8. november 2025 · udskrift

OM DENNE EPISODE

What comes first in your mind when you hear "AI and ethics"?

For Mark, it's a conversation with his teenage son about driverless cars choosing who to hurt in an accident. For Stephan, it's data privacy and the question of whether we really have a choice about what we share. For Niko, it's the haunting question: when AI makes the decision, who's responsible?

Niko anchors a conversation that quickly moves from sci-fi thought experiments to the uncomfortable reality—ethical AI decisions are happening every few minutes in our lives, and we're barely prepared. Joining him are Mark (reflecting on how fast this snuck up on us) and Stephan (bringing systems thinking about data, privacy, and the gap between what organizations should do and what governments are actually doing).

From Philosophy to Practice
Mark's son thought driverless cars would obviously make better decisions than humans—until Mark asked what happens when the car has to choose between two accidents involving different types of people. The conversation spirals quickly: Who decides? What's "wrong"? What if the algorithm's choice eliminates someone on the verge of a breakthrough? The philosophical questions are ancient, but now they're embedded in algorithms making real decisions.

The Consent Illusion
Stephan surfaces the data privacy dimension: someone has to collect data, store it, use it. Niko's follow-up cuts deeper: "Do we really have the choice what we share? Can we just say no, and then what happens?" The question hangs—are we genuinely consenting, or just clicking through terms we don't read because opting out isn't really an option?

Starting Conversations Without Creating Paralysis
Mark warns about a trap he's seen repeatedly—organizations leading with governance frameworks and compliance checklists that overwhelm before anyone explores what's actually possible. His take: "You've got to start having the conversations in a way that does not scare people into not engaging." Organizations need parallel journeys—applying AI meaningfully while evolving their ethical stance—but without drowning people in fear before they've had a chance to experiment.

Who's Actually Accountable?
The hosts land on three levels: individuals empowered to use AI responsibly, organizations accountable for what they build and deploy, and governments (where Stephan is "hesitant"—Switzerland just imposed electronic IDs despite 50% public skepticism). Stephan's question lingers: "How do we make it really successful for human beings on all different levels?"

When Niko asks for one takeaway, Mark channels Mark Twain: "It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so. My question to you is, what do you know about AI and ethics?"

Stephan reflects: "AI is reflecting the best and the worst of our own humanity, forcing us to decide which version of ourselves we want to encode into the future."

Niko's closing: "Ethics is a socio-political responsibility"—not compliance theater, not corporate governance alone, but something we carry as parents, neighbors, humans.

This episode doesn't provide answers—it surfaces the questions practitioners should be sitting with. Not the distant sci-fi dilemmas, but the ethical decisions happening in your organization right now, every few minutes, while you're too busy to notice.

Administration

Teknologi

Forretning

Uddannelse

Selvudvikling

Engelsk

USA

BrugAPItil å hente denne episode iJSON

I DENNE EPISODE

# mark twain

Redigér disse tags

UDSKRIFT 🔗

1
00:00:00,120 --> 00:00:02,700
Mark Richards: G'day and welcome
to another episode of SBCs
2
00:00:02,700 --> 00:00:07,620
unleashed, almost in the process
of renaming it. We kind of
3
00:00:07,620 --> 00:00:10,080
agreed last week that it was
going to become agents
4
00:00:10,080 --> 00:00:13,980
unleashed. But, well, agents
unleashed, change unleashed. We
5
00:00:13,980 --> 00:00:16,200
had a few, but it looks like
it's going to be agents
6
00:00:16,200 --> 00:00:19,860
unleashed. So coming soon, but
we've got to go and do all the
7
00:00:19,860 --> 00:00:24,260
admin that sits behind the
decision. In the meantime, the
8
00:00:24,260 --> 00:00:28,340
boys were very excited to talk
about ethics, and I don't know
9
00:00:28,340 --> 00:00:31,940
whether Niko, Niko was just
predicting the conversations he
10
00:00:31,940 --> 00:00:35,240
was going to have with his kids
this week, or whether it was a
11
00:00:35,240 --> 00:00:38,000
bit of chicken and egg, but over
to you, Niko. Take us away.
12
00:00:39,320 --> 00:00:41,800
Nikolaos Kaintantzis: Thank you.
Then that then that agents
13
00:00:41,800 --> 00:00:46,540
unleashed. Yeah, talking about
ethics. Great. I prepared
14
00:00:46,540 --> 00:00:49,540
something, and I thought it
wouldn't be cool if some AI
15
00:00:49,840 --> 00:00:54,100
persona would read this. And
then I ran out of time, so I'm
16
00:00:54,100 --> 00:00:56,860
now AI, and I'm reading what
I've prepared. If it's okay with
17
00:00:56,860 --> 00:01:00,040
you, with a sexy voice, it was
the idea. So let me prepare my
18
00:01:00,040 --> 00:01:05,520
sexy voice. So and then this
show, three voices, one
19
00:01:05,520 --> 00:01:10,680
question, AI and ethics. Stephan
speaks of responsibility that
20
00:01:10,680 --> 00:01:15,360
should never hide in a black
box. Mark. Mark wants awareness
21
00:01:15,480 --> 00:01:19,860
without reflection, just another
algorithm running blind. And
22
00:01:19,860 --> 00:01:23,480
finally, Niko, that's me,
reminds us ethics is not
23
00:01:23,480 --> 00:01:27,680
compliance. Is a socio political
responsibility, one stretches
24
00:01:27,740 --> 00:01:31,100
far beyond the goal itself. The
riddle remains, is
25
00:01:31,100 --> 00:01:35,300
accountability? Accountability a
burden we carry, or the compass
26
00:01:35,300 --> 00:01:40,160
that carries us forward? This
summary was provided by AI, so I
27
00:01:40,160 --> 00:01:44,200
just feed in all the things I
thought you will say, or I've
28
00:01:44,200 --> 00:01:47,320
seen the preparation, and there
was a summary, and they asked
29
00:01:47,320 --> 00:01:50,980
repair something, yeah, yeah,
still not catchy. I think I
30
00:01:50,980 --> 00:01:52,300
tried myself again. I know
31
00:01:53,140 --> 00:01:55,900
Mark Richards: you needed some
background music.
32
00:01:56,860 --> 00:01:59,860
Nikolaos Kaintantzis: Oh, for
sure. Let me. Let me call the
33
00:01:59,860 --> 00:02:08,880
kids. No, exactly so we really
agreed on AI and ethics, and I
34
00:02:08,880 --> 00:02:12,660
think it's a broad topic, and
the first thing I would like to
35
00:02:12,660 --> 00:02:17,220
know for you, and not not what,
what is your passion, or what
36
00:02:17,220 --> 00:02:20,360
scares you, or whatever, but I
would like to know what comes
37
00:02:20,360 --> 00:02:24,140
first in your mind when you hear
AI and ethics.
38
00:02:24,380 --> 00:02:28,640
Mark Richards: So when I read
that, I went without any control
39
00:02:28,640 --> 00:02:31,340
whatsoever, my mind went
straight back to like six or
40
00:02:31,340 --> 00:02:36,800
seven years ago, and it was
actually a moment where my son
41
00:02:36,800 --> 00:02:39,500
had decided that he was not a
Math Science Kid, he was a
42
00:02:39,500 --> 00:02:42,280
humanities kid, and he was
starting to think about what
43
00:02:42,280 --> 00:02:45,280
possible futures might look
like. And we wound up in this
44
00:02:45,400 --> 00:02:49,840
one hour long discussion about
ethics and AI and driverless
45
00:02:49,840 --> 00:02:53,680
cars. And he at the time, and,
you know, his early teens at
46
00:02:53,680 --> 00:02:56,800
this point, and life was very
black and white to him, and I
47
00:02:56,800 --> 00:02:59,380
was taking him down this, you
know, very long discussion to
48
00:02:59,380 --> 00:03:02,520
go. Well, let's imagine how
there could be ethical dilemmas
49
00:03:02,520 --> 00:03:06,840
for a driverless car. And he's
like, Well, surely not, surely
50
00:03:06,900 --> 00:03:09,900
it's just going to make better
decisions than humans. And I
51
00:03:09,900 --> 00:03:13,080
said, Well, what if it is
choosing between two accidents,
52
00:03:13,380 --> 00:03:16,740
and the accidents involve
different types of people? And
53
00:03:16,740 --> 00:03:19,380
he went, Well, you know, you
pick the ones that have got the
54
00:03:19,380 --> 00:03:25,040
least remaining lifespan. So had
this huge tunnel we dug down to
55
00:03:25,040 --> 00:03:28,940
go, you know, how do you get to
a point where the AI is making a
56
00:03:28,940 --> 00:03:33,140
good decision about life and
death choices, controlling a
57
00:03:33,140 --> 00:03:36,380
car? And, of course, just then,
there were starting to be people
58
00:03:36,380 --> 00:03:40,720
actually doing PhDs and schools
of study emerging around, you
59
00:03:40,720 --> 00:03:44,260
know, the whole notion of the
psychology of AI and ethics. But
60
00:03:44,260 --> 00:03:47,320
I think the thing that struck me
when I finished reminiscing on
61
00:03:47,320 --> 00:03:52,000
that hour of torturing my son in
the car was back then, when I
62
00:03:52,000 --> 00:03:55,900
thought about AI and ethics, it
was big, grand choices like, you
63
00:03:55,900 --> 00:04:02,220
know, who does the car choose to
hurt in a crisis, as opposed to
64
00:04:02,460 --> 00:04:07,080
actually today, for most of us,
there are ethical AI decisions
65
00:04:07,080 --> 00:04:10,680
happening every few minutes in
our lives. It's moved from a
66
00:04:10,680 --> 00:04:15,360
kind of almost science fiction
imagination thing into a day to
67
00:04:15,360 --> 00:04:19,020
day reality for us, and it's
like that's kind of snuck up.
68
00:04:19,259 --> 00:04:22,039
Nikolaos Kaintantzis: Yeah, I
had something similar in my mind
69
00:04:22,219 --> 00:04:27,139
when I had to answer this
question myself. Was exactly the
70
00:04:27,139 --> 00:04:30,199
question not, can AI be
responsible, but who is
71
00:04:30,199 --> 00:04:34,039
responsible? So if AI chooses
that somebody trained that is
72
00:04:34,039 --> 00:04:39,019
responsibility of the ones who
train the AI, or Yeah, who is
73
00:04:39,019 --> 00:04:42,519
responsible for that? So if he
chooses the one who was older,
74
00:04:42,819 --> 00:04:46,479
maybe he was one day away from a
breakthrough curing diseases,
75
00:04:46,599 --> 00:04:50,139
and the other one was just a
mean kid that will never do
76
00:04:50,139 --> 00:04:53,079
anything society, if we can
judge it, I don't know. Maybe
77
00:04:53,319 --> 00:04:56,559
this kid will change in 10
years. So I really thought if
78
00:04:56,559 --> 00:05:00,659
you drive yourself, you do this
decision instantly. And you have
79
00:05:00,659 --> 00:05:03,479
to live with the choice if
somebody else is doing this
80
00:05:03,479 --> 00:05:06,659
decision, or if AI is doing
decision who is really
81
00:05:06,659 --> 00:05:11,279
responsible. So the first thing
that came in my mind now was not
82
00:05:11,339 --> 00:05:16,439
if a car has to decide, but, uh,
when he decides, or it decides
83
00:05:16,739 --> 00:05:21,979
who is responsible and what is
wrong. So deciding wrong is also
84
00:05:22,039 --> 00:05:26,179
a nice set. But what is really
wrong? Killing the old, killing
85
00:05:26,179 --> 00:05:30,559
the young. What is what is old?
So you see, I hadn't read an
86
00:05:30,559 --> 00:05:34,699
answer, so I already started
talking with my family, since I
87
00:05:34,699 --> 00:05:37,639
prepared this episode, the
questions on Saturday after this
88
00:05:37,639 --> 00:05:42,219
show, and we had the whole week
discussions at home, about about
89
00:05:42,219 --> 00:05:46,359
ethics, about the future, and
was really cool discussions.
90
00:05:47,139 --> 00:05:49,839
Stephan, your what comes first
in mind when you hear AI and
91
00:05:49,839 --> 00:05:51,579
ethics? Yeah, I kind
92
00:05:51,580 --> 00:05:54,280
Stephan Neck: of have mixed
feelings when I read your
93
00:05:54,280 --> 00:05:58,780
question or your trigger. As an
Information Systems Engineer,
94
00:05:59,140 --> 00:06:04,620
I'm aware that systems need
data, and my mixed feelings are
95
00:06:04,620 --> 00:06:11,280
about privacy of data. Someone
will collect data. Someone it
96
00:06:11,280 --> 00:06:15,180
has to be stored somewhere, the
way it's stored, and then the
97
00:06:15,180 --> 00:06:20,960
way it is used. So this popped
up when you came up with this
98
00:06:20,960 --> 00:06:26,120
trigger. And looking back a few
of our past episodes, we were
99
00:06:26,120 --> 00:06:31,820
dealing with data, we need that
data. And yep, that's why two
100
00:06:31,820 --> 00:06:33,380
cents to this trigger.
101
00:06:33,620 --> 00:06:36,980
Nikolaos Kaintantzis: Wow, wow.
Data. Talking about data, do we
102
00:06:36,980 --> 00:06:41,680
really have the choice what we
share? So can we say, just know,
103
00:06:41,680 --> 00:06:43,060
and then what happens?
104
00:06:45,460 --> 00:06:46,600
Unknown: Very good question.
105
00:06:46,780 --> 00:06:48,940
Mark Richards: Why don't you
keep the ball rolling, Stephan,
106
00:06:48,940 --> 00:06:51,520
since you opened it, because I
think I'm going to be a little
107
00:06:51,520 --> 00:06:53,800
bit more flippant, and you're
probably going to be more
108
00:06:53,800 --> 00:06:54,400
systematic.
109
00:06:56,260 --> 00:07:00,400
Stephan Neck: Yeah, you know,
I'm, I'm actually, I'm actually
110
00:07:00,400 --> 00:07:07,260
in Scotland for a short holiday
trip, and we just experienced
111
00:07:07,440 --> 00:07:12,480
storm Amy. And you can be
structured as much as looking
112
00:07:12,480 --> 00:07:14,640
over regarding our flights,
113
00:07:14,880 --> 00:07:16,860
Mark Richards: it's Oh, you are
hacking
114
00:07:16,860 --> 00:07:21,180
Stephan Neck: over how much
debris you have in front on of
115
00:07:21,180 --> 00:07:24,920
the house on your lawn all the
day branches from the trees,
116
00:07:24,920 --> 00:07:25,220
right?
117
00:07:25,219 --> 00:07:28,879
Mark Richards: So looks like
you're coming and going with Wi
118
00:07:28,879 --> 00:07:33,319
Fi. By the way, Stephan, you
kind of were there and here, but
119
00:07:33,619 --> 00:07:34,219
I'll jump in.
120
00:07:35,000 --> 00:07:36,800
Nikolaos Kaintantzis: I think
your wife should stop watching
121
00:07:36,800 --> 00:07:40,600
Netflix and Disney plus and give
you the one bandwidth back. Just
122
00:07:40,600 --> 00:07:44,860
Just kidding. We just missed the
last, the last sentence of
123
00:07:44,860 --> 00:07:45,340
years.
124
00:07:46,180 --> 00:07:49,600
Stephan Neck: Yeah, what I said
is, what I said is, can you
125
00:07:49,600 --> 00:07:53,920
control everything now? Storm
Amy has taken over in Scotland,
126
00:07:53,980 --> 00:07:59,440
and this morning is kind of
under restructuring, how we
127
00:07:59,800 --> 00:08:03,180
behave, what we do and how it
affects our lives,
128
00:08:03,540 --> 00:08:06,480
Nikolaos Kaintantzis: yeah, but
it's not the same. So until now,
129
00:08:06,600 --> 00:08:11,520
I was in charge of my data, my
decisions. So storm is not
130
00:08:11,520 --> 00:08:14,760
really something I decide. And
then I realized, oh, I have no
131
00:08:14,760 --> 00:08:19,500
choice. It's something that just
happens. What's about data I
132
00:08:19,500 --> 00:08:23,780
have? So there comes a window
and says, log in with your that
133
00:08:23,780 --> 00:08:26,480
that account, and with that, you
give your soul and your privacy
134
00:08:26,480 --> 00:08:31,340
away. You can say yes or no.
Will it be okay to say no, and
135
00:08:31,340 --> 00:08:33,680
what happens if no? So what
happens with that decision?
136
00:08:34,459 --> 00:08:37,459
Stephan Neck: Yeah, let me be
picky on that one. Are we really
137
00:08:37,459 --> 00:08:44,139
in control, or are we just
thinking we are in control. I
138
00:08:44,139 --> 00:08:48,819
think we have the same basic
questions to answer with AI as
139
00:08:48,819 --> 00:08:54,939
well, but it's speeding up. You
get more opportunities to deal
140
00:08:54,939 --> 00:09:00,239
with data. The question about is
it, is it proper data? Is a
141
00:09:00,239 --> 00:09:05,759
clean data we are using stuff
like that is adding to the
142
00:09:05,759 --> 00:09:11,039
overall challenges we still
have. I don't think if I look
143
00:09:11,039 --> 00:09:16,379
back my last 3035, years in the
trade, yeah, we thought we could
144
00:09:16,499 --> 00:09:22,519
decide we could master
everything, but yeah, it opens
145
00:09:22,519 --> 00:09:23,899
up a new whole field for me.
146
00:09:24,860 --> 00:09:27,920
Mark Richards: I think for me, a
lot of people think they're in
147
00:09:27,920 --> 00:09:31,460
control today, but they don't
necessarily have the
148
00:09:31,460 --> 00:09:35,780
mindfulness. Like, what control
do I have? What kind of data am
149
00:09:35,780 --> 00:09:39,200
I sharing? Do I really have
awareness? And it's one of those
150
00:09:39,200 --> 00:09:42,760
things where I'm going to do two
kind of generational looks.
151
00:09:42,880 --> 00:09:45,520
There's a generational look for
me, which is, I've got a 20 year
152
00:09:45,520 --> 00:09:49,060
old son who grew up in the
digital age, and if I look at
153
00:09:49,060 --> 00:09:53,020
the way he treats data about
himself, he is so much more
154
00:09:53,020 --> 00:09:56,980
mindful about his personal data
than I am. He's mindful
155
00:09:56,980 --> 00:09:59,920
everybody. Somebody takes a
photo with him in it. He's mind.
156
00:10:00,060 --> 00:10:03,660
Full of his ability to control
people's access, right? He
157
00:10:03,660 --> 00:10:06,000
turned around to me a little
while ago and said, Dad, nobody
158
00:10:06,000 --> 00:10:09,000
shares phone numbers with each
other anymore. You share instas
159
00:10:09,300 --> 00:10:14,640
Because you can block somebody
from your Insta, right? And you
160
00:10:14,640 --> 00:10:19,260
get control. So these kids of
today, they grow up with layered
161
00:10:19,560 --> 00:10:22,640
views of their personal data and
how they're exposing it to the
162
00:10:22,640 --> 00:10:25,580
world. It's just built into
them, almost just a part of
163
00:10:25,580 --> 00:10:28,820
being society. But us older
people didn't grow up like that.
164
00:10:29,120 --> 00:10:35,240
You know, of course, I share my
phone number and and, you know,
165
00:10:35,240 --> 00:10:38,060
do I have different emails that
I use for different purposes?
166
00:10:38,060 --> 00:10:41,440
Well, I've often wished that I
did, but I don't. But any kid of
167
00:10:41,440 --> 00:10:45,400
today is going to have that, and
I think that that same level of
168
00:10:45,640 --> 00:10:49,900
am I aware, has to start to
creep in to questions I'm
169
00:10:49,900 --> 00:10:53,920
running with AI, right? Do I
know whether I have decisions
170
00:10:53,920 --> 00:10:58,120
about whether or not to share
information with chat GPT or
171
00:10:58,120 --> 00:11:01,080
with Claude? Do I know the
repercussions of different
172
00:11:01,080 --> 00:11:04,560
choices I make. You know, am I
even checking? There was
173
00:11:04,560 --> 00:11:07,320
recently a big furor because
Claude put out an update that
174
00:11:07,320 --> 00:11:11,160
ship switched the toggle from
automatically not sharing to
175
00:11:11,160 --> 00:11:14,520
automatically sharing, and there
was such a controversy blew up
176
00:11:14,520 --> 00:11:18,000
about it that a day later, they
reversed the decision. And we've
177
00:11:18,000 --> 00:11:21,620
got this world where people are
becoming more and more conscious
178
00:11:21,620 --> 00:11:24,500
of the commercial value of data,
right? We've been talking about
179
00:11:24,500 --> 00:11:30,080
it for years, and companies like
Google and Amazon and Facebook
180
00:11:30,200 --> 00:11:33,860
have built multi billion,
potentially multi trillion
181
00:11:33,860 --> 00:11:37,280
dollar empires on the basis of
the commercial value of data.
182
00:11:37,280 --> 00:11:40,360
Most of us don't, but you look
around and what you're seeing at
183
00:11:40,360 --> 00:11:43,360
the moment, and all of the
companies, let's say LinkedIn
184
00:11:43,360 --> 00:11:46,180
works very hard to make sure the
only way you get at LinkedIn
185
00:11:46,660 --> 00:11:50,260
data is logged in manually and
tracking. Instagram, it's pretty
186
00:11:50,260 --> 00:11:53,140
easy to get at their data, but
they're busy locking it down, so
187
00:11:53,140 --> 00:11:56,140
it's harder to get at their
data. People are getting a lot
188
00:11:56,140 --> 00:11:59,980
more focused about putting
messages on their websites.
189
00:11:59,980 --> 00:12:02,880
Right there's a file, I think
it's called llms, dot text,
190
00:12:03,180 --> 00:12:06,420
which actually communicates what
you are and aren't willing to
191
00:12:06,420 --> 00:12:11,220
share from your website to llms.
Is everybody using that? Do you
192
00:12:11,220 --> 00:12:14,100
have the awareness to know what
you're sharing? Because if you
193
00:12:14,100 --> 00:12:16,860
don't know what you're sharing,
you've never got any choices
194
00:12:16,860 --> 00:12:19,860
going on, because you're not
making choices. You're just
195
00:12:19,860 --> 00:12:20,480
letting it
196
00:12:20,480 --> 00:12:24,200
Stephan Neck: fly by, but that
that leads to a situation of
197
00:12:24,200 --> 00:12:29,240
take it or leave it right, and
you're deciding in that context
198
00:12:29,240 --> 00:12:33,380
with your data, because your
data is like has a price tag as
199
00:12:33,380 --> 00:12:39,320
well. Am I sharing it? Am I? Am
I blocking it? And if I block
200
00:12:39,320 --> 00:12:44,140
too much, I'm probably out of
the inner circle. I'm not within
201
00:12:44,140 --> 00:12:50,560
the fence. I can't use it. So
that makes me think, and for me,
202
00:12:50,560 --> 00:12:55,240
the producers, the creators,
have a huge responsibility
203
00:12:55,240 --> 00:13:01,560
regarding the usage of data,
collection of data, bringing it
204
00:13:01,560 --> 00:13:07,080
in, into the systems, and where
do you position this Take it or
205
00:13:07,080 --> 00:13:11,880
leave it right? Is it a binary
thing, or is it kind of an area
206
00:13:11,880 --> 00:13:17,460
where you really are in control,
where you start deciding, am I
207
00:13:17,460 --> 00:13:24,560
in and to what kind of range Am
I in, in a system, or in an in
208
00:13:24,560 --> 00:13:26,300
an even in an ecosystem,
209
00:13:27,559 --> 00:13:29,839
Nikolaos Kaintantzis: the more I
think, the less I think I have
210
00:13:29,839 --> 00:13:33,379
it under control. So let's go
into WhatsApp. When you have
211
00:13:33,379 --> 00:13:36,199
WhatsApp and somebody has your
number, you're already part of
212
00:13:36,199 --> 00:13:40,579
the system, even if you never
agreed, and they are saying bye,
213
00:13:40,579 --> 00:13:44,619
bye, if you are part of
WhatsApp, you took care that
214
00:13:44,619 --> 00:13:46,899
everybody on your phone book
agree that they are on the phone
215
00:13:46,899 --> 00:13:50,919
book. So what the hell question
is that? So I'm sure it's not
216
00:13:50,919 --> 00:13:53,859
really legal to say your
responsibility that everybody in
217
00:13:53,859 --> 00:13:58,539
your phone book agrees that your
phone book now is public. Yet, I
218
00:13:58,839 --> 00:14:01,139
think Stephan, you're right. You
convinced somehow we are not
219
00:14:01,139 --> 00:14:06,659
really under control of the
data. And Mark, I think it's
220
00:14:06,659 --> 00:14:13,499
also the way you or we handle
data, because I see also kids.
221
00:14:13,799 --> 00:14:16,979
Let's go to my youngest
daughter. They have no idea
222
00:14:16,979 --> 00:14:20,419
about the stuff, so the older
ones, I think they're much more
223
00:14:20,419 --> 00:14:26,539
careful. I remember last time
when my middle kid watched me
224
00:14:26,539 --> 00:14:30,259
how I just accept all cookies
without reading it, and she was
225
00:14:30,319 --> 00:14:34,039
really, how can you? And
thought, oh, that's just a dummy
226
00:14:34,039 --> 00:14:36,679
browser which just does
everything. And then I close it,
227
00:14:36,679 --> 00:14:40,659
and then I'm a completely fresh
new person. Ah, okay, because
228
00:14:40,659 --> 00:14:45,279
she really tries to handle what
she wants to give away or not.
229
00:14:45,639 --> 00:14:50,679
But Stephan, you're right if
you're anyway part of it. And
230
00:14:50,679 --> 00:14:53,199
the other question is how much
you want to interact with that.
231
00:14:53,199 --> 00:14:56,679
So for instance, our youngest
doesn't want to be part of
232
00:14:56,679 --> 00:15:02,519
what's up. Just she's missing a
lot of. Loss gossip, which, by
233
00:15:02,519 --> 00:15:05,519
the way, I think it's good that
she misses, but that's another
234
00:15:05,519 --> 00:15:08,939
point she maybe misses, also,
also important, important stuff.
235
00:15:09,000 --> 00:15:11,220
Mark Richards: So what's such a
great example to me of a
236
00:15:11,220 --> 00:15:15,480
generational divide? Thing we
all know about the fact that,
237
00:15:15,480 --> 00:15:18,660
basically, by continuing to use
it, we're giving Facebook a
238
00:15:18,660 --> 00:15:24,560
stupid amount of our data. What
my children know is that their
239
00:15:24,560 --> 00:15:27,920
parents are old and they want to
use WhatsApp. Therefore, because
240
00:15:27,920 --> 00:15:30,740
they don't view family
interaction as particularly data
241
00:15:30,740 --> 00:15:33,380
sensitive, they're willing to
use it for family interaction,
242
00:15:35,000 --> 00:15:38,180
but they would never use
WhatsApp, right? They literally,
243
00:15:38,180 --> 00:15:40,840
and I watch, I watch them talk
about these things, make these
244
00:15:40,840 --> 00:15:44,680
calls, they literally go, what
are the privacy concerns I've
245
00:15:44,680 --> 00:15:47,140
got with different ways of
communicating, and which way am
246
00:15:47,140 --> 00:15:50,380
I going to communicate based on
the privacy and I care about my
247
00:15:50,380 --> 00:15:55,540
ownership of that data
awareness, so
248
00:15:55,540 --> 00:15:59,260
Nikolaos Kaintantzis: you guys
read the whole stuff. So before
249
00:15:59,260 --> 00:16:01,320
you sign up, sign in, sign up
for something.
250
00:16:02,040 --> 00:16:04,380
Mark Richards: I had a friend
who did a PhD on whether people
251
00:16:04,380 --> 00:16:04,920
read those.
252
00:16:08,100 --> 00:16:10,620
Nikolaos Kaintantzis: You do the
meta reading, okay? Stephan,
253
00:16:12,240 --> 00:16:18,000
Stephan Neck: No, obviously not.
But I the past few years, maybe
254
00:16:18,000 --> 00:16:22,040
once or twice, I really took the
time to read it. First of all,
255
00:16:22,100 --> 00:16:25,100
do I really understand
everything? No, you should be a
256
00:16:25,100 --> 00:16:31,040
legal person to understand
every, every part of such a such
257
00:16:31,040 --> 00:16:37,400
a text, in the end, and I must
confess, I don't care, because I
258
00:16:37,400 --> 00:16:40,420
want to take part in that
service. I want to have this
259
00:16:40,420 --> 00:16:44,500
action. So I have to agree, it's
again this, take it or leave it
260
00:16:44,500 --> 00:16:49,480
right, and within seconds, or
pretty fast, you have to make a
261
00:16:49,480 --> 00:16:51,400
decision. But I
262
00:16:51,760 --> 00:16:53,680
Mark Richards: think, like
there's a little bit deeper
263
00:16:53,680 --> 00:16:56,080
there for us, right? So if you
talk about what you just said
264
00:16:56,080 --> 00:16:59,680
there, Stephan was, I'm going to
make a mindful choice. Yeah, I
265
00:16:59,680 --> 00:17:02,940
know what I'm giving up, and I'm
willing to pay it. But if you
266
00:17:02,940 --> 00:17:05,580
think about being in a world
where you're trying to be an AI
267
00:17:05,580 --> 00:17:07,920
change agent, you're trying to
help a company, think about
268
00:17:07,920 --> 00:17:11,700
their AI strategy, what data
they're potentially giving away
269
00:17:12,000 --> 00:17:14,640
when they use chat GPT to help
them write an email, what
270
00:17:14,640 --> 00:17:17,160
personally identifiable
information are they potentially
271
00:17:17,160 --> 00:17:22,220
sharing with the world in doing
that? And do you have an
272
00:17:22,220 --> 00:17:25,640
awareness of the potential
exposure of the types of
273
00:17:25,640 --> 00:17:28,700
information that we tend to
share in an AI setting? If you
274
00:17:28,700 --> 00:17:31,400
don't have that, then you're not
going to help your conversation
275
00:17:31,400 --> 00:17:33,800
client think about making
mindful choices. So I think
276
00:17:33,800 --> 00:17:37,640
there's a real duty of care to
us if we're coaching clients in
277
00:17:37,640 --> 00:17:41,200
the AI space to be actually
highly aware of some of the
278
00:17:41,200 --> 00:17:45,580
implications around data and the
various ways that data can get
279
00:17:45,580 --> 00:17:48,640
out of your world from an
ethical perspective, Yeah,
280
00:17:48,700 --> 00:17:49,060
agreed.
281
00:17:49,059 --> 00:17:53,559
Stephan Neck: Mark as a
consultant, I get all of a
282
00:17:53,559 --> 00:18:00,279
sudden different NDAs to sign
right extended ones, or now
283
00:18:00,339 --> 00:18:03,659
companies really being aware, as
you mentioned, they're giving
284
00:18:03,659 --> 00:18:07,619
away more than they think.
Right? Talking to a customer
285
00:18:07,619 --> 00:18:11,399
about strategy, talking to a
customer about a lean portfolio
286
00:18:11,399 --> 00:18:17,819
management, it opens up a whole
new range of possibilities, but
287
00:18:17,819 --> 00:18:23,119
also access to data, which then,
with the help of AI, gets a
288
00:18:23,119 --> 00:18:28,339
whole new a whole new sphere of
options and possibilities in my
289
00:18:28,339 --> 00:18:28,759
trade.
290
00:18:29,360 --> 00:18:31,880
Mark Richards: But that's what
makes it really fascinating to
291
00:18:31,880 --> 00:18:35,300
me, right? Is every potential
internal application of AI is
292
00:18:35,300 --> 00:18:39,560
gonna be fueled by data, right?
So you want more and more data
293
00:18:39,560 --> 00:18:42,400
available to AI because it opens
up more and more possibilities
294
00:18:42,400 --> 00:18:45,220
to you, but at the same time, it
opens up more and more risk
295
00:18:45,220 --> 00:18:48,520
surface. And I think, you know,
the world is caught between
296
00:18:48,520 --> 00:18:52,120
companies that are just blindly
taking an awful lot of risks and
297
00:18:52,120 --> 00:18:54,640
companies that are so concerned
about the risk they're not using
298
00:18:54,640 --> 00:18:56,920
AI at all. And I think there's
gonna be a really interesting
299
00:18:56,920 --> 00:19:00,160
time for us in the next couple
of years as we try and find the
300
00:19:00,160 --> 00:19:01,620
space to hold between those two.
301
00:19:01,620 --> 00:19:06,240
Nikolaos Kaintantzis: I would
like to add on that the leaders
302
00:19:06,240 --> 00:19:09,840
think the people don't use AI
because they just restricted
303
00:19:09,840 --> 00:19:13,020
everything. But the other thing
is, we have aI on our pocket, on
304
00:19:13,020 --> 00:19:16,800
our on our mobile phone, so they
use just different ways, which
305
00:19:16,800 --> 00:19:19,620
is much more dangerous because
now I use it as a private
306
00:19:19,620 --> 00:19:22,640
person. And we just heard how
Stephan act as a private person,
307
00:19:22,640 --> 00:19:26,840
different as a consultant. So so
now we're doing it on our mobile
308
00:19:26,840 --> 00:19:32,000
phones, taking pictures of our
screens and feeding to a free
309
00:19:32,000 --> 00:19:36,980
AI, which has no limitations at
all. So instead of having a
310
00:19:36,980 --> 00:19:42,344
proper governance, we took
decision, Oh, it's nothing for
311
00:19:42,344 --> 00:19:46,364
us, but people still using it so
and that's one of the huge
312
00:19:46,364 --> 00:19:50,504
problems, or one of the we
talked about blind spots, I
313
00:19:50,504 --> 00:19:54,284
think two episodes ago, it's
something people don't or
314
00:19:54,284 --> 00:19:57,764
management, manager, manager,
don't realize, if you just
315
00:19:58,784 --> 00:20:04,020
banish. AI, it's still there,
just on the legal way or in the
316
00:20:04,020 --> 00:20:07,680
private way. Yeah, so in my
case, what I'm doing with with
317
00:20:07,680 --> 00:20:11,640
the whole privacy part. When I
was younger, I read everything,
318
00:20:11,700 --> 00:20:17,940
and I remember, I remember days
when it was on paper and they
319
00:20:17,940 --> 00:20:22,040
just take a pen and remove parts
I didn't like before I signed a
320
00:20:22,040 --> 00:20:25,340
contract, which in Switzerland
is completely legal if the other
321
00:20:25,340 --> 00:20:29,180
person says nothing, that's the
new contract. That's the new new
322
00:20:29,180 --> 00:20:33,080
part of the contract. And it
changed because now too much
323
00:20:33,080 --> 00:20:37,820
reading, it's too many tools,
too many things. And what I'm
324
00:20:37,820 --> 00:20:42,040
doing now is really asking AI to
summarize what they want from
325
00:20:42,040 --> 00:20:45,700
me, and ask also, how is it
compared to something else? You
326
00:20:45,700 --> 00:20:49,780
see, things I shouldn't sign, at
least I'm aware of, but I'm with
327
00:20:49,780 --> 00:20:53,320
Stephan, have no choice. I have
to be in or out, and at least I
328
00:20:53,320 --> 00:20:55,120
try to find out what it means
when I'm in.
329
00:20:55,120 --> 00:20:58,180
Mark Richards: And then, just to
top it off, what you're doing at
330
00:20:58,180 --> 00:21:01,020
that moment is you're giving AI
data about your internal
331
00:21:01,020 --> 00:21:04,020
contractual arrangements with
your clients that may well be
332
00:21:04,020 --> 00:21:04,800
confidential.
333
00:21:05,940 --> 00:21:10,500
Stephan Neck: Yeah, exactly. It
sounds like the start of a
334
00:21:10,500 --> 00:21:11,580
vicious circle, right?
335
00:21:13,200 --> 00:21:15,720
Nikolaos Kaintantzis: Yeah, I
didn't say it was NDAs, but
336
00:21:15,720 --> 00:21:19,020
you're right, yeah, if you
extend it on this way, you
337
00:21:19,020 --> 00:21:23,600
realize, okay, you give it a lot
of information, yes, yes. But
338
00:21:24,500 --> 00:21:27,080
even if you just send it per
email, you give the information
339
00:21:27,080 --> 00:21:30,140
also the whole world, because
it's just a postcard from one
340
00:21:30,140 --> 00:21:35,060
person or the other. But wow,
let's, let's go to a little
341
00:21:35,060 --> 00:21:38,120
different topic. But on a
section about future
342
00:21:38,120 --> 00:21:40,780
generations, we talked about
kids, about they are different.
343
00:21:40,780 --> 00:21:45,820
Their handle data, how aware
they are. I personally think
344
00:21:45,880 --> 00:21:50,080
they are aware because parents
were aware too. Because I'm also
345
00:21:50,080 --> 00:21:53,020
sure some parents don't care,
and the kids will also not care.
346
00:21:53,440 --> 00:21:56,800
So by the way, if I imagine how
many people, how many parents
347
00:21:56,860 --> 00:22:02,040
just post pictures of naked
babies and naked young kids. I'm
348
00:22:02,820 --> 00:22:05,760
not sure if these kids are more
aware or later, if they have a
349
00:22:08,760 --> 00:22:13,440
if it comes back again when they
are older, but this brought me
350
00:22:13,440 --> 00:22:17,340
to the question about future
generations. What more
351
00:22:17,340 --> 00:22:20,960
obligations do we have when
we're deploying AI today? What
352
00:22:20,960 --> 00:22:24,440
can we destroy? What can we
what's Robert G obligate,
353
00:22:24,440 --> 00:22:27,500
obligate. Ah, come on, English
obligations.
354
00:22:29,000 --> 00:22:32,480
Mark Richards: Who wants to
start? I drew a blank. So it's
355
00:22:32,480 --> 00:22:36,080
over to you guys. So first time
ever I sat and thought five
356
00:22:36,080 --> 00:22:38,420
minutes I went, I really don't
know what I'm going to answer to
357
00:22:38,420 --> 00:22:39,020
this. So you're
358
00:22:39,020 --> 00:22:43,000
Nikolaos Kaintantzis: too old
for that shit or what to know.
359
00:22:43,060 --> 00:22:47,740
Stephan Neck: I like your
question. Niko and I scribbled
360
00:22:47,740 --> 00:22:51,580
down a few things, and then you
mentioned your family, and
361
00:22:51,820 --> 00:22:57,100
thinking about my family, my
kids, my grandkids, and when I
362
00:22:57,100 --> 00:23:04,080
summarized my sketches and my
notes again using AI. It came up
363
00:23:04,080 --> 00:23:09,600
with the term of stewardship,
right? I'm not telling my kids
364
00:23:09,960 --> 00:23:14,880
how to live their life, but what
is my stewardship, towards the
365
00:23:14,880 --> 00:23:20,240
family, towards my kids, the
next few generations, and I'm
366
00:23:20,240 --> 00:23:24,560
not using the word obligations,
although I feel like that. I do
367
00:23:24,560 --> 00:23:28,700
have my obligations for the
future, although I'll disappear
368
00:23:28,700 --> 00:23:32,840
one day, and that's it. And I
think it's about, we already
369
00:23:32,840 --> 00:23:36,260
talked about that. How do you
control environments? How do you
370
00:23:36,320 --> 00:23:39,680
control certain things, right?
And some are control freaks.
371
00:23:40,040 --> 00:23:43,780
Some don't even care. There's
the whole range and and behind
372
00:23:43,900 --> 00:23:49,420
this, this whole range for me is
I want to prevent harm, for my
373
00:23:49,420 --> 00:23:53,620
family, for other people, for my
friends. So this is, this is one
374
00:23:53,620 --> 00:23:57,400
of the obligations I probably
have as a producer, as a creator
375
00:23:57,400 --> 00:24:03,660
of AI stuff. The other one is,
every culture has values, and we
376
00:24:03,660 --> 00:24:11,400
as human beings, what I cherish
is dignity, right to live a life
377
00:24:11,760 --> 00:24:16,020
in dignity. And if you look at
this world, first world, second
378
00:24:16,020 --> 00:24:19,560
world, third world, there's not
the same dignity for everyone.
379
00:24:19,560 --> 00:24:23,840
And now a new technology comes
in, new stuff comes in, which
380
00:24:23,840 --> 00:24:31,160
probably shakes up those values
and how we we keep, yeah,
381
00:24:31,160 --> 00:24:36,200
dignity, and with that comes
what I like. If I look back in
382
00:24:36,200 --> 00:24:40,120
my life, it's what is a fair
share. A fair share I know
383
00:24:40,120 --> 00:24:44,080
something. How do I share my
knowledge I do have something.
384
00:24:44,080 --> 00:24:48,400
How do I share with my friends?
Am I a good host at home? Is
385
00:24:48,400 --> 00:24:54,160
there an open door policy? Do we
have guests over for a coffee or
386
00:24:54,400 --> 00:25:00,660
for an event? Or is it No, this
is my this is my castle. Uh,
387
00:25:00,660 --> 00:25:06,480
everyone else is out. And the
funny thing, AI came back with
388
00:25:06,900 --> 00:25:12,600
an idea, calling it a user
manual for the future. So when
389
00:25:12,600 --> 00:25:18,780
it comes to the stewardship, are
we not only leaving traces and
390
00:25:18,780 --> 00:25:25,040
systems? Are we using it with a
manual so people understand why
391
00:25:25,040 --> 00:25:29,420
we built stuff the way we built
it. How should it be used?
392
00:25:29,480 --> 00:25:33,380
Because that's that's, for me,
what the user manual is, right?
393
00:25:34,460 --> 00:25:38,060
My brother eats user manuals. I
don't. I just fiddle around with
394
00:25:38,060 --> 00:25:42,580
the new stuff, and then I
probably hit the wall, and I
395
00:25:42,580 --> 00:25:46,060
have to go back to the user
manual. Why? While my brother
396
00:25:46,180 --> 00:25:51,220
reading the user manual is much
cleverer than I am, and he gets
397
00:25:51,220 --> 00:25:57,040
faster in no time understanding
what the system is, how it works
398
00:25:57,160 --> 00:25:59,860
and how it should be used. And
this leading back to the
399
00:25:59,860 --> 00:26:02,880
stewardship we have as as
people,
400
00:26:02,999 --> 00:26:04,319
Nikolaos Kaintantzis: my
thoughts is going now in a
401
00:26:04,319 --> 00:26:07,679
direction. If it becomes too
political, please stop me and we
402
00:26:07,679 --> 00:26:10,439
do something else, because I
still want to visit some
403
00:26:10,439 --> 00:26:15,299
countries in the next years.
What really touched me, and I
404
00:26:15,299 --> 00:26:20,999
agree with that, is the idea of
fair share, and we have now a
405
00:26:21,059 --> 00:26:25,999
technology that takes away,
maybe jobs, but that can makes
406
00:26:25,999 --> 00:26:29,659
us much more wealthier, and now
we can share. But when I look
407
00:26:29,659 --> 00:26:34,279
what happens in the world in
politics, I'm just wondering,
408
00:26:34,879 --> 00:26:37,639
will it be a nice share, or we
would just be the richer
409
00:26:37,639 --> 00:26:42,459
Richards, so what I'm with you
is I really want to avoid some
410
00:26:42,639 --> 00:26:48,699
in reverse, in reversible, in
reversible damage. But I don't
411
00:26:48,699 --> 00:26:52,659
know if those people who are in
charge really think that way, or
412
00:26:52,659 --> 00:26:55,119
they're just playing now
whatever they have to play the
413
00:26:55,119 --> 00:26:57,999
role. But there's a wife, and
that's what makes me a little
414
00:26:57,999 --> 00:27:01,019
bit sad. But I think now we are
going we derailed somewhere in a
415
00:27:01,019 --> 00:27:05,459
discussion which is not part of
agent unleashed, which is just a
416
00:27:05,699 --> 00:27:09,359
criticism on society or
criticism of the riches I don't
417
00:27:09,359 --> 00:27:10,619
know. Mark help me.
418
00:27:12,480 --> 00:27:17,100
Mark Richards: I was just about
Ali on the side track because so
419
00:27:17,100 --> 00:27:22,340
my son is very much an
environmental activist, and he
420
00:27:22,340 --> 00:27:25,400
is working on making me more
environmentally aware every
421
00:27:25,400 --> 00:27:30,140
week. So we have a lot of
discussions about, you know, the
422
00:27:30,140 --> 00:27:35,000
challenge of environmentalism is
that some countries may be doing
423
00:27:35,000 --> 00:27:39,980
something about it, while a lot
of the world is not. And how do
424
00:27:39,980 --> 00:27:42,040
you deal with that, because it's
the same kind of thing, right?
425
00:27:42,340 --> 00:27:45,820
If you're ethical
environmentally, but your
426
00:27:45,820 --> 00:27:49,300
neighbors aren't, and we've got
a great example with Australia,
427
00:27:49,300 --> 00:27:51,640
we've been trying to be more
ethical about managing our fish
428
00:27:51,640 --> 00:27:56,680
stocks and putting a lot of
licensing controls over it to
429
00:27:56,680 --> 00:28:00,040
allow fish stocks to replenish.
But we have neighboring
430
00:28:00,040 --> 00:28:03,300
countries who are poor, and they
come and they fish in our waters
431
00:28:04,620 --> 00:28:08,460
because they are not living the
same things. And I think you get
432
00:28:08,460 --> 00:28:13,260
that same question with ethics,
becomes really quite
433
00:28:13,260 --> 00:28:16,260
fascinating, but I reckon we
should have a crack at Harm's
434
00:28:16,260 --> 00:28:20,960
answer before we get distracted
by jiggles. So harm came in and
435
00:28:20,960 --> 00:28:24,380
said, What about a scenario with
using AI to enforce ethics? And
436
00:28:24,380 --> 00:28:26,900
actually, the thing that came
straight to mind was, I was
437
00:28:26,900 --> 00:28:32,180
watching a video a couple of
days ago, somebody was going
438
00:28:32,180 --> 00:28:34,820
through, of course, they leaked
the news Claude 4.5 system
439
00:28:34,820 --> 00:28:38,300
prompt. And somebody was going
through doing a bit of a
440
00:28:38,300 --> 00:28:40,900
dissection about interesting
things they found in the system
441
00:28:40,900 --> 00:28:45,040
prompt, and one of the things
they found was an awful lot of
442
00:28:45,100 --> 00:28:48,520
anti plagiarism prompting. So
there are a whole bunch of
443
00:28:48,520 --> 00:28:52,480
sections in the new 4.5 system
prompt to actually stop it from
444
00:28:52,480 --> 00:28:58,600
doing direct cited quotes that
might breach copyright. And you
445
00:28:58,600 --> 00:29:01,560
can see they're starting to try
and get to, well, people are
446
00:29:01,560 --> 00:29:05,520
using it, and it's doing a whole
bunch of copyright breaches. You
447
00:29:05,520 --> 00:29:08,220
know, the suspicion, of course,
is they're protecting themselves
448
00:29:08,220 --> 00:29:11,280
from lawsuits. People are
actually starting to get into,
449
00:29:12,000 --> 00:29:14,820
how do I prompt, whether it's in
a system prompt, or whether it's
450
00:29:14,820 --> 00:29:19,200
in an Asian definition, Agent
definition, how do I prompt for
451
00:29:19,200 --> 00:29:23,120
ethical behavior and or, How do
I prompt to detect unethical
452
00:29:23,120 --> 00:29:27,020
behavior? So people are starting
to do it. I think the really
453
00:29:27,020 --> 00:29:30,320
interesting question becomes,
how do you codify your ethics in
454
00:29:30,320 --> 00:29:32,480
a way that a prompt is capable
of interpreting?
455
00:29:32,900 --> 00:29:36,800
Stephan Neck: Let me pick up
this one mark, going back to
456
00:29:37,520 --> 00:29:41,320
Stephan beers, viable system
model. There's this system
457
00:29:41,320 --> 00:29:45,220
number five, which is the
normative system, ring fencing,
458
00:29:45,220 --> 00:29:51,460
the other systems in a viable
system. And I high, I'm a high
459
00:29:51,580 --> 00:29:56,140
advocate on this question that
comes from harm. It's about the
460
00:29:56,140 --> 00:30:03,900
scenario that allows the. A
company to from onboarding
461
00:30:03,900 --> 00:30:07,500
people during their work until
they leave a company to
462
00:30:07,500 --> 00:30:11,400
accompany them in a very
supportive way, right? Even
463
00:30:11,400 --> 00:30:15,600
saying no to certain actions, as
you mentioned. Am I allowed to
464
00:30:15,600 --> 00:30:20,840
use such a prompt and before the
prompt gives back a result, you
465
00:30:20,840 --> 00:30:25,280
probably get back a message,
this is not how we think as a
466
00:30:25,280 --> 00:30:29,840
company. And what you just, what
you just mentioned, mark for me
467
00:30:30,140 --> 00:30:34,460
is, I would say, and I'm being
very blunt about this, 190, 9%
468
00:30:34,880 --> 00:30:39,200
of all companies don't have
ethics. They can't express it.
469
00:30:39,500 --> 00:30:43,540
How can you bring it into a
scenario, and bring it into the
470
00:30:43,540 --> 00:30:47,800
AI world if you don't know what
you are talking about, if you
471
00:30:47,800 --> 00:30:53,140
don't know how to how to
describe it and to even script
472
00:30:53,140 --> 00:30:57,520
or code it in a system. But the
scenario, yes, we have to deal
473
00:30:57,520 --> 00:31:02,100
with that one. And it's again,
going back to leadership, this
474
00:31:02,100 --> 00:31:07,200
is one of the big obligations
you have as a C level to make it
475
00:31:07,200 --> 00:31:08,940
happen in your organization.
476
00:31:09,240 --> 00:31:12,180
Nikolaos Kaintantzis: I'm now
sad and sorry, mad and happy in
477
00:31:12,180 --> 00:31:15,420
the same time, because I just
realized it's something we have
478
00:31:15,420 --> 00:31:18,240
to do as companies, as
individuals. I just looked the
479
00:31:18,240 --> 00:31:21,080
whole time for an answer. How
can the whole society enforce
480
00:31:21,080 --> 00:31:24,140
this. And that's much more
difficult, because I still think
481
00:31:24,140 --> 00:31:28,640
some companies, some crazy
people, will not do that. So,
482
00:31:29,480 --> 00:31:33,260
but yeah, being there as a
company, as a person, as a
483
00:31:33,260 --> 00:31:36,500
human, and saying this, this is
our ethic, and I want to enforce
484
00:31:36,500 --> 00:31:40,180
this in our AI, that makes
absolutely sense. I haven't seen
485
00:31:40,180 --> 00:31:44,140
a solution, how we can enforce
it worldwide. So one discussion
486
00:31:44,140 --> 00:31:48,340
we had with the family was, what
happens to masks AI, which was
487
00:31:48,340 --> 00:31:53,200
somehow half an Nazi, half
whatever. So where was there the
488
00:31:53,260 --> 00:31:57,340
ethics, or is it normal? What is
what it's doing? So I think some
489
00:31:57,340 --> 00:32:01,020
companies will will have, as I
said, have ethics and should and
490
00:32:01,020 --> 00:32:04,320
will enforce it, and others will
just take whatever they have. I
491
00:32:04,740 --> 00:32:07,320
also thought about it, what's,
what's with open source and AI?
492
00:32:07,560 --> 00:32:11,340
If an AI is open source, I can
remove all the control parts.
493
00:32:11,340 --> 00:32:13,980
That was where I'm going with
the questions I'm really happy
494
00:32:13,980 --> 00:32:16,380
you answered in a different way.
There was the happy part and the
495
00:32:16,380 --> 00:32:18,780
sad part was what I just said,
sorry if I make the day.
496
00:32:18,780 --> 00:32:21,680
Everybody said it wasn't,
wasn't, the intention.
497
00:32:21,740 --> 00:32:22,820
Mark Richards: That's not your
job. Niko,
498
00:32:26,360 --> 00:32:29,180
Nikolaos Kaintantzis: clever
guy, thank you so much. Let's
499
00:32:29,180 --> 00:32:32,480
use also, let's mess up with our
brains. So let's, let's go to
500
00:32:32,480 --> 00:32:38,180
jiggle and where, each time I
have the chance to give you
501
00:32:39,080 --> 00:32:42,760
Yeah, to activate other parts of
the brain, and this time, it's a
502
00:32:42,760 --> 00:32:46,840
hard one. Well, also hard for
me, if AI and ethics were an
503
00:32:46,840 --> 00:32:51,640
exhibit in a museum, again, if
AI and ethics was an exhibit in
504
00:32:51,640 --> 00:32:56,260
a museum, what does it look
like, and which museum exhibits
505
00:32:56,260 --> 00:32:56,620
it
506
00:32:56,860 --> 00:32:59,080
Mark Richards: so I'm going to
go first. And when I first
507
00:32:59,080 --> 00:33:01,800
looked at this one, I went, I've
got absolutely no idea. I left
508
00:33:01,800 --> 00:33:04,560
it alone, and then I picked it
up again, and it was like I got
509
00:33:04,560 --> 00:33:08,040
an instant answer. So here's my
instant answer. It's a very
510
00:33:08,040 --> 00:33:11,400
lightly re flavored Mongols
visit. So if you think about the
511
00:33:11,400 --> 00:33:14,940
story of the Mongols in a
museum, they came, they saw they
512
00:33:14,940 --> 00:33:18,300
conquered. They became the thing
that they conquered. And the
513
00:33:18,300 --> 00:33:21,260
question that I leave with you
is, in this scenario, are the
514
00:33:21,260 --> 00:33:23,300
humans, the Mongols, or the
Chinese?
515
00:33:24,020 --> 00:33:24,860
Unknown: That's a good one.
516
00:33:26,060 --> 00:33:28,040
Nikolaos Kaintantzis: Oh yeah.
But by the way, before, before I
517
00:33:28,040 --> 00:33:34,400
hand over to Stephan, each time
we we use Mark's custom chat GPT
518
00:33:34,400 --> 00:33:37,520
to generate an image. And this
time, I really love the image,
519
00:33:37,520 --> 00:33:40,720
because it showed exactly what
you're seeing. So who is which?
520
00:33:40,780 --> 00:33:46,000
So who is? Who is the origin,
who is the master, who is giving
521
00:33:46,000 --> 00:33:50,860
whom the torch? What the heck is
the kids doing? So I think,
522
00:33:50,860 --> 00:33:53,920
unfortunately, it just
disappears from LinkedIn as soon
523
00:33:53,920 --> 00:33:57,040
as we, as we, as we did the
show. So I think maybe we should
524
00:33:57,040 --> 00:34:00,780
have, like a gallery walk around
all the images this time I
525
00:34:00,780 --> 00:34:04,980
really liked what the co
creation with AI was. And at the
526
00:34:04,980 --> 00:34:09,480
end, I think this time AI was,
was, was the master of the
527
00:34:09,480 --> 00:34:12,660
ideas, and I just tweaked
around. So I really like this,
528
00:34:12,660 --> 00:34:16,620
this picture and this one just
symbolize your story. Mark at
529
00:34:16,680 --> 00:34:19,380
the end, I like this very much.
Stephan,
530
00:34:19,740 --> 00:34:23,420
Stephan Neck: yeah, well, I You
said, Where would you see it? In
531
00:34:23,420 --> 00:34:27,980
a museum. And if you, if you go
to museums, you look at things.
532
00:34:28,040 --> 00:34:34,460
And the idea behind magical was,
what if you look at something
533
00:34:34,460 --> 00:34:37,400
and it looks back and you
recognize, oh, it's a mirror.
534
00:34:37,460 --> 00:34:41,080
It's an algorithmic mirror,
because it's about math, it's
535
00:34:41,080 --> 00:34:45,160
about statistics, it's about
data and how we deal with that
536
00:34:45,160 --> 00:34:49,360
so reflecting our digital
selves. And where would I go to
537
00:34:49,420 --> 00:34:55,900
the Deutsches Museum in Munich,
which probably has written marks
538
00:34:56,560 --> 00:35:01,740
exhibit, they show the part.
Past, but they also show what
539
00:35:01,740 --> 00:35:06,420
could the future look like? So
they really span ages, and
540
00:35:06,420 --> 00:35:11,520
according to your interest, you
go and you have a look. And so
541
00:35:11,760 --> 00:35:16,980
the algorithmic mirror. It's our
biases we already talked about
542
00:35:16,980 --> 00:35:20,900
that, our blind spots, our
hopes, our flaws, but also our
543
00:35:20,900 --> 00:35:26,600
values, because we said it's
about ethics. And if I could
544
00:35:26,600 --> 00:35:30,860
create such an exhibit, it
should be highly interactive. No
545
00:35:30,860 --> 00:35:34,100
matter who is in front of the
mirror, it acts differently,
546
00:35:34,100 --> 00:35:38,060
right? Oh, and instead of the
signature of Yeah, and instead
547
00:35:38,060 --> 00:35:42,340
of the signature of an artist,
there will be a quote that says,
548
00:35:42,340 --> 00:35:46,180
The future is not yet written.
You are the architect. How do
549
00:35:46,180 --> 00:35:46,960
you deal with it?
550
00:35:47,260 --> 00:35:53,500
Nikolaos Kaintantzis: Sounds
like matrix, no kind of so I was
551
00:35:53,500 --> 00:35:57,700
somehow just thinking and then
starting writing levels
552
00:35:57,700 --> 00:36:01,140
thinking. So I thought it has to
be something transparent,
553
00:36:01,440 --> 00:36:04,620
because we talk about ethics, it
should be transparent. And then
554
00:36:04,620 --> 00:36:07,860
I realized transparent has
different meanings in German and
555
00:36:07,860 --> 00:36:11,400
in English. So if in English,
something is transparent, you
556
00:36:11,400 --> 00:36:15,000
can look through it like a
glass. And that's the picture I
557
00:36:15,000 --> 00:36:17,400
had first in my mind. And then I
realized in German, it's
558
00:36:17,400 --> 00:36:22,160
something you look how it works,
so you see all the gears and
559
00:36:22,160 --> 00:36:24,980
everything there, because it's
transparent. You can see how the
560
00:36:24,980 --> 00:36:28,400
machine is working. It's not a
black box, it's you see
561
00:36:28,400 --> 00:36:30,860
everything, and that's
transparent. And then English is
562
00:36:30,860 --> 00:36:34,040
something I don't care. I just
look through and then I realize
563
00:36:34,400 --> 00:36:37,040
I would like to see the gears
somehow. But still, it's now
564
00:36:37,040 --> 00:36:42,280
some gears in the glass that's
now the art the exhibit I'm
565
00:36:42,280 --> 00:36:45,700
having there, and they're
working there, two little girls
566
00:36:45,700 --> 00:36:49,960
just working, working in a huge
glass. And the museum is
567
00:36:50,200 --> 00:36:54,460
everywhere. There are just
drones carrying this glass
568
00:36:54,820 --> 00:36:59,800
around our heads and flies
around. And hopefully it has
569
00:36:59,800 --> 00:37:04,200
enough ethics not to let it fall
down on my head. That was what I
570
00:37:04,200 --> 00:37:09,120
came with. So was a cool journey
in my head. At least, sounds
571
00:37:09,120 --> 00:37:13,920
good about to continue is what
to continue. Let's go to the
572
00:37:13,920 --> 00:37:19,980
story. Mark you started with the
story about the car and deciding
573
00:37:20,340 --> 00:37:24,200
who to kill or who to hurt more
if you have to do an accident
574
00:37:24,200 --> 00:37:28,280
you cannot break on time. I
would like to know over what you
575
00:37:28,280 --> 00:37:32,060
think AI should be allowed to
decide, and if so, who's
576
00:37:32,060 --> 00:37:33,860
accountable Mark, what do you
577
00:37:33,860 --> 00:37:37,460
Mark Richards: think so? Look,
it's all for me about the risk
578
00:37:37,460 --> 00:37:41,620
and the consequence of the wrong
decision in any decision. But
579
00:37:41,620 --> 00:37:44,860
then I think if I put an ethics
over on it, because that's like,
580
00:37:44,860 --> 00:37:49,000
that's the default answer if you
are using AI in some fashion for
581
00:37:49,300 --> 00:37:53,200
automation of a capability, or
automation of a decision or a
582
00:37:53,200 --> 00:37:56,920
workflow, or anything else, you
got to look at. Well, how much
583
00:37:56,920 --> 00:37:59,740
human control do I need? Is a
combination of the
584
00:37:59,740 --> 00:38:03,420
sophistication of my ability to
implement the AI and the risk
585
00:38:03,420 --> 00:38:06,720
level of something being made as
a wrong decision. Ethics just
586
00:38:06,720 --> 00:38:09,180
happens to be a branch, and I
think that's where it gets
587
00:38:09,180 --> 00:38:12,060
really interesting, because
again, it goes back to that
588
00:38:12,060 --> 00:38:16,980
awareness. And if you again, if
you look a little bit more
589
00:38:16,980 --> 00:38:19,920
seriously at the way people are
approaching engineering with AI
590
00:38:19,920 --> 00:38:23,540
these days, you'll hear a term
that you know the hardware guys
591
00:38:23,540 --> 00:38:26,540
love, and occasionally you hear
it in software, which is failure
592
00:38:26,540 --> 00:38:29,780
modes. And I still remember back
when I was an architect, I would
593
00:38:29,780 --> 00:38:32,660
occasionally have to write a
failure modes effect analysis.
594
00:38:33,200 --> 00:38:36,800
How might this thing fail? What
would be the impact? How might I
595
00:38:36,800 --> 00:38:41,800
get mitigate for it? Etc, etc,
etc. And everybody looking at
596
00:38:41,800 --> 00:38:46,780
engineering AI solutions
stresses failure modes analysis
597
00:38:47,560 --> 00:38:53,320
through a behavioral evaluation
perspective. And the challenge
598
00:38:53,320 --> 00:38:55,660
to you is think of the different
ways it might fail. It might
599
00:38:55,660 --> 00:38:58,180
fail because it gets the wrong
answer. It might fail because it
600
00:38:58,180 --> 00:39:00,960
uses too many tokens. It might
fail because it takes too long.
601
00:39:01,500 --> 00:39:04,980
Goodness knows. But I think the
question is, you ask yourself,
602
00:39:04,980 --> 00:39:08,700
is when I'm applying AI in a
situation, how might it fail
603
00:39:08,760 --> 00:39:11,940
ethically, and what am I going
to do to defend against that
604
00:39:12,000 --> 00:39:15,900
potential ethical failure? Now,
is it potentially that it's
605
00:39:15,900 --> 00:39:19,260
going to have bias? But let's
say I'm using I'm an HR team,
606
00:39:19,260 --> 00:39:22,520
and I'm using it as part of
screening resumes. When I
607
00:39:22,700 --> 00:39:27,140
advertise job listings, really
logical productivity application
608
00:39:27,140 --> 00:39:29,300
of it, well, what's the chance
that it's going to have bias
609
00:39:29,300 --> 00:39:32,180
about how it picks job listings?
It doesn't mean you can't use it
610
00:39:32,180 --> 00:39:35,600
to screen resumes, but it does
mean that when you think about
611
00:39:35,600 --> 00:39:39,620
using it to screen resumes,
you've got to consider the way
612
00:39:39,620 --> 00:39:43,240
in which bias or unethical
decisions might enter that
613
00:39:43,240 --> 00:39:46,300
decisioning process and be
explicit how you're going to
614
00:39:46,300 --> 00:39:50,020
actually validate that it hasn't
done so. And that's where I
615
00:39:50,020 --> 00:39:55,240
think the combo, for me is about
going, on the one hand, what's
616
00:39:55,240 --> 00:39:59,140
the risk associated with a wrong
decision, and recognizing that
617
00:39:59,140 --> 00:40:02,220
ethical risk is. Very much a
part of that potential risk. But
618
00:40:02,220 --> 00:40:06,120
the other is, what is my current
ability to safeguard against
619
00:40:06,120 --> 00:40:10,140
that risk? What sophistication
do I have in the way that I
620
00:40:10,140 --> 00:40:14,700
currently employ and deploy AI
to enable me to effectively
621
00:40:14,700 --> 00:40:17,880
manage it? And if my
sophistication is low and my
622
00:40:17,880 --> 00:40:21,780
risk is high, I probably
shouldn't be doing it at all.
623
00:40:22,560 --> 00:40:26,300
But if my sophistication grows,
then realistically, my risk
624
00:40:26,300 --> 00:40:29,720
becomes more effectively
controlled. So you know, who's
625
00:40:29,720 --> 00:40:31,700
accountable? Well, at the end of
the day, the company's
626
00:40:31,700 --> 00:40:34,700
accountable. They're taking a
stance, and it's a risk posture,
627
00:40:36,140 --> 00:40:39,680
but it's not black and white.
It's about awareness, asking
628
00:40:39,680 --> 00:40:42,700
yourself questions, and
evaluating the trade offs, and
629
00:40:42,700 --> 00:40:46,000
every discussion I have with
anybody in business about
630
00:40:46,000 --> 00:40:51,220
responsible AI full stop, right?
Is you are choosing a risk
631
00:40:51,220 --> 00:40:54,700
posture. Make sure you're making
an informed choice.
632
00:40:56,200 --> 00:40:59,740
Stephan Neck: And this mark will
be related for me to what are
633
00:40:59,740 --> 00:41:04,440
the stakes? Right? Is it? Do I
have low stakes? I would say,
634
00:41:04,440 --> 00:41:09,240
Okay, go for it. Help me, right?
You have high autonomy. My
635
00:41:09,240 --> 00:41:13,260
stakes are low, so I probably
have low risks in it. The
636
00:41:13,260 --> 00:41:18,000
question is, when it comes to a
medium stake, like we talked
637
00:41:18,000 --> 00:41:23,180
about PILOTs and CO pilots. The
last episode, right? You're a
638
00:41:23,180 --> 00:41:29,900
pilot in a plane. Your plane is
full of AI helping you. There's
639
00:41:30,560 --> 00:41:36,500
an automatic navigation system.
There's GPS whatever. There is
640
00:41:36,500 --> 00:41:42,280
kind of a co working space
where, based on my stakes, I
641
00:41:42,280 --> 00:41:48,280
have to decide, where do I
interact? Where do I set my
642
00:41:48,280 --> 00:41:53,500
stakes? And say, now I decide.
And the other one where I have a
643
00:41:53,500 --> 00:41:56,860
high stake, let's think about
I'm in a hospital in palliative
644
00:41:56,920 --> 00:42:01,620
care. I don't want the machine
to decide if I live or not, it's
645
00:42:01,620 --> 00:42:06,360
still what I've written down on
a sheet of paper as a patient
646
00:42:06,480 --> 00:42:10,980
that someone, for me, will
decide if I'm out of mind, if
647
00:42:10,980 --> 00:42:15,060
I'm not able to do that. So for
me, it's about stakes, and as
648
00:42:15,060 --> 00:42:19,080
you said, Mark, yeah, in the
end, those who create those
649
00:42:19,080 --> 00:42:24,980
systems and are using AI, they
have the primary responsibility
650
00:42:24,980 --> 00:42:27,920
also for the future, as you
mentioned, Niko, for the future
651
00:42:27,920 --> 00:42:32,000
generations. And again, this
leads me back to, as you said,
652
00:42:32,000 --> 00:42:36,140
Mark fitness for use, right? You
have an architecture that is fit
653
00:42:36,140 --> 00:42:42,100
for use, and on top comes
regulatory aspects, government,
654
00:42:42,280 --> 00:42:46,900
aspects that have to be baked in
as well, which is part of
655
00:42:46,900 --> 00:42:50,860
ethics, where there are certain
values behind it. Regulator
656
00:42:50,860 --> 00:42:56,500
tries to, tries to protect the
majority of the people against
657
00:42:56,500 --> 00:43:01,860
harm, and how you deal with
certain risks. So, yeah, there's
658
00:43:01,860 --> 00:43:05,400
a there's a huge range of
stakes, and based on the
659
00:43:05,400 --> 00:43:09,360
context, again, based on the
stakes, I would answer the
660
00:43:09,360 --> 00:43:13,500
question, who's responsible,
who's dealing with it, and how
661
00:43:13,500 --> 00:43:14,100
you deal with
662
00:43:15,000 --> 00:43:16,620
Nikolaos Kaintantzis: it? Let's
make the practical. I like the
663
00:43:16,620 --> 00:43:21,860
discussion. We call the episode
unleashed agents. Unleashed
664
00:43:22,520 --> 00:43:26,120
change, unleashed whatever, but
unleashed. What should we
665
00:43:26,300 --> 00:43:31,280
unleash to make this practical,
or let's ask differently. What
666
00:43:31,340 --> 00:43:34,700
cultural shifts do we need to be
prepared for this
667
00:43:34,700 --> 00:43:39,320
accountability, for ethics to be
part of our daily lives? So what
668
00:43:39,320 --> 00:43:45,100
can we do? What can our
spectaculars do? Our listeners
669
00:43:45,760 --> 00:43:47,260
do? Destroy the
670
00:43:53,980 --> 00:43:57,880
Stephan Neck: whole flow. Niko
as a moderator, is fully
671
00:43:57,880 --> 00:44:02,520
unleashed, and we're following,
okay, so let me throw in two
672
00:44:02,520 --> 00:44:06,060
things that are still the same.
It shouldn't change the culture
673
00:44:06,060 --> 00:44:09,300
that we have. And if we say
culture is the sum of our
674
00:44:09,300 --> 00:44:14,040
behavior, we still should build
carefully. And if things are
675
00:44:14,040 --> 00:44:16,800
broken, you have to fix it.
That's how I was brought up,
676
00:44:16,800 --> 00:44:22,280
right? You fix it. If you break
it or you clean up. Second one
677
00:44:22,280 --> 00:44:26,180
we talked about it is, data is
still a liability, right? You
678
00:44:26,180 --> 00:44:32,120
leave traces, you give your
data, and now what changes
679
00:44:32,120 --> 00:44:38,300
drastically, and I as an old
dog, compared to mark son, who's
680
00:44:38,360 --> 00:44:42,820
a digital native, right? What is
a the question for me is, what
681
00:44:42,820 --> 00:44:46,300
is the digital digital
citizenship? We are all
682
00:44:46,300 --> 00:44:51,160
citizens. We all belong to a
certain group, to a client, to a
683
00:44:51,160 --> 00:44:56,620
city, to people. We have
different languages. So how does
684
00:44:56,620 --> 00:44:59,500
that evolve? That would be one
question that's for sure, a
685
00:44:59,500 --> 00:45:07,500
digital. Little shift. And the
other one is, yeah, still, I
686
00:45:07,500 --> 00:45:12,540
think that should be the same,
like in the old days. Nowadays.
687
00:45:12,540 --> 00:45:16,980
It's healthy skepticism. There's
always an option. There's new
688
00:45:16,980 --> 00:45:23,660
stuff to use, but, yeah, staying
skeptical, skeptic a little bit.
689
00:45:24,860 --> 00:45:29,060
Don't just jump into the pool.
It might not be deep enough, and
690
00:45:29,060 --> 00:45:29,720
you will hit your
691
00:45:30,620 --> 00:45:32,360
Nikolaos Kaintantzis: head. I
lacked your positivity. How you
692
00:45:32,360 --> 00:45:35,060
started and say, we are doing
this anyway. You cannot change
693
00:45:35,060 --> 00:45:37,940
the culture. We are in an
environment where we'll care for
694
00:45:37,940 --> 00:45:41,860
it. You just have to continue
caring. I love, I love this part
695
00:45:41,860 --> 00:45:45,700
of the answer from for me, I
think what we have to change or
696
00:45:45,700 --> 00:45:50,860
do differently is I still think
leadership management thinks the
697
00:45:50,860 --> 00:45:54,280
whole AI thing is an IT thing,
and ethics should be a
698
00:45:54,280 --> 00:45:59,560
compliance issue. So let's move
it to a compliance department
699
00:45:59,620 --> 00:46:03,060
and or HR department, if there
are some other rules there and
700
00:46:03,060 --> 00:46:06,720
that will solve the problem. And
as you heard before, I'm out,
701
00:46:06,720 --> 00:46:10,920
and I'm now the sad guy this
episode, the warning guy, I
702
00:46:10,920 --> 00:46:14,940
think, no, not. I think, I know
it's a responsibility of the
703
00:46:14,940 --> 00:46:20,280
whole society. So we need, we
need models who are traceable.
704
00:46:20,280 --> 00:46:24,680
We need to know where data come
from. We need to know what is
705
00:46:24,680 --> 00:46:28,940
fake or not. Is it generated by
AI or not? So I'm deeply
706
00:46:28,940 --> 00:46:33,020
concerned seeing pictures that
look like original pictures from
707
00:46:33,020 --> 00:46:36,860
the past, and now we are faking
and changing history, and I
708
00:46:36,860 --> 00:46:42,220
think this is not continue as we
are doing now. It's really being
709
00:46:42,220 --> 00:46:45,340
aware that the society issue and
the society have to do
710
00:46:45,340 --> 00:46:48,640
something, and it's not just the
IT department or department or
711
00:46:48,640 --> 00:46:52,660
compliance department. So I
think we need not a culture
712
00:46:52,660 --> 00:46:55,540
shift, but we need a shift in
thinking it's our all
713
00:46:55,540 --> 00:46:59,920
responsibility, and not the tech
moguls that will take care for
714
00:46:59,920 --> 00:47:02,160
us. They will not at the moment,
what I see,
715
00:47:02,520 --> 00:47:05,460
Mark Richards: I'm gonna go
slightly different way, just for
716
00:47:05,460 --> 00:47:10,260
change. Have either of you seen
Mad Men in the streets
717
00:47:10,260 --> 00:47:15,840
everywhere? But there's a movie.
It was a TV series, and it was
718
00:47:15,840 --> 00:47:20,720
set in advertising in the world
of the 60s, and fundamentally,
719
00:47:20,720 --> 00:47:23,240
it told stories about the
craziness of the world of the
720
00:47:23,240 --> 00:47:27,020
60s through the lens of somebody
in advertising. So of course,
721
00:47:27,020 --> 00:47:31,280
this is still the days when
cigarettes paid for an awful lot
722
00:47:31,280 --> 00:47:35,840
of advertising. So big tobacco
was was a part of the game. It
723
00:47:35,840 --> 00:47:39,140
was a world where women's place
in the office was very
724
00:47:39,140 --> 00:47:44,140
different. And I watched it with
my son two or three years ago
725
00:47:44,140 --> 00:47:49,000
now, and we sat there every
episode, and I would sit going,
726
00:47:49,420 --> 00:47:52,060
like, I know the workplace has
changed a lot since I was young,
727
00:47:52,120 --> 00:47:56,380
but it's a startling to see how
it was then. And and he would
728
00:47:56,380 --> 00:47:59,440
look at it and go, Dad, is that
actually possible? Like, is this
729
00:47:59,440 --> 00:48:02,760
real? Surely they've got to be
inventing this because the way
730
00:48:02,760 --> 00:48:07,020
they're treating different types
of people and and you look at
731
00:48:07,500 --> 00:48:11,340
the way our workplace has
changed in our lifetimes, I know
732
00:48:11,340 --> 00:48:14,100
when I started working, nobody
got training about sexual
733
00:48:14,100 --> 00:48:18,480
harassment or bullying, talk
about privacy and security and
734
00:48:18,480 --> 00:48:22,040
personal information. All the
things that we have learned in
735
00:48:22,040 --> 00:48:25,400
our businesses are important to
establish expectations for with
736
00:48:25,400 --> 00:48:29,480
people and to teach them about
because we needed to change our
737
00:48:29,480 --> 00:48:33,500
workplace, right? And a lot of
those changes were really about
738
00:48:33,500 --> 00:48:36,320
changing our workplace to be
more inclusive for everybody who
739
00:48:36,320 --> 00:48:40,040
was a part of it. And you know,
they took years, because I would
740
00:48:40,040 --> 00:48:43,240
periodically go, hey, the life
of women in the corporate world
741
00:48:43,240 --> 00:48:46,240
is much better now than it used
to be. And then I would coach
742
00:48:46,240 --> 00:48:50,140
women who would go, No, it's not
despite years and years and
743
00:48:50,140 --> 00:48:53,260
years of mindfulness and
training and awareness campaigns
744
00:48:53,260 --> 00:48:57,280
and all the rest of it. And I'm
not going to suggest that AI
745
00:48:57,280 --> 00:48:59,500
agents being part of the
workplace is anything like the
746
00:48:59,500 --> 00:49:02,880
issues we had with equality in
the workplace, but I'm going to
747
00:49:02,880 --> 00:49:05,100
go it's really disruptive,
because now we're going to have
748
00:49:05,100 --> 00:49:09,000
a workplace that's got people,
humans and agents, and it's
749
00:49:09,240 --> 00:49:12,180
going to have a whole bunch of
expectations raised about
750
00:49:12,180 --> 00:49:16,200
awareness, whether it's
awareness of agents for ethics
751
00:49:16,680 --> 00:49:19,920
or awareness of people for the
way in which ethical concerns
752
00:49:19,920 --> 00:49:24,500
arise In the world of AI, it's
going to take years of real
753
00:49:24,740 --> 00:49:31,580
education, mindfulness and clear
action, because I know, I don't
754
00:49:31,580 --> 00:49:34,280
know how many times I've been
through sexual harassment
755
00:49:34,280 --> 00:49:36,860
training to make sure that I'm
aware of it, but every now and
756
00:49:36,860 --> 00:49:39,620
again, I'll see something in it
while I'm doing it for the 20th
757
00:49:39,620 --> 00:49:43,420
Time at a different company, and
I go, Oh, that's a really
758
00:49:43,420 --> 00:49:45,760
interesting scenario. I had
never thought of that as
759
00:49:45,760 --> 00:49:49,480
potential bullying or
harassment, and it's actually
760
00:49:49,480 --> 00:49:52,240
awakening me to the possibility.
And I think right now for a lot
761
00:49:52,240 --> 00:49:54,940
of people, it's about just being
awakened to the possibility. But
762
00:49:54,940 --> 00:49:56,860
Nikolaos Kaintantzis: you said
we needed years, and still, the
763
00:49:56,860 --> 00:50:01,320
situation isn't better. Now we
are much more faster. So what
764
00:50:01,320 --> 00:50:02,580
can we do different this time?
765
00:50:02,580 --> 00:50:04,140
Mark Richards: Well, that's,
that's the really interesting
766
00:50:04,140 --> 00:50:07,500
question about it, right? I
mean, there are wells in which
767
00:50:07,800 --> 00:50:11,040
you know the fundamental deal
with AI is, the more that you
768
00:50:11,040 --> 00:50:13,740
try and make your company AI
powered, the more you organize
769
00:50:13,740 --> 00:50:16,860
your information. And as part of
organizing your information, you
770
00:50:16,860 --> 00:50:19,140
find things that need to be
codified, and whether that's
771
00:50:19,140 --> 00:50:23,780
coding standards or values or
ethical principles. You know,
772
00:50:23,780 --> 00:50:26,300
there's all kinds of things that
need to be codified. Once you
773
00:50:26,300 --> 00:50:30,680
get them codified, you can get
agents to scan for them. I
774
00:50:30,680 --> 00:50:34,220
could, for example, have some
kind of ethical behaviors,
775
00:50:34,520 --> 00:50:38,120
ethical interactions. Agent that
scans the transcript of every
776
00:50:38,120 --> 00:50:42,400
meeting that's held on teams
looking for potential ethical
777
00:50:42,400 --> 00:50:46,480
violations and raising red flags
to go it seems like we're seeing
778
00:50:46,480 --> 00:50:49,420
patterns of a particular type of
ethical violation. So I can get
779
00:50:49,480 --> 00:50:52,840
AI to help me to monitor it and
be aware of it, and design
780
00:50:52,840 --> 00:50:55,900
action and response if that
happens to be something I
781
00:50:55,900 --> 00:50:58,900
choose. But of course, to choose
that, I have to also look at the
782
00:50:58,900 --> 00:51:01,620
ethical dilemma of allowing an
agent to scan every interaction
783
00:51:01,620 --> 00:51:06,720
they've had on teams between
people for analysis. So you can
784
00:51:06,720 --> 00:51:10,440
develop a system that starts to
develop a certain amount of self
785
00:51:10,440 --> 00:51:15,060
monitoring and self policing
there, but then it even just
786
00:51:15,060 --> 00:51:18,060
raises the potential complexity
even more, because what if your
787
00:51:18,300 --> 00:51:21,200
ethical self monitoring agents
start to monitor for the wrong
788
00:51:21,200 --> 00:51:23,960
type of behavior, because, like,
if you're using Croc, there's a
789
00:51:23,960 --> 00:51:24,800
fair chance of it,
790
00:51:27,920 --> 00:51:31,100
Stephan Neck: but doesn't it
start with how you are brought
791
00:51:31,100 --> 00:51:34,820
up. Because if I look at my
kids, if I look at my
792
00:51:34,820 --> 00:51:40,220
granddaughter, how do I
exemplify what good behavior is?
793
00:51:40,760 --> 00:51:45,880
So they can follow, they can
learn and and, yes, the future
794
00:51:45,880 --> 00:51:49,480
will be totally different after
I'm gone, but it's about
795
00:51:49,480 --> 00:51:56,020
behavior of human beings. And
does AI help? Yes, of course,
796
00:51:56,260 --> 00:52:02,220
does it maybe enforce? Yes,
let's, let's build it in. But,
797
00:52:02,220 --> 00:52:06,840
but the question will be, how do
we create an environment,
798
00:52:06,960 --> 00:52:10,920
starting in a family, in
partnerships, going through our
799
00:52:10,920 --> 00:52:14,880
workspaces, as you mentioned,
Mark, where we where we are kind
800
00:52:14,880 --> 00:52:19,140
to each other, where we treat
each other like equals, like
801
00:52:19,140 --> 00:52:22,940
human beings. Because this is,
this is the behavior I would
802
00:52:22,940 --> 00:52:27,740
like to see, the degree of
support by AI. It depends right
803
00:52:28,220 --> 00:52:31,160
again, it depends on the
context, it depends on the
804
00:52:31,160 --> 00:52:35,240
stakes we have. But just be kind
and do the good job.
805
00:52:35,480 --> 00:52:38,240
Nikolaos Kaintantzis: Great. I
would like to close with much
806
00:52:38,240 --> 00:52:41,020
more open questions than I
prepared. What can we do
807
00:52:41,020 --> 00:52:44,980
differently? How can this look
concrete in practice? We still
808
00:52:44,980 --> 00:52:49,900
called unleashed agents
unleashed. How can successful AI
809
00:52:49,900 --> 00:52:52,960
accountability look like in
practice? You said already some
810
00:52:52,960 --> 00:52:56,020
parts of it, I know, yeah,
811
00:52:56,080 --> 00:52:58,780
Mark Richards: and look, I'd
really stay with if I boil it
812
00:52:58,780 --> 00:53:01,860
all back down. And you know,
it's like a broken record
813
00:53:01,860 --> 00:53:06,720
mindfulness. You've got to start
having the conversations. And I
814
00:53:06,720 --> 00:53:09,540
think it's really important you
start having the conversations
815
00:53:09,540 --> 00:53:13,860
in a way that does not scare
people into not engaging. You
816
00:53:13,860 --> 00:53:16,680
know, there's a there's a world
which is there are people there.
817
00:53:16,680 --> 00:53:19,200
Some people are excited about
AI, some people are scared about
818
00:53:19,200 --> 00:53:22,340
AI, some people who are worried
that it's going to take their
819
00:53:22,340 --> 00:53:24,920
job. Some people are worried
it's going to take over the
820
00:53:24,920 --> 00:53:27,920
world. Like there's all kinds of
attitudes of people towards AI.
821
00:53:27,920 --> 00:53:30,740
Then there's people
understanding how AI might help
822
00:53:30,740 --> 00:53:34,160
them in different types of jobs.
But then there's also, what are
823
00:53:34,160 --> 00:53:37,520
the risks of AI? And I think
there's a risk, and this was
824
00:53:37,520 --> 00:53:42,640
kind of, I'll go there right for
at least a year, the surface for
825
00:53:42,640 --> 00:53:48,280
AI in safe was responsible AI,
so the whole surface that safe
826
00:53:48,280 --> 00:53:50,680
happened, obviously, they've
changed a lot over the last six
827
00:53:50,680 --> 00:53:54,700
to eight months, but for a good
while, if you went AI in safe,
828
00:53:54,760 --> 00:53:57,640
all you saw was responsible AI,
which is fundamentally, how do
829
00:53:57,640 --> 00:54:02,640
we govern AI and ethical AI,
right? And it was missing the
830
00:54:02,640 --> 00:54:07,260
what's exciting about AI, you
just read this long laundry list
831
00:54:07,260 --> 00:54:09,540
of, oh my goodness, these are
all the corporate governance
832
00:54:09,540 --> 00:54:12,240
measures you can have to put in
place before you can responsibly
833
00:54:12,240 --> 00:54:16,860
use this new thing. And that
overwhelms and it drowns
834
00:54:17,040 --> 00:54:20,660
evolution where, for me, I
think, just as you're for most
835
00:54:20,660 --> 00:54:24,080
organizations, you're entering a
mindfulness journey to say, how
836
00:54:24,080 --> 00:54:27,920
can we apply AI in our
organization in a meaningful way
837
00:54:27,920 --> 00:54:30,800
where we get return for our
investment? I think we've got to
838
00:54:30,800 --> 00:54:33,200
have a parallel conversation
that also says, Well, how can we
839
00:54:33,200 --> 00:54:36,560
do that ethically? What are the
implications, and how are we
840
00:54:36,560 --> 00:54:41,800
starting to evolve and take on
an organizational stance in
841
00:54:41,800 --> 00:54:45,580
terms of our belief about AI and
ethics and let it, let it
842
00:54:45,580 --> 00:54:45,940
unfold.
843
00:54:46,300 --> 00:54:47,980
Nikolaos Kaintantzis: Sounds
like you're advocating becoming
844
00:54:47,980 --> 00:54:50,620
teenagers again, being
responsive, responsible,
845
00:54:50,620 --> 00:54:54,100
responsible, responsible, but
sometimes little bit stretch of
846
00:54:54,100 --> 00:54:56,620
boundaries, and then say Sorry,
sorry, sorry, and do it the next
847
00:54:56,620 --> 00:55:00,840
time better and learn and try.
Sounds great. I. I had many,
848
00:55:00,840 --> 00:55:03,480
many bingo, so I listened to
you. For me, it's really get
849
00:55:03,480 --> 00:55:06,720
people involved who are
affected, like mindfulness, you
850
00:55:06,720 --> 00:55:10,500
said before, and build in these
correlations in the code that we
851
00:55:10,500 --> 00:55:14,760
can be sure, yeah, that we have
ethics, yeah, those who are
852
00:55:14,760 --> 00:55:18,840
affected, and some kind of what
harm suggested in beginning to
853
00:55:18,840 --> 00:55:22,460
use AI to enforce this ethics.
Stephan, you have to add
854
00:55:22,460 --> 00:55:24,500
something. You said some things
about about this,
855
00:55:25,160 --> 00:55:29,900
Stephan Neck: yeah, just suing
what we just discussed. People
856
00:55:29,900 --> 00:55:36,020
will redress with AI, how do we
empower them? Organizations,
857
00:55:36,080 --> 00:55:40,340
corporations are responsible for
what they built, how they deploy
858
00:55:40,340 --> 00:55:45,400
it, how it's used. So this is
this accountability. The third
859
00:55:45,400 --> 00:55:51,640
level, where I'm a bit hesitant,
is government. Governments, for
860
00:55:51,640 --> 00:55:55,240
me, at the moment, worldwide,
are not doing a good job when it
861
00:55:55,240 --> 00:55:58,960
comes to this item. We just had
an election in Switzerland about
862
00:55:58,960 --> 00:56:04,440
an electronic ID, almost 50% of
the people are not yet
863
00:56:04,440 --> 00:56:09,360
convinced. But hey, now it's
imposed. It's we have to do it.
864
00:56:09,360 --> 00:56:14,160
So how do we deal with that? And
how do we make it really
865
00:56:14,160 --> 00:56:18,540
successful for human beings on
all different levels?
866
00:56:19,260 --> 00:56:22,280
Nikolaos Kaintantzis: Just one
takeaways, my friends, just one
867
00:56:22,280 --> 00:56:25,640
takeaway. Which one would it be?
Wrapping up, wrapping up the
868
00:56:25,640 --> 00:56:26,240
episodes
869
00:56:26,600 --> 00:56:29,660
Mark Richards: I'll jump in. I'm
going to steal from Mark Twain.
870
00:56:30,020 --> 00:56:32,420
It ain't what you know don't
know that gets you into trouble.
871
00:56:32,420 --> 00:56:35,600
It's what you know for sure that
just don't soak so my question
872
00:56:35,600 --> 00:56:38,480
to you is, what do you know
about AI and ethics?
873
00:56:39,140 --> 00:56:43,840
Stephan Neck: Wow, yeah, that's
that's a strong one mark. So I'm
874
00:56:43,840 --> 00:56:47,140
still looking into the mirror.
AI is reflecting the best and
875
00:56:47,140 --> 00:56:51,460
the worst of our own humanity,
forcing us to decide which
876
00:56:51,460 --> 00:56:56,200
version of ourselves we want to
become we want to encode into
877
00:56:56,260 --> 00:56:56,800
the future.
878
00:56:57,160 --> 00:56:59,800
Nikolaos Kaintantzis: Oh, that's
so great. So some mind now loses
879
00:56:59,800 --> 00:57:04,200
up. My mind sounds like it's
responsibility of our society,
880
00:57:04,260 --> 00:57:07,980
of us as parents, us as
neighbors, us as humans. So
881
00:57:07,980 --> 00:57:12,180
ethics is a socio political
responsibility. And with that, I
882
00:57:12,240 --> 00:57:16,140
give you to the hands of Mark,
and he can wrap up and tell you
883
00:57:16,140 --> 00:57:20,480
what's next week you want to
mute somehow mark on my side, at
884
00:57:20,480 --> 00:57:20,960
least,
885
00:57:21,080 --> 00:57:25,160
Mark Richards: I had to mute
myself to now I'm back and some
886
00:57:25,160 --> 00:57:30,140
post production editing. Next
week, I think will be the first
887
00:57:30,140 --> 00:57:32,960
of a few episodes where we're
going to look at different roles
888
00:57:33,980 --> 00:57:36,560
and talk about and I think for
me, this is kind of inspiring,
889
00:57:36,560 --> 00:57:38,540
because it's what I'm finding a
lot of my coaching
890
00:57:38,540 --> 00:57:42,820
conversations. Is that beginning
of saying to someone, how might
891
00:57:42,820 --> 00:57:45,880
AI be relevant to you and your
job, right? What are the
892
00:57:45,880 --> 00:57:49,480
possible places you could use
it? And, you know, we kicked off
893
00:57:49,480 --> 00:57:52,780
season three with a look at AI
and user stories, and how I
894
00:57:52,780 --> 00:57:55,960
might be disrupting user
stories. We got a bit broader
895
00:57:55,960 --> 00:57:58,300
next week, and we're going to
talk about AI and the product
896
00:57:58,300 --> 00:58:01,860
management role. So looking
forward to that one, and we will
897
00:58:01,860 --> 00:58:05,640
be in stephan's capable hands.
In the meantime, thank you for
898
00:58:05,640 --> 00:58:09,780
joining us, and hopefully you
don't have too many ethical
899
00:58:09,780 --> 00:58:11,640
dilemmas between now and next
week.

Rediger udskrift

SØK BLANT TIDLIGERE EPISODER

Søk etter gamle episoder av Agents Unleashed.

Integrer denne søgefelt på ditt nettsted

Ansvarsfraskrivelse: Podkasten og kunstværket, det finnes indlejret på denne siden, er fra Stephan Neck, Niko Kaintantzis, Ali Hajou, Mark Richards, som tilhører dens eier og ikke er tilknyttet eller godkjent av Listen Notes, Inc.