Transcripción completa de Who's Responsible When AI Decides? Navigating Ethics Without Paralysis

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.