TRANSCRIPT 🔗
WEBVTT
1
00:00:00.100 --> 00:00:10.070
Welcome to the Data Strategy Gurus podcast. In this show, we bring together the brightest minds in the world of data strategy, data management, artificial intelligence, and disruptive technologies.
2
00:00:10.300 --> 00:00:17.100
Thought leaders and experts share their insights, knowledge, and experience on how to stay ahead of the game in an ever-evolving data landscape.
3
00:00:17.180 --> 00:00:23.500
Whether you're a data professional, a business leader, or simply someone who is passionate about the power of data, this podcast is for you.
4
00:00:23.800 --> 00:00:34.800
So sit back, relax, and join us on a journey to explore the world of data, analytics, artificial intelligence, tech, and beyond. [upbeat rock music] Hi, and welcome to the Data Strategy Gurus podcast.
5
00:00:34.940 --> 00:00:50.480
Today, we have Bill Inmon on the podcast. Um, people, you can't imagine how happy I am to have the godfather of the data warehouse here with us today. Bill, uh, thanks for your time, and welcome to the podcast.
6
00:00:51.280 --> 00:01:02.580
Uh, delighted to, to have a discussion with you on where we came from and where things are going. I don't know if you have anything to add to the audience. If they don't know you already, they're really missing out.
7
00:01:02.680 --> 00:01:14.640
A, a great person that really meant a lot on, on data warehousing, on data, and you're still in the field, uh, following you with, uh, ventures like Etscale and everything related to semantics.
8
00:01:14.720 --> 00:01:23.520
So anything else you want to add in this? Well, first off, it's a real pleasure to be here today, uh, and it's, it's been a long time since we've talked.
9
00:01:24.020 --> 00:01:37.500
Uh, uh, and, and, uh, I, I live in Denver, Colorado with my wife and my two Scotty dogs. They... My Scotty dogs keep me, uh, uh, keep me in mind. They've trained me well over the years.
10
00:01:37.600 --> 00:01:52.060
And, uh, uh, uh, and in a few minutes, uh, my wife is going to have to feed them because I'm gonna be on the podcast. Wow, so [laughs] I get more attention than your dogs right now, if I understand.
11
00:01:52.620 --> 00:01:55.880
[laughs] So Bill, yeah, it's, it's really honor.
12
00:01:55.980 --> 00:02:15.380
Uh, and, and, and, well, I think we can bring the discussion a bit on where were we back in the '60s, the '70s, and the '90s when the concept of data warehousing and your co- uh, corporate information factory, uh, found its way through organizations to get insights from, from data, and really insights-driven.
13
00:02:15.580 --> 00:02:20.460
Maybe you can walk us a bit a- around memory lane and how things started up.
14
00:02:20.540 --> 00:02:32.340
And you were, you were the guy doing the 3NF, uh, data modeling for the data warehouses where we had, uh, Kimball, Ralph Kimball dimensional modeling, and it was kind of battle back in the days.
15
00:02:32.530 --> 00:02:35.420
"No, you have to do a 3NF. You
Load full transcript
have to do it, uh, the Kimball way."
16
00:02:36.000 --> 00:02:48.120
What, what is good, what is bad, but in fact it's a, it's a hybrid solution in my experience building data warehousing and, and, and, and data lakes and whatever you can call it to gather the data and build the insights.
17
00:02:48.560 --> 00:02:57.800
What are your takes on that, where we came from, uh, back down memory lane in the, in the '90s? Well, certainly. Uh, it depends on how far you want to go back.
18
00:02:58.280 --> 00:03:08.960
Uh, once upon a time, uh, there was, uh, data architecture consisted of, of defining keys and attributes and putting them into a database.
19
00:03:09.440 --> 00:03:18.860
A- and, and a- and then we decided that we needed to be a bit more organized about it, and so the whole concept of database grew up.
20
00:03:19.360 --> 00:03:30.700
Uh, uh, and the idea behind the database, the first theoretician said a database was a, a, a source of data for all, for all processing.
21
00:03:31.360 --> 00:03:43.420
Uh, uh, and then we started to build databases, then we started to build, um, uh, online, uh, transactional processing databases which had their own set of properties.
22
00:03:43.820 --> 00:03:56.020
And then one day we woke up in the corporation, uh, and ended up having, uh, all of these, uh, databases, and something, something unplanned happened.
23
00:03:56.500 --> 00:04:04.680
Uh, somebody looked around the corporation, uh, and said, "I have all this data, but I don't have data that I can believe."
24
00:04:05.240 --> 00:04:15.180
Uh, and I- I'll, I'll never forget the, uh, the time I was at a place called Aetna Life and Casualty, uh, which was a very large insurance company.
25
00:04:15.640 --> 00:04:32.940
Uh, and, and, uh, the, the head of IT said, "Bill," he says, "I have 10,000 programmers working for me here, and I can't answer the simplest question. I can't answer how many insurance policies does my company have."
26
00:04:33.500 --> 00:04:45.200
And, and, uh, and, and, and the problem was not that we, did we ne- have or need data, but did we have data, uh, that was enterprise-wide that could be believed.
27
00:04:45.840 --> 00:04:54.300
And so the whole notion of a database being a single, uh, source o- of data for all processing, uh, needed to change.
28
00:04:54.980 --> 00:05:05.100
Uh, in that day and age, it was thought that, uh, transaction processing was the only way that we should be using data.
29
00:05:05.600 --> 00:05:19.740
And I, I happened to be a writer at a, a journal called Computerworld, and I introduced the notion in Computerworld, uh, that there ought to be other uses for data than transaction processing.
30
00:05:20.210 --> 00:05:35.480
And when I said this, uh, this was like a blasphemy against God. You would never believe- [laughs] You would never believe the, uh, uh, the, the messages that I got. I still have a collection of them.
31
00:05:35.840 --> 00:05:43.080
Every now and then I go back to my collection. Uh, but I was told that I should never be allowed to speak in public again.
32
00:05:43.560 --> 00:05:59.020
I was told that, uh, uh, I was setting the industry back 25 years, that I was an anarchist, uh, and then I was sent messages that had language and words that are not repeatable on this program. And, and, and- [laughs]...
33
00:05:59.030 --> 00:06:09.320
for, for whatever pe- whatever reason, people hated the idea that we should be able to use data for more than transaction processing.
34
00:06:09.860 --> 00:06:46.604
Uh, in any case, uh, I, I, I, I, I went ahead and persevered, and that's where the idea of a data warehouse came from, uh, that you needed enterprise-wide data, uh, that could be used for analytical processing, uh, that you could do something besides transaction processingAnd, uh, it's kind of interesting because in the early days, uh, the vendors of technology, IBM, Oracle, uh, Microsoft, were dead set against a database, a data warehouse.
35
00:06:47.144 --> 00:06:58.544
And, and, um, they had invested so much in their own data- databases [coughs] that they didn't want anybody interfering with what the- what they said.
36
00:06:59.044 --> 00:07:12.104
So, so the first people that started to support and recognize the value of data warehouse were the, uh, marketing people of the world.
37
00:07:12.164 --> 00:07:24.304
The marketing and, uh, salespeople of the world recognized the value. And I used to do conferences where, uh, everybody in the conference was from marketing.
38
00:07:24.364 --> 00:07:40.564
The IT people had totally rejected the idea of, of a data warehouse. So from a, from, uh, an organizational standpoint, uh, there was an organization in California, uh, called Pactel Cellular.
39
00:07:41.164 --> 00:07:51.264
And we were working as a consultant with Pactel Cellular, and for whatever reason, uh, Pactel Cellular decided to build a data warehouse.
40
00:07:51.844 --> 00:08:08.784
Now, this was in the day and age when, uh, cellular telephone was just beginning, that, that there... People don't realize once upon a time there were no cell phones. Well, uh, uh, uh, this was in the very early days.
41
00:08:09.344 --> 00:08:15.084
And Pactel Cellular built a very successful, uh, data warehouse.
42
00:08:15.524 --> 00:08:28.524
And the next thing you know is the other cellular telephone companies, uh, heard, "Oh my gosh, uh, Pactel Cellular's got this thing called a data warehouse. We have to have one too." So it's kind of interesting.
43
00:08:28.984 --> 00:08:40.754
Uh, top management went down to the IT people and said, "Build us a data warehouse." Uh, uh, it, it wasn't the IT people, uh, that supported data warehousing.
44
00:08:41.124 --> 00:08:54.384
It was management, uh, wanting to, uh, maintain, uh, uh, uh, their, their lead in the, the marketplace of cellular telephones. Uh, that was the start of data warehouse.
45
00:08:55.024 --> 00:09:16.684
Uh, and then shortly thereafter, uh, a, a guy named,
uh, Walton, who was head of, uh, Walmart, uh, came along, and he wrote a book in which he talked about, not, not extensively, but it was in the book, he talked about, uh, data warehousing in his book and how they were using it at Walmart.
46
00:09:17.084 --> 00:09:30.904
Well, once the cellular telephone companies and Walmart started using data warehouse, then it started to spread like wildfire, that soon everybody needed and wanted to start to use data warehouse.
47
00:09:31.444 --> 00:09:37.244
And so that was the, uh, uh, that was the origins of data warehouse.
48
00:09:37.464 --> 00:09:49.564
Uh, uh, the vendors, uh, in particular IBM, Oracle, and Microsoft never supported, and to this day, Mi- IBM does not support data warehouse.
49
00:09:50.124 --> 00:10:04.164
I think it's really kind of interesting how, how, how, how an organization can make millions of dollars off of something that they don't support, but, but that's life and, and, uh, and I...
50
00:10:04.264 --> 00:10:11.724
And, and, and, and, and, and that's the way it happened. Uh, from- Yeah, it's intriguing to, to, to hear, Bill, sorry to interrupt you.
51
00:10:11.964 --> 00:10:18.804
Uh, IBM having Cognos now and building those, uh, insight solutions, they have Watson to, to do these insights.
52
00:10:18.824 --> 00:10:30.364
But a takeaway from this is that the vendors were more interested in the transactional systems and only wanted to sell those type of solutions instead of the, the data warehouse and the analytics.
53
00:10:30.764 --> 00:10:37.464
And at the same time as well, the business saw the value, and they were driving, but IT didn't understand the business value.
54
00:10:37.544 --> 00:10:50.083
So still back in those days already was the misalignment between business strategy and IT and technology strategy and what you can do with that. So taking it a bit, uh, uh, hop and, and how it go, it goes further.
55
00:10:51.444 --> 00:11:06.124
Absolutely. [coughs] The vendors were making so much money off of transaction processing that they didn't want to change anything, and, and the world changed, but the vendors didn't change.
56
00:11:06.604 --> 00:11:22.044
And, and, uh, a- and to some extent, we still see that today. Uh, O- Oracle has gotten to be a little bit better about supporting data warehouse. [laughs] Uh, uh, data warehouse at IBM is like a swear word.
57
00:11:22.124 --> 00:11:38.244
It, it, it's like a word that you can't use. Um, anyway, that's their problem because IBM has, uh, lost market share and, and, and continued to lose market share. And so anyway, um, uh, it didn't stop there.
58
00:11:38.384 --> 00:11:53.144
So, uh, when data warehouse first came out, uh, uh, the only thing people used data warehouse for, uh, were for structured systems, 'cause that's what we had at the time. It was really more a historical thing.
59
00:11:53.644 --> 00:12:09.344
Um, and so data warehouse ended up, right, wrong or indifferent, uh, data warehouse ended up being, uh, something that was used for structured systems. Uh, in today's world, we have something quite different.
60
00:12:09.884 --> 00:12:17.764
Uh, in addition to structured systems, and structured systems, don't get me wrong, are still very valuable.
61
00:12:18.264 --> 00:12:34.224
But, uh, uh, in today's world, uh, we have text, that people are starting to recognize the value, uh, that's, that's found in text. Uh, we have analog IoT data, uh, data coming from machines all over.
62
00:12:34.704 --> 00:12:55.692
And so the world of data today is a much more diverse, much more complex world, uh, than the world that, uh, data warehouse, uh, uh, came into being.And so, uh, so today, uh, you, you have a different kind of architectural structure.
63
00:12:56.212 --> 00:12:58.812
Uh, you have something called the data lake house.
64
00:12:59.492 --> 00:13:12.412
Now, the, the origins of the data lake house are kind of interesting, that a while back, somebody, I don't know who, uh, uh, came along and said, "We need something called a data lake."
65
00:13:13.092 --> 00:13:25.472
And, and so they said, "What you do is you take all this data, you put it into a data lake, and what happens is, is you can then go use the data for analytical processing."
66
00:13:26.092 --> 00:13:43.352
And in case after case, a very short order, a data lake turns into a data swamp, that people found out that they can't use the data in the data lake, and, and for, um, for many valid reasons.
67
00:13:43.792 --> 00:13:58.652
I have yet to find one person anywhere to come up to me and say, "Oh, we have a very successful, uh, data lake." Data lake architecturally was an amateurish, very- [laughs]
68
00:13:58.752 --> 00:14:09.152
I, I, I don't like to use this word, but it was a stupid idea. And I don't know who came up with it, but it was a really poorly thought out, dumb idea.
69
00:14:09.812 --> 00:14:25.932
So people have these things called data lakes that's not doing them any good. Uh, uh, they need to go beyond data lake. They need to turn the data lake into something that is useful.
70
00:14:26.532 --> 00:14:39.052
Uh, in order to turn the data lake into something that's useful, uh, they, they, they need to do two things. Uh, they need to create what's called an analytical infrastructure.
71
00:14:39.632 --> 00:14:56.132
Uh, the analytical infrastructure allows you to find data, uh, in your data lake, and then they need to integrate the data in the data lake, that, uh, that just throwing data into a data lake does no good.
72
00:14:56.592 --> 00:15:11.312
You've got to put the data in the data lake in a, uh, a form that, uh, can be used for analytical processing. So, uh, in today's world-- uh, in yesterday's world, we had data warehouse.
73
00:15:11.432 --> 00:15:27.052
In today's world, with the diversity of data that we have out there, uh, we have a need for, um, uh, uh, something, a-an architectural structure that's different. We need the data lake house.
74
00:15:27.752
--> 00:15:44.152
Now, the data lake house, a question I get asked all the time, she says, "Gee, Bill, isn't a data lake house the same thing as a data warehouse?" And that is actually a very complex question.
75
00:15:45.192 --> 00:15:48.892
So let me, let me go through the question.
76
00:15:49.392 --> 00:16:08.312
If you take the definition of a data warehouse, uh, uh, what a data warehouse is and was today, uh, uh, then yes, a data lake house is the same thing as a data warehouse from a, from a definitional standpoint.
77
00:16:08.992 --> 00:16:23.112
But from the standpoint of content and structure and an organization, no, a, a data lake house and a data warehouse are two very different things. Why?
78
00:16:23.572 --> 00:16:31.992
And the reason why is, is because the data warehouse really only ever contained structured data.
79
00:16:32.432 --> 00:16:44.112
But the data lake house contains textual data, uh, uh, and contains, um, um, uh, analog IoT data itself.
80
00:16:44.732 --> 00:16:58.591
And you don't treat textual data and analog data the same way you treat structured data. It-- they, they are, uh, they are completely different environments.
81
00:16:59.072 --> 00:17:14.812
So, so in some regards, a data lake house is a data warehouse. In other regards, a data lake house is not a data ware- a, a data lake house is not a data warehouse at all.
82
00:17:15.732 --> 00:17:31.492
When people say, "Gee, can you explain this better?" I, okay, l-l-let me ma- paint an image for you. Suppose you have two young cousins, two ten-year-old cousins sitting in the same room.
83
00:17:31.632 --> 00:17:45.712
One is a young boy, and one is a young girl. When you look at the cousins, you see a, a facial similarity. You see the color of their hair, the color of their skin, their body type.
84
00:17:46.092 --> 00:17:59.212
It's obvious that there is a relationship between the two cousins. But if you were to look at the DNA of the cousins, you indeed would find out that they're, they have different DNA.
85
00:17:59.952 --> 00:18:17.292
And, and for, for one thing, uh, one person is a female, and the other is a male. So, uh, uh, so are these people related? Yes, these cousins are related. Are they twins? No, they're not twins.
86
00:18:17.872 --> 00:18:32.692
And, and, and that's, is the best I can do to describe the relationship between a data warehouse and a data lake house. Yes, there are some similarities. No, they are not the same thing.
87
00:18:34.432 --> 00:18:44.772
So is that more on, on the technical level? I understand the data lake, like you said, it was a data swamp. It was a mistake, but it came with, with big data where, well, we had open source.
88
00:18:44.872 --> 00:18:59.012
We had, uh, software for storing data on commodity hardware. It was very cheap to keep all data aside, and we put everything in there, like you say, unstructured data, log data, everything, but it was not integrated.
89
00:18:59.132 --> 00:19:07.012
And now understanding that it needs to be... easier in a consumable way, that's where the concept from the, the lakehouse comes from.
90
00:19:07.372 --> 00:19:21.332
You integrate it in a better way, you try to align that, and depending on what type of data you put in there, you can refer it more to the data warehouse than to the data swamp, if I understand what you were, uh, explaining, Bill.
91
00:19:22.012 --> 00:19:37.232
Let me, let me kind of back up and, and, and, and recast what you said, that the vendors of the world, when they looked at data warehouse, thought that, "Oh my gosh, what we need to do is manage volumes of data."
92
00:19:37.792 --> 00:19:53.592
And, [coughs] and indeed, volumes of data are, are one of the characteristics of a data warehouse. But what the vendors of the world [coughs] neglected to do was to talk about the whole notion of integration.
93
00:19:54.052 --> 00:19:57.492
You don't just put data into a data warehouse, you have to integrate it.
94
00:19:57.612 --> 00:20:13.272
So the vendors of the world supported the idea of, uh, m- managing volumes of data, but when you, when it comes to integration, they, they, they, they have a blind spot in their brain.
95
00:20:13.632 --> 00:20:24.432
They, they, they, they don't, they don't, uh, they don't get it, uh, uh, w- when it comes to the whole notion of integration. And if you don't believe me, take a look at Hadoop.
96
00:20:24.872 --> 00:20:39.232
Once upon a time, a Hadoop was what everybody had to have, Hadoop. Let's just put... But, but Hadoop was, w- was... Take a look at where Hadoop is today. It's nowhere. And, and, and why?
97
00:20:39.572 --> 00:20:48.372
Is because Hadoop solved the problem of volumes of data, but they didn't solve the problem of integration of data.
98
00:20:48.852 --> 00:20:58.832
And, and, um, the, that, the whole notion of integration of data is, is a very complex subject.
99
00:20:59.492 --> 00:21:17.972
When I, when I talk about integration of data, I typically tell people integrating data is like planting tomatoes in your backyard in the springtime, that you can't plant your tomatoes in your backyard without getting your hands dirty.
100
00:21:18.412 --> 00:21:28.332
That, that, you can't do- I don't know, m- maybe somebody can do it, but I, I certainly can't do it- [laughs]... and my wife can't do it. When you plant tomatoes, you get your hands dirty.
101
00:21:28.772 --> 00:21:41.112
Well, vendors and consultants hate getting their hands dirty. They just don't like dirty hands. Well, uh, if you're gonna integrate your data, you gotta get your hands dirty.
102
00:21:43.412 --> 00:21:56.572
Yeah, I mean, speaking from, from experience as well, doing a lot of data integration and ETL stuff, uh, at all the technical level, and still trying to match, uh, all the data together, I completely get it, uh, uh, what, what you're saying, Bill.
103
00:21:56.672 --> 00:22:02.752
It's kind of we have the magic sauce, and you throw in all your data, and
it aligns together in such a way.
104
00:22:03.292 --> 00:22:16.252
You once told us as well, the one version of the two, of the truth as well to multiple versions of the truth, which I believe more depending on the context where you're working in, this is the, the version you're trying to align.
105
00:22:16.312 --> 00:22:26.282
You, you referred it in the beginning of our conversation as well of how many policies do I have, that insurer you've been working with, and how many subscribers, uh, do we have.
106
00:22:26.292 --> 00:22:36.392
Those type of, of definitions, that, that's very important on integrating your data. You do a lot of work as well in, into more the semantic space and the text analytics.
107
00:22:36.572 --> 00:22:42.752
Is that something where you see that is the next step up onto the data lakehouse or the lakehouse concept?
108
00:22:44.152 --> 00:23:04.652
Um, it's strange that you mention that because, uh, uh, uh, from the standpoint of business value, if you look at it from the standpoint of business value, there is tremendous business value, uh, in corporations wrapped up in text that, uh, everyone is ignoring.
109
00:23:05.072 --> 00:23:22.661
That, that for whatever reason, uh, they, they, they just don't want to, uh, uh, look into text. Uh, and, and, um, uh, they instead want to go to, um, things... O- okay.
110
00:23:23.052 --> 00:23:39.012
There's been recent, some, some recent issues and things that have come up in text. People say, "Gee, Bill, what about ChatGPT? Isn't ChatGPT- [laughs]... uh, uh, uh, uh, re- relating to text?"
111
00:23:39.432 --> 00:23:48.272
And indeed, indeed, don't get me wrong, ChatGPT is elegant. It does some very interesting things.
112
00:23:48.672 --> 00:24:04.092
It does things that haven't been done before, uh, uh, uh, and, a- and it does operate on text, and, and no question about it. However, uh, ChatGPT looks at external data.
113
00:24:04.492 --> 00:24:20.272
ChatGPT is good for looking at data out on the internet, uh, uh, and other places. And, and, n- na- and some corporations undoubtedly do need to look at external data.
114
00:24:20.332 --> 00:24:36.732
There's nothing wrong with that, and [coughs] ChatGPT has certainly opened that door up. But the value in the corporation is not external data, it's internal data. It's, it's the sales that you have.
115
00:24:37.092 --> 00:24:47.842
It's what customers are saying about you. It's the contents of medical records. There, there's many, uh, uh, places in the corporation where there's- Mm...
116
00:24:47.882 --> 00:25:05.812
tremendous business value, uh, that needs to be wait- uh, opened up. And, and ChatGPT can operate on data internally, but the cost of using ChatGPT internally, uh, on your internal data is prohibitive.
117
00:25:05.912 --> 00:25:15.672
It, it, it, it... The training that ChatGPT has got to go through to get to your internal data is, is, is quite frankly beyond belief.
118
00:25:16.292 --> 00:25:30.636
Uh, uh, uh,
a- and, and so-Uh, when I look at when people say, "Gee, Bill, hasn't ChatGPT opened the door up to text?" Well, in a way they have. In a way they, in a way the answer is yes.
119
00:25:31.156 --> 00:25:55.795
Uh, but, but in a way it's still a great mystery as to what's going on in the corporation, uh, with, with the internal data and, and quite frankly, in terms of business value, ChatGPT does not drive business value unless ChatGPT, uh, unless your corporation is looking at external data.
120
00:25:56.216 --> 00:26:07.696
And, and again, I'm sure there are a few corporations in the world that do look at external data, and if that's the case, then ChatGPT is a wonderful tool.
121
00:26:08.196 --> 00:26:20.076
But ninety-nine percent of the corporations in the world aren't concerned with what's going on on the internet. They're concerned with the sales their organization is making.
122
00:26:20.496 --> 00:26:35.196
They're concerned with what the customer is saying about their product. Uh, they need the internalized version of, of, uh, of, of what's going through text. So there, there is this...
123
00:26:36.196 --> 00:26:51.446
I, I liken text to California in 1848. We are told by the historians that in 1848 you could walk into the streams of California and pick up gold. The, the...
124
00:26:51.796 --> 00:27:06.696
and, and according to the historians, you could actually do that. Well, people discovered gold in California in 1848, and in 1849 the whole population of the United States started running to California.
125
00:27:07.236 --> 00:27:19.016
And, and, and so, uh, uh, the, the world of text is like California in 1848, uh, that, uh, it's just waiting there.
126
00:27:19.496 --> 00:27:34.296
Right now you can go pick up these, uh, uh, wonderful business value, uh, gems that are gonna make your... You know, and then you say, "Okay, Bill, give me some idea about the value of internal text in the corporation."
127
00:27:34.856 --> 00:27:35.096
Um,
128
00:27:36.436 --> 00:27:57.316
one value, uh, that, that makes a, a lot of sense is going out to the internet and going to, um, uh, your corporate emails and going to your call centers in your corporation and taking those sources of information and hearing the voice of your customer.
129
00:27:57.876 --> 00:28:06.736
Uh, being able to say, "Gee, my customer is saying, uh, this about my product. My customer would like to have this."
130
00:28:07.236 --> 00:28:29.236
And believe me, corporations that listen to their customers are corporations that are gonna be very successful because they-it's gonna increase their revenue, it's gonna increase the number of customers they have, uh, it's gonna take away from their competitors, and, and they're gonna end up being the winners in the marketplace.
131
00:28:29.996 --> 00:28:37.736
If a corporation doesn't listen to their customers, uh, their business is, is on shaky ground.
132
00:28:38.436 --> 00:28:48.056
And so, uh, uh, one, one way in which text plays
a great role is in terms of listening to the voice of your customer.
133
00:28:48.776 --> 00:28:57.576
Uh, let me give you another example of where text plays a, uh, a, a great role, and that is in medical records.
134
00:28:58.136 --> 00:29:21.716
The current medical record that we have in the world, uh, is written in the form of text, and that's good for the one doctor and that's good for one patient, that when a doctor goes to read a patient's records, uh, reading it in the form of text is, is, is, is what the doctor expects, and that's good.
135
00:29:22.316 --> 00:29:30.716
However, uh, if you want to look not at one patient, but at 100,000 patients, like in COVID.
136
00:29:31.176 --> 00:29:42.656
When COVID came around, we, the world, discovered that, oh my gosh, we've got to be able to look at not at 100,000, we've got to look at millions of customers.
137
00:29:43.096 --> 00:29:57.856
And what do we find when we can look at millions of customers? Uh, we find that we can now do correlative analytics. We can say, how does COVID relate to smoking? How does COVID relate to cancer?
138
00:29:58.356 --> 00:30:05.486
How does COVID relate to gender and race- Yeah... and age and, and all of the important factors?
139
00:30:05.556 --> 00:30:16.816
And these are, if you're, if you're gonna get a grip on COVID, you've got to be able to look at millions of records o-o-o-o-of, of patients.
140
00:30:17.276 --> 00:30:33.416
However, as long as those patient records are wrapped up in the form of text, there's not really a whole lot you can do with them because you can't get a human being to sit there and read the text for 100,000 records.
141
00:30:33.476 --> 00:30:46.216
You just, that, that just doesn't work. And, and so, uh, uh, another really important case for, uh, reading text is, is, is in terms of medicine and healthcare.
142
00:30:46.696 --> 00:30:56.396
And, and, and I could-- And if you, you probably don't want me to get started on this topic because I can go on for forever.
143
00:30:56.996 --> 00:31:08.216
Uh, let me give you one more, one more use case.Uh, uh, the, and, and one that is kind of my pet peeve in life and, and, and, and, and, and that's in corporate contracts.
144
00:31:08.716 --> 00:31:19.756
I can't tell you how many times I've talked to groups of executives and say, "Okay, how many executives here know down to the penny what's in your corporate contracts?"
145
00:31:19.776 --> 00:31:31.436
And to date, not one executive has ever said, "Oh, yeah, we know what's in our corporate contracts." And, and then I say, "Well, I guess I'm kind of confused. You guys are executives, right?"
146
00:31:31.606 --> 00:31:44.036
"Oh, yeah, we're executives." "And, and you're in charge of corporate liability?" "Oh, yeah, we understand risk and liability." "Well, don't you think there's liability in your corporate contracts?"
147
00:31:44.096 --> 00:31:50.216
And, and at the end of the day, that's one of the purposes of a corporate contract.
148
00:31:50.316 --> 00:32:07.976
And, and, and, uh, um, so and, and at this point in time, the executives, uh, they look, kind of look down at the ground and kind of mumble, and they say, "Well, uh, you see, Bill, we've got millions of corporate contracts, and each one is different.
149
00:32:08.356 --> 00:32:20.956
We can't look at every contract." Or, or they say, "Well, we've been running our corporation for 50 years this way, and, and, uh, uh, we haven't had to look at them yet, so we'll just continue."
150
00:32:21.446 --> 00:32:35.016
And, and so for whatever reason, uh, executives haven't discovered that indeed, once you can start to process text, you can look at all of your corporate contracts.
151
00:32:35.156 --> 00:32:51.276
But I gave up on that one a long time ago, because I, I, I talked to I don't know how many executive groups, and I couldn't find one executive that was actually interested in finding out what was going on in their corporate contracts.
152
00:32:51.356 --> 00:33:00.596
So, uh, uh, it's, it's much more, uh, fertile to look at, uh, marketing people. Marketing people,
153
00:33:01.916 --> 00:33:17.996
the ones that are not asleep, uh, marketing people are, are, are interested in knowing what their customer is saying. Yeah, fully, fully understand. [laughs] You, you gave me some takeaways on, on, on that Bill.
154
00:33:18.216 --> 00:33:29.236
Uh, liability contracts make me think of, well, we have enough lawyers. We pay them enough money, and they take care of the liability. We don't need text, text analytics. That's the feeling I get from there.
155
00:33:29.616 --> 00:33:39.336
But I fully understand what you say on, on finding those hidden gems and, and tightening those contracts by being able with the technology and, and the semantics what we have in-
156
00:33:40.296 --> 00:33:43.436
available right now, this day, you can do that.
157
00:33:43.896 --> 00:34:00.156
If I'm thinking about generative AI, that you can apply to these contracts in such a way, where you said most companies looking onto a, with ChatGPT, onto external data, but you have the large language models the- which you already can apply onto your local data.
158
00:34:00.166 --> 00:34:12.156
There is a risk in there because they're not so precise as special narrow, uh, AI models to find those, those, uh, that information and those hidden gems. So I think we're on the path to get there.
159
00:34:12.516 --> 00:34:23.656
But another thing is the semantics and the ontology, trying to understand what is really in that business data, the business glossary as such, which is still a hard way, which we did back in the days manually.
160
00:34:23.716 --> 00:34:50.856
You understood the business, and you, you knew that you had to build a product dimension, you had to build a, a customer dimension, and, and based upon that, you were able to dissect the organization and give those insights of how many customers do we have, how many invoices, how much is our revenue per invoice, per customer, per region, and those type of questions which we, uh, could answer back, back in the days already, but with a lot of manual, uh, labor.
161
00:34:50.876 --> 00:34:57.236
And I think with, with generative AI and the large language models, we're coming close to be able to, uh, to do that.
162
00:34:57.276 --> 00:35:03.496
Maybe still a few years to go, but at the rapid speed what these models are developing, uh, there's, there's still... Yeah.
163
00:35:03.996 --> 00:35:19.155
Uh, let's hope we can crack all the, that information in, in the text, uh, which is lying down and finally get to your insights, which you've been preaching for [laughs] the last 40 years already, and getting those insights from, from the hidden, uh, text in there.
164
00:35:19.215 --> 00:35:36.296
And, and what, what you say as well on top of that is within, uh, call centers, where you have all the audio that has been recorded, we now can transform that into text and understand what is the, the problem of the customer, and you can do that even with generative AI in real time.
165
00:35:36.396 --> 00:35:44.836
So that's, that's, that's a real big evolution where I, uh, feel that, that your vision from back in the days is, is finally coming to fruition as such.
166
00:35:45.776 --> 00:35:58.616
When you think about, you said you, you talk to the, the, the executives as well, where do they really mishit the ball trying to find, uh, the hidden gems into their data, getting the value out of that?
167
00:35:58.996 --> 00:36:11.116
What is the next step they could do? So first maybe where are they, uh, mishitting the ball to get the insights from the data? Let me, let me tell you a little story. A- a- as you were talking, I thought of a...
168
00:36:11.216 --> 00:36:25.816
I say this is a humorous story. This is not humorous at all. This is a very sad story. Uh, the company in the United States, uh, that has the most lawsuits against them is a large company called Johnson & Johnson.
169
00:36:26.176 --> 00:36:30.975
They are a pharmaceutical company on the East Coast of the United States.
170
00:36:30.996 --> 00:36:51.156
And, and, uh, so one day, uh, a while back, uh, we approached Johnson & Johnson and said, "Uh, gee, how would you like to, uh, be able to listen to what your customer is saying and be able to anticipate a lawsuit rather than have a lawsuit?
171
00:36:51.516 --> 00:37:07.896
Because if you can anticipate a lawsuit, it's much less expensive, much easier to head off the lawsuit than to have the actual lawsuit."So, so we presented this idea to Johnson & Johnson.
172
00:37:08.576 --> 00:37:11.836
And, and by the way, I'm not making this up. This is the truth.
173
00:37:12.376 --> 00:37:26.336
And so Johnson & Johnson lawyers came back and said, "Bill, Johnson & Johnson has got a whole army of lawyers, and if you were able to head off lawsuits, what are these lawyers gonna do?
174
00:37:26.776 --> 00:37:36.296
These lawyers are well
employed, and we- they don't wanna become unemployed, so, uh, so we like these lawsuits." And, and I, I thought, you know,
175
00:37:37.376 --> 00:37:50.216
uh, this is ridiculous because, uh, uh, um, uh, lawy- a, a company telling me that it likes lawsuits is, is, is absurd. It's just plain absurd.
176
00:37:50.876 --> 00:38:08.075
And so anyway, um, uh, uh, th- that, that's, that's one story. Uh, what, what can people do today? Uh, I'm happy to report to you that there are some organizations in the world...
177
00:38:08.106 --> 00:38:26.136
By the way, one of the things that, that, that, that strikes me really, really strange is w- we do, Forestrim Technology does a lot of work, but nearly all of our customers are outside the United States.
178
00:38:26.676 --> 00:38:40.416
Uh, the United States, and I-- and as an American citizen, uh, it, it bothers me to say this, but it's the truth. Um, uh, uh, the American corporations have gone to sleep.
179
00:38:40.956 --> 00:39:02.896
And, and, and the, the, the people that are the forward-thinking leaders of the people o- o- o- of the world are outside the United States, uh, places like Abu Dhabi, uh, France, Australia, uh, um, um, uh, South America, Mexico, uh, of all the places.
180
00:39:03.456 --> 00:39:08.896
Everywhere but the United States. And, and, and, and I, I can't explain that.
181
00:39:08.976 --> 00:39:19.816
And, and again, as an American citizen, that really, really bothers me, because that does not portend well for, for the future for the United States.
182
00:39:19.936 --> 00:39:31.476
But, uh, at, at any rate, um, um, the-- there are corporations out there that are starting to unlock the secrets of, uh, of text.
183
00:39:31.616 --> 00:39:42.676
And, and what's gonna happen is the same phenomenon is going to occur that occurred at Pactel, Cellular and Data Warehouse.
184
00:39:43.016 --> 00:39:52.136
There are gonna be a few companies that start to unlock text and start to move forward with it and have success with it.
185
00:39:52.496 --> 00:40:04.976
Then the next thing you know is, uh, a, a few people in their industry are gonna say, "Oh my gosh, look, look what XYZ company's doing. Uh, we need to find out how that happens."
186
00:40:05.006 --> 00:40:17.676
And so, uh, uh, it's gonna be the forces of the marketplace, uh, that pry open, uh, the, the minds of people, uh, to, to, uh, uh...
187
00:40:17.716 --> 00:40:30.116
We, uh, we recently had, and I'm not gonna name the name of the company, but we had an experience with an oil company, uh, that, uh, uh, the...
188
00:40:30.176 --> 00:40:39.256
We, we showed the oil company a, a glimpse of what could be done with text, and the data scientist shut it down.
189
00:40:39.636 --> 00:40:51.196
The data scientist came along, and they were-- they had been working on this problem for five year-- I'm not making this up, by the way. They'd been working on this problem for five years' time.
190
00:40:51.736 --> 00:41:03.136
Uh, we came in
and, uh, with the right technology and the right approach. In two weeks' time, uh, we showed the corporation what could be done and how to do it.
191
00:41:03.576 --> 00:41:18.496
And instead of being thrilled with the success, the data scientists were embarrassed. They said, "Oh my gosh." [laughs] Uh, uh, um, uh, uh, and, and, and so, um, so, so there is that.
192
00:41:18.576 --> 00:41:28.856
And again, I'm not gonna name the name of the, uh, company or the country, but, uh, uh, the, the data scientists, uh, uh, not everywhere.
193
00:41:29.016 --> 00:41:44.176
There, there are data scientists that do want to have genuine progress for their corporations, but data scientists are like a jealous brother and sister says, "Well, you can't do that. I'm the one that gets to do that."
194
00:41:44.586 --> 00:42:11.856
[laughs] And, and, uh, uh, so, uh, so it is gonna happen, that there's going to be, uh, marketplace success with text, and then once that happens, like the day that, uh, Walton, uh, wrote his book, uh, uh, on Walmart, uh, uh, about the, uh, success Walmart was having, uh, that, that, that was the day that data warehousing really started.
195
00:42:12.276 --> 00:42:24.956
And the same thing's gonna happen, uh, with text. Yeah. Uh, well, wonderful worlds, [laughs] uh, Bill. With all the knowledge, how, how do you stay updated?
196
00:42:25.396 --> 00:42:37.376
Do you have any sources that you frequently visit, uh, a lot of papers that you read, or you just, uh, relax and enjoy, uh, your retirement and, [laughs] and just see what is happening with the world?
197
00:42:37.476 --> 00:42:44.016
But I, I, I feel you're still very passionate about all the data and the hidden gems which we can get out of there.
198
00:42:44.496 --> 00:42:56.156
You talked about the technology as well that you're using to get these insights in two weeks instead of the people already crunching it for five years and not getting to the same results you were able to get in two weeks.
199
00:42:56.196 --> 00:43:07.496
So that, that triggered me to understand what technology are you using and which approach are you using? Because technology alone can't solve all, all these issues, uh, and bring the insights.
200
00:43:07.856 --> 00:43:16.912
It's a combination of various factors.Of, of, of, f-first off, I'm going to retire when I die.
201
00:43:17.412 --> 00:43:34.452
Uh, when, when I'm six feet under the earth, uh, is when Bill Inmon will retire, and, uh, uh, so I have no intention of retiring. Uh, second off, how do I get my information? I talk with corporations all the time.
202
00:43:35.212 --> 00:43:43.292
I talk with, uh, probably 15 corporations a week, and by talking to corporations,
203
00:43:44.472 --> 00:43:57.952
listening to what they say, hearing what their problems are, uh, uh, th-that's actually how I stay abreast of, of the marketplace. Um, so, so but in terms of retiring,
204
00:43:58.872 --> 00:44:11.272
I have no intention, I have no intention o-of retiring. By the way, my, the, the, the, the cough that I have in my throat, in Denver we've had something that's kind of...
205
00:44:11.292 --> 00:44:30.572
We've never had it before, but there are these wildfires that are occurring in Canada and, and the Canadians are sending the smoke from Canada down to Denver, and the air that we breathe here, uh, uh, is, is, is bad, bad air.
206
00:44:31.032 --> 00:44:45.172
And, and, um, they, they tell me that breathing in Denver is the equivalent of smoking two packs of cigarettes a day, and, and but, but, you know, you know, what are you gonna do? Stop breathing?
207
00:44:45.632 --> 00:44:59.152
And, and so, u-uh, uh, the, the, the air that I breathe, uh, uh, is, uh, uh, irritating my throat, so I... And I, I can't help it, and, uh, I can't help it.
208
00:44:59.672 --> 00:45:13.892
So anyway, in terms of the technology that, that is there, that is available, that does have large language modules, that we've worked out those problems, is something called my company, Forestrim Technology.
209
00:45:14.512 --> 00:45:24.812
And, uh, Forestrim Technology is able to take text, uh, turn text into a database with context.
210
00:45:25.172 --> 00:45:40.152
It's one thing to go in and look at text and just, uh, tag the text and pick out the, the words of the text, and that needs to be done, and that's important. But text is kind of funny.
211
00:45:40.232 --> 00:45:50.232
Uh, it doesn't do you any good to have text unless you have context. That context is an equal partner with text.
212
00:45:50.352 --> 00:46:03.952
And we've been doing this now for a number of years, and I can tell you that, that ninety-five percent of the battle is not in handling text, but in handling context.
213
00:46:04.012 --> 00:46:17.732
Con-- And, and the reason why, la-language is, is, is more complex than people ever imagine. Take the word fire. What does the word fire mean?
214
00:46:18.212 --> 00:46:29.412
Well, it can mean that your house is burning and it's on fire. Uh, uh, it can mean that your boss wants to, uh, get you out of the company and you've got fired.
215
00:46:29.992 --> 00:46:40.832
It means that you're holding a gun and pulling the trigger, and you've now fired the gun. And, and, and, and if we look further, I'm sure there are other meanings of the word fire.
216
00:46:41.312 --> 00:46:55.312
And, and so words in and of themself, language is very complex. And so when you go to do language processing, uh, uh, it's not enough just to look at the text.
217
00:46:55.712 --> 00:47:12.752
Uh, you've also got to look at the context, and the context is immensely more complex than the text is itself. And so we at Forestrim, we've been working on this pro-problem now for going on twenty-three years.
218
00:47:12.832 --> 00:47:21.372
I'm, I... People say, "Bill, twenty-three years to work on a problem? Uh, you must be a stubborn person."
219
00:47:21.732 --> 00:47:32.012
Well, let me tell you, the two most stubborn people
that ever walked this earth were my mother and my father, so I, I, I, I come by stubbornness honestly.
220
00:47:32.652 --> 00:47:48.872
But, uh, uh, but, but the, the, the problem of how do you handle text, people talk about large language mo-mo, uh, modules. Uh, they talk about all this text. Uh, um, I, I, I don't
221
00:47:50.452 --> 00:47:58.952
like to get into the subject a lot, but, uh, they're, they're, they're using what's called the NLP approach, natural language processing approach.
222
00:47:59.482 --> 00:48:10.292
The problem with natural language processing is that natural language processing was never designed to be a commercial product.
223
00:48:10.592 --> 00:48:25.992
It was designed as a research product, uh, to understand language, and for that it serves its purpose. It's great at that. But in terms of doing commercial processing against text, that's why, uh, um,
224
00:48:27.432 --> 00:48:38.872
NLP has set the world back by decades by taking the approach that they have. And so if you want to find out, uh, how to...
225
00:48:38.912 --> 00:48:46.652
A commercial approach to doing it, you need to look at Forestrim Technology, because we do it every day.
226
00:48:47.192 --> 00:49:06.412
And, and we do it for oil companies, uh, we do it for, uh, uh, call center organizations, uh, we do it for manufacturers, and, and, and so, uh, uh, and, and, and we, we've cracked the code on how do you handle...
227
00:49:07.032 --> 00:49:13.082
Now you say, "Gee, Bill, what's the secret to Forestrim, uh, doing what it does?"
228
00:49:13.532 --> 00:49:26.655
Well, there's really several secrets, but, uh, one of the secrets is ontologies.We know how to create and manage and handle ontologies better than anybody on this planet.
229
00:49:27.076 --> 00:49:43.916
Uh, th-that's the heart of being able to, uh, handle and manage text. Plus, uh, we have the techn- uh, the technology, uh, that does the interpretation. Uh, so, so I liken it to the human body.
230
00:49:44.296 --> 00:49:58.436
There's the brain and the heart. You need the brain to tell you what to do, but you need the heart to pump the blood to make your body work. And, and, and so, uh, Forestrim has gotten those...
231
00:49:58.516 --> 00:50:18.336
It's, again, it's taken us 23 years or so to develop it, uh, but, uh, we can sit down with a straight face, we can sit down and tell somebody, "Yeah, uh, we can take your, uh, customer sentiment and do it with, uh, context, uh, and we can do it."
232
00:50:18.676 --> 00:50:32.516
Uh, it typically takes us a couple of hours to do, uh, the work, uh, for a customer, uh, and the, the same amount of work, uh, would take years to do with NLP.
233
00:50:33.076 --> 00:50:44.236
So if you want a commercial, uh, a-approach to, to handling text, you need to look at textual ETL. Yeah, and that's the ontology part.
234
00:50:44.356 --> 00:50:49.636
I mean, you can do it now in a few hours, where you spent 23 years upon being able to do that.
235
00:50:49.766 --> 00:51:05.776
And the ontologies and, and the approach what you're using, that's, that's industry agnostic because sometimes you can find ontologies which are specific for insurance, uh, companies, but you can't use them for pharmaceutical companies, or you can't, uh, use that for any other industry.
236
00:51:05.816 --> 00:51:20.116
And I think that's, that's one of the, the hidden tricks you f- you find out, and crack the code, as you said. Uh, I'm not gonna discuss in public here- [laughs]... uh, exactly how we do it. But, uh, we have
237
00:51:21.316 --> 00:51:35.076
the world's largest library of ontologies, industrial ontologies. Uh, we can do airlines, we can do manufacturing, we can do banking, we can do finance.
238
00:51:35.556 --> 00:51:51.776
Uh, now, I'm sure somewhere out there there's something we can't do, but, uh, in terms of doing 99% of the world's normal business, uh, we already have the ontologies built for that.
239
00:51:52.416 --> 00:52:05.236
And we have a- and how, how long does it take me to get an ontology for a new b-business? It takes me about, uh, a minute or two. Wow, that's [laughs] that's amazing.
240
00:52:05.716 --> 00:52:16.436
Can't imagine if you, if you have those, but there are very likely a lot of analogies between industries, and that's, that's part of the secret. Bill, we can go on for a few hours again, I think.
241
00:52:16.536 --> 00:52:28.136
Uh, data connects us all, as I say, to, to, to wrap up, but music connects us as well. What is your favorite band or favorite type of music, next to all the data stuff what you're doing?
242
00:52:28.736 --> 00:52:43.556
Well, I'm going to date myself, and fi- you'll find out how old I really am. Uh, but I, I, I still like the Eagles and Fleetwood Mac. Yeah, it's... There is nothing wrong with those.
243
00:52:43.696 --> 00:52:54.896
Uh, well, they still go along, and, um, they're still performing. It's really amazing, and these are people really passionate about music as well, and they go a long way with, with a, a lot of generations.
244
00:52:55.396 --> 00:53:07.296
Bill, it was so great talking to you and understanding more about text analytics, Forestrim, your company, which is really able to do that with the ontologies across industries. That's my really, uh, great takeaway.
245
00:53:07.656 --> 00:53:16.916
Thanks for your time, uh, so much. And- Can I say one thing? There, there is one thing- Yes, yes... that you might be interested in, and it's free. We don't make any money on it.
246
00:53:17.216 --> 00:53:29.836
But if you go to the, the textual ETL, the Forestrim website, we have a book on text analytics. Uh, if you're interested in how to make text analytics simplified, uh, the book is free.
247
00:53:30.296 --> 00:53:45.136
All you have to do is give us your name, and we'll send you the book electronically, uh, no charge. And again, forestrimtech.com. Great one. I'll put it in the comments, uh, for this, uh, podcast.
248
00:53:45.596 --> 00:53:55.556
Bill, great, uh, great so much. [upbeat music] Thanks for your time, and see you next time. Thank you. Thank you for joining us on this awesome podcast.
249
00:53:55.836 --> 00:54:03.496
As senior executives, data and analytics architects, and AI professionals, your time is valuable and we appreciate you choosing to hang out with us.
250
00:54:03.556 --> 00:54:09.836
If you like what you heard, please give us a thumbs up, hit the subscribe button, and leave a comment. We love hearing from our audience.
251
00:54:10.056 --> 00:54:25.556
Don't forget to spread the word on social media, and let's continue to drive innovation in the industry together. Thanks for listening, and we'll catch you on the next episode. [outro music]