0:00

Note, this podcast does not discuss

0:02

nor endorse the idea of discussing stupid

0:04

ideas, because we all know there are no

0:06

stupid ideas. Hello

0:14

and welcome to discussing stupid, the

0:17

podcast where we will tackle everything digitally stupid

0:19

from stupid users and the crazy things they

0:21

do, just stupid practices and the people

0:23

who use them for the stupid

0:25

things we all do and maybe even come

0:28

up with a few ideas on how to do things better.

0:31

And now that I got your attention, let's

0:33

start discussing stupid.

0:37

Hello and welcome to the first broadcast

0:39

of the podcast. I'm Virgil Carroll, your

0:42

host and principal human solutions

0:44

architect at High Monkey living in

0:46

the beautiful city of Minneapolis, Minnesota.

0:48

Welcome to my first show. I'm really excited

0:50

to finally get this off the ground and really

0:53

excited about a lot of the topics that we have coming

0:55

up. For our first topic I really wanted

0:57

to talk about something a little bit next

0:59

generation, something a little newer that kind of

1:01

hits it. One of the things that came up in a

1:03

very big way is the

1:05

world of artificial intelligence,

1:07

bringing artificial intelligence into the

1:10

digital customer experience and I thought one of the ways that we could

1:12

really look at this was to look at it from

1:14

a search perspective because obviously when you

1:16

look at services such

1:19

as Alexa, Siri or Hey Google, and

1:21

that's fundamentally what they are...is search engines. You ask them

1:23

a question, they try to give you back an answer.

1:25

So today we're gonna talk about

1:27

artificial intelligence in search and to help me out with

1:29

that, I have joining me is Ben Tilley,

1:32

who is vice president of sales and

1:35

marketing at Funnelback. Funnelback is an enterprise search

1:37

engine company based out of Australia and Ben's going

1:39

to bring some of his experience in this area as well. Well

1:46

Ben, welcome to episode one.

1:48

If you can actually believe it, you're going to be the

1:50

first person anybody's ever gonna hear

1:53

on discussing stupid, which I'm sure since this is going

1:55

to be like one of the most popular podcasts

1:57

of all time. You're going to be quite

1:59

famous after this. Why don't we start

2:01

out by maybe giving a little bit more

2:03

background about yourself and kind of what

2:05

makes you an expert in the world of search.

2:08

Hi everybody. Thanks Virgil. Well I guess for

2:10

a podcast about stupid people probably helps

2:12

to have one of the most stupid of ones leading it all. So

2:14

I'm glad to be here. A little bit of a background about

2:16

me. I worked for a company called Funnelback. We

2:18

are an enterprise search company.

2:20

I head up a US operation and then I

2:22

also look after sales and marketing across the organization

2:25

in our various global offices. So we're working

2:27

with search technologies and customers

2:30

using search for almost eight years

2:32

now, I think. A fair bit of time

2:34

in the, in the world of search.

2:36

All right, great. Let's

2:38

go ahead and just kind of get into it and you know, we,

2:40

you and I kind of talked before this about

2:42

what we thought would be kind of an interesting first

2:45

episode and we definitely thought about

2:47

artificial intelligence and kind of how that's

2:49

starting to affect our lives and all the different pieces

2:51

that are out there and kind of the digital marketing

2:53

world. Especially. One of the things I think you

2:55

and I both thought really resonate today as

2:57

artificial intelligence and searching and

3:00

what it really does when we start talking about

3:02

that in the world of search. What can artificial

3:05

intelligence really help? What is this stupid

3:07

it could really fix

3:11

I think it could help almost every aspect of search. I think

3:13

when we look at search and we look at the challenges that

3:15

we faced with search, quite often the challenges

3:18

can be either related to the

3:20

content that's being searched over, so the

3:22

information that's actually in place, it's just not written

3:25

in a way or structured in a way that actually

3:27

supports being able to easily find it. And

3:30

the other side of that is the people who are actually

3:32

running the searches. They don't know what to search

3:34

for. They might be searching for terms that have

3:36

nothing to do with what they actually want. So being

3:38

able to aid them in

3:41

the way in which they search is

3:43

also very, very helpful when it comes

3:45

to improving the search experience.

3:47

So wait, I just want to understand

3:49

what you're saying. So you're actually saying

3:52

that most people don't know how to

3:54

search.

3:56

I would say there are better ways

3:58

to search than what most people do.

4:02

You know, we'll often get reports back

4:04

and people going, I on this piece of information

4:06

and it's because they entered into a generic one, one

4:09

keyword entity in one keyword to try and

4:11

find it and really

4:13

actually the keyword had nothing to do with what they were looking for.

4:15

So I would say that quite often people

4:18

don't know how to search, but we in

4:20

the world of search and people who

4:22

use such technologies, we should be doing everything we

4:24

can to make it easier for them. But

4:26

I don't have to know how to search so we

4:28

can aid them in their search.

4:34

You know, obviously there was a lot of sarcasm in that question. (Ben laughs) Oh, I tell you what, I can think of how many

4:37

times I've done usability tests around

4:39

search and just the crazy,

4:42

crazy things that I see from that. You know,

4:44

it's not that people are stupid, they just

4:46

don't understand it. You know, I mean, they

4:48

don't necessarily have the concept of everything that

4:50

goes behind a search engine. How many times have I

4:52

seen somebody go to Google and make the most

4:55

basic search as well? The problem is there, when you go to

4:57

something like Google, you get a result

4:59

and you don't tend to have a lot of expectations

5:01

around that, but from a corporate standpoint

5:03

or if you have it on your public website or

5:05

if it's part of your digital workplace or

5:07

whatever it is, people do the same type of bad

5:10

searches, but then they also have these

5:12

expectations that go along with it and how many times have I

5:15

seen somebody look for a

5:17

particular type of form on an intranet

5:19

and go into the search box and type in the word

5:22

form, hit enter and not really understand what they're

5:24

doing. Basically what we're talking about is then

5:26

all you really need to do is you need. You just

5:28

have artificial intelligence in your search and you'll

5:31

never have to worry about that again. Right?

5:33

Yeah, absolutely. You just tick the artificial

5:35

intelligence box in the setup of your search and that it works.

5:39

Wouldn't that be nice? Yeah.

5:42

Yeah. Well, the reality is that's

5:44

what every business owner thinks is that, you

5:46

know, I mean, they have no concept of technology.

5:49

It must be 'Well, don't you just turn that on and all of a sudden

5:51

everything's working well?'

5:51

And that's a real

5:54

challenge to try and overcome. There

5:56

is a lot of marketing by a lot of organizations

5:58

out there at the moment that puts the message

6:01

forth that is actually all you need to do and maybe

6:03

that might be true for their platforms and the way in which they're

6:05

using various artificial intelligence and machine learning

6:08

technologies, but certainly from search

6:10

it requires a little bit more effort than that. You know, there's,

6:13

there's certainly some time that needs to be spent

6:16

in terms of tuning and training

6:18

the various machine learning and artificial

6:21

technologies out there.

6:22

Yeah, well we all know that's

6:24

not really real in their worlds either. I

6:26

mean there's a lot of work that goes behind artificial

6:29

intelligence and let's be honest, I mean,

6:31

sorry to have you on

6:33

here and to pick on this, but sales and marketing folks

6:35

will make it seem like, yeah, you just flip a switch

6:38

and that's all it takes. But overall there are

6:40

some benefits of it from that

6:42

standpoint, and so when you talk

6:44

to organizations and you're talking to them, uh,

6:46

you know, let's, let's Kinda tackle public

6:50

website search and that we're, can AI really helped.

6:52

We kind of talked about that. It might be able

6:54

to better understand what two people

6:56

are trying to find when they're bad searchers,

6:58

but how can it help an organization better

7:00

understand and do something about the way

7:03

it services searches on its

7:05

website?

7:05

Yeah. Well, I think one of the first places to look is

7:07

to look at the content that's actually being indexed

7:10

instead of documents as a part of search and that's often

7:13

the first point where search can start

7:15

to fail. If your content is

7:17

not structured in a way that

7:19

works with the search engine that you've got in place,

7:22

then it's going to be very, very challenging for people

7:24

to find things. I was also quite a challenging

7:26

conversation to have with an organization.

7:29

Organizations might go out there and put a search

7:31

engine in place and they have

7:33

fairly realistic expectations that that search engine

7:35

is going to improve the ability of their users

7:37

to find things, which I think is completely fair.

7:40

You're purchasing a search engine. You're going to to

7:42

use the search, but then they go out and they get

7:44

a search engine and they put it in place and

7:47

they set it all up and then it stops working or it's not finding

7:49

relevant information and then they get told, oh

7:51

well actually the problem is not

7:53

the fact that the search engine is not working.

7:55

If the fact that your content is terrible and so you're going to

7:58

have to go and fix all of that content.

8:00

Search engine works and I think

8:02

that's where ai machine learning can start to

8:04

help. Because you can sit down with an organization

8:06

and go, okay, look. Your content is

8:08

terrible and your content is really having a

8:10

negative impact on the effectiveness of your search.

8:12

But actually there are all these different

8:15

automated ways that we are able to

8:17

improve your content for you or

8:19

help you understand what you need to do to improve your

8:21

content. So I think that's a great way in which

8:24

machine learning can help people right away. I

8:26

hope people today,

8:26

well I got a first step back and say that your

8:29

statement there about loading up search

8:31

and turning it on and it doesn't work out. I actually

8:33

think that'd be a great episode in and of itself.

8:35

We might have to think about doing an episode

8:38

around that at some point because that isn't that so

8:40

true. I mean, you know, people have no concept of

8:42

search and then you start layering something like artificial

8:45

intelligence and some of the things that it does

8:47

on top of that. And you're not necessarily.

8:49

Well maybe eventually if you put

8:51

the work in, you're making things easier, but overall

8:54

you're making a more complex situation

8:57

more complex on top of that and you're

8:59

kind of adding these layers after layers after

9:01

layers of different things. And it's like you said,

9:04

if you don't understand your content and you don't really

9:06

understand what you're trying to search, probably one

9:08

of the biggest scenarios I see around this

9:10

when it comes from a search perspective, is

9:13

people on their websites. When people

9:15

search on things, they want their users to

9:17

land on kind of their landing pages. You know,

9:19

the kind of these pages that are a

9:22

conglomeration of all these different topics in that

9:24

that maybe have more like one to two

9:26

sentences or maybe a paragraph about a particular

9:28

topic. Just to kind of do an introduction. From a

9:31

marketer's perspective that makes a lot of

9:34

sense because they want you to kind of get to the

9:36

high end fluff of stuff and

9:38

kind of learn that and then potentially work

9:40

your way. And as a matter of fact, from an information

9:43

architecture standpoint, that can make

9:45

sense as well because sometimes we

9:48

can overload people with information

9:50

upfront when you're navigating through a website

9:52

and we want to make sure that they

9:54

really kind of have an opportunity

9:56

to learn kind of the basics and then dig

9:59

deeper. But when we actually look at search and the

10:01

way search works, it

10:03

doesn't really work the same way because

10:05

from a search perspective, the first and

10:07

foremost a search works about words,

10:09

it works about words and locations of

10:11

words and the words that you use and if you have

10:13

a great detail about a topic

10:16

versus a brief detail about a topic

10:19

overall, fundamentally

10:22

search should always show

10:24

that great detail about that topic

10:26

more because there's more instances of the

10:29

word. If you typed in a phrase in the search

10:31

engine is probably seeing those words more

10:33

in close proximity and

10:35

numerous times and all that kind of stuff. One of the things I

10:37

think about Ai, and you can tell me if

10:39

you feel like this is the direction that it's

10:42

going, is that ai can

10:44

help learn that, yeah, okay. You

10:46

were searching on this,

10:48

but actually we found that most people want

10:51

to actually go here and get the

10:53

basics first.

10:53

Yeah, for sure. And certainly that's

10:56

something that actually already being used at the moment, but I

10:58

think it's important to point out, but I always

11:00

such a broad term and

11:02

such a broad category and

11:04

in fact there are elements of ai

11:06

and elements with machine learning that have already

11:08

been in search engines

11:12

for years, you know, and in our search engine for years as well as the ability to do that

11:14

is certainly one thing that

11:16

does exist at the moment and could

11:19

easily be categorized as AI or machine

11:21

learning, but yeah, being able to understand

11:23

more about what the user should

11:25

be searching for. So not what they're searching for,

11:27

what they should be searching for based on how

11:30

people who exhibited similar behavior to them.

11:32

I have searched in the past.

11:34

Yeah, I mean I think one of the things

11:36

that, let's be honest, the

11:38

way we use the phrase ai

11:40

or artificial intelligence isn't

11:42

probably really the way that

11:45

when anybody first heard the phrase,

11:47

it actually is being used. You don't have

11:49

robots that are full consciousness

11:52

going through and looking through search results. All

11:54

be it, that would actually be really cool. You

11:56

know, if, if you could have that, that would be kind

11:58

of fun if they were doing it, I think about the superbowl

12:01

ad that we just saw where they had the

12:03

people actually answering in place of Alexa

12:05

because Alexa had a cold. How much fun would that

12:07

be if you actually had. Yeah, robot

12:09

on the other end answering,

12:11

but that goes on to a complete different topic over there.

12:13

But from that side, I mean, you

12:16

know, for me the cynical

12:18

side of me just says, well, isn't this just another

12:20

way of saying behavioral analytics?

12:22

Because behavioral analytics has been around for

12:24

a really long time. You know, obviously companies

12:26

like Amazon have almost perfected that where

12:28

they understand from what you purchased,

12:31

what you might else want to purchase or what

12:33

you're looking at and you know, constantly

12:35

dilute you with all this information that

12:37

you need a frankly information

12:39

you don't need. That goes along the sides.

12:42

And isn't that really what this ai

12:44

stuff is doing is, is just a better way

12:46

of saying behavioral analytics and that's really

12:48

all it is.

12:50

I think it's broader than that. I certainly agree with

12:52

you, that AI is a term

12:54

that is used for lots of different things that had

12:57

completely useful terms beforehand, so

12:59

they had terms that actually made sense beforehand

13:01

that are now being grouped under AI and definitely what you've described is a really

13:04

big part of it, but there are other elements

13:06

of it as well. Certainly some of the more complex

13:10

things like image recognition. I think Google

13:13

vision, the Google Vision Api still

13:15

considered part of AI and definitely

13:17

different for that. So I'd say it's part of it,

13:20

but it becomes

13:22

so broad these days it covering so many different areas.

13:24

Yeah, I suppose I can see that in, you

13:26

know, one of the areas that I know that is

13:29

really up and coming these days is

13:32

voice-based activities and

13:34

I know that's actually getting to be a huge thing

13:36

in the search environment where now

13:38

not only can you search from

13:40

the standpoint of just looking

13:43

at typing words in a search box,

13:45

but now you could actually, you know,

13:48

ask Alexa a question and have Alexa

13:50

get you the answer. Which fundamentally is search

13:53

in that and so are you seeing that

13:55

in your world you're starting

13:57

to need to recognize that potentially

13:59

one of the futures and that is

14:01

going to be voice recognition and

14:03

I got to imagine that's where one of the areas

14:06

where AI would become very important

14:08

because everybody speaks differently.

14:11

Absolutely, and certainly voice is now starting

14:15

a requirement that our customers are having of us

14:17

and it's something that we're seeing organizations use

14:19

on their site, so we actually worked with a bank

14:21

in Australia, one of the largest banks in Australia,

14:23

and they have voice search that hooked into their

14:25

Funnelback on their website. So if you go to their website

14:28

and you want to find information on how to

14:30

open an account, or how to apply for a credit card,

14:32

you can actually just talk to it and it

14:34

will return you a set of search results. I don't know

14:36

when you're planning on publishing this podcast, Virgil,

14:38

but potentially when this podcast is published

14:41

the Funnelback website will also have voice search on it we're actually

14:43

looking at putting it on the our site just to demonstrate how

14:45

it can operate. So I think it's something at

14:47

the moment it's, it's a cool nice

14:49

to have. I yet had a customer

14:52

say to me, we absolutely

14:54

must have voice search. It is a mandatory requirement,

14:57

but I think soon as more and more people start

14:59

putting it on their site and then more and more people get

15:01

used to the convenience of searching with your voice.

15:04

It is going to become a mandatory requirement and everyone's

15:06

going to be expected to have.

15:09

I just kinda think of that and I think of

15:11

standing in line at the grocery store and listen

15:13

to the people talk on their phones. So is that

15:15

what we have next is people are going to be searching

15:18

through their corporate websites

15:20

on their phone. I mean it's as if we need

15:23

another reason to listen to other people talk

15:25

about their stuff. But I hate to say

15:27

it. You're right. And you know, I mean obviously there's

15:29

other aspects as well, I've seen some

15:31

things out there about organizations

15:33

that are looking at bringing, you know, kind of a,

15:36

also these practices with, with virtual

15:38

reality and some of the virtual reality technology

15:40

and it is kind of crazy where

15:43

it's all going and, and what's it doing

15:45

in there. So, but I, I have one question

15:47

before I kind of switch gears here

15:49

a little bit. You kind of

15:51

said it right there. Other people are

15:53

doing it, therefore we're going to need to do it

15:55

too. And I think that's one of the big

15:58

things that kind of push a lot of these trends

16:01

forward. But one of the questions I have for

16:03

you that I think is really interesting is,

16:05

okay, so I'm an organization and I've

16:07

invested heavily in

16:09

artificial intelligence. That's great. Well the

16:12

reality is there's a cost

16:14

associated with that and there's time and effort

16:16

associated with it. And for a lot of organizations that's

16:19

not maybe the most realistic thing.

16:21

So do you in your kind of

16:23

world of search, see any negative effects of

16:25

this? Otherwise now you've got the big

16:28

organizations doing this and people are starting to

16:30

use this and expecting basically search to

16:32

kind of fix itself instead of

16:34

you having to be a good searcher. To

16:36

me that kinda is going to propagate across the

16:38

Internet and basically mean if you don't

16:41

participate in fixing people's stupid

16:44

practices and you just don't have the time, money or

16:46

effort to be able to do that, then

16:48

basically you're going to be left behind because people are going

16:50

to go to your website and never find

16:53

anything.

16:53

Yup. Absolutely. I

16:55

think that'll, that'll definitely happen. It will

16:58

become an expectation and if people are not providing

17:00

it then they will, uh, will

17:02

just get left behind. You know, there will be overtaken taken by their

17:04

competition. It's, it's interesting.

17:06

So when I, when I moved to America about

17:09

two years ago now, I

17:11

had to pretty much restart my life

17:13

so I had to get a bank account, get health insurance,

17:15

all that sort of stuff. And when I was looking to

17:17

open bank accounts, there were a whole bunch

17:19

of banks that I shortlisted as being

17:21

ones that I was potentially interested in. I

17:24

then started getting into each of their websites to start looking

17:26

at them and finding a little bit more information out about

17:28

them and I can tell you the ones that had a

17:31

bad user experience once it had bad websites,

17:33

had bad search and just didn't even get

17:35

considered by, but I opened it up and I typed in

17:37

how do I open an account with different types of accounts

17:39

available. If it was a terrible

17:41

experience, I just discounted it to the side and that

17:43

was because there were so many other

17:45

banks that were based up here in Seattle that

17:48

were easier for me to and more convenient for

17:50

me to find the information that I needed

17:52

in order to make a decision about which bank

17:54

to become a member of and I think we'll

17:56

see the same thing happening with technologies like

17:58

voice search and it will literally be. Someone

18:00

will be looking at and evaluating

18:03

organizations and it might be again looking to open up a bank account

18:05

and I go, okay, well this website, I can't

18:07

search on search on their site with my voice,

18:09

so why would I bother? I'll go to the next bank because at least

18:11

I can search with my voice there and that's

18:13

going to make it easier for me to complete the

18:16

task that I came to this site to do, which is

18:18

to become a customer of that bank.

18:19

And actually I remember actually

18:22

I probably should look to see if it's still around, but there was

18:24

a website a number of years ago, an organization called Wolfram

18:27

Alpha and they were kind of the first ones that were supposed

18:29

to be doing intelligence search in that

18:31

side. And so you could, you know, they had those

18:33

kitschy things like you could go on and search

18:35

what is the meaning of life and it gives you 42

18:38

and different things like that. But I think you're right.

18:40

I think that it's going to be, you're

18:42

either going to be left behind or you're

18:44

not going to care. I think that's really

18:46

your, your two options that are, which is, you

18:48

know, maybe good but also bad and you

18:51

know, but I think the digital world

18:53

is starting to move at such a fast pace.

18:55

We're just having to constantly kind of readjust

18:57

and adapt and that kind of stuff. So I

18:59

kind of want to switch gears here and I kind of would

19:01

like to talk a little bit about what we've talked about ai

19:03

and obviously maybe this is important.

19:06

The question is if you

19:08

have bad search right now, how

19:11

do you get started in that area? I mean, what? What

19:13

do you even start looking at? I mean it's not like we

19:15

said you just, you just purchased a product,

19:17

flip on the AI switch and you go,

19:20

but where would you recommend that an organization

19:22

starts looking first or

19:24

starts doing first before they actually

19:26

even start looking at how they could bring some of

19:28

this additional resources into

19:30

their search systems?

19:34

I said the first thing you need to do is try

19:37

and work out what is bad

19:39

about the church because they can. It can be a number

19:41

of things that can present themselves

19:43

as bad search it could be

19:45

that you'll use as just a not a tech

19:47

savvy group. They just do not know how

19:50

to really use a search engine

19:52

and to really search well, it could be

19:54

that you have a very technically

19:56

proficient group of users who are using your search

19:58

and can maybe they're too technically savvy who

20:00

are using a search engine searching really

20:02

well and your content just not up

20:05

to scratch. It's just not coming back like it should be.

20:07

Your relevance is all over the place because you've got

20:09

varying levels of quality

20:11

amongst your content, so I took first

20:14

place I would recommend starting is just getting

20:16

into some analytics or actually just sitting

20:18

there and spending some time with the people

20:20

that have identified your search as being bad and trying

20:22

to identify a little bit more about why is it bad.

20:25

If it's the users well then there's a one

20:28

path where you need to take to start trying to either educate

20:30

them on how to search better

20:33

or trying to put in place

20:35

tools and AI and machine learning

20:37

to improve the way they search. If it's

20:39

your content, well then you need to start looking at,

20:41

okay, how can we either improve our

20:43

content ourselves? How can we put in place tools

20:45

to improve our content, but I think the

20:47

first step is you got to work out why such

20:49

bad because there can be a number of different

20:51

reasons why your search is bad.

20:54

Yeah. I've got to add on there that when

20:56

when you search, when

20:58

you test your search in that, which is

21:00

usually the way people do this as they test

21:02

it. Again,

21:04

you want to just talk about crazy scenarios

21:06

and of course the people are most

21:08

notorious about this are IT people, but I've

21:10

also seen marketers do this. They go up and

21:13

or search by test to, you know, typing in

21:15

the specific name of a page or

21:17

the specific name of a document or something. And you

21:20

know, in my experience, even some of the worst

21:22

search engines out there in the world. If you

21:24

type in the exact wording of the exact

21:26

title of a document, you probably have

21:28

a decent chance of getting it at least within

21:31

the first five results. And what they don't

21:33

do is they don't look at those other avenues,

21:35

which is where I think

21:37

AI can really help from that side is

21:39

when that group of searches that we

21:42

know we need to get something, we're

21:44

not really sure what it is or maybe where

21:46

it is, but we have a general idea. We

21:48

may not know the word we need to use, but we

21:50

need to kind of get there, but we're not really

21:53

sure how to get there. And

21:55

I think first and foremost, if

21:57

you aren't testing your search that way, then you

21:59

don't know really well whether it works or

22:01

not. And like you said, kind of looking into

22:03

analytics, a lot of times we'll give you that. I think one of

22:06

the more entertaining things I ever see

22:08

when I look at our Google analytics is why

22:10

people get to us. We have a huge

22:12

amount of traffic to our blog site

22:14

because of a One

22:17

Note and Kanban

22:19

board blog post. One of our project managers

22:21

did like seven or eight years ago. I mean, we

22:24

literally have thousands and thousands of visits

22:26

a month to our site because of that

22:28

one post. Does that really have anything to do

22:30

with our business? No, it. But it was something that

22:32

that person decided as the moment to do, but when

22:34

you look at that and you look at the reason people get to

22:36

it, it's amazing how many people actually

22:39

visit highmonkey.com because they search on one

22:41

note and project management and that kind of stuff. So

22:44

I think that's a very important side of

22:46

that. So let's say that I go through there and I kind

22:48

of have some understanding

22:51

of what maybe the issues are

22:53

and then what kind of technologies out there, you

22:55

know, when we start talking about bringing in an

22:57

artificial intelligence, what does that really mean? What kind of

22:59

technologies can we bring into play

23:01

really to help us get to that next

23:04

step of being able to help guide

23:06

the user to the correct information.

23:08

So on the content side, I'd

23:10

say one of the easiest things to do is to start looking

23:12

at how you can bring into natural language processing

23:14

to extract entities from

23:17

unstructured content, so a

23:19

very common issue with

23:21

content is that it's structured poorly and what I

23:23

mean by being structured poorly is that metadata

23:25

is either useless or

23:28

nonexistent or on or part

23:30

of your content might have great metadata and

23:32

the rest of the content might have okay metadata,

23:35

so being able to use some form of natural

23:37

language processing to extract information

23:40

out of your out of your written text and then turn

23:42

that back into structure is a really

23:44

great way that you can start to improve search

23:46

and that also goes to helping improve

23:49

the actual output of search

23:51

to so there's the relevance side of things, so

23:53

improving the result ranking, but when

23:55

you've got structured like that in place, most modern search

23:57

engines will take that structure and use that to improve

23:59

how search is displayed. So

24:02

making it easier for people who were looking at your search

24:04

results to identify that, yes, actually result

24:06

number three is the result that I'm after.

24:08

So that's one thing that can be done and it can be done. We'd been

24:10

with relative ease these days. I would also

24:12

recommend looking at tuning the ranking

24:15

algorithm that's in place. This is where

24:17

various forms of machine learning really do come into

24:19

play, so thinking from the Funnelback side of

24:21

things we have at tuning interface within

24:23

our product where the organization would

24:25

sit down and they would actually start to put in place

24:27

a training set of data. That training

24:30

set of data would consist of

24:32

search queries and then the results that should be

24:34

coming back with number one and

24:36

in fact that's actually a

24:38

great opportunity for an organization to really sit down and have a

24:41

think about the different use cases for this

24:43

search and to look at their analytics as to how

24:45

people searching right now and then once you've put the

24:47

training set in, there were various algorithms that

24:50

takeover and start to look

24:52

at all the different ranking

24:55

parameters that exist within the search engine and they

24:57

start to alter them until it finds

24:59

the combination that best suits the

25:01

training set that you've put in place. I know

25:04

within Funnelback we have

25:06

72 individual parameters that can be used

25:08

to determine relevance. Each one of them can

25:10

be set to any number of settings so you can

25:12

start to imagine all the different combinations

25:15

that are possible and really it's only with

25:17

a machine learning algorithm that

25:19

you can actually have something like this performed

25:22

and performed well. If you go there and try to do it

25:24

manually yourself, it would take you on, I don't know

25:26

how much time it would take you to take a very long time

25:28

to be able to actually tune

25:31

an algorithm yourself to the best possible setting.

25:33

So those are two easy ways you can get started on

25:36

the content side of things right away.

25:37

It's not just turn it on. Well, that's just

25:39

crazy in that, but I think.

25:41

I think on the other side, one of the important

25:44

pieces there is the

25:46

flip side. You can do all this stuff

25:49

and you could actually make your search worse.

25:51

I mean, I've seen a lot of organizations that, you know,

25:53

they take to actually tune in their results, but they don't

25:55

really have an understanding what they're doing and they're like,

25:57

oh, well, you know, I really want this to show

25:59

up first in this, to show up second and this to show

26:01

up in the search results. But the reality

26:04

is, like you said, the content isn't very good.

26:06

Therefore you can almost screw

26:08

up the search engine by saying that

26:10

something that has really bad content

26:13

you want to show up first. So

26:15

I think it's a lot like anything

26:17

when we start talking about automation,

26:20

which is really what artificial intelligence

26:22

is, automation and bringing in automation

26:24

to things. There has to be that human

26:27

element that's brought into there in that human

26:29

element has to be smart. They

26:31

can't just do things. You can't just walk

26:33

in, click a few buttons and hope for the

26:35

best. And, and that's kind of one of the things that

26:37

concerns me when you start bringing together these

26:40

type of technologies into our

26:42

ecosystem is typical marketer or the typical

26:45

business owner or even the typical IT person, they

26:47

don't necessarily have the understandings

26:49

around this stuff to really understand

26:51

it and I and I think we

26:53

need to continue to strive to make

26:55

our interfaces better so that

26:57

when somebody is doing an activity like

27:00

tuning or better understanding how

27:02

natural language processing works in that is

27:04

giving us the type of feedback that people

27:06

really can use. Not the type of feedback

27:08

that says this is what's affecting the search

27:10

engine type of feedback that like when

27:12

you change this,

27:14

this is going to change it like this

27:17

and here could be the consequence one way or

27:19

the other. Now I know

27:21

from a product perspective that it's like, well that's

27:23

crazy maybe to do it, but overall

27:26

isn't that kind of what we're asking people to

27:28

do in the first place is make all these

27:30

assumptions and leaps of faith and giant

27:32

decisions around kind of how things are going to do and yet at

27:34

the same time, from a product

27:37

standpoint, we're providing them the

27:39

resources to do this, but we're not

27:42

necessarily giving them the understanding to be

27:44

able to do it successfully.

27:46

Yeah. That's a really good point and I do agree

27:48

with you. I think it is an ongoing

27:50

process and certainly ensuring

27:53

that people do have the resources to

27:55

do it successfully is a big part of ensuring

27:57

that the actual outcome is what you want.

28:00

Yeah, and I think one of the things I always see

28:02

is that you know that search in lot

28:04

of organizations, I'm sure you've seen this as well,

28:06

tends to be an afterthought. It tends to be

28:08

something that we think of after we've done everything

28:11

else and it's like, oh yeah, we should probably have a search

28:13

engine for our website, or oh yeah, we should probably

28:15

have some type of enterprise search for

28:17

our document management solution. Whatever

28:19

it might be. It tends to be that afterthought

28:22

and going along with an afterthought, it tends to be a highly

28:24

underfunded afterthought so they

28:26

don't put the resources in it and the reality is

28:28

to do this stuff, to implement what we've been

28:30

talking about today, you have

28:32

to put a lot of time, a lot of effort and

28:35

you can't just rely on the software to do

28:37

it for you, but the good news is

28:39

that with artificial intelligence, getting more

28:41

integrated into search and having

28:44

a better understanding about these things, we're

28:46

getting to that point where

28:49

AI is going to help start and do more

28:52

of the heavy lifting for us, but

28:54

not necessarily replace us. Would

28:56

you agree?

28:58

I would agree. Yeah.

28:59

Yeah. Well, Ben, I really appreciate you joining

29:01

me today, especially during my very

29:03

first episode. I appreciate you taking time

29:06

out of your busy day. Before we go, can you tell

29:08

us if people want to kind of learn more about artificial

29:10

intelligence? Is Search and in particular

29:13

how funnel back kind of does this and

29:15

some of the things there, how could they learn more about funnel

29:18

back and kind of what you guys are doing in that area?

29:21

Yeah, absolutely. So we've got a fair bit of information

29:23

up on our website. We will call

29:25

it out now. We have deliberately been very

29:27

careful about promising

29:30

AI and machine learning aspects of that product and

29:32

and by that I mean we're not using

29:34

messaging a language which talks about

29:36

having Ili available as soon as you click

29:38

the button out of the box, but we certainly have a lot

29:40

of blog posts and a lot of content about at

29:43

where we see AI influencing search

29:45

and and the ways in which we're using it with our

29:47

product at the moment. So funnelback.com

29:49

is our website. I would definitely recommend

29:52

going to the blog section thats probably where we put

29:54

the bulk of the information right now because it's

29:57

fair to say that while we do

29:59

use aspects of machine learning and

30:01

ai within our product, there's still a long

30:03

way to go and so most of the information

30:06

that I think is going to be useful and interesting to people

30:08

is actually an blog section where members

30:10

of our staff were just talking about what

30:12

they see as being the

30:14

direction that it's going to take or areas in which

30:16

they're experimenting with it rather than a full product

30:19

section on our site that talks about our AI

30:21

product that you can just turn on with the click of a button.

30:24

Well, great. Well again, Ben, thank you for joining

30:26

me today and hopefully

30:28

our discussion gave people some new insight

30:30

into artificial intelligence in search.

30:33

Absolutely. Fantastic. Thank you for having me here.

30:40

In my final segment of the podcast,

30:42

I want to talk about stupid buzzwords

30:44

and I call this segment the stupid

30:47

buzz. So what do I mean by stupid

30:49

buzzwords? I basically mean words that

30:51

used to have meaning that we've

30:53

kind of in the digital world, taken and made

30:55

meaningless. Now this one arguably

30:58

may not have had a lot of meaning prior,

31:01

but overall it is something that is completely

31:03

misunderstood and that is chatbot.

31:06

What is a chatbot? Do you know

31:08

what a Chatbot is? Does anybody really

31:11

know what a Chatbot is? In very

31:13

simple, plain human language?

31:16

A chatbot means an automated system

31:18

that I asked a question and it responds

31:20

back. How does that relate to artificial

31:22

intelligence? Well, they try and take it

31:24

a little bit further and say that it has

31:27

an understanding. Maybe it's using search,

31:29

natural language processing, all

31:31

those kinds of things and it's actually providing

31:33

some type of meaningful response back

31:36

to you. The problem is that

31:38

nobody seems to really understand what these things

31:40

are for yet it's really a fad

31:43

that everybody seems to think they need. So

31:45

now instead of having some type of chat

31:47

module on top of your ecommerce platform

31:50

or having some type of discussion forum

31:52

for your employees to get something, you want

31:54

to have an automated system that basically doesn't

31:57

make you work and takes care

31:59

of this for you, but there's a few things

32:01

you should know about chatbots really that are very

32:03

important. Number one, chabots

32:06

are really only as sophisticated as you

32:08

make them. They don't start

32:10

out. There is no chatbot here

32:13

that is already smart. There is something

32:15

that you have to build yourself and

32:17

actually do it. Now, there are some commercial products

32:19

out there, but again, you still have to teach them what they

32:21

need to learn. They don't learn phrases on

32:23

their own. You have to give them their phrases. Those

32:26

that do learn phrases on their own, you're

32:28

probably not going to want to use in your instance.

32:31

Number two, does your business actually

32:33

need a chatbot? You know,

32:35

this again is kind of a fad that people

32:37

have been talking about, but do you really need one? Do

32:40

you have a business purpose for this?

32:42

And what is that business purpose if

32:44

you have maybe a business where you're communicating

32:47

with your clients on a regular basis or

32:49

customers need to constantly communicate. Maybe

32:51

there's some benefit to have an automated processes

32:54

versus having an actual human

32:56

at the other end. But overall what tends

32:58

to happen is we tend to deploy these without really

33:01

much need or understanding

33:03

of why we need these things and that.

33:06

And so when you kind of look at that and you

33:08

kind of look at what this is, chatbots

33:10

to me is very much a fad. It was a huge

33:12

thing in 2017 and you're actually already

33:15

starting to hear a lot less of it in 2018

33:17

because the reality is it's

33:19

kind of cool and new and thinking everybody

33:22

could get an automated answer that they could actually

33:24

understand. But in reality

33:27

there's just not a lot behind it and a lot

33:29

of people can't do it. And the one thing I'll leave you

33:31

with before I sign off here is why

33:33

do we keep trying to replace people

33:37

and the meaning of words when they're so

33:39

much more meaningful when they come out of somebody's

33:41

mouth. And with that, we

33:43

officially end our first episode.

33:46

Thank you for joining me. I hope you enjoyed

33:48

the topic. We have a lot more interesting

33:50

discussions coming up around a whole wide

33:52

variety of topics and

33:54

I'm really excited for where this podcast might

33:56

go. You can subscribe to us

33:59

through many of the most popular podcast directories

34:01

such as Itunes, stitcher, and soundcloud,

34:04

or you can visit us on the web at

34:07

discussingstupid.com. If you have any questions or comments, feel

34:09

free to email

34:12

us at [email protected] or visit us on twitter with

34:14

the handle @discussingstupid. So thank

34:16

you again for joining. And then tell the next time

34:18

we do, you can just discuss stupid

34:21

on your own.