0:00

Welcome everyone to another episode of

0:02

Dynamics Corner . Is AI

0:05

a necessity for the survival of humanity

0:07

? That's my question . I'm your co-host

0:09

, chris , and this is Brad .

0:11

This episode was recorded on December 18th 2024

0:15

. Chris , chris , chris . Is

0:17

AI required

0:20

for the survival of humanity ? Is

0:23

humanity creating the requirement for

0:25

AI for survival ? That's

0:27

a good question . When it comes to AI , I have

0:30

so many different questions and there's so many points

0:32

that I want to discuss about it With

0:34

us . Today we had the opportunity to speak with Zoran Fries-Alexanderson

0:36

and Christian Lenz about some of those

0:39

topics . Good

0:58

morning , good afternoon . How are ?

0:59

you doing there , we , there we

1:02

go good day

1:04

good afternoon over the pond

1:06

.

1:06

How are you doing ? Good

1:08

morning , well , good good good

1:10

, I'll tell you , soren , I love

1:13

the video . What did you do

1:15

? You have the nice , the nice blurred

1:17

background , the soft lighting yeah

1:21

, it's uh .

1:23

You can see great things with a great camera

1:25

.

1:27

It looks nice , it looks really nice , christian

1:30

. How are you doing ?

1:31

Fine , thank you very much .

1:35

Your background's good too , I like it

1:37

, it's real .

1:38

Back to the future .

1:41

It is good , it is good , but

1:43

thank you both for joining us this afternoon , this

1:45

morning , this evening , whatever it may be

1:47

been looking forward to this conversation . I was talking

1:49

with chris prior to this . This is probably

1:52

the most prepared I've ever been for a discussion

1:54

. How well prepared I am we'll see

1:56

. Uh , because I have a lot

1:58

of things that I

2:00

would like to bring up based on some individual

2:03

conversations we had via either

2:05

voice or via text . And before

2:09

we jump into that and

2:11

have that famous topic , can

2:14

we tell everybody a little bit about yourself , soren ?

2:18

Yes , so my name is Soren

2:20

Alexandersen . I'm a product

2:23

manager in the Business Central engineering

2:25

team working on finance

2:28

features basically rethinking finance

2:30

with co-pilot and AI .

2:33

Excellent , excellent Christian .

2:37

Yeah , I'm Christian . I'm a

2:39

development facilitator at CDM

2:41

. We're a Microsoft Business Central partner

2:44

. Development facilitator at CDM . We're a

2:46

Microsoft Business Central partner and

2:49

I'm responsible for the education of my colleagues in all the new topics , all

2:51

the new stuff . I've been a

2:53

developer in the past and a project manager

2:55

and now I'm taking care of taking

2:57

all the information in that

2:59

it leads to good solutions

3:02

for our customers .

3:04

Excellent excellent and thank you both for joining

3:06

us again . You're both veterans and I appreciate

3:08

you both taking the time to speak with

3:10

us , as well as your support for the podcast

3:13

over the years as well . And

3:15

just to get into this

3:17

, I know , soren , you work with

3:19

AI and work with

3:21

the agent portion I'm simplifying

3:24

some of the terms within

3:26

Business Central for the product group and

3:29

you know , in our conversations you've

3:31

turned me on to many things . One thing you've turned me on

3:33

to was a podcast called

3:35

the Only Constant , which I was pleased

3:37

I think it was maybe at this point

3:40

a week or so ago , maybe a little bit longer

3:42

to see that there was an episode where

3:44

you were a guest on that podcast talking

3:47

about AI , and

3:49

you know Business Central , erp

3:52

in particular . I mean , I think you referenced

3:54

Business Central , but I think the conversation that you had was

3:56

more around ERP software and

3:59

that got me thinking

4:01

a lot about AI , and

4:04

I know , christian , you have

4:06

a lot of comments on AI as well too , but

4:09

the way you ended that

4:12

with you know nobody wants to do the dishes is

4:14

wonderful , which got my mind thinking about

4:17

AI in

4:19

detail and what AI

4:21

is doing and how AI is shaping . You

4:24

know business , how AI is shaping how

4:26

we interact socially , how

4:28

AI is shaping the world , so

4:30

I was hoping we could talk a little bit about

4:33

AI with everyone today . So

4:37

with that , what are your thoughts

4:39

on AI ? And also , maybe , christian

4:42

, what do you think of when you hear of AI or

4:44

artificial intelligence ?

4:46

I would say it's mostly

4:49

a tool for me Getting

4:53

a little bit more deeper into what

4:55

it is . I'm not an AI expert

4:57

, but I'm talking to

5:00

people who try

5:03

to elaborate how to use AI

5:05

for the good of people . For

5:08

example , I had a conversation with

5:11

one of those experts from Germany just

5:13

a few weeks before directions and

5:16

he told me how to make

5:19

use of custom GPTs and

5:22

I got the concept and tried

5:24

it a little bit custom

5:28

GPTs and I got the concept and tried it a little bit and when I got to Directions

5:30

EMEA in Vienna in the beginning of November , the agents topic

5:32

was everywhere , so it was

5:34

co-pilot and agents and it

5:36

prepared me a lot how this

5:38

concept is evolving and how

5:40

fast this is evolving . So I'm

5:43

not able to catch up everything , but

5:46

I have good connections to people

5:48

who are experts in

5:50

this and focus on this , and the conversations

5:52

with those people , not only

5:54

on the technical side but also on how

5:58

to make use of it and what to keep in mind when using

6:00

AI , are very

6:02

crucial for me to make

6:05

my own assumptions and decide

6:08

on the direction where

6:10

we should go as users

6:12

, as partners for our customers

6:14

, and to consult our

6:17

customers and

6:19

on the other side . With the evolving

6:22

possibilities and capabilities of

6:24

AI , generating whole

6:28

new interactions with

6:30

people , it gets

6:33

much more harder to

6:35

have this barrier in mind . This is a machine

6:37

doing something that I receive

6:40

and this is not a human

6:42

being or a living being

6:45

that is interacting with me . It's

6:48

really hard to have

6:51

a bird's eye view of what is really happening

6:54

here , because it's so

6:56

like human interaction

6:59

that we have with AI

7:01

, that is hard to

7:03

not react as a human

7:05

on this human interaction and

7:08

then have an outside

7:11

view of it . How can I use it and where is

7:13

it good or bad , or something like that , that moral

7:15

conversation we're trying

7:17

to have . But

7:20

having conversations about

7:22

it and thinking about it helps

7:25

a lot , I think .

7:27

Yeah , it does , Saren . You

7:30

have quite a bit of insight

7:32

into the agents and working with AI

7:34

. What is your comments

7:36

on AI ?

7:38

I think I'll start from the same perspective as Christian

7:40

. From the same perspective as

7:42

Christian , that

7:45

for me , ai is also

7:47

a tool in the sense that when

7:51

looking at this from a business perspective , you have

7:53

your business desires , your business goal , your business

7:55

strategy and

7:58

whatever lever you can

8:00

pull to get you closer to that business

8:02

goal you have AI might be a tool you can pull to get you closer to that

8:04

business goal you have . Ai might be a tool you can utilize for

8:07

that . It's

8:09

not a hammer to hit

8:12

all of the nails . I mean it's not the tool

8:14

to fix them all . In

8:16

some cases it's not at all the right tool . In

8:19

many cases it can be a fantastic tool

8:21

. So that depends a lot on the scenario . It depends a lot on the goal . It can be a fantastic tool . So that depends

8:23

a lot on the scenario . It

8:27

depends a lot on the goal

8:29

. I will say that I'm fortunate in the

8:31

way that I don't need to know the intricate

8:33

details of every new

8:35

GPT model that comes out and

8:37

stuff like that . So that's

8:39

too far for me to go

8:41

and I could do nothing

8:44

else . And to your point , christian . So you

8:46

said you're not an ai expert . So

8:48

but I mean by

8:50

by modern standards and the

8:53

ai that we typically talk about these days . Well , lms

8:55

, it's only been out

8:57

there for such a short while . Who who can actually

9:00

be an ai expert yet ? Right

9:02

, I mean , it's been out there for

9:04

a couple of years . In this modern incarnation

9:06

, no one is an

9:08

expert at this point . I mean , you have people

9:11

who know more than me and

9:13

us , maybe given in this audience

9:15

here , but we

9:18

all try to just learn every day . I

9:20

think that's how I would describe it . There's

9:27

some interesting things . I mean from

9:29

my perspective as a product manager

9:31

. What

9:34

I'm placed in this world to do is to

9:36

basically

9:38

rank customer opportunities and

9:40

problems . That's my

9:42

primary job . Whether

9:45

or not AI can help solve some of those opportunities or problems that's my primary

9:47

job . Whether or not AI can help solve some of those opportunities or problems

9:49

great . So

9:52

that's what I'm about to

9:54

do , like reassess all

9:57

those things that I know about our customers , our

9:59

joint customers and partners , and how

10:01

can AI help those ?

10:05

Yeah , just when

10:08

you started speaking about the

10:10

dishwasher , it made me chuckle

10:12

and say how can you relate that to

10:14

why AI

10:16

was invented ? And I

10:19

had to look it up . I looked up , you

10:21

know why was the dishwasher invented

10:23

? So I thought it was pretty interesting

10:26

to share to the listeners . One

10:29

was to Josephine

10:32

Cochran , who invented the dishwasher

10:34

, and her

10:36

reasoning was to protect her china dishes

10:40

and she didn't want to

10:43

hand wash and then free

10:45

up time . And how

10:47

relatable is that with AI

10:49

? Is that we want

10:52

to free up

10:54

our time to do other things and

10:57

use AI to . In

10:59

this case , she

11:01

had noted that hand washing , avoiding

11:04

hand washing , she wanted to create a machine that could wash

11:06

dishes faster and more carefully

11:08

than she could . So , in a sense

11:11

, when

11:13

AI is invented , you

11:15

kind of want to have a

11:17

tool in this case an AI tool

11:19

to do other things for you , maybe

11:22

better than you can and

11:25

maybe more carefully in

11:27

feeding you information . I don't know , but

11:29

I thought that was pretty interesting .

11:31

The relatable component

11:34

there and that

11:36

makes total sense to me . That makes

11:38

sense in the sense that AI

11:41

is very good at paying attention to detail

11:43

that a human might overlook if

11:45

we're tired or

11:47

it's end of the day or early morning

11:49

. Even so

11:52

, there's so much relatable

11:54

things to what you just said that applies for AI

11:56

, or even just technology , I

11:58

mean , and automation . It's not just AI , because

12:01

IT is about

12:03

automating stuff . Ai just

12:05

brings another level of automation .

12:08

You could say it

12:12

is a beneficial tool . But , chris

12:14

, to go back to your point with the

12:16

invention of dishwasher and maybe even the invention

12:18

of AI , I think I

12:20

don't know the history of AI and I'm not certain

12:23

. If you know , I'm sure you could use AI to

12:25

find the history of AI . But is AI

12:27

one of those tools ? I have so many thoughts

12:29

around AI and it's tough to find

12:31

a way to get into unpack all of the

12:34

comments that I have on it . But

12:37

a lot of tools get

12:39

created or invented

12:42

without the intention of them

12:44

being invented . You

12:51

know it's sometimes you create a tool or you create a process or something comes

12:53

of it and you're trying to solve one problem . Then you realize that you

12:56

can solve many other problems by either

12:58

implementing it slightly different , you

13:01

know , working on it with another

13:03

invention or a tool that was created

13:05

. So where does it end

13:07

? And with

13:11

AI , I think we're

13:13

just I don't know if we'll ever or

13:15

we can even understand where it will go or where it will end

13:17

. We see how individuals are using it now , such

13:19

as creating pictures . Right , I'm looking at

13:21

some of the common uses of it outside of the analytical

13:24

points , points of it people creating pitches you

13:26

know a lot of your search engines now will primarily give you

13:28

the ai results of the search engines , which is

13:30

a summary of sources that they cite

13:32

. Uh , ai gets used

13:35

, you know , from that way , from

13:37

like the language model points of view , but then ai

13:39

also gets used from a technical point of view . Um

13:42

, I'm also reading . I started

13:44

reading a few weeks ago a book

13:46

uh , moral ai and how we get there

13:48

which is by pelican books and I think

13:50

it's borg , synod , armstrong and

13:53

contents I'm so bad with names which

13:55

also opened

13:57

up my eyes to ai

14:00

and how ai impacts

14:02

everybody in

14:05

the world .

14:07

I think it creates different iterations , right with

14:10

AI . You know , clearly

14:12

, you see AI

14:14

in almost practically anywhere

14:17

you had mentioned . You know creating

14:19

images for you and

14:23

started with that and then followed with creating

14:25

videos for you now and

14:28

and so much more , and then you

14:30

know , uh , sorted . You know I was trying

14:32

to . I mean , I was listening to your episode um

14:35

, you know , where does ai come into play in erp

14:37

and where does it go from there

14:39

? Right , I'm sure a lot of people

14:41

are going to create different iterations

14:43

of AI and Copilot

14:46

and Business Central , and

14:48

that is where I'm excited about

14:50

. We're kind of scratching the surface

14:52

in the ERP and

14:55

what else can it do for you in the

14:57

business sense ? Of course , there's different

14:59

AIs with M365

15:01

and all the other Microsoft ecosystem

15:03

product lines . What's

15:07

next for businesses

15:09

, especially in the SMB space ? I

15:11

think it's going to create a level

15:14

playing field for SMBs

15:16

to be able to compete better

15:19

and where they can focus more on strategy

15:21

and

15:24

be more tactical in the way they do business . So

15:26

that's where I'm excited about and and I think a

15:28

lot of us here in this call we're

15:32

the , I guess , curator and

15:35

and that's where we become

15:37

more of business consultants in a

15:39

sense of how you would run your business

15:41

utilizing all these Microsoft tools

15:43

and AI .

15:46

I think yeah .

15:46

I think , Go ahead .

15:48

Christian .

15:49

Okay , I think that we

15:52

see some processes

15:54

done by AI or agents

15:56

which we never

15:59

thought would be possible without doing

16:01

the human . What

16:03

was presented is really mind

16:06

what level of steps

16:08

and pre decisions

16:11

AI can make and offer a more

16:13

, better

16:18

result into the process until

16:20

a human needs to interact to

16:22

that . And I think that

16:24

will go further and further and further . What

16:28

I'm thinking is where is

16:31

the point where the human says okay

16:33

, there is a new point where

16:35

I have the

16:38

feeling that now I have

16:40

to grab into this

16:42

process because the AI is not

16:44

good enough and that point

16:47

is , or this

16:49

frontier is

16:51

, leveraged on and on

16:53

and on , something like that . But

16:57

to have this feeling , to have

16:59

in mind this

17:01

is the thing AI

17:04

cannot do . I

17:06

have to be conscious and

17:08

cautious and

17:11

I think , on the one hand side

17:13

, with AI we can

17:15

make more processes

17:17

, we can make more

17:20

decisions easily , and on

17:22

the other side , the temptation

17:25

is high that we just accept

17:27

what the AI is

17:29

prompting to us or offering us

17:31

. I like the concept of

17:33

the human in

17:35

the loop . So at

17:38

least the human at some

17:40

point in this process has to say , yes

17:43

, I accept what the AI is suggesting

17:45

, but

17:47

having more time to

17:50

process . More communication

17:52

is also critical

17:54

. Just to click yes , okay , okay

17:57

, okay . I

17:59

think we should implement

18:03

processes where we just

18:05

say , okay , let's look

18:08

at how we use AI here

18:10

and take a little bit back

18:12

and say , wow , what number

18:16

of steps AI can make for us . But

18:19

just think where it

18:21

just goes too far .

18:25

I think that's an interesting line of thinking

18:27

, christian , and I think so . Before

18:30

we go deeper , let me maybe just say

18:32

that some of

18:34

the stuff that we talk about in this episode like

18:37

, if nothing else is mentioned

18:39

, these are my personal opinions

18:41

and may not reflect the opinions

18:43

of Microsoft . Let's sort of get into product-specific

18:45

stuff , but I would

18:47

like to take sort of a product's eye view on what

18:49

you just said , which

18:52

is when we look at agents these days and

18:54

what an agent can do and what should

18:56

be the scope of a given agent and

18:58

what should be its name , and so now

19:01

we've released some information about

19:03

the sales order agent

19:05

and described how

19:07

does it work and actually

19:09

being fairly transparent about what it intends

19:11

to do and how it works , which I think is great

19:14

. We

19:17

actually start by drawing up in

19:21

the process today , before

19:23

the agent . How would this process look

19:26

? Where are the human interactions

19:28

between which parties ? Now

19:31

bring in the agent ? Now

19:34

, how does that human in the loop let's

19:37

say flow look like ? Are

19:39

there places where the human actually

19:41

doesn't need to be in the loop ? That's

19:44

the idea . Don't bring in the human unless it's need

19:46

to be in the loop . That's the idea . Don't bring in the human unless it's really

19:48

necessary or adds value . So that's the line , that's

19:51

the way that we think about it , to try

19:53

to really apply . You know , if

19:55

that A to Z process

19:58

can remove the human like

20:03

can automate a piece We've always been trying to automate

20:06

stuff right for many years . If

20:08

AI can do that better now , well

20:10

, let's do that . But of course

20:12

, whenever there's a risk situation

20:14

or wherever there's a situation where

20:16

the human can add value to a decision

20:18

, by all means let's bring in the human into

20:20

the loop . So that's the

20:23

way that we think about the agents and

20:26

the tasks that they should perform in whatever business

20:28

process . And to

20:30

your point , chris , I think

20:32

that the

20:35

cool thing about AI in

20:37

ERP , as in Business Central

20:40

these days , is that it

20:42

becomes super concrete . Like

20:45

we take AI from something that is very sort

20:47

of fluffy and marketing

20:50

and buzzwords that we all see online

20:52

and we make it into something that's very

20:54

concrete . So

20:57

the philosophy is that in BC unless

20:59

, of course , you're an ISV that needs to build something

21:01

on top of it , or a partner , a customer wants

21:04

to add more features AI

21:06

should be ready to use out of the box . You

21:09

don't have to create a new AI

21:11

project for your business , for your enterprise

21:14

to start leveraging AI ? No , you

21:16

just use AI features that are already there , immersed

21:19

into the UI and among all

21:21

other feature functions in Business Central and among

21:23

all other feature functions in Business Central

21:25

. So , because

21:27

small medium businesses , many of them don't even have the budget to

21:29

do their new AI project and hire

21:32

data scientists and what have you and

21:34

all these things create their own models . No

21:37

, they should have AI ready to use . So

21:39

that's another piece of our philosophy .

21:44

AI is . I look at that as more as AI as a function

21:46

, because if you have AI

21:49

as a function , you can get the

21:51

efficiencies . I think , to some

21:53

of the comments from the conversations that

21:55

we've had and the conversations that I've heard , you look

21:57

for efficiencies so that you can do

21:59

something else . People want

22:01

to use the word something else or something

22:04

that they feel is more productive and let

22:06

automation or AI or

22:09

robots I use

22:11

the word quote do the tasks

22:13

that are mundane or some

22:15

would consider boring or repetitive

22:17

. And

22:20

we do use AI on

22:22

a daily basis and a lot of the tools that we

22:25

have . To your point , Soren , that it's just embedded

22:27

within the application If you

22:29

buy a vehicle , a newer vehicle now , they

22:31

have lane avoidance , collision avoidance , all

22:33

of these AI tools that you

22:35

just get in your vehicle . You either turn it on

22:38

or turn it off , depending upon how

22:40

you'd like to drive , and

22:42

it works and it helps the , the function

22:44

, uh , be there for you . But

22:47

to kind of take a step back from

22:50

um ai

22:52

in that respect . But

22:54

a couple things that I come with ai

22:56

we . We talk about the vehicle . Um

22:58

, I'll admit I have a tesla . I

23:01

love the fsd and I used

23:03

it a lot and it just seems to

23:05

improve and improve and improve to

23:07

the point where I think sometimes

23:10

it can see things I

23:12

use the word see or detect things faster

23:15

than I can as a human right

23:17

Now . Ai may

23:19

not be perfect and AI makes mistakes

23:22

. Humans make mistakes . Humans get into car crashes

23:24

and have accidents right

23:26

for some reason , and we

23:28

have accepted that . But if AI

23:30

has an accident

23:33

, we find fault or find

23:35

blame in that process , instead

23:37

of understanding that . You

23:39

know , in essence , nothing is perfect , because

23:41

humans make mistakes too and we accept

23:44

it . Why don't we accept it when AI

23:46

may be a little off

23:48

?

23:51

That's such a great question and

23:54

the fact is , I think right now is

23:56

that to a point that we

23:59

don't accept it , like we don't give machines that

24:01

same benefit

24:03

of the doubt , or like

24:05

if they don't work it's

24:07

crap and we throw them out , like I mean that's like

24:10

, but humans like we , we're

24:12

much more forgiving , like we give them a second chance

24:15

. And oh , maybe I didn't teach you uh

24:18

well enough how to do it , or so

24:20

, but that's a good point and I , I , I

24:23

love your example with the Tesla . So I also

24:25

drive a Tesla , but I'm

24:27

not in the US , so I can't use the full self-driving

24:29

capability , so I use

24:32

the what do you call it ? The semi-autonomous , so it can keep

24:34

me within the lane . It reacts

24:36

in an instant if something drives

24:38

out in front of me much faster

24:41

than I can do . So I love that mix

24:43

of me being in control

24:45

but just being assisted by these

24:47

great features . That uh makes

24:50

me drive in a much safer way . Basically

24:52

, uh , I'm not sure I'm a proponent

24:55

of sort of full self-driving . I don't know

24:57

, I'm still torn about that , but

24:59

uh , that could lead us into a good discussion

25:01

as well , um I

25:03

think you have that trust because that I'm .

25:05

I'm the same way with brad , you know , I love , I

25:08

love it , um , as

25:10

as I , you know , continue to use

25:13

it . But in the very beginning I could not trust

25:15

that thing . I had my hand in the

25:17

steering wheel . Um , you know , a

25:19

white knuckle on on

25:21

the steering wheel . But uh , eventually

25:24

I come to accept

25:26

it and I was like , oh , that's a pretty good job

25:28

, uh , getting me around . Uh

25:30

, am I still cautious ? Absolutely , I

25:33

still want to make sure that I can quickly control

25:35

something if I don't believe it's doing the

25:37

right thing .

25:38

So I , I think , um , actually

25:41

my reason for not being

25:43

a sort of full believer in in sort of full

25:45

self-driving , like complete autonomy with

25:48

cars is is not

25:50

so much because I don't I mean , I actually

25:52

do trust the technology to a large extent

25:54

. It's more because of many of the reasons

25:57

that are now in that book that I pitched

25:59

to all of you that moral AI like

26:02

who has , like if something

26:04

goes wrong . And there's this example in the book

26:06

where , where an uber car

26:09

like you would think it was a volvo they

26:11

, they test an uber car , some self-driving

26:13

capabilities in some state and

26:15

it accidentally runs over a , a

26:18

woman who's who's passing the street in in

26:20

an unexpected place and it was dark

26:22

and things of that nature , and the driver

26:25

wasn't paying attention , and there was all these

26:27

things about who has the responsibility

26:30

for that end of the day . Was it the

26:32

software ? Was it the driver who wasn't

26:34

paying attention ? Was it the , the

26:37

government who allowed that car

26:39

to be on that road in the first place ? But

26:41

while testing it out all of these

26:43

things and if

26:46

we can't figure that out or

26:48

all those things need to be figured out first

26:50

before you allow a technology

26:53

loose like that , right , and so that and

26:56

I wonder if we can do that . If

26:58

we can , we

27:00

like

27:03

we don't have a good track

27:05

record of of doing that , uh

27:08

. So I wonder I I'm

27:10

I'm fairly sure the technology will , will

27:12

get us there , if we can live with the

27:14

uh , uh

27:17

when it doesn't work well . So what

27:20

happens if a self-driving car kills

27:22

20 people per year , or

27:25

cars multiple ? Um , can

27:28

we live with that ? What if 20 people is a lot better

27:30

than 3000 people from from human

27:32

drivers Like yeah , that is

27:35

.

27:35

I think in the United States there's 1.3

27:37

. I don't don't quote me on the statistics . I think

27:39

I heard it again with the all these

27:41

conversations about self-driving

27:44

and you know the Moralei

27:46

book and listen to some other tools . I

27:48

think in the United States is one point three million fatalities

27:50

due to automobiles a year . You

27:52

know I forget if it's a specific type , but it's

27:55

a lot . So , to get to

27:57

your point , you know not to focus

27:59

on the you

28:01

know , the driving portion , because a lot of topics

28:04

we want to talk about . Is

28:07

it safer ? In a sense , because you may

28:09

lose 20 individuals

28:11

tragically in an accident per

28:13

year , right , whereas before it was

28:15

a million because AI

28:18

? You know I joke

28:20

and I've had conversation with Chris talking about the Tesla

28:22

. I trust the FSD a lot driving around

28:24

here in particular , I trust the FSD a

28:26

lot more than I trust other people . And

28:28

to your point of someone losing

28:31

their life tragically , crossing in the

28:33

evening at

28:36

an unusual place and

28:39

having a collision with a vehicle , that

28:41

could happen with a person doing it as well

28:43

, and

28:46

I've driven around and the

28:49

Tesla detected something

28:51

before I saw it . So the reaction time is

28:54

a little bit quicker because if you're driving right

28:57

and it goes up to a couple

28:59

points I want to talk about , which I'll bring up to is

29:01

, you know , too much trust and de-skilling

29:03

. I want to make sure we get to those points . And

29:06

then also , if we're looking at analytics

29:08

, some you know harm bias as well , and

29:11

then also , if we're looking at analytics , some you

29:13

know harm bias as well , no-transcript

30:02

. And then to Christian's point and

30:04

even your point where the humans are involved

30:06

. Are the humans even capable

30:08

with the skilling ? Because you don't have

30:10

to do those tasks anymore

30:12

to monitor the AI ? You know

30:14

, if you look back , I'm going to go on a little tear in

30:16

a moment . In in education , when

30:18

I was growing up , we learned a lot of math

30:21

and we did not , you

30:23

know , use calculators . I don't even know when the

30:25

calculator was invented , but we weren't allowed

30:27

to . You know , they taught us how to use a slide rule . They

30:29

taught us how to use a slide rule . They taught us how to use even believe it or not , when I was

30:31

really young an abacus , and now

30:34

and then I could do math really , really well . Now

30:37

, with the , you know , ease

30:39

of using calculators , ease of using your phone

30:42

or ease of even using AI

30:44

to do math equations ? can

30:47

you even do math as quickly as you used to

30:49

? So how can you monitor a tool that's supposed

30:51

to be calculating math , for example ?

30:54

I , I , I think you're , I

30:56

mean , you have very good points about the like

30:58

. Just coming back to the car for a second , because

31:01

, uh , I mean , technology

31:03

will speak for itself and what it , what it's capable

31:05

of , I think . I think

31:07

where we have to take some decisions

31:09

that we haven't had to before

31:11

is when we dial up the autonomy to

31:14

100% and the car drives completely

31:16

on its own , because then

31:19

you need to be able to question how

31:22

does it make decisions ? And get

31:24

insights into how does it make decisions based

31:26

on what ? Who determines how

31:29

large an object has to be before the car

31:31

will stop if it runs ? So I

31:33

think back in the old days in Denmark , insurance

31:41

companies wouldn't cover if the object you ran over was smaller than a small

31:43

dog , something like that . So

31:45

who set those rules

31:48

? And the same thing for the technology

31:50

too Should I just run that pheasant

31:52

over or should I stop ? For the pheasant

31:55

? Those

31:57

kind of decisions . But if it's a human

31:59

driving in control , we can always just

32:02

point to the human and say , yeah , you need to follow

32:04

the rules , and here they are . But if it's a machine

32:06

, all kinds of things , and

32:09

eventually if the machine fails or we end up in

32:11

some situation where there's a dilemma who's

32:15

responsible , who's accountable and that

32:17

just becomes very hard questions . I

32:19

don't have the answer , but I think

32:22

when we dial up the autonomy to that level

32:24

, we need to be able to have you

32:26

know and we need to talk about what level of transparency

32:28

can I demand

32:30

as a user or as a bystander

32:33

or whatever ? So there's just so many

32:35

questions . That opens up , I think .

32:39

And if you are allowed

32:41

to turn off AI assistance

32:43

, will , at some point in

32:45

time , when a failure is occurring

32:47

, you be be responsible

32:50

for turning that assistance off .

32:53

That's a very good point .

32:55

Someone could say . So

32:57

you

32:59

have to keep in mind that with assistance

33:02

you're better . Like in

33:04

the podcast episode you mentioned

33:06

, a human together

33:08

with a machine is better than the machine . Other

33:12

ways you could say a human with

33:14

a machine is better than another human or

33:17

just a human . And I

33:20

think at some point in time , companies

33:22

who are looking for accountability

33:25

and responsibility will

33:27

increase the level of you

33:30

have to turn

33:32

on AI assistance . You

33:35

could imagine when you get into a car

33:37

that is recognizing you as

33:40

a driver your facial

33:42

expression or something like that that

33:44

it can recognize if you're

33:46

able to drive or not , and

33:48

then the question is will it allow

33:50

you to drive or

33:52

will it decide no , don't

33:55

touch the wheel , I will drive

33:57

, or something like that . Or if something

34:01

pops up you're not able to

34:03

drive , I decide that for you and

34:05

I won't start the engine . Will

34:08

you override it or not ? That

34:10

are those scenarios that pop

34:12

up in my mind . And and how will

34:14

you decide as a human when

34:17

you have something , uh , emergent

34:21

happening ? You have to drive someone

34:23

to the , to the hospital or something like that ? You

34:26

will override , but will the

34:28

system ask is it really

34:30

an emergency ? Or something like that ? You say

34:32

I just want to do this

34:35

. How are you

34:37

reacting in this moment ?

34:40

I think that's super interesting

34:42

. And coming back to the transparency

34:44

thing , one of my favorite examples

34:46

is if

34:51

I go to the bank and I need

34:53

to borrow some money , for

34:55

many years , and even before AI , there's

34:57

been some algorithm that

34:59

the bank person don't

35:02

even know about how it works , probably

35:04

, but can just see a red or green light

35:06

after I ask so

35:09

, okay , how much money do you want to borrow ? Oh , I want to borrow

35:11

100K . No , you can't

35:13

do that , sorry . Uh , machine says

35:15

no , right . And and

35:17

uh , even before ai

35:19

, if something is complex enough , uh

35:22

, it doesn't really matter if it's ai or not

35:24

. But in these sort of life impacting

35:27

situations , do I

35:29

have a right for

35:32

transparency ? Do I have a right to know

35:34

why they say no to lend

35:36

me money , for example ? The

35:39

same if I get rejected for a job interview

35:41

based on some decision made by an

35:43

algorithm or AI . These

35:45

are very serious situations

35:48

where that will impact my life and

35:52

of course , they don't go . You can't claim transparency

35:54

everywhere , but I think

35:56

there are some of these situations where , as

35:59

humans , we do have a right for transparency

36:01

and to know how do these things know ? And

36:03

there is a problem if the person who's conveying

36:07

the information to us . The bank bank person doesn't

36:09

even have that insight , doesn't even know how it works

36:11

. They

36:14

just push the button and then the

36:16

light turns red or green . So

36:21

that's yeah , but

36:23

again , so many questions , and

36:25

that's why I'm actually happy that today I

36:27

don't know if you saw it we released

36:29

a documentation article for BC

36:32

about the sales audit agent that

36:34

, in very detailed way , describes what

36:36

this agent does , what it tries to

36:38

do , what kind of data it

36:41

has access to , what kind of permissions

36:43

it has , all these things . I think that's

36:45

a very , very transparent way of describing

36:48

a piece of AI and I'm actually very , very

36:50

proud of that . We're doing that . Yeah

36:52

, just want to make that , doesn't make that segue

36:55

.

36:56

Yeah , it's

36:58

filling the need of humans

37:00

to know how

37:02

does the system

37:04

work or does the system make decisions ? To

37:06

proceed to the next step , Because

37:11

I think there's a need to have

37:13

a view on is what

37:16

has happened before and has

37:19

an influence on me as a human is

37:21

judged in a way that is doing

37:24

good for me or not ? Like

37:26

your example , what is evaluated

37:28

when you ask for a back credit

37:30

or something like that . And

37:34

having this transparency brings us

37:37

back to yes , I have an influence

37:39

on how it is needed , Because

37:42

I can override the AI , because

37:44

I can see where it makes

37:46

a wrong decision or wrong step

37:48

or something like that . Make

37:51

the wrong decision or wrong step or something

37:54

like that , Like I would do when I talk to my bank

37:56

account manager and say , hey

37:59

, does it have the

38:01

old address ? I moved already . Oh

38:04

no , it's not in the system . Let's

38:10

change that and then make another evaluation or something like

38:12

that . And I think this autonomy

38:14

for us as users to keep

38:16

this in play , that

38:18

we can override it or we can add

38:20

information , new

38:23

information , in some kind of way . We

38:26

can just do it when we know where

38:29

is this information taken . We

38:31

can just do it when we know where is this information

38:33

taken , how old is it and how is it processed

38:35

. So I like

38:37

that approach very much . I don't

38:39

think every user is looking

38:41

at it , but

38:45

as an ERP system owner like I'm

38:47

in our company as well needs

38:54

to have answers to those questions from our users when we use

38:56

these features , but it's true and just so

38:58

.

38:58

Yeah , coming back , just come back to the banking

39:00

sample just again . So the bank person

39:03

probably doesn't know if

39:05

their AI or algorithm takes into

39:07

account how many pictures they can find

39:09

with me on it on Facebook

39:11

where I hold a beer , like

39:13

would that be an

39:15

influencing factor on if

39:18

they want to lend me money ? So

39:20

all these things . But we just don't have that

39:22

insight and I think that's a problem

39:25

in many cases . You

39:27

could argue I don't know how the

39:30

Tesla autopilot

39:33

does its . You know whatever

39:35

influences it to take decisions

39:38

, but that's why I like

39:40

the semi-autonomous piece

39:43

of work right now .

39:45

No , it is , I think . But

39:48

listening to what you're saying , I do like

39:50

the transparency , or at least the understanding

39:53

. I like the agent

39:55

approach because you have specific functions

39:57

. I do like the transparency so that you understand

40:00

what it does , so you know

40:02

what it's making a decision on . So if you're going to trust

40:05

it in a sense or you want to use the information

40:07

, you have to know where it came from . Ai

40:11

or computers in general can process data much

40:13

faster than humans . So

40:15

, being able to go back to

40:17

your bank credit check example , it

40:20

can process much more information than

40:22

a person can

40:24

. I mean a person could come up to the same

40:27

results , but it may not be as

40:29

quick as

40:32

a computer can , as

40:34

long as that information is available to it . But

40:36

I do think for certain functions the transparency

40:39

needs to be there because in the case of bank credit

40:41

, how can you improve your credit

40:43

if you don't know what's being evaluated to maybe

40:46

work on or correct that ? Or , to

40:48

Christian's point , there may be some misinformation in

40:50

there that , for whatever reason is in there , that's impacting , so that . Or to Christian's point , there may be some misinformation in there that you

40:52

know , for whatever reason was in there , that's impacting so

40:54

that you need to force it

40:56

. Some other things , to

40:59

the point that Christian also made . You

41:01

know humans with a machine is

41:03

better than a human . You know

41:05

, potentially in some cases , because

41:07

the machine can

41:09

be the tool to help you do

41:11

something , whatever it may be . You referenced

41:13

the hammer before and I use that example a

41:15

lot . You have hammers , you have screwdrivers , you have air guns . Which

41:18

tools do you use to do the job ? Well , it depends on what you're

41:21

trying to put together . Are you doing some rough work on a

41:23

house where you need to put up the frame

41:25

, so maybe a hammer or an air gun will work , and

41:27

if you're doing some finish work , maybe you need a screwdriver

41:29

. You know , with a small screw to do something . So there

41:32

does have to be a decision made . And

41:37

at what point can AI make that decision versus a human make that decision ? And , to

41:39

your point , where do you have that

41:41

human interaction ? But

41:43

I want to go with the human

41:45

interaction of de-skilling , because

41:48

if you have all these tools that

41:50

we rely on . To go back to the calculator , and

41:52

you know we've all been

41:54

reading , you know I think we all read the same book

41:56

and I think we all listened to some of the same episodes

41:58

. But you look at pilots and

42:00

planes with autopilots right same

42:03

thing with someone driving a vehicle like , do you

42:05

lose the skill to ? You know ai

42:07

does so much portion of flying a plane . I didn't even really think about that

42:09

. You know AI does so much portion of flying a plane . I didn't even really think about that . You

42:12

know the most difficult or the most

42:14

most dangerous is what ? The taking off and landing

42:16

of a plane , and that's where AI gets used

42:18

the most . And then a human

42:21

is in there to take over in the event that AI fails

42:23

. But if the human isn't doing

42:25

it often right

42:28

, even with the reaction time , okay well , how

42:30

quickly can a human react , you

42:32

know , to a defense system ? Same thing , you

42:34

know , if you look at the Patriot missile examples , where you

42:37

know the Patriot missile detects a threat

42:39

in a moment and then will

42:41

go up and try to , you

42:44

know , disarm the threat . So

42:47

at what point do

42:49

we as humans lose

42:52

a skill ? Because we

42:54

become dependent upon these tools and we

42:56

may not know what to do in

42:58

a situation because we

43:00

lost that skill .

43:04

That's a good point . Sorry

43:06

, go ahead . No , it's a really good point .

43:08

Sorry , go ahead . No , it's a really good point . I like that example from

43:10

I think it was from the Moral AI book as well

43:12

, where there's this example

43:14

of some military people

43:16

that you know they sit in their

43:18

bunker somewhere and

43:21

handle these drones like

43:23

day in and day out and

43:25

, because they're so autonomous

43:28

, everything happens without their

43:30

. You know they don't need to be involved , but

43:37

then suddenly a situation occurs . They need to react in sort of a split second and take

43:39

a decision , and I think one of the outcomes was you know

43:41

, their manager says that

43:43

. Well , who can blame

43:45

them if they take a wrong decision at that point

43:47

? Because

43:52

it's three hours of boredom and then it's three

43:54

seconds of action . So they , they're just not feeling it . Where

43:57

, to your point , right , if they were like they're , they're

43:59

, they're being de-skilled for two hours

44:01

and 57 minutes and now there's three minutes

44:03

of action where everything happens . Right

44:06

, who can , who can

44:08

expect that they keep up the level of

44:10

you know , skills and what have you

44:12

if , if they're just not involved . So it's

44:14

super interesting point . Um

44:16

, yeah

44:19

, so many , so

44:21

many questions that it raises .

44:23

Uh this , it goes

44:25

on , it goes on , it goes on , it's , and

44:27

it is in that moral a book is , and it was the

44:29

patriotot missile example . Because the

44:32

Patriot missile had two failures

44:34

, one with a British jet and one with an American jet

44:36

shortly thereafter . And that's what they were talking

44:38

about is how do you put human

44:41

intervention in there , you know , to reconfirm

44:43

a launch ? Because in the event , if it's a

44:45

threat , it will use the word threat . How

44:48

much time do you have to immobilize

44:51

that threat ? Right , you may only have a second

44:54

to two . I mean , things move quickly in the . In the

44:56

case of the patriot missile , again , it was

44:58

intended to disarm , uh , you

45:01

know , and again , missiles that are coming at you

45:03

, that are being launched , you know , over the pond

45:05

, as they say , so they can take them

45:07

down , and that's the point with

45:10

that .

45:11

And if

45:13

I could step back for a second . You

45:17

know when we're having a conversation about the usefulness

45:19

of AI is based upon the source

45:21

that it has access to and

45:25

you know understanding where

45:27

it's getting its source from and

45:30

what access it has . If

45:33

you're limiting the source

45:35

that it can consume to

45:37

be a better tool , are we

45:40

potentially limiting

45:42

its capabilities as

45:45

well , because we wanna control

45:47

it so much , in

45:49

a sense , to where it's more focused , but

45:51

are we also limiting its potential

45:54

, right ? Yes , so

45:57

yeah , go

45:59

ahead , sorry .

46:01

Yeah , no , I think that's very well put

46:03

and I think that's a consequence

46:05

and I think that's

46:07

fine . I mean , just take the sales auto

46:10

agent again as an example . We

46:13

have railed it very

46:15

hard . We put many constraints

46:18

up for it , so we can only do a

46:21

certain number of tasks . We

46:23

can only do task A , b

46:25

and C , d , e , f . It cannot do

46:27

. We had to set some guardrails

46:29

for what it can do . It's not

46:32

just about and I think this is a misconception

46:34

sometimes people think about agents and say here's

46:37

an agent , here's my keys

46:39

to my kingdom . Now

46:41

, agent , you can just do anything in

46:43

this business , in this system , and user

46:46

will tell you what to do or we've given you a task

46:48

. That's not our

46:50

approach to agents . In BC . We

46:52

basically said here's an end-to-end process

46:55

or a process that has sort of a natural beginning

46:57

and a natural ending . In

46:59

between that process you can trigger

47:01

the agent in various places , but the agent

47:03

has a set instruction

47:06

. You

47:08

receive inquiries for products

47:10

and eventually you'll create a sales order

47:13

. Like everything in between there could be

47:15

all kinds of you know human in the loop and

47:17

discussions back and forth , but

47:19

that's the limit of what that agent can do and

47:22

that's totally fine . It's not fully

47:24

autonomous . You can't just now go and say , oh

47:27

, by the way , buy more inventory

47:29

for our stock , that's

47:32

out of scope for it , and at that

47:34

point I think that's

47:36

totally fine . And it's

47:38

about finding those good use cases where there

47:41

is a process to be automated , where the

47:43

agent can play a part , and

47:46

not about just

47:48

creating a let's call it a super agent that

47:50

can do anything

47:53

with like . So I

47:55

think that's it's a very natural development

47:57

.

47:58

So you don't aim

48:00

for a T-shape profile

48:03

agent like it is in many

48:05

job descriptions Now . You want a T-shape

48:07

profile employee with

48:10

a broad and deep knowledge . We

48:13

as human can develop this , but

48:15

the agent

48:18

approach is different . I

48:20

would more say it's not

48:22

limiting the agent or

48:24

the AI of the

48:27

input or the capabilities . It is more

48:29

like going more deep , having

48:32

deep knowledge . In this specific

48:34

functionality , the AI agent

48:37

is assisting . That

48:39

can be more information

48:41

and it can go deeper than

48:43

a human can be

48:45

. For example , I was

48:47

very impressed by one

48:49

AI function I

48:52

had in my future

48:55

leadership education . We had an alumni

48:57

meeting in September

49:00

and the company set up an AI

49:02

agent that is behaving like

49:04

a conventional business

49:06

manager . Because we

49:08

learn how to set

49:11

up businesses differently

49:14

and when you have something new

49:16

you want to introduce to an

49:18

organization , often you

49:20

are hit by the cultural barriers

49:23

and just to

49:25

train that more

49:27

without humans , they invented

49:30

an ai model where

49:32

you can put your ideas in and you

49:34

have a conversation

49:36

with someone who has traditional

49:39

tayloristic business

49:41

thinking and something like that

49:43

. So you can train how

49:46

you um put your ideas

49:48

to such a person and

49:50

what will the reactions will

49:52

be just to train your ability

49:54

to be better

49:57

when you place these new ideas

49:59

to a real person in a traditional

50:01

organization or something like

50:03

that and that

50:05

had such a deep knowledge about

50:08

all these methodologies and thinking

50:10

and something like that . I

50:12

don't know who

50:14

I could find to

50:16

be so deep in this knowledge

50:18

and have exactly this profile

50:21

, this

50:28

deep profile that I needed to train

50:30

myself on .

50:31

That is a really interesting use case . I think then it becomes

50:33

to continuing a conversation about

50:35

maybe there's a misconception

50:37

or misunderstanding in the business

50:39

space , because right now , you

50:42

know , I've had several conversations

50:45

where AI is going to solve their problems

50:47

. Ai is going to solve their business

50:50

challenges

50:52

, but they , you

50:54

know , from a lot of people's perspective

50:57

, it's just this one entity of

50:59

, like it's going to solve all my business

51:01

problems , whereas for

51:03

us engineers , we understand that you

51:05

can have specific AI

51:07

tool that would solve a specific

51:09

problem or a specific process

51:11

in your business . But right now a

51:13

lot of people believe , like I'm just going to install

51:16

it , it's going to solve everything for me , and

51:18

so not realizing that there are different

51:20

categories for that , you know different areas

51:22

and I

51:25

think having these kinds of conversation in

51:27

hopes that know it's it's not just a one-size-fit-all

51:30

um kind of solution

51:32

out there , yeah , and indeed , and when

51:34

you see , like the um

51:37

industrial work developed

51:39

in the first phases , it's like going

51:41

back to um having

51:44

one person just

51:46

fitting is a

51:48

bold or a school or something like that

51:50

.

51:51

That is the agent at the moment , just

51:53

one single task it can do . But

51:56

it can do

51:58

many , many things into this

52:00

task at the moment and

52:04

what I think it will

52:07

take some time to develop is

52:10

developing this T-shape

52:13

from the ground of the T to have this

52:15

broad knowledge and broad capabilities

52:18

out of one agent

52:20

, or the

52:22

development of the network of agents . So

52:24

in some sessions in Vienna that was presented

52:27

, the team of agents , that

52:29

was presented , the team of agents . So

52:32

you have a coordinator that coordinates the agents and then

52:34

brings back the proposal from the agent

52:36

to the user or something like that . That

52:38

will look like the

52:41

one agent can do all of these capabilities

52:44

for the user . That is presented

52:46

. But in the deep functionality

52:50

there is a team of agents

52:52

and a variety of agents doing

52:54

very specific things .

52:57

I like that case . It goes to

52:59

, chris , to your point of sometimes

53:03

it's just a misunderstanding of what AI is

53:05

, because I think there's so many different levels of

53:07

AI and we talked about that before

53:09

. You know what is machine learning , what is large language

53:11

models . I mean , that's all in AI . A

53:13

lot of things you know can

53:15

fall into AI . But to the point of the

53:17

agents to go into ERP software

53:19

, even Christian , to your point , maybe even

53:21

in an assembly line or manufacturing

53:24

, I'd like the agents in

53:26

the business aspect to

53:28

have a team of agents

53:30

together so they all do specific

53:32

functions . To Soren's point of where

53:35

do you have some repetitive

53:38

tasks or some precision tasks , or

53:41

even , in some cases , some skilled tasks that

53:43

need to be done , and then you can chain

53:45

them together . Because even if you look at an automobile

53:48

we talked about an automobile there isn't

53:50

an automobile , that just appears . You

53:53

have tires , you have engines

53:55

, you have batteries , you have right

53:57

. The battery provides the power , the wheel provides

54:00

, you know the , the ability

54:02

to easily move right . The engine

54:04

will give you the force to push . So putting that all

54:06

together see , this is how I start to look at putting

54:08

that all together now gives you a vehicle

54:11

. So the same thing if you're looking at erp software

54:13

. That's why when I first heard about the agent

54:15

approach when we talked some months ago , soren , that

54:18

having an agent for sales orders

54:20

or having an agent for finance

54:23

or having an agent for purchase orders or

54:25

something , a specific task , you

54:27

can put them all together and then use the

54:29

ones you need and

54:31

then have somebody administer those agents

54:33

, so you have like an agent administrator .

54:35

That is where

54:40

the human comes back into the loop

54:42

, because at some point you have to

54:44

put these pieces together . I think

54:46

at the moment , this is the

54:48

user that needs to do this , but

54:51

this will develop further

54:53

in the future . So

54:56

you have another point where you end

54:59

in or where you need

55:01

ideas or something like that , because

55:03

that is also what I learned and found

55:05

very interesting . When

55:07

you see an AI

55:10

suggesting something to you

55:12

, this feeling

55:14

this is a fit for

55:17

my problem is inside your

55:19

body and at the moment

55:21

, you cannot put this into a machine . So

55:24

the idea , if the suggestion is right

55:26

and you decide to take it and

55:28

to use it , you

55:31

need a human to make this

55:33

decision , because you need the human

55:35

body , the brain and everything together

55:37

seeing and perceiving this , to

55:40

make this decision if it is wrong

55:43

or good for

55:45

this use case .

55:48

I think that depends

55:51

a bit Christian , if I may . So

55:54

there are places where , let's say

55:56

, one AI could you

55:58

could give it a problem to tackle and it will come

56:00

with some outcomes . And there could

56:03

then be another AI and

56:05

now I use the term loosely but another process

56:07

that is only tasked

56:09

with assessing the output

56:11

of the first one within

56:14

some criteria , within some

56:16

aspects . So that has been , say

56:19

loosely , now trained , but its only

56:21

purpose is to say , okay , give

56:23

me the outcome here and

56:26

then assess that with complete fresh

56:29

eyes like it was a different person

56:31

. Of course it's not a person and we should

56:33

never make it look like it's a person but

56:36

one machine can assess

56:38

the other .

56:38

Basically , that's what I'd say to a certain

56:41

degree , right , if

56:45

we can frame the problem , right

56:47

, yeah , and you had mentioned about from the human aspect

56:49

, to take over and said you know that's wrong

56:51

. Right , like , oh , it's wrong

56:53

, I know it's wrong , I'm going to take over . It

56:57

reminds me of a story when I did a

57:00

NAV implementation a while back

57:03

where we had demand forecasting

57:05

and

57:11

when we introduced that to the organization it does like tons of calculation and it's going

57:13

to give you a really good output of what you need based

57:16

upon information and data that you have

57:19

. And I had this individual

57:21

person that I was working with , or that

57:23

person was working for this organization , where that's

57:26

not right , that's wrong

57:28

, and I would

57:30

ask can you tell me why it's

57:32

wrong ? I'd love to know

57:34

, like , how are you feeling ? Like , what

57:36

made you feel like it was wrong ? Do you have

57:38

any calculations ? No , I just

57:40

know it's wrong because typically we do

57:42

it , you know we , typically it's this number right

57:45

, but they couldn't prove it . So

57:47

that's also a dangerous component

57:50

where a person could take over

57:53

and then whatever decision , whatever they

57:55

feel like it's wrong , it

57:57

could . Where they think it's wrong , they

57:59

can also be wrong . Right

58:01

, it's just like the human aspect of

58:03

it . But , but they can . But

58:06

they can .

58:07

Yes , but they can , yeah , yeah

58:09

and I think I mean and that . So

58:11

the first time when I learned

58:13

more about sort of ai , like these recent years

58:15

, was some eight , nine

58:17

years ago when we we did some of the classic

58:19

sort of machine learning stuff for some

58:22

customers and what

58:24

was an eye-opener for me was that

58:26

it didn't have to be a black box . So back then

58:28

, let's say , you had a data

58:30

set . I think the specific customer wanted

58:32

to predict which

58:35

of their subscribers would churn

58:37

right , and

58:39

there was a machine learning model for that on

58:42

Azure that they

58:44

could use for that . I don't know the specific

58:46

name of it and the

58:48

data guy that helped us one

58:51

of my colleagues from Microsoft back then showed

58:55

them data because they had their

58:57

ideas on what were the influencing factors

59:00

that made consumers churn . These

59:02

were , these were magazines that

59:05

they were subscribing to , and

59:07

when he told them , show them

59:09

the data , and then said

59:11

uh , and showed

59:13

them because they could do that with with the

59:15

machine learning tools they could , he could show them

59:17

these are the influencing factors

59:20

, like actually determine

59:22

based on the data that you just see and

59:26

he had validated against their historic data

59:29

. They were just mind-blown

59:31

. So it turned out I'm just paraphrasing

59:33

now that people in the western

59:35

part of the country were the ones

59:37

who churned the most . So the geography was

59:39

the predominant influencing factor

59:41

to predict churn . They

59:45

were just mind-blown because they had never

59:47

seen that data . They had other

59:49

ideas of what it means to churn . Like to your

59:51

point , chris , like . But that was just so

59:53

cool that we could bring that kind of transparency

59:56

and say this is how the model calculates

59:58

, these are the influencing factors that it has

1:00:00

found by looking at the data

1:00:02

. So I just thought that was a great

1:00:04

example of bringing that transparency when humans

1:00:07

, like you say , are just

1:00:09

being stubborn and saying no , it doesn't

1:00:11

work , it's not right .

1:00:15

That's definitely another factor , because we've

1:00:19

all come into those situations where that just doesn't feel

1:00:21

right and in some cases it

1:00:23

could be correct .

1:00:25

But it depends on the skills . That's

1:00:27

what I want to go back to is the skills

1:00:29

. It's the skills .

1:00:31

How , if

1:00:33

we're going to keep creating AI

1:00:35

tools to help

1:00:37

us do tasks

1:00:40

okay , one

1:00:42

, I'm going to go off

1:00:44

on a tangent a little bit . One how do

1:00:46

we ensure we have the skills to

1:00:48

monitor the AI ? How

1:00:50

do we ensure that we have the skills to

1:00:53

perform a task ? Now I understand . The

1:00:56

dishwasher Chris you talked about was invented . Now

1:00:58

we don't have to wash dishes manually

1:01:01

all the time to save us time to do other

1:01:03

things . We're always building these tools

1:01:05

to make things easier for us and

1:01:07

, in essence , up the required skill

1:01:09

to do a function , saying we need to work on more valuable

1:01:12

things . Right , we shouldn't have to

1:01:14

be clicking post all day long . Let's

1:01:17

have the system do a few checks

1:01:19

on a sales order . If it meets those checks

1:01:21

, let the system post it . But

1:01:24

when is there a point

1:01:26

where we lose the ability to have

1:01:28

the skill to progress forward ? And

1:01:31

then with this , with all of these tools that help

1:01:33

us do so much , because now

1:01:35

that we have efficiency with tools

1:01:37

, oftentimes it takes

1:01:40

a reduction of personnel . I'm not trying to say

1:01:42

people are losing their jobs . It's going to take a reduction

1:01:44

of personnel to do a task . Therefore

1:01:48

, relieving the dependency on

1:01:50

others . Humans are communal . Are

1:01:52

we getting to the point

1:01:54

where we're going to lose skill and

1:01:56

not be able to do some complex tasks

1:01:58

because we rely on other tools ? And

1:02:01

if the tools

1:02:03

are to get more complex and we

1:02:05

need to have the skill to determine that complexity

1:02:07

, if we miss that little middle

1:02:09

layer of all that mundane building

1:02:12

block stuff , how do we have the skill

1:02:14

to do something ? And two , if

1:02:16

I can now I

1:02:18

see AI images , I see AI

1:02:21

videos being

1:02:23

created all the time . It does a great job . Before

1:02:26

we used to rely on artists

1:02:28

, publishers , other individuals

1:02:30

to create that content for the

1:02:35

videos , for

1:02:37

brochures , pictures , images , the

1:02:40

B-roll type stuff we'll call it . If

1:02:42

we don't need any of that stuff and we're doing it all

1:02:44

of ourselves , what are we doing to us

1:02:46

to be able to work together as a species

1:02:48

if now I can do all the stuff myself

1:02:51

with less people ? So I have many points

1:02:53

there . One , it's the complexity of

1:02:55

the skill . And how do we get that skill if

1:02:57

we immediately cut out the

1:03:00

need , for we no longer need someone

1:03:02

to put the screw on that bolt

1:03:04

. As you pointed , christian , we need someone to come in

1:03:06

and be able to analyze these complex results

1:03:09

of ai . But if nobody

1:03:11

can learn that by

1:03:13

doing all those tasks , what does that give

1:03:15

us ? So that's my little , so

1:03:17

two points so what is ?

1:03:19

yeah , no , that's great , great questions

1:03:21

. So what you're saying is how do we

1:03:23

determine if this car is built right

1:03:25

if there's no drivers left to

1:03:28

to to test it , like no

1:03:30

, no one has the skill to drive anymore . So

1:03:32

how ? How can they determine if this car is built

1:03:34

up to a certain quality standard and what have

1:03:36

you ? Well , the other answer would be

1:03:39

you don't have to because it

1:03:41

will drive itself . But until we get that point

1:03:43

, like in that time

1:03:46

in between , you need someone

1:03:48

to still be able to validate and probably

1:03:50

for some realms of our

1:03:52

work and jobs and society , you will

1:03:54

always need some people to validate . So what do you do ? I

1:03:56

think those are great questions and

1:03:59

I certainly don't have the answer to it .

1:04:01

I would say I've had

1:04:03

this conversation with Brad for a couple

1:04:05

of years , I think him and I , you

1:04:07

know , we just we love where I

1:04:11

love where AI is coming and I pose

1:04:13

the question about , you know , is AI becomes a necessity

1:04:15

for the survival of humanity . Becomes

1:04:17

a necessity for the survival of humanity

1:04:20

Because , as

1:04:22

you all pointed out , that

1:04:24

eventually you'll lose some of those skills

1:04:26

because you're so dependent . Eventually

1:04:29

you'll lose it . And I've had

1:04:31

tons of conversation Right

1:04:33

now we don't need AI . We

1:04:37

don't need AI for the survival of humanity , but

1:04:40

as we become more dependent

1:04:43

, as we lose some of those

1:04:45

skills , because we're giving it to AI

1:04:47

to do some tedious tasks sometimes it

1:04:49

could be in the medical field

1:04:52

or whatnot it

1:04:54

becomes a necessity in the

1:04:56

future . It will eventually

1:04:58

become a necessity in the future for humanity's survival

1:05:01

, but we're forcing it

1:05:03

. Right now we don't need it .

1:05:03

We are forcing the dependency by

1:05:06

losing this Because . I'm not saying it's right or wrong

1:05:08

, but I'm listening to what you're saying , saying

1:05:10

that we are going to be dependent

1:05:12

on machine for the

1:05:15

survival of the human race . I

1:05:18

mean , humans have been around for how long ?

1:05:23

But we're already dependent on machines . Right , we've been around for how long ? But we're already dependent

1:05:25

on machines . Right , we've been there for a long time . We're forcing ourselves to be dependent

1:05:28

upon it .

1:05:29

That's why I use the word machine , because

1:05:31

we force ourselves to

1:05:33

be dependent upon that right

1:05:35

. We force ourselves to lose

1:05:38

the skill or use something so

1:05:40

much that it's something that we must have to

1:05:43

continue moving forward

1:05:45

.

1:05:47

Yeah , my point was that that's not

1:05:49

new . I mean , we've done that for 50

1:05:52

years like force dependency

1:05:55

of some machines , right ? So without them we wouldn't

1:05:58

even know where to begin where to do

1:06:00

that task . So AI is just probably

1:06:03

accelerating that in

1:06:05

some realms now , I think

1:06:07

.

1:06:07

Yeah , it is , Because , you know , as

1:06:09

humans' desire is to improve

1:06:12

quality of life , expand

1:06:14

our knowledge and mitigate

1:06:16

risk . It's not improving quality of life

1:06:18

.

1:06:18

It's to be lazy ? I hate to tell

1:06:20

you it's . Humans take the path of least resistance

1:06:23

and I'm not trying to be there's a little levity

1:06:25

in that comment . But

1:06:27

why do we create the tools to do the things

1:06:29

that we do ? Right ? We create

1:06:31

tools to harvest fruits

1:06:34

and vegetables from the farm , right

1:06:36

, so we can do them quicker and easier and

1:06:39

require less people , right

1:06:41

? So it's not necessarily

1:06:43

, you know , we do it because

1:06:45

to make things better . We do it because

1:06:47

, well , we don't want someone to have to go

1:06:50

to the field and , you know , pick the

1:06:52

cucumbers from the cucumber vine , right

1:06:54

, we want , you know , they shouldn't have to do that , they

1:06:57

should do something else . We're kind of , in my opinion , forcing

1:06:59

ourselves to go that way . It is necessary

1:07:01

to harvest the fruits and the vegetables

1:07:03

and the nuts to eat , but

1:07:06

, you know , is it necessary

1:07:08

to have a machine do it ? Well , no , we just said it would

1:07:10

be easier , because I don't want to go out

1:07:12

in the hot sun all day long and you

1:07:14

know harvest .

1:07:16

You can do the dishes by hand if you like , right

1:07:18

yeah ?

1:07:20

If you like , yeah , if you choose

1:07:22

to . No one wants to . No one wants to do the dishes

1:07:24

.

1:07:24

trust me I will never live in a place without a dishwasher

1:07:26

. I mean , it's the worst

1:07:29

that can happen .

1:07:31

It is , and the pots and the pans forget

1:07:33

it right .

1:07:35

If you take this , further at

1:07:38

some point in time . If you have a new colleague

1:07:40

and you have to educate him or her

1:07:43

, do

1:07:46

you educate him to make these steps

1:07:48

the sales order agent is doing by

1:07:51

him or herself , just

1:07:54

to have the skill to know

1:07:56

what you're doing . Or if

1:07:59

you are saying , just

1:08:01

push the button .

1:08:06

Yeah , but I think what ? Eventually , as

1:08:09

you continue to build upon these

1:08:11

co-pilots in AI , eventually you just

1:08:13

have two ER pieces and talk to each other . And

1:08:16

then what then ? Where

1:08:19

are we then ?

1:08:23

Yeah , super interesting . What then ? Where are we then ? Yeah , super

1:08:25

interesting . I mean , who knows ? I

1:08:28

think it's so hard to predict where we'll

1:08:30

be even just in 10 years .

1:08:34

I don't think we'll be able to predict where we'll be in two years , I

1:08:37

think it's . Will

1:08:40

we ever be able to press a button Like

1:08:42

right now ? I can create video images

1:08:45

and still images . I'm using that

1:08:47

because a lot of people relate to that , but I can

1:08:49

create content , create things . I've

1:08:52

also worked with AI

1:08:54

from programming in a sense

1:08:56

, to create things . I was listening to a podcast the other

1:08:58

day . In the podcast they said

1:09:00

within 10 years , the most common

1:09:02

programming language is going to be the human language . Because

1:09:06

it's getting to the point where you can say create

1:09:08

me this . It needs to do

1:09:10

this , this and this , and an

1:09:12

application will create it , it will do the test

1:09:14

and produce it . You wake up in the morning and now you have an app

1:09:17

. So it's going to get to the point where what

1:09:20

happens now ? Let's move fast forward a little bit , because

1:09:22

you even look at github , co-pilot

1:09:24

for coding , right . You look at the sales agents

1:09:26

chris's point erp systems can just talk

1:09:28

to each other . What do you need to do ? Is

1:09:31

there going to be a point where that's what I was getting

1:09:33

at where we don't need other people because

1:09:35

we can do everything for ourselves ? And

1:09:38

then how do we survive if we don't

1:09:40

know how to work together

1:09:42

because we're not going to need to ?

1:09:45

that is so how we go

1:09:47

yeah , I'm sorry .

1:09:49

Sorry , now , that's so . To go to your point , how

1:09:51

is ai going to help progress

1:09:53

, the human civilization

1:09:55

, right , or the species , if

1:09:57

we're going to get to the point where we're

1:09:59

not going to need to do anything , we're all just going

1:10:02

to sit in my house because

1:10:04

I can say make me a computer and

1:10:07

click a button , it will be , you know

1:10:09

there and

1:10:12

that's you know where I come

1:10:14

from with I would in that

1:10:16

other podcast show that you mentioned , where I

1:10:18

quote james burke when he says that we

1:10:20

will have these nanofabricators , that in

1:10:22

60 years , everyone will have everything they

1:10:24

need , and just produce it from air , water and

1:10:27

dirt .

1:10:27

Basically right , so and uh

1:10:29

, so that that's the end of scarcity . So all the stuff

1:10:31

that we're thinking about right now are just temporary

1:10:34

issues that we don't need to worry about

1:10:36

in 100 years . So that

1:10:38

that's just impossible to even imagine . But

1:10:40

because , as one of you said just

1:10:42

before , we'll probably always just move the needle and

1:10:45

figure out something else to desire , something

1:10:48

else to do . But I

1:10:50

think it is a good question to ask but

1:10:52

what will we do with this productivity that

1:10:54

we gain from AI ? Where

1:10:57

will we spend it ? So now you're a company , now you

1:10:59

have saved 20% cost because you're a company . Now you save 20%

1:11:01

cost because you're more efficient in some

1:11:03

processes due to AI or

1:11:05

IT in general . What will

1:11:07

you do with that 20% ? Do

1:11:11

you want to give your employees more time off ? Do you

1:11:13

want to buy a new

1:11:15

private jet ? I don't know . You

1:11:18

have choices right , but

1:11:21

as a humanity , I definitely personally my's . Uh , you have choices , right , um , but as a but as a humanity

1:11:23

, I definitely . I personally , my personal opinion is I

1:11:26

mean , I would welcome a future

1:11:28

where we would , where we could work less

1:11:30

, where we could have machines to do

1:11:32

things for us . But it requires that we have a conversation

1:11:35

, start thinking about how will we interact

1:11:37

in such a world where we don't

1:11:39

have to work the same way we do today . What ? What will

1:11:41

our social lives look like ? Why do we need

1:11:43

each other ? Do we need each other ? We

1:11:46

are social creatures , we are communal creatures . So

1:11:49

, yes , I think we do . But how , what

1:11:52

will that world look like ? I think this keeps

1:11:55

me up at night sometimes .

1:12:04

I can't imagine , and nor did I imagine , there'd be full self-driving vehicles within a short

1:12:06

period of time , as it had to . I mean , I think , as you made a great

1:12:08

point , soren , I don't think anyone can know what

1:12:11

tomorrow will be or what tomorrow will bring

1:12:13

with this , because it's advancing

1:12:16

so rapidly . And go back to the points I said I

1:12:18

had mentioned you talked about the podcast

1:12:21

with James Burke , which was a great podcast as well

1:12:23

too . That was the You're

1:12:25

Not so Smart episode I think

1:12:27

it was 118 on connections , which

1:12:29

talked a lot about that . And

1:12:32

yes , it was a great episode . That's another

1:12:34

great podcast , and a

1:12:37

lot of this stuff is going to be building blocks that

1:12:39

we don't even envision what it's going to build

1:12:41

. You know , look at the history of the engine . You look at the history

1:12:43

of a number of inventions . They

1:12:45

were all made of small little pieces

1:12:47

. So we're building those pieces now . But

1:12:50

also our mind is going to need to be I

1:12:52

use the word stimulated . If we're going to get to the

1:12:54

point where we don't have to do anything , how

1:12:57

are we going to entertain

1:12:59

ourselves ? We're

1:13:04

we going to entertain ourselves ? We're always going to find something else right to have to do , but

1:13:06

is there going to be a point where there is nothing else because

1:13:08

it's all done

1:13:10

for us ?

1:13:12

yeah , just want to comment on that one

1:13:14

thing . You said there like that you referenced that

1:13:16

no one , no one just

1:13:18

imagined the car , like you

1:13:20

. You know , people did stuff

1:13:22

, invented stuff , but suddenly some other people

1:13:25

could build on that and invent other stuff and

1:13:27

then eventually you had

1:13:29

a car , right ? Or anything else

1:13:31

that we know in our life . And

1:13:33

I think James Berg also says that innovation

1:13:36

is what happens between the disciplines

1:13:38

, and I really love that

1:13:40

. I mean , just look at agents today

1:13:42

. Like four years ago

1:13:44

, before LLMs were such a big

1:13:47

thing . I know they were in a very niche

1:13:49

community , but with sort

1:13:51

of the level of LLMs

1:13:53

today , no one

1:13:55

said let's invent LLMs

1:13:57

so we could do agents . No

1:13:59

, I mean , LLMs was invented Now because we have LLMs , so we can do agents

1:14:02

. No , I mean , llms was invented Now because we

1:14:04

have LLMs . Now we think , oh , now we can

1:14:06

do this thing called agents and

1:14:08

what else comes to mind in six months , right ? So

1:14:11

it just proves that no

1:14:13

one has this sort of five-year plan of

1:14:15

, oh , let's , in five years , do this and this . No

1:14:18

, because in six months someone will have invented

1:14:21

something that , oh , we can use

1:14:23

that and oh , now we can build this

1:14:25

entirely new thing . So that's what's just

1:14:28

super . It's both super exciting , but it's also

1:14:30

a bit scary . I mean I can , I can speak

1:14:32

for as as a product

1:14:34

developer . It's definitely

1:14:36

challenged me to rethink

1:14:39

my whole existence as a product

1:14:41

person , because now I

1:14:44

don't actually know my toolbox

1:14:46

anymore . Two years

1:14:48

ago I knew what AL could do Great

1:14:51

. I knew the confines of what we could build

1:14:53

. I knew the page types in BC and stuff

1:14:55

. So if I had a use case , I

1:14:57

could visualize it and see how we can

1:14:59

probably build something . If we need a new

1:15:02

piece from the client , then we could talk

1:15:04

to them about it and we can figure that out . But

1:15:06

now I don't

1:15:09

even know if we can build it until we're

1:15:11

very close to having built it . I mean , so

1:15:13

it's . There's so much experimentation that

1:15:16

, yeah , we're building

1:15:19

the airplane where we're flying it in that sense , right

1:15:21

and so that also challenges our whole testing

1:15:24

approach and testability and frameworks . But so

1:15:26

, which is super exciting in itself

1:15:28

, so it's just a mindset change

1:15:30

, right , um , but , but definitely

1:15:32

challenge your product people oh

1:15:35

, it definitely does .

1:15:36

I I think uh ai is

1:15:39

um , it's definitely

1:15:41

changing things and it's here to stay

1:15:43

. I guess you could say . I'm just wondering

1:15:45

, you

1:15:48

know . I say I think back of a movie

1:15:50

was it from the 80s , called Idiocracy . You

1:15:54

know , if you haven't watched it it's a mindless

1:15:56

movie , but it is . It's the same type of thing where

1:15:58

a man from the past goes into the future

1:16:01

movie , but it is . It's

1:16:03

the same type of thing where a man from the past goes into the future

1:16:05

and you know what happens to the human species in the future

1:16:07

and how they are . It's pretty comical

1:16:09

. It's funny how some of these movies are

1:16:12

some of these circling back . Yeah , they

1:16:14

circle back , you know with .

1:16:16

You know star trek , star wars I'm

1:16:18

wondering when we will be there .

1:16:25

That already happened . I just hope we won't get to the

1:16:27

state where I think you said that cartoon or

1:16:30

that animated movie Wall-E where

1:16:34

the people are just lying back

1:16:36

all day and eating and their bones

1:16:38

are deteriorating because they don't use

1:16:40

their bones and muscles anymore . So the skeleton sort

1:16:43

of turns

1:16:45

into something like they just become like wobbly

1:16:48

creatures that just lie there .

1:16:50

As I don't know seals

1:16:52

, or consuming what

1:16:56

was really interesting with Back to

1:16:58

the Future is this thing here , because

1:17:02

Doc Brown made

1:17:04

this time

1:17:06

machine using a

1:17:09

banana to

1:17:11

have the energy of

1:17:15

1.2.1 gigawatts

1:17:17

or something like that . You

1:17:19

don't have to wait for a thunderstorm to

1:17:21

travel into time a bit . This

1:17:23

idea was mind-blowing back then and

1:17:26

and I I'm

1:17:28

dreaming of using my

1:17:30

using free time as as a human

1:17:32

to to make this leaps . Because

1:17:34

we are . We

1:17:37

have this scarcity in resources and

1:17:40

, even if this goes further and further

1:17:42

and further , I assume that we

1:17:44

don't have enough resources to make this

1:17:46

machine computing power to

1:17:49

fulfill all that . I

1:17:51

think there will be limitations

1:17:53

at some point in time , and

1:17:56

most of what

1:17:58

is AI freeing us up is to

1:18:00

have ideas on how are we using

1:18:02

our resources that is sustainable

1:18:05

.

1:18:07

I like that . I

1:18:09

have no way to say what

1:18:11

you fear will become

1:18:13

true or not , but I like the idea

1:18:16

of using whatever productivity

1:18:18

we gain for more sort of

1:18:20

humanity-wide purposes , and I also

1:18:22

hope that whatever we do with technology and

1:18:25

AI will

1:18:27

reach a far audience and also help the people who

1:18:29

today don't even have access to clean

1:18:31

drinking water and things like that . So

1:18:34

I hope AI will benefit most

1:18:36

people and

1:18:38

, yeah , let's see how that goes .

1:18:41

Yeah , I think it's going to redefine human identity

1:18:44

. Yeah .

1:18:44

I'd like to take it further and I'd say the planet . I

1:18:47

think you know , with the

1:18:50

AI , I hope we gain some efficiencies

1:18:53

, to go to your point , christian , that

1:18:55

we don't . We can have it

1:18:57

all sustainable so we're not so destructive

1:18:59

, because you know

1:19:01

the whole circle of life , as they say . You

1:19:04

know it's important to have all

1:19:06

of the species of animals

1:19:09

. You know plants , water , you

1:19:11

know anything else is on the planet . It's

1:19:14

an entire ecosystem that needs to work together . So

1:19:16

I'm hoping , with this AI

1:19:19

, that's something that we get out of . It is

1:19:21

how to become less destructive

1:19:23

and more efficient and more sustainable

1:19:25

, so that everything benefits

1:19:27

, not just humans because

1:19:29

we are heavily dependent upon everyone else

1:19:31

.

1:19:32

That's the moral aspect of it . So

1:19:34

if we use it to

1:19:37

use all of the resources , then

1:19:42

it is moral aspects

1:19:44

bad because it is not

1:19:47

sustainable for us as

1:19:49

a society and as human beings

1:19:51

on this planet . So

1:19:54

, as I see , moral is

1:19:56

a function

1:19:58

of keeping

1:20:00

the system alive , because

1:20:03

we use the distinction between

1:20:05

good and bad in that way that it is

1:20:07

not morally good

1:20:09

to use all the resources . So

1:20:16

if we could extend anything that we can do with AI using all of the resources , that

1:20:18

is not really good and that

1:20:21

what we can use with our brains

1:20:23

is think ahead when

1:20:25

will this point in time will

1:20:28

be and label

1:20:30

it as bad behavior . So

1:20:33

the discussion we are having now and

1:20:36

I'm very glad that

1:20:38

you brought this point , sorin is

1:20:41

that we have this discussion now to think

1:20:43

ahead . Where will the

1:20:45

use of AI be

1:20:47

bad for us as a society

1:20:49

and as human beings

1:20:51

and for the planet ? Because now

1:20:54

is the time we can think ahead what we

1:20:56

have to watch out in the next

1:20:58

month or years or something

1:21:00

like that , and

1:21:02

that is the moral aspect I think

1:21:04

we should keep in mind when we

1:21:06

are going further with AI .

1:21:09

I think there are so many aspects there to your

1:21:11

point , christian . So one is of course the whole

1:21:13

, like we all know , the

1:21:16

energy consumption of AI in itself , of AI in itself

1:21:18

. But there's also the other side , I

1:21:23

mean the flip side

1:21:25

, where AI could maybe help us spotlight or shine a

1:21:27

bright light on where can we save on

1:21:30

energy in companies and

1:21:32

where can AI help us , let's

1:21:35

say , calibrate our moral compasses

1:21:38

by shining a light on

1:21:40

where we don't behave as well today

1:21:42

as a species

1:21:44

. So I think there's a flip side

1:21:46

. I'm

1:21:50

hoping we will make some good decisions along the way

1:21:52

to have AI help us

1:21:54

in that .

1:21:58

There's so many things I could talk about with AI

1:22:00

and we'll have to have I

1:22:03

think we'll have to schedule another discussion to

1:22:05

have you on , because I did . I had a whole list of notes

1:22:07

of things that I wanted to talk about when

1:22:10

it comes with AI , not just from the ERP point

1:22:12

of view , but from the AI point of view , because , you

1:22:15

know , after getting into the more AI book

1:22:17

and listening to several

1:22:19

podcasts about AI and humanity

1:22:22

, there's a lot of things that I

1:22:25

wanted to jump into . You

1:22:27

know we talked about the de-skilling . We talked about

1:22:29

too much trust . I'd like to get into harm

1:22:32

bias and also , you know how

1:22:34

AI can analyze data . You

1:22:37

know that everyone thinks anonymous because , reading

1:22:41

that Morley , I booked some statistics they put in

1:22:43

there . I was kind of fascinated . Just

1:22:45

to throw it out , there is that

1:22:47

87% of the

1:22:49

United States population can be identified by

1:22:52

their birth date , gender and their zip

1:22:54

code . That was mind

1:22:56

blowing . And then 99.98%

1:23:00

of people can be identified with 15

1:23:02

data points . So all of this anonymous

1:23:05

data . You know , with the data sharing

1:23:07

that's going on , it's very easy to make

1:23:09

many pieces of anonymous

1:23:11

data no longer anonymous

1:23:13

. Is what I got from that . Um . So

1:23:16

all that data sharing with those points , that um

1:23:18

, the , the birth date , gender

1:23:20

and five digit us zip code here again

1:23:22

, that's in the united states was one that that

1:23:25

shocked me , and now I understand why those

1:23:27

questions get asked the most because it's going to give

1:23:29

, with a high probability , 87 percent

1:23:31

.

1:23:32

Uh who you are maybe

1:23:36

just for the audience , uh , watching

1:23:38

this or listening to this . So so

1:23:40

the book that we're talking about is this

1:23:42

one Mole AI . I

1:23:45

don't know if you can see it . Does it get into focus ? I

1:23:47

don't know if it does .

1:23:48

Yeah now it does .

1:23:50

So it's this one , mole , ai and

1:23:52

how we Get there . It's really

1:23:55

a great book that

1:23:57

goes across fairness , privacy

1:23:59

, responsibility , accountability , bias

1:24:03

, safety , all kinds

1:24:05

of and it tries to take sort of a pro-con

1:24:07

approach . You know , because I

1:24:10

think maybe this is a good way to end the discussion

1:24:12

, because I have to go . I

1:24:15

think one cannot just say AI

1:24:17

is all good or AI is

1:24:19

all bad , like it depends on

1:24:21

what you use it for and how we , how

1:24:24

we use it and how we let

1:24:27

it be biased or not , or how

1:24:29

we implement fairness into algorithms , and

1:24:31

so there's just so many things that we could talk about for an hour

1:24:33

. But that's what this book is all about and

1:24:36

that's what triggered me to to share

1:24:38

a month back . So just thank

1:24:40

you for the , for the chance to talk about

1:24:42

some of these things , and I'd be happy to jump on

1:24:44

another one .

1:24:46

Absolutely , We'll have to schedule one up , but

1:24:48

thank you for the book recommendation . I did

1:24:51

start reading the Moral AI book that you just mentioned

1:24:53

. Again , it's Pelican Books . Anyone's looking

1:24:55

for it . It's a great book

1:24:57

. Thank you , both Soren

1:24:59

and Christian , for taking the time to speak with us this

1:25:02

afternoon , this morning , this evening , whatever it may

1:25:04

be anywhere . I know where I have the time zones

1:25:06

and we'll definitely have to schedule to

1:25:08

talk a little bit more about AI and some of the other aspects

1:25:10

of AI . But if you would

1:25:12

, before we depart , how can

1:25:14

anyone get in contact with you to

1:25:17

learn a little bit more about AI , learn a little bit more

1:25:19

about AI , learn a little bit more about what you do and

1:25:21

learn a little bit more about all the great things that

1:25:23

you're doing ?

1:25:26

Soren , so the best place to find me is probably

1:25:28

on LinkedIn . That is my only

1:25:30

media that I participate in

1:25:32

these days . I deleted all the other accounts and

1:25:35

that's a topic for another discussion .

1:25:37

It's so cleansing to do that too .

1:25:38

Yeah , and

1:25:42

for me it's also on LinkedIn and

1:25:44

on Blue Sky . It's Curate Ideas

1:25:46

excellent

1:25:48

, great .

1:25:48

Thank you both . Look forward to talking with both of you again soon

1:25:50

.

1:25:51

Ciao , ciao thanks for having us . Thank you so

1:25:53

much bye , thank you guys .

1:25:57

Thank you , chris , for your time for

1:25:59

another episode of In the Dynamics Corner Chair

1:26:01

and thank you to our guests for participating . Thank you for your time for another episode of In the Dynamics Corner Chair and thank you to our guests

1:26:03

for participating .

1:26:04

Thank you , brad , for your time . It is

1:26:06

a wonderful episode of Dynamics Corner

1:26:09

Chair . I would also like to thank

1:26:11

our guests for joining us . Thank

1:26:13

you for all of our listeners tuning in as well

1:26:16

. You can find Brad at

1:26:18

developerlifecom , that

1:26:20

is D-V-L-P-R-L-I-F-Ecom

1:26:24

, and you can interact with them via

1:26:27

Twitter D-V-L-P-R-L-I-F-E

1:26:30

. You can also find

1:26:32

me at matalinoio

1:26:35

, m-a-t-a-l-i-n-oi-o

1:26:38

L

1:26:46

I N O , dot I O , and my Twitter handle is Mattelino16 . And see , you can see those links down

1:26:48

below in their show notes . Again , thank you everyone . Thank you and take care

1:26:50

.