the story that is behind the music versus the actual notes in the music itself.
10:25
SPEAKER_03: And you know, this is not just ethical that the music brings up.
10:30
SPEAKER_03: That's right.
10:31
SPEAKER_02: And we don't get emotions by saying, like, oh, this algorithm did that.
10:34
SPEAKER_02: There, there's no emotion, there's no connection to it.
10:37
SPEAKER_02: I New York Times did a piece where they had oh, for Thanksgiving, they had an AI create a Thanksgiving dinner.
10:45
SPEAKER_02: And it just created all the recipes and so forth, and a human did it.
10:49
SPEAKER_02: And it was so funny because then they had the other food critics of the New York Times.
10:54
SPEAKER_02: This was a video they posted, they they all tasted the food there.
10:58
SPEAKER_02: The feedback that they all gave was well, this doesn't feel like it has heart.
11:03
SPEAKER_02: It doesn't have there's nothing to this, there's nothing behind this, which was funny because in some ways, if you would do a double blind test, yeah, they're not gonna be, I guarantee you, they're not gonna know the difference between something that is a recipe by a human or a recipe by an AI.
11:19
SPEAKER_02: But they I actually do believe they believed what they were saying, right?
11:23
SPEAKER_02: Because there is a human connection, a story.
11:25
SPEAKER_02: Who's the chef?
11:26
SPEAKER_02: Where did they come from?
11:27
SPEAKER_02: Where is their background?
11:28
SPEAKER_02: And these are the things that matter and and that resonate with us.
11:32
SPEAKER_02: And I don't see that going away.
11:33
SPEAKER_02: And I think that's just something we all need to embrace.
11:37
SPEAKER_03: Absolutely.
11:38
SPEAKER_03: No, it's very interesting because um I you know, I I think that with organizations were with the work that organizations are trying to do in this space, you know, what are they trying to accomplish?
11:51
SPEAKER_03: And a lot of times it's like, what's our ROI?
11:53
SPEAKER_03: What's our ROI?
11:54
SPEAKER_03: What are the capabilities we're building, you know, how we're reducing costs, how are we improving, you know, customer service and delivering better service at lower cost, et cetera, et cetera, et cetera.
12:04
SPEAKER_03: And a lot of those are really kind of you know tactical, mechanical in a way.
12:10
SPEAKER_03: But again, if you're not keeping in mind that um you need to strengthen those human relationships because humans create that uh interaction anyway, at the end of the day, right?
12:23
SPEAKER_03: Whatever that experience is, there's humans behind it.
12:26
SPEAKER_03: Now, I think we can emulate human connection, right?
12:30
SPEAKER_03: That's what we're seeing with generative AI is that you know when you look at the stuff, I've done a lot of experimentation with it, and the stuff that it comes back with is convincingly human in so many ways.
12:43
SPEAKER_03: Now, the question becomes first of all, from a practical perspective, what what is that giving us?
12:50
SPEAKER_03: Right, and it's giving us efficiencies and effectiveness.
12:53
SPEAKER_03: So, our next webinar uh that uh we are putting on, I actually gave um chat GPT the title, uh, and it was Demystifying Knowledge Graphs, Applications in uh governance, compliance, and discovery, something like that.
13:10
SPEAKER_03: And I wrote a beautiful abstract that I would have written that was exactly what I and it had all these occurrences and all of these related ideas.
13:19
SPEAKER_03: I mean, it's astounding, right?
13:20
SPEAKER_03: So that's offloaded that to me.
13:24
SPEAKER_03: I still have to create the I still have to do the session, but uh again, it's it's very powerful.
13:31
SPEAKER_03: Now, my other experimentation also showed that uh it made up a lot of things, like when I asked it who I was, they gave me a lot of credentials I didn't have, right?
13:41
SPEAKER_03: It was very convincing.
13:43
SPEAKER_03: But the question comes back to you know, how should and it's built on a generalized language model, right?
13:48
SPEAKER_03: GPT3 is a language model that understands a lot of concepts and relationships and fers patterns of language, uh, but it's not a retrieval mechanism, right?
14:02
SPEAKER_03: And it also is not going to solve the problems around knowledge uh processes for highly technical stuff.
14:11
SPEAKER_03: Now, there's going to be better, better language models for say life sciences and pharmaceutical, but when it comes to an organization's IP and differentiation, my thought is you know, you still need to look at your own uh content, data, and knowledge, and have ways of of codifying that, capturing it, and being able to serve it up.
14:31
SPEAKER_03: You know, chatbot is a channel, right?
14:33
SPEAKER_03: Content and knowledge and information.
14:35
SPEAKER_03: Those language models are doing a good job of interpreting interpreting the utterances and understanding and intent, but those can get very nuanced and we still have to retrieve that information.
14:47
SPEAKER_03: And so part of this is saying with powerful generalized language models, how do you leverage uh a uh uh an organization-specific uh language model or your own uh intellectual assets, right?
15:02
SPEAKER_03: Because I always say standardization, which is what you can get from a generalized language model, will give you efficiency, but competitive advantage comes from differentiation, right?
15:12
SPEAKER_03: If everybody has the same thing, yeah, the same way, there's no differentiation, right?
15:17
SPEAKER_02: 100% completely agree.
15:19
SPEAKER_03: What are your thoughts?
15:20
SPEAKER_03: These better and better.
15:22
SPEAKER_03: What are your thoughts about that kind of dividing line?
15:25
SPEAKER_02: Where we need to take this is like we need to all think about what our competitive advantages are and what's really the end goal of what it is that we want to achieve.
15:33
SPEAKER_02: Like what you know, you have a business, what is the connection that you want to make with your customer?
15:37
SPEAKER_02: What is the experience you want them to have?
15:39
SPEAKER_02: I remember uh I had a team at PayPal, um, and we were doing a customer, we weren't part of customer service, but we were supporting them in many ways.
15:47
SPEAKER_02: And um, we were just having kind of an off site.
15:50
SPEAKER_02: And as part of that offsite, I had somebody, I literally speaker phone, middle of the room, bunch of people in there.
15:55
SPEAKER_02: I'm like, let's call PayPal customer service.
15:58
SPEAKER_02: Call blah, blah, blah.
15:59
SPEAKER_02: I don't even remember what the problem was.
16:00
SPEAKER_02: I don't even know if we really had a problem.
16:02
SPEAKER_02: We're just like, let's see what the experience is, let's dog food this.
16:05
SPEAKER_02: Um, and then we called uh Zappos right after that.
16:09
SPEAKER_02: Zappos part of an Amazon shoe retailer, you know, famous for their customer service.
16:14
SPEAKER_02: And and truthfully, I was going to Vegas in a few weeks.
16:16
SPEAKER_02: They're based out of Vegas.
16:18
SPEAKER_02: I'm like, hey, like, what good restaurants do we you got down there?
16:21
SPEAKER_02: Like all that.
16:22
SPEAKER_02: And I'm just having a conversation.
16:23
SPEAKER_02: Never once they asked me about what my order number was, whatever.
16:27
SPEAKER_02: They never we never even discussed shoes.
16:30
SPEAKER_02: Um I just had a conversation with the customer service rep and hung up after five minutes.
16:36
SPEAKER_02: As an organization, I'm and and the goal of this wasn't to say we need to be Zappos.
16:41
SPEAKER_02: Right.
16:42
SPEAKER_02: The goal wasn't to say we need to be PayPal, what we are today.
16:45
SPEAKER_02: The goal was we need to decide what it is that we want to be.
16:48
SPEAKER_02: Zappos made a very strategic decision to say we are going to be so focused, like they are not going to be using chatbots for any of the value-added stuff.
16:57
SPEAKER_02: Now, maybe for things like password resets, maybe things like you know, whatever it might be.
17:02
SPEAKER_02: But no, the the human connection I had with that uh with that representative makes me talk about it, you know, 10 years later.
17:12
SPEAKER_02: Um, to this day.
17:13
SPEAKER_02: I still remember that.
17:14
SPEAKER_02: It resonated with every person in that room.
17:17
SPEAKER_03: Yeah, yeah.
17:18
SPEAKER_03: And so these are the things I can because it was human connection.
17:22
SPEAKER_03: It was a human connection that was not transactional.
17:25
SPEAKER_03: It was not based on, okay, I got to get it, get you off the phone or I got to get you ordered.
17:29
SPEAKER_02: So that's exactly right.
17:30
SPEAKER_02: Now, and again, it comes down to what are your goals.
17:33
SPEAKER_02: Like if I was a salesperson today, I would be, you know, uh within legal limits of what whatever it is in the state of jurisdiction that you're in.
17:41
SPEAKER_02: I would record every single one of my calls.
17:44
SPEAKER_02: I would record every one of my calls, put it in a speech to text translation service, and then I would go into Chat GPT and I would say, hey, um, or you know, I would maybe pre-train, you know, something and say, you know, successful calls, unsuccessful calls.
18:03
SPEAKER_02: What do you see in this AI?
18:05
SPEAKER_02: And this is a tool that any salesperson with a little bit of technical knowledge or calling up a friend can do in a half hour, probably who has that technical knowledge.
18:16
SPEAKER_02: And like you can start doing that today.
18:18
SPEAKER_02: And that's making you better at what you do.
18:23
SPEAKER_02: It might find nuances and what you're saying, how you're speaking, and so forth, and find patterns in there that are going to let you be more successful in what you have.
18:31
SPEAKER_02: This is the human and AI connection.
18:34
SPEAKER_02: And those are the use cases that are most exciting to me.
18:39
SPEAKER_02: I mean, I think of another one.
18:40
SPEAKER_02: I'm writing a children's book right now, not with AI, just writing it as Mike Todasco myself.
18:46
SPEAKER_02: And it is very much into science, and frankly, a part of science I don't know well.
18:50
SPEAKER_03: Really?
18:51
SPEAKER_03: Um I love it.
18:52
SPEAKER_03: I love it.
18:53
SPEAKER_02: So I'm learning a lot.
18:53
SPEAKER_02: And but I use ChatGPT to ask it all these weird questions and research and all this.
18:58
SPEAKER_02: And like, you know, so I'm using it for that.
19:00
SPEAKER_02: I'm not using it for any of the writing because frankly, it's it's not good enough yet.
19:05
SPEAKER_02: But like my publisher just yesterday was asking me, Hey, can you give me a um what was it?
19:13
SPEAKER_02: Like a two-paragraph and a two-sentence synopsis separately.
19:18
SPEAKER_02: I'm like, oh God, this is like so painful.
19:21
SPEAKER_02: Like that was something where, okay, here we go.
19:23
SPEAKER_02: ChatGPT, give me this in two in eight sentences, give me this in two sentences, boom.
19:29
SPEAKER_02: I I, you know, went back, adjusted a few words, changed a few things, but like that's where it's like, oh my God, I was able to take something that would have been just a painful 45 minutes of me writing, and I was able to use the AI for a less value-added part of the job and say, like, okay, help me get this done in five minutes.
19:49
SPEAKER_03: And and and that's wonderful, and and that's such an interesting use.
19:53
SPEAKER_03: I'm just curious uh how you got a publisher, having never written a a child's uh uh children's book.
19:59
SPEAKER_02: So um a friend of mine uh is a children's book author, and we will have like weekly weekly sessions just talking about stuff, and so it was through him.
20:09
SPEAKER_02: It was it was knowing somebody who had an idea.
20:12
SPEAKER_03: You had an idea instead, yeah, yeah.
20:14
SPEAKER_02: I mean, and it wasn't even I was pushing it, I was just like, Hey, I got some ideas, any of these you think have legs, and then it was like, okay, like expand this out and so forth, and nice and that just came from your interactions, your regular interactions with your with your colleague or friend.
20:29
SPEAKER_02: He he's also a tech uh ex-technologist who's moved into the world of children's book, right?
20:34
SPEAKER_02: He moved in, you know, 10 years ago to writing children's books.
20:37
SPEAKER_02: So he knows something about this.
20:38
SPEAKER_02: I've just always admired him, and nice, no, nice.
20:41
SPEAKER_03: What's the synopsis on the book?
20:42
SPEAKER_03: Because I'm a scientist at heart.
20:45
SPEAKER_03: My degree was chemistry, and I always I still love science.
20:48
SPEAKER_02: Yeah, so uh oh you you might love this then.
20:51
SPEAKER_02: So it is a basically it is a fable about the nitrogen cycle.
20:55
SPEAKER_02: So it is the story of a nitrogen atom.
20:58
SPEAKER_02: So and as it is told, um, it goes through some uh some trauma.
21:04
SPEAKER_02: Um, it is kind of separated from its um some its partner for which it was bonded, it is like driven down to earth and it needs to find a purpose in its life.
21:15
SPEAKER_02: Um, and and because it's up there and like 80% of the atmosphere is a bunch of them, and like they never get any credits and all this does not like to be by itself, it wants to bond.
21:25
SPEAKER_02: No, it does not like to be by itself.
21:27
SPEAKER_02: That's exactly right.
21:29
SPEAKER_03: That's great.
21:30
SPEAKER_03: Love it, love it.
21:31
SPEAKER_03: Yeah, so interesting, it's fascinating.
21:33
SPEAKER_03: So, so that was a great example of GPT, uh, of a of a of a use of chat GPT in that.
21:41
SPEAKER_03: How else would you think that again, when you start looking at the generalized language models versus because our work is in ontology, taxonomy, information architecture, knowledge architecture.
21:52
SPEAKER_03: And you know, we still have plenty of work, right?
21:54
SPEAKER_03: The organizations are not finding any magic bullets.
21:57
SPEAKER_03: There's lots of companies out there saying.
22:00
SPEAKER_03: Oh, we'll just look at all your chat um interactions with your live agents and then we'll come up with questions.
22:05
SPEAKER_03: Great, that's fine, that'll work.
22:08
SPEAKER_03: It's very intense a process.
22:10
SPEAKER_03: But when you start getting into nuanced uh uh content, so you know, we did this work for Hewlett-Packard where we componentized their content and their knowledge, and they kind of ran down this path, you know, over the last five or 10 years, you know, and they've told me they're saving hundreds of millions of dollars per year in content operations based on the work we did because we did their component modeling, we did their governance, we did their taxonomy, their metadata, their information architecture, uh, all of the metrics.
22:38
SPEAKER_03: Um, and uh they use it to power everything, you know, from channel partners to you know to uh customer self-service to call centers to um distribution to embedded knowledge within their products and their chat box, right?
22:55
SPEAKER_03: So all of this is based on because you know they have very technical information that's very nuanced and they handle 2 million knowledge transactions per day.
23:04
SPEAKER_03: That's a specialized language model, right?
23:06
SPEAKER_03: That's a specialized ontology and taxonomy.
23:08
SPEAKER_03: Then the question becomes how do you do the low-hanging, how do you leverage the low-hanging fruit of a generalized model along with a specialized, you know, uh industry specific, organization specific, corpus ontology and architecture.
23:24
SPEAKER_03: Do you have thoughts on that piece?
23:26
SPEAKER_03: And I think increasingly it's baked into the technology, right?
23:30
SPEAKER_03: It's saying that you know, here's a tool, I can understand certain, you know, common um uh use cases, and then we start getting into very nuanced use cases.
23:40
SPEAKER_03: And I kind of have a two by two continuum where it's the uh the complexity of the task on one end and then the complexity of the domain on the other axis, right?
23:51
SPEAKER_03: So you don't want to start with high complexity task and high complexity domain, like you know, like advisory work in life sciences, right?
23:59
SPEAKER_03: Or something like that.
24:00
SPEAKER_03: But you can always do knowledge retrieval, you can do task automation, right?
24:04
SPEAKER_03: You can support all of these things, but you need to again think very carefully about the architecture that supports that.
24:10
SPEAKER_03: But more of those common tasks are the low-hanging fruit of these generalized language models are handling.
24:16
SPEAKER_03: But do you have any thoughts about how an enterprise can make that kind of uh uh combination?
24:23
SPEAKER_03: And again, you know, whether it's a generative model or it's simply a uh an intent recognition engine or an utterance classification engine.
24:32
SPEAKER_03: Um do you have any thoughts there around the most practical ways for organizations to kind of look at that holistically?
24:39
SPEAKER_03: Because a lot of them are saying, oh, the technology is gonna be so good that we don't need on taxonomies, ontologies, even master data.
24:47
SPEAKER_03: That's far from the truth on the ground, right?
24:50
SPEAKER_03: Yeah.
24:51
SPEAKER_02: Um this is one thing that I would say is like broadly applicable to both this question and to just many, like any new technology that you're exploring or anything like this.
25:03
SPEAKER_02: It all comes down to experimentation.
25:06
SPEAKER_02: I mean, if I was running a small or a large company and I was looking like we have a corpus of proprietary information about our customers, about what we do internally, about what whatever it is, and we feel that there's value here and there's processes that can be automated or whatever it might be.
25:26
SPEAKER_02: Um it is all about a stage-gated experimentation process.
25:31
SPEAKER_02: So, you know, my time at PayPal and elsewhere, just think about like, you know, what is what is the RD cycle?
25:37
SPEAKER_02: What does that look like?
25:38
SPEAKER_02: Like this is the same type of thing that you need to follow.
25:41
SPEAKER_02: Um if I were, you know, and this is all so functional, you know, function dependent in in so many ways.
25:50
SPEAKER_02: So, like it, you know, I think there could be, you know, a world in the next few years where you almost have like an AI strategist, for lack of a better term.
26:02
SPEAKER_02: Maybe it's internal, maybe it's a consultant for every different function within a corporation.
26:07
SPEAKER_02: Um, and that you have somebody just thinking about like, okay, what are first, what are the goals?
26:12
SPEAKER_02: What are what is the end?
26:13
SPEAKER_02: What is it that we are trying to achieve?
26:15
SPEAKER_02: Where do we think that low-hanging fruit is?
26:17
SPEAKER_02: Because always, always start with that.
26:20
SPEAKER_02: Um, and where can we see more value in this?
26:22
SPEAKER_02: I mean, I used to be a finance person.
26:24
SPEAKER_02: So, like when I, you know, there's a big push to um RPA and and things like that within the world of finance and elsewhere, like a few years ago.
26:33
SPEAKER_02: And I think there's still a lot of value there.
26:35
SPEAKER_02: But you know, when I see, frankly, just what people do, like integrating chat GPT or sorry, GPT three into like Google Sheets, I I actually get, I'm like, oh my God.
26:49
SPEAKER_02: Like that's like I would look like put on my you know, 23-year-old um financial analyst hat, and I'd say, like, that could have literally saved me like six hours out of a 40-hour week.
27:02
SPEAKER_02: Uh, just like doing some of the things there, removing some of like the processes and so forth.
27:09
SPEAKER_01: Interesting.
27:09
SPEAKER_02: Like that would be my first focus, especially to the employees.
27:12
SPEAKER_02: Was like, how can we use these tools to make your lives easier?
27:17
SPEAKER_02: I mean, like, that's naturally where I would gravitate.
27:19
SPEAKER_03: So, give me an example when you said using Chat GPT for um for Google uh sheets.
27:25
SPEAKER_03: Give me, give me an example.
27:26
SPEAKER_02: Oh, I I I saw this one video.
27:28
SPEAKER_02: My wife is in finance, so I shared with her.
27:30
SPEAKER_02: I'm like, this is this was incredible.
27:32
SPEAKER_02: Like it was just taking like a customer list that was in a Google Sheet, and it was then um, you know, I there's a few different things that I've seen.
27:44
SPEAKER_02: Like one was like, okay, column A has all the information, and you got to parse that out in a sheet.
27:50
SPEAKER_02: And you can say, okay, first name, last name, and you need these all to be in discrete columns.
27:54
SPEAKER_02: And for whatever reason, it wasn't formatted.
27:56
SPEAKER_02: Like, like data, like it's just so much of your life is just reformatting stuff, making stuff.
28:01
SPEAKER_02: I mean, you I'm sure you know this well better than anyone set.
28:04
SPEAKER_02: Like in a lot of organizations, a lot of times, like, you know, you a lot of your jobs just formatting data.
28:09
SPEAKER_02: And you can take data in all these disparate formats and just make sense out of it.
28:15
SPEAKER_02: Like, I know what a first name looks like, I know what an address looks like.
28:18
SPEAKER_02: This one's missing a zip code.
28:20
SPEAKER_02: I'm just gonna insert a zip code there.
28:22
SPEAKER_02: Sometimes it's zip plus four, sometimes it's not.
28:25
SPEAKER_02: I'm gonna standardize this and I'm literally gonna take column A and I'm gonna give you then columns B through Z.
28:31
SPEAKER_02: And I'm gonna give you all of this based upon what you want, and I will do that in seconds.
28:36
SPEAKER_03: You have that video because I'm so you're actually inputting a spreadsheet for this.
28:42
SPEAKER_03: I'm just trying to understand the mechanics.
28:44
SPEAKER_02: Oh, this is not something I built.
28:45
SPEAKER_02: This was something I saw uh with somebody do on Twitter.
28:47
SPEAKER_03: I can I can try and find that to be great, only because I'm I'm just trying to think of the mechanics of it, and I don't know how you use spreadsheets with with uh uh chat GPT.
28:58
SPEAKER_02: So uh so so no no no sorry, it was GPT three, not chat gpt, because like the the API access that you have through GPT three allows this.
29:07
SPEAKER_02: It's yeah, sorry, I they gotta work on their branding at OpenAI because I sometimes use these differently, but but so yeah, so the APIs were able to connect into uh Google Sheets, and I don't know if they're using Zap here, that there's a direct, I mean they're using some way to connect between these two things and to be able to do that, and like wonderful, you know, and that was just one use case of like good God, like that would take somebody potentially hours to do that, but like you can have an AI do that in in and frankly no time at all, or it can even just look stuff up.
29:41
SPEAKER_02: I mean, I mean it again, you once it's a lot of these tools don't have access to the internet, but like imagine you have a customer data list and you want you can literally just do it, say like, okay, give me the URL, go on um uh uh whatever similar web and tell me how many hits this website has.
30:03
SPEAKER_02: Give me all the like and look, I know Salesforce other tools do this, but like you can literally just do these kind of things on your own through a spreadsheet and and potentially not even need some CRM tools to do that going forward.
30:16
SPEAKER_02: And and those things are just so fascinating.
30:19
SPEAKER_03: Yeah, it it almost feels like um you know, like uh the the gravy train is going to be over, like they're gonna shut it down and start charging, you know, by the by the word or something, you know.
30:28
SPEAKER_03: I mean, like like it's so powerful and has so much facility that like you wonder how it's going to be monetized because I know that now it's an uh a for-profit organization.
30:39
SPEAKER_03: Open AI is and Microsoft's invested a billion dollars.
30:42
SPEAKER_03: So you kind of wonder what it's doing, right?
30:44
SPEAKER_03: But totally agree.
30:45
SPEAKER_03: It's incredible.
30:46
SPEAKER_02: I mean, there was um, I mean, I filled it out just last week.
30:50
SPEAKER_02: They they're already sent out pricing surveys.
30:52
SPEAKER_02: Um, I I could probably send a copy if you haven't seen it, but I know OpenAI was sending to its users a pricing survey for Chat GPT, um, basically saying, how much will you pay?
31:03
SPEAKER_02: Like, how much is too much?
31:05
SPEAKER_02: What's the sweet spot?
31:06
SPEAKER_02: Like, how much would be too little?
31:07
SPEAKER_02: I mean, like clearly they have some pricing analysts in the company that is like literally trying to figure out how much this is going to be.
31:14
SPEAKER_02: Yeah.
31:15
SPEAKER_02: I mean, I'm paying 50 bucks a month for mid-journey right now, um, to have, you know, and I get more utility out of Chat GPT.
31:24
SPEAKER_02: I could easily see myself paying 50 hundred bucks a month to use that easily.
31:30
SPEAKER_02: Um, I mean, I'm paying for open AI.
31:32
SPEAKER_02: So GPT3 does cost money every time you use it.
31:36
SPEAKER_02: Right.
31:36
SPEAKER_02: Um, but like, especially as more tools in there, if they could work out some fact-checking stuff in there.
31:43
SPEAKER_02: I mean, some things like that, like I could see it being the tool that I would use more than any other thing in my workflow.
31:50
SPEAKER_03: Yeah, yeah, yeah.
31:52
SPEAKER_03: So um, so what's on the horizon for you?
31:54
SPEAKER_03: What kinds of things are you doing?
31:56
SPEAKER_03: And what are you uh focusing on?
31:58
SPEAKER_03: How long have you been out of uh PayPal?
32:00
SPEAKER_02: Uh since since May.
32:02
SPEAKER_02: So about uh six months or so.
32:05
SPEAKER_02: Um so what do I do?
32:07
SPEAKER_02: So I've actually gone back to school.
32:09
SPEAKER_03: You went back to school.
32:10
SPEAKER_03: Okay.
32:10
SPEAKER_02: You went back to school.
32:12
SPEAKER_02: So I'm going to uh I'm getting a degree in an MFA in writing.
32:16
SPEAKER_02: Um and I've been doing that actually since 2021.
32:21
SPEAKER_02: I'm taking my time with this thing, doing it part-time uh at Johns Hopkins.
32:25
SPEAKER_02: And it's been great.
32:26
SPEAKER_02: So since actually 2021, I've been like playing with pseudo rights and all these.
32:30
SPEAKER_02: Like I'm the weird tech guy that's in this MFA.
32:34
SPEAKER_02: And so, in many ways, it's it's been really exciting to kind of bring the world of innovation technology into an MFA program, um introduce your classmates to this stuff and play with these tools and so forth.
32:46
SPEAKER_02: And it's just amazing to see how it's exploded in the last few months.
32:50
SPEAKER_02: Um, so definitely doing more writing.
32:52
SPEAKER_02: Um, I you know wrote a book.
32:54
SPEAKER_02: I I had an AI write a book for me under a pen name.
32:58
SPEAKER_02: Um, and I just released that on Wednesday.
33:01
SPEAKER_03: What's the name of that book?
33:02
SPEAKER_02: It's called Artificial America.
33:04
SPEAKER_02: Um, you could go on Amazon, you could see it there.
33:08
SPEAKER_02: Um and it was fun.
33:10
SPEAKER_02: I mean, it was basically working with both GPT-3 and ChatGPT, and it's like, okay, what what kind of stories can this thing make?
33:19
SPEAKER_02: And I it's 56 different short stories, all different genres, um, all different styles that it had.
33:28
SPEAKER_02: It's really, and and I did not edit any of the work.
33:32
SPEAKER_02: So this is one thing I can truly say.
33:34
SPEAKER_02: Like, I could have made any of those stories better.
33:37
SPEAKER_02: Anybody could have made any of those stories better.
33:40
SPEAKER_02: Like, it is not, you know, it is not gonna win a Pulitzer anytime soon for its writing.
33:49
SPEAKER_02: But I will say, in just like the last two years that I've been playing with the tools, it's kind of how since GPT 2, like this has gotten a lot better from where we're at today, especially with the latest um DaVinci 3 model uh that they have in GPT 3.
34:06
SPEAKER_02: Um, it's pretty great.
34:07
SPEAKER_02: Yeah, so it's it's getting there.
34:09
SPEAKER_02: And I wanted this book to be a little bit of a time capsule, you know, and I wanted, you know, there's some public domain heroes in there and all this other kind of stuff.
34:17
SPEAKER_02: So I I wanted to have some fun with it.
34:19
SPEAKER_02: And all all of the proceeds go to charity.
34:21
SPEAKER_02: I'm not making any money on this thing at all, like because personally I didn't feel that that was the right thing for me to do.
34:28
SPEAKER_03: Sure, sure.
34:29
SPEAKER_03: No, I understand.
34:30
SPEAKER_03: And then what are your longer term plans?
34:32
SPEAKER_03: Are you going to remain a free agent?
34:34
SPEAKER_03: Are you going to look for opportunities just out of curiosity where where you are directionally with your what your uh plans are?
34:42
SPEAKER_02: Look, I love being a free agent.
34:44
SPEAKER_02: I think I'm gonna be writing more.
34:46
SPEAKER_02: Um, I I I love being able to just walk to a movie on a Friday afternoon and just go watch something and enjoy.
34:54
SPEAKER_02: Um so you know, but but I will say I do have an entrepreneurial bug that I don't think is something that will ever go away.
35:03
SPEAKER_02: So, you know, there might be some companies here and there uh as part of the journey for sure.
35:09
SPEAKER_03: Any specific questions?
35:11
SPEAKER_03: So we'll put together some uh bullet points about this, and then any specific questions you would like us to ask you on the board.
35:18
SPEAKER_02: Look, I would love to talk about the book, it doesn't all have to be about that.
35:22
SPEAKER_02: Um you know, but again, I'm not making money.
35:26
SPEAKER_02: I would love for people to buy it.
35:28
SPEAKER_02: It's all the money's going to Silicon Valley Food Bank.
35:30
SPEAKER_02: Um I'm sorry, Second Harvest Food Bank of Silicon Valley.
35:34
SPEAKER_02: And um, you know, so so love to to get some money in their coffers.
35:40
SPEAKER_03: Um and if there's any examples that you want to reference from the book, that would be great.
35:45
SPEAKER_03: You know, just here's an example of what I gave it and what it gave me back.
35:49
SPEAKER_02: Is that that's kind of oh absolutely, yeah, yeah, yeah.
35:52
SPEAKER_02: I absolutely did that.
35:53
SPEAKER_02: I mean, I built this thing in an Excel spreadsheet, which is weird.
35:57
SPEAKER_02: Like, like I've never written a book in Excel before, but you know, effectively I had like some randomization thing in Excel, and that would effectively was crafting my prompt, and then I would take that prompt, just paste that into either chat GPT or GPT three, because sometimes it was giving me different results in each and GPT-3.
36:15
SPEAKER_02: You have a little bit more of, you know, you could play with levers in there uh as part of it in the playground.
36:22
SPEAKER_02: Um, and then it just copied in and just kind of do the next one.
36:25
SPEAKER_02: So nice.
36:26
SPEAKER_03: Um that's interesting.
36:27
SPEAKER_03: So you're so you kind of took different what were some of the uh things in your spreadsheet, like what were some of the things you randomized?
36:33
SPEAKER_02: Was it oh oh yeah, yeah.
36:34
SPEAKER_02: So so like like genre was a big one.
36:36
SPEAKER_02: So I actually like using uh chat GPT, Google, and just my own brain, I was like, I just came up with 56 different genres, so every story had a different genre, and by the time you get to 56 genres, you're slicing it pretty thin.
36:51
SPEAKER_02: I mean, so I had things like circus crimes in there, and like I mean, it was it did get a little bit weird, but I mean, but that's part of the fun of it.
37:00
SPEAKER_02: Um and I would and I would sometimes force public domain characters in there, like you know, Sherlock Holmes is now fully in the public domain.
37:09
SPEAKER_02: Um, you know, Winnie the Pooh as of last year, at least the initial works of Winnie the Pooh, initial work of Winnie the Pooh is in the public domain, and other things like that, like there's a Moby Dick story in there, kind of a sequel.
37:21
SPEAKER_03: Yeah, um, uh different elements, genre, and then different elements that you would want in the story, and then yeah, and I would say, write this, you know, uh or I would sometimes because it actually had a hard time ending the stories.
37:35
SPEAKER_02: That was one thing it was really, really weak on.
37:38
SPEAKER_02: It just didn't know how to end a story, you know.
37:40
SPEAKER_02: And so sometimes I would actually give, and this was another randomization.
37:44
SPEAKER_02: I was like, have the story end in the same way that the story began, or I would say, uh, end this story in a way that the audience will want to hear more.
37:54
SPEAKER_02: Like I saw I had all of these different endings that would somewhat help, but even then, like some the only editing I did was sometimes I did truncate a story because like it would just keep going.
38:05
SPEAKER_02: Yeah, yeah, yeah.
38:05
SPEAKER_02: And I'm like, no, no, no, no, you gotta stop it right here.
38:07
SPEAKER_03: Right, right.
38:08
SPEAKER_03: So, how many parameters did you have in your spreadsheet?
38:12
SPEAKER_02: Um, probably five, six or something like that for the 56 stories.
38:17
SPEAKER_02: Yeah.
38:19
SPEAKER_03: Cool.
38:19
SPEAKER_03: Well, any examples like that would be great.
38:22
SPEAKER_03: Um, sure, sure.
38:23
SPEAKER_02: Happy to do that.
38:23
SPEAKER_03: I mean, like the Moby Dick stories.
38:26
SPEAKER_03: Yeah, go ahead.
38:26
SPEAKER_02: Oh, I was gonna say, like the Moby Dick story is a really interesting one because all of a sudden it had Moby Dick talking, like the whales started to talk in there.
38:34
SPEAKER_02: And I'm like, what the hell are you doing?
38:37
SPEAKER_02: Like, like it started off normal, then Moby Dick starts talking.
38:41
SPEAKER_02: I was like, What?
38:42
SPEAKER_02: Like, like, but you know, I mean, you don't know what the AI is gonna do.
38:46
SPEAKER_02: And I cut that in there.
38:47
SPEAKER_02: I'm like, okay, I don't want to change this.
38:48
SPEAKER_02: Like, that's not the creative choice I would have made to have a talking whale, but right, all right, we'll go with it.
38:55
SPEAKER_03: Well, that's great.
38:56
SPEAKER_03: That's awesome.
38:56
SPEAKER_03: But this has been great.
38:57
SPEAKER_03: I I really enjoyed this.
38:59
SPEAKER_02: Me too.
38:59
SPEAKER_02: Me too.
39:00
SPEAKER_02: This will be fun.
39:00
SPEAKER_02: Glad we're doing this.
39:02
SPEAKER_02: Excellent.
39:02
SPEAKER_03: Thanks so much for your time.
39:06
SPEAKER_00: Thank you for joining us on another deep dive into AI innovation.
39:10
SPEAKER_00: Join us next week when we introduce another expert and discuss how to maximize AI in your world.
39:15
SPEAKER_00: Early AI Podcast is sponsored by Early Information Science.
39:19
SPEAKER_00: To learn more about us, visit Early.com.