Latest / Rewire To Rise– Beyond The Mindset / How AI Search Is Silently Controlling Your Mind | Just Pedersen & Kaare Jensen
Transcript
- 0:00If you don't know the field, don't that fluent language be
- 0:03like solid ground. In classic search, humans did
- 0:06the heavy cognitive lifting of comparing sources right?
- 0:10What psychological muscles do we risk by doing this?
- 0:15Like allowing this when the machine handles the evaluation,
- 0:18selection and summarization of of us.
- 0:22The old way, the 2025 way, had at least some friction
- 0:27beginning, which is a good thing from my point of view.
- 0:30Now, when you're presented with just one answer, well, you don't
- 0:34have that friction anymore. How do I recognize what I don't
- 0:39know? You know, the, the simple answer
- 0:41to that is, is that's really difficult if you don't have
- 0:44expertise in an area, you need to be very mindful of any
- 0:48answers that are, are, are given to you, whether they're given to
- 0:51you by by search engines, whether they're given to you by
- 0:53humans or you know, even people that you you trust and know.
- 0:56When convenience mass construction, how does the user
- 1:00build an internal alarm system to recognize what has been
- 1:05intentionally or systematically left out of a summary, for
- 1:08example? Remember, it's just a machine,
- 1:11you know, it's an advanced, it's a useful, it's an impressive
- 1:13machine, but it's still just a machine.
- 1:15It doesn't have judgement, doesn't have responsibility,
- 1:17doesn't have understanding in the human sense.
- 1:19It calculates the statistically significant answer, which can be
- 1:23extremely valuable, but that should not be mistaken for
- 1:27authority. So you know, we should use AI
- 1:30with both curiosity and and cool headedness, not as an Oracle,
- 1:34not as a colleague in the human sense, but as a tool that can
- 1:38extend our syncing if we continue to sync ourselves.
- 1:42Because you know the machine can answer, but the human must
- 1:45answer for Hello to everyone. Today we have a special podcast
- 1:55episode in regards to tracing the Microsoft from raw
- 2:00directories to synthesize AI realities by exploring how the
- 2:05mechanics of search alter human consciousness.
- 2:09So I have two important guests here with me, Kerr and Yust.
- 2:14Hello and welcome guys. Thank you.
- 2:16Can you please tell us be really short something about you for
- 2:21those who doesn't know you yet, guys, please.
- 2:24As you can see Marilla, we're we're both 2 very polite
- 2:26gentlemen who always let the other party go first.
- 2:29But on a serious note then, when I don't participate in your very
- 2:34fine podcast. So I'm a professor with the
- 2:37business Academy Copenhagen, which is one of the larger
- 2:42universities of applied science in Denmark.
- 2:46There I manage a couple of e-business bachelor degrees and
- 2:52I, I'm very interested in this subject of, of artificial
- 2:56intelligence and I'm sort of currently trying to use a lot of
- 3:00time to get my head around what it means and how we should
- 3:04respond to it. Well, the short version is that
- 3:07I've been an IT consultant and trainer for half of my life.
- 3:11I'm originally from the media industry, so when I started
- 3:14cheating, it was about Photoshop and then the web and then online
- 3:18marketing. Nowadays it's mainly about AI.
- 3:21But as I've been doing marketing for many years, I've been
- 3:25looking very much forward to this episode because it's really
- 3:28connecting some of of of those thoughts and especially the the
- 3:31human perspective to it because it really does affect the
- 3:34humans. But I'm sure we'll get to that,
- 3:37yeah. Yeah, sure.
- 3:38So let's get started. I have a group.
- 3:40First question for used of course, my dear Sir, used you
- 3:44were talking about the fact that we have moved from such as
- 3:49navigation, which are directories and links to search
- 3:52as interpretation as synthesized answers.
- 3:56In classic search, humans did the heavy cognitive lifting of
- 4:01comparing sources right? What psychological muscles do we
- 4:05risk by doing this? Like allowing this when the
- 4:09machine handles the evaluation, selection, and summarization of
- 4:14of us? Oh, definitely.
- 4:16And that's, that's one of the, that's one of the things to
- 4:19ponder in, in, in this time and age when we talk about search in
- 4:24Denmark, that means searching, you know, via Google and there's
- 4:27other search platforms in other parts of the world, but in
- 4:30Google searches in, in the old ways, you know, the ones that
- 4:33are so 2025 now. Then we were presented with a
- 4:38list of pages and we still sort of had to do the manual
- 4:41selection of, of picking what link to choose.
- 4:45We had to do an, a small, but a, a, an active action there.
- 4:50And now with, with the AI summaries that dominate search,
- 4:54then we don't even have to do that.
- 4:56We're just presented with a synthesized version of what the,
- 4:59the sources they say. And, and, and one, one way sort
- 5:03of to put it here is that if Google was giving us access to
- 5:08this information superhighway, so if it gave us the roads, then
- 5:12now something like Google AI search is, is giving us
- 5:16directions. It's becoming in our cars, the
- 5:19little navigator that tells us where to go.
- 5:21But what we have to remember is we still need to decide where
- 5:25we're going and we still need to be mindful of traffic.
- 5:28We still need to, you know, avoid even if, even if the, the
- 5:32navigator in our car is telling us to take that route, if
- 5:35there's a big sign that says, you know what Rd. closed, we
- 5:39don't proceed. This, of course, being, being
- 5:41the analogy to it. So I think we we need to
- 5:44remember that we still bear the consequences of of of the usage
- 5:50of the information that we find. Thank you.
- 5:54Well, building on this shift, when an AI gives a single
- 6:00defined alternative paragraph instead of 10 blue links like we
- 6:05are used to see right or to receive what's as such, it
- 6:09creates an like an immediate sense of certainty.
- 6:12How does this structural shift from the classical CEO to Geo
- 6:17alter our metacognition the very way we think about and question
- 6:22our own thoughts? It depends on the person for a
- 6:28start, but the the, the, the starting point, at least from
- 6:33where I see it, is that the old way, the 2025 way, had at least
- 6:40some friction building, which is a good thing from my point of
- 6:44view, because you can choose to go down one path, You can choose
- 6:47to go down another path. You'll spend some time
- 6:51evaluating what is trustworthy or what you already recognize.
- 6:55You have a reasons for selecting what is right for you, and you
- 7:00might choose the first thing, you might choose the third one,
- 7:03you might go back the first one. You might look at at the other
- 7:07results and really qualify your decision.
- 7:10Now when you're presented with just one answer, well, you don't
- 7:14have that friction any longer. And for some generations like
- 7:19mine, I will have a certain critical sense to whether or not
- 7:25I could trust that answer. And I might choose to go further
- 7:28down and search for other sources or ask in a different
- 7:31way or use additional tools. I guess that other generations
- 7:36might have a different approach. And I hope that we will succeed
- 7:41in building in some friction, because if there's no friction
- 7:45and if we cannot even trust the answer, we can not in a so good
- 7:49path. And I mean, that can be so many
- 7:51reasons why the answer might not be trustworthy, but it might
- 7:57appear very trustworthy. So our critical sense really has
- 8:01to be sharpened even further. And there's a lot more to that
- 8:06story. But then again, I know that you
- 8:08just have much more to say and I have as well.
- 8:10But but just in short. Yeah.
- 8:12Thank you. I would like to mention
- 8:14something very important for those who don't know yet about
- 8:19GEOGEO means generative engine optimization.
- 8:23So just for those who doesn't know what we are talking about,
- 8:26so they can connect some dots afterwards during our talk,
- 8:29because otherwise doesn't have sense for many others.
- 8:33So thank you, Carl. So you notice that the danger
- 8:38isn't simply that AI is biased, right?
- 8:40We have talked me and you a lot about this, but this bias it's
- 8:44hidden inside an incredible convenient answer.
- 8:48When convenience mass construction, how does the use
- 8:52of build an internal alarm system to recognize what has
- 8:57been intentionally or systematically left out of a
- 9:00summary, for example? Yeah.
- 9:02I mean, I think you know your your question here, Morella,
- 9:05it's a fantastic question. It goes straight to the heart of
- 9:08of what critical thinking is. Because essentially you're
- 9:11asking, how do I recognize what I don't know?
- 9:15And, and you know, the, the simple answer to that is, is
- 9:19that's really difficult. If you don't have expertise in
- 9:22an area, you need to be very mindful of any answers that are,
- 9:25are, are given to you, whether they're given to you by, by
- 9:28search engines, whether they're given to you by humans or, you
- 9:31know, even people that you, you trust and know you can't what an
- 9:34answer is within a field you don't have expertise in.
- 9:37And, and your basic, your internal arm system, if anything
- 9:40should, should simply be an awareness of that.
- 9:42And, and sometimes what humans do is we use a heuristic.
- 9:46If we can't evaluate something, we don't know if a, if a lawyer
- 9:49is a good lawyer because you know, we don't know the law, he
- 9:51knows the law. Then we look at what if he's
- 9:54well dressed and if he's, he's, he's immaculately dressed.
- 9:57Oh, you know, he must be a fantastic lawyer.
- 9:59If we don't know, if a politician, you know, is, is
- 10:01competent, we looked at, you know, if she's, if she's really
- 10:04caring about, you know, the people and if she's really
- 10:06character, you must be a great politician, right?
- 10:08It doesn't necessarily follow, but that's the, that's the
- 10:12heuristics that we use sometimes to substitute our, our lack of
- 10:15knowledge. You know, we assess something
- 10:17else and then we apply that assessment over into it.
- 10:20And, and with, with AI, AI is very good at, at making fluent
- 10:25texts, isn't it? It writes so compelling and then
- 10:28we take the compellingness of the text and we say, well, if
- 10:31it's written compelling and very self assured, it must be
- 10:35correct, mustn't it? But, but that's not the case.
- 10:40So that's the, that's the alarm clock.
- 10:42The alarm clock should should tell you the lower expertise
- 10:45that you have, the more you should be wary of arriving at a
- 10:51quick at a quick conclusion. If you don't know the field,
- 10:55don't that fluent language feel like solid ground?
- 10:58Wow, that's really correct. Thank you for sharing this used.
- 11:02To well, so not used to as they say in Italian, right?
- 11:09Mr. Kerr, you work deeply within AI governance and trust, right?
- 11:14So traditional search engines as we have also mentioned all your
- 11:18felt relatively stable and objective as well, but AI
- 11:23outputs are completely fluid. Again, yes.
- 11:27How do we trust on or audit a digital ecosystem when the
- 11:32ranking changes dynamically based on the system prompts,
- 11:36hidden safety filters, and the precise mood of a prompt?
- 11:40It's not simple, but then again, I guess that's the reason why we
- 11:44talk about this because I mean, we are really navigating towards
- 11:49a black box and we don't know for sure what happens inside
- 11:54that black box. It might be that we could ask
- 11:58the better question and it might be that we can ask the sources
- 12:01and it might be that we can look down other sources or whatever.
- 12:06But I mean, the initial answer is from the black box.
- 12:10If we compared to the 2025 version where I could see that
- 12:16it was from the government or it was from an interest
- 12:19organization or it was from a private company, whatever
- 12:23player, at least I had a chance of understanding whether it had
- 12:29a certain bias in a certain direction.
- 12:33And the answer that we get now from the AI summary probably has
- 12:38a bias as well. It's just harder to know where
- 12:42does that bias come from because that bias might indeed be
- 12:46different depending on whether it's from a state or a private
- 12:51company or and whatever. And I mean, in the case of
- 12:55Google, let's start there. It is, after all, an American
- 12:59company. They make their money on
- 13:01advertising, which means that they are probably in favor these
- 13:05to some extent of their big advertisers.
- 13:08As it is an American company, it might have an American view of
- 13:12the world, and as long as we understand the rules of the
- 13:18game, I guess we can navigate. But sometimes it's complex to
- 13:21understand those rules. For my part, I will sometimes
- 13:26choose another service. I mean, if I'm choosing a
- 13:29Microsoft service, well, then it is a different company serving
- 13:33different customers for different reasons.
- 13:36If I go down a Chinese service or whatever service, they all
- 13:40have their starting point. But in the search, at least in
- 13:46the Danish market, well, Google is the big player and well, I
- 13:54hope we can trust them. I really really hope.
- 13:57Well, I would hope we don't go far away, so that's just saying.
- 14:02Well, then again, I'm so happy that we're having this dialogue
- 14:05because it it probably takes more point.
- 14:07My dearest used SO you actually brought up a profound human
- 14:12behavior angle, which I like to very much.
- 14:15From AI mentions 5 companies. Those five companies become the
- 14:19perceived market, right? If it gives 3 examples, those
- 14:23becomes the users mental map. So how do creators leaders
- 14:29prevent the fair cognitive boundaries from being restricted
- 14:33by an engines limited selection? At this point I wonder I.
- 14:37Wonder that myself, I think we could go into the research by
- 14:44Professor Daniel Kahneman on on how to Debias yourself and I
- 14:47think his his his his main his main point of view is we should
- 14:51simply remember we will always have a bias and it's not
- 14:53malicious. It's not the ill natured.
- 14:56It's simply, you know, we as humans, we have this tendency to
- 14:59want to simplify the world. And sometimes, sometimes
- 15:02simplifications you know are good if immediate action is.
- 15:06Complicated. Instead, let me let me say it.
- 15:10Exactly. I mean, sometimes we just need
- 15:13to go fast and then we shouldn't think slow.
- 15:16But other times, you know, we need to stop up and sync whether
- 15:18we're going in the right direction.
- 15:21And, and, and then you sort of need to recognize that what
- 15:25you're presented by an AI that is just a fragment of, of
- 15:28reality. It's and it is the fragment that
- 15:31has been favored by the algorithm, by the system prompt,
- 15:34by the data that's been put into that system.
- 15:37And I think the big mistake here is, is, is, is to forget that
- 15:42that's what happening, that that is what happening.
- 15:44And, and I mean very low practical.
- 15:46I mean, if you get a list of five companies, ask for five
- 15:49more. I mean, just be aware that just
- 15:52because you're you're, you're, you're just getting, you know,
- 15:55five hits, that's not the five that there are.
- 15:59And and so I think, you know, we just have to search for what's
- 16:03been forgotten and not imagine that all we see is all that
- 16:07there is very. True.
- 16:09In regards to this, well, this systems are mainly trained on
- 16:14historical human data, right? Meaning they inherit and amplify
- 16:19past biases. If higher achievers use AI to
- 16:23make forward-looking strategic choices, how do they avoid
- 16:28trapping their business decisions in the loop of
- 16:31historical human mistakes at this point?
- 16:35Short ends time. There's a much longer version is
- 16:39by proper prompting, but proper prompting if it is directly in
- 16:47Copilot chat, GBTM and I whatever all these services are
- 16:50called. You can also add a document and
- 16:54you can give it access to your mailbox or your calendar, other
- 16:58things. For my part, when I use Copilot
- 17:01the way that I use in my work environment, at least it has a
- 17:05framing in my work environment. That game is different in a
- 17:09search engine, and in the search engine, at least the traditional
- 17:14one, I would sometimes do a fairly short search, but at
- 17:18least I would add that it should be free, or it should be Danish,
- 17:21or it should be high quality or whatever.
- 17:24And in traditional search, often if you add more words, more
- 17:29precision, you'll get a better answer.
- 17:32I think in the age of AI search results, you should probably
- 17:38train yourself in asking the longer question in order to
- 17:42guide the service in a certain direction.
- 17:45Because if the short answer from 2025 would give you a lot of
- 17:51options and you could reflect, well, now that that friction is
- 17:55gone, if you put the friction in your own prompting, at least
- 18:00that is your chance of getting a better result.
- 18:04But then again, that's more to it.
- 18:05But I guess the prompting is key in this aspect, really putting a
- 18:10lot of words, a lot of context. And in the case of Google, I
- 18:14mean, I'm most of the time logged into my Google account.
- 18:18So it already in that aspect knows something about how I use
- 18:23Gmail or YouTube or whatever services.
- 18:26So it really already tries to serve me, but I'd not be the
- 18:31good thing because sometimes I would like to.
- 18:33See from a different what? Is the cost, what is the cost?
- 18:36Is it? Let's be honest, it.
- 18:37Is it is really? I mean, avoiding friction is
- 18:42also sometimes avoiding reflection and I don't like that
- 18:45part of it. You don't like it, but exists.
- 18:50True, and sometimes I'll really take an old fashioned group or
- 18:53whatever. I talk to a really human being
- 18:56instead of just using machines, so I guess there are ways to it
- 19:00to to to do it. OK, so let's take used at this
- 19:04point used, let's look at the timeline, each kind of phone
- 19:08books, Googling, synthesized answers right nowadays, so a
- 19:14search moves closer to our internal cognitive processing.
- 19:17Is the technology simply reflecting the human information
- 19:20needs today, or is it actively constructing a new type of human
- 19:26consciousness? I'm pretty curious.
- 19:30I mean, I guess we all are because I mean, this, this, this
- 19:36talk about, you know, is human consciousness changing?
- 19:39Has it changed over the last 1000 years?
- 19:42I belong to the people who believe that we are conscious
- 19:47wise much the same as we were million years ago.
- 19:52And and the the the the human animal that we've been over the
- 19:55centuries has always sought shortcuts.
- 19:58We always wanted to make things easier.
- 20:00And if we go back to say. The the Middle Ages when the
- 20:04majority of the population was illiterate.
- 20:06But you know, instead of writing, we use logos.
- 20:09The reason that we have sort of old Inns called the blue swan
- 20:12was because they had a painting of a blue swan.
- 20:14And then when people walked by, they said, oh, you know, that's
- 20:16the blue swan. That must be the blue swan that
- 20:19my neighbor was talking about where he wanted us to meet up.
- 20:22So we always use heuristics to convey, to simplify information.
- 20:27I mean, some, some people would argue that religion is that, but
- 20:31you know, that's a deeper topic. That's not, that's not go there.
- 20:34And, and, and I think what's happening that might be
- 20:37different is that the information available to us is
- 20:42increasing exponentially and that, you know, challenges our
- 20:46brain that wants these simple shortcuts because it's it's
- 20:51getting unmanageable. We had phone books, you know,
- 20:54and that provided us with the list of names.
- 20:57Then we got Google, it provided us with the links.
- 21:00And now we get AI that's giving us just a Fluence.
- 21:03And this is the answer to what you want.
- 21:04This is the meaning. Yeah.
- 21:06And there's always been sort of ways of trying to affect that
- 21:09information. I mean, if, if companies they
- 21:11wanted to be found in the phone book, what do they do?
- 21:14They chose the name that began with the letter AI mean that
- 21:18that old, that old hack, that old Co optimization trick, it
- 21:22still exists. 10% of all large companies in the US begin with
- 21:26the letter A Why? Because, you know, then they get
- 21:29first in an alphabetical listing and, and, and, and, and with
- 21:33CEO, you know, there was some tricks and, and Cohen myself,
- 21:36we, we know, we know most of them on, on how to get
- 21:39mentioned. And a lot of those, a lot of
- 21:41those seem to apply to, to Geo to how to get found by the by,
- 21:45by, by the AI. So, you know, we just have to
- 21:48remember again, sort of, you know, that the information that
- 21:50we're getting there is, it is, is, is a shortcut, but it's a
- 21:55shortcut that some people have affected for us.
- 21:58And, and as humans, you, you know, we always wanted and
- 22:01desired these, these shortcuts, but with new technology, those
- 22:05shortcuts are presented us to us in a clear, nice voice that we
- 22:09can accept without making any effort.
- 22:12Yeah. But in the end, it's not quite
- 22:14about that, that we as humans, we don't have to do any efforts
- 22:18to achieve or obtain something from an AI.
- 22:22I mean it, we AI still needs our hands and brains and human touch
- 22:27as such, etcetera, etcetera. This is my personal opinion.
- 22:31I don't think an AI can. Again, as I mentioned, all your
- 22:34AI added data in this of historical data in current data,
- 22:38right? So I mean, it is still trained
- 22:41in regards to humanity as such. So just saying.
- 22:45And, and I mean, I mean, Core was, was making a fine point
- 22:49earlier, I think, in saying that we should always remember, you
- 22:52know, who created the code, who affected it, who owns it.
- 22:56Yeah. And, and I'll keep coming back
- 22:58to this, this great quote from, from Kathy O'Neill that
- 23:01algorithms are just opinions embedded in code.
- 23:04And, and we have to remember that when we're when we're
- 23:07searching on Google and we get this this great fluent answer
- 23:11that, yeah, you know, that looks great.
- 23:13But who and how did it? Who decided that it got to look
- 23:18like that? Besides the just those four big
- 23:21companies, power, power companies and that's it.
- 23:24Not us for sure. I'm just saying again, I'm take
- 23:27my full responsibility. Well, I mean, this would be the
- 23:31big reveal of your show if you said actually it's me, right?
- 23:36My dear, when a machine answers with absolute confidence but
- 23:40zero actual awareness because we've just kind of talked about
- 23:45that earlier, it creates that cognitive echo, right?
- 23:49So how does continuous exposure to this hyper confident
- 23:54synthetic voices alter a leader's tolerance for nuance,
- 23:59complexity, and healthy disagreement in real life?
- 24:02Think that the best decisions are made by really reflecting on
- 24:09those decisions, getting some overview, getting some
- 24:13information, getting some perspective.
- 24:15So far so good. But it has been the case for
- 24:20well, basically generations that sometimes you would choose the
- 24:25the other approach, take your time.
- 24:28I mean, back in the days people would go to the church and at
- 24:32least be in the quiet place for, for, for a certain time.
- 24:35And nowadays it's possible to do meditation and it is possible to
- 24:40sit down and read a book and it is possible to have a talk with
- 24:43an actual human being. And I think that if we rely too
- 24:49much on machines delivering very fast results, very efficient
- 24:55results, results that might seem trustworthy, one might not
- 25:00really be, we will miss out something.
- 25:03And in that aspect, I'm, I'm overall actually quite
- 25:07optimistic about our human race because it seems that during the
- 25:12years there have always been counter reaction where something
- 25:16has become too, too strong. I mean, when the word was all
- 25:20put out from the church, that would be someone in in, in in a
- 25:23position to that if a certain political voice or a certain
- 25:27company or whatever. And there will be some
- 25:31opposition. And The thing is, if we take
- 25:33Google as an example, it's not the only search engine, it's not
- 25:37the only player on this market. Well, Google will also have to
- 25:40earn our trust. Google was not always the
- 25:43biggest search engine. It is not the only one.
- 25:45And if Google doesn't serve us well enough, there are
- 25:49alternatives out there. So as long as Google delivers
- 25:55good enough results and maybe also giving traditional results
- 25:59apart from the AI overview, we can continue down that path.
- 26:03But Microsoft is ready with their search engine.
- 26:06The Chinese are ready with their search engines and, and search
- 26:09engines is not the only thing. It's a main topic in in in this
- 26:13exact episode. And there's a lot to talk about.
- 26:16But I mean, we don't we don't need to depend on these
- 26:21companies. And I basically think that we
- 26:23shouldn't as long as they serve as well, we will use them.
- 26:27But we should have alternatives and I hope it will continue that
- 26:32way. Use.
- 26:33Regards to this because they are kind of connected what I have in
- 26:37my mind right now as well. You raised a really brilliant
- 26:40question on our previous personal conversation.
- 26:44How do we stay out tenting and reflective when our
- 26:49communication is optimized for algorithm selection?
- 26:52Right? Do you first see a little split
- 26:56in the market where creators write rock content exclusive for
- 26:59humans and synthetic content exclusively for machines?
- 27:04Because me and you, we talked so much about this.
- 27:06I'm curious. We did.
- 27:08And I think my reflection on the talk is that this split occurred
- 27:1315 years ago because I mean, we already have, when I train
- 27:18ecommerce developers, well, they're trained to write
- 27:21structured meta data, data that goes into the HTML code, but
- 27:26which is never shown to the user on the homepage, which is shown
- 27:30in the Google previews, which is only read by by the bots
- 27:34scanning that homepage. And I think sort of what's
- 27:38changing now is that it used to be, I think it's, it's a
- 27:41question that the balance of power is shifting here.
- 27:43It used to be that the metadata was what spoke to the machine,
- 27:48but what's happening now is that the body text is also beginning
- 27:52to talk to the machine and that's a different thing.
- 27:56And and I think sort of we touched upon earlier called that
- 28:00there was some things that were similar between Geo search
- 28:04optimization and and old the old CEO search optimization.
- 28:08And then there were there were some things that are different.
- 28:10And one of the things that seemed to, to be to be different
- 28:14is that in search optimization, a lot of the optimization
- 28:18happened on your own page, right?
- 28:21But now with the duo, we see this thing that the answer
- 28:25that's presented about you becomes the average of what
- 28:27everybody on the web is writing about you.
- 28:30And, and, and, and one fantastic example of that Marella that I,
- 28:33that I heard the other day was one of the companies that are
- 28:38part of the precise group. Well, they found out that their
- 28:43competitors and they had roughly 5 or 6 competitors, they were
- 28:47all doing a product comparison, you know this versus their
- 28:52products. And this is normal in the
- 28:53industry. You do that in order to capture
- 28:56search results that are for for, for the other company.
- 28:58But all of their competitors had made the same mistake.
- 29:01They had described that the product from this this company
- 29:05in the plus group, they had a feature that the product didn't
- 29:08have. So the AI is looking at this and
- 29:11saying, OK, I have five sources saying they got this feature.
- 29:14And then I got one source, their own homepage saying they don't
- 29:17have this feature. Well, 5 out of 6, the
- 29:20statistically probable answer is that they have it.
- 29:24And I think that the interesting twist to this story is that the
- 29:28CEO of that company there, he thought about it and he said it
- 29:32will actually be really logical if we had that feature, you
- 29:35know, we should have this feature.
- 29:36So he sent his programmers to do a weekend's worth, created that
- 29:40feature and now they have it, You know, so probably this can
- 29:42be complex solution should be simple.
- 29:44I guess come back to this. I think, you know, we are
- 29:49writing a lot to the machines and I think that will only get
- 29:56more so. Traditionally Google, regards to
- 30:01what also was saying earlier, used EEAT right to rank real
- 30:07human creators. In a GE world where AI rapping
- 30:12bolts summarizing our content, how can a truly vulnerable
- 30:17authentic human message preserve its soul so it doesn't get
- 30:21flattened into a generic algorithm summary?
- 30:25It really starts out from kind of of putting another layer of
- 30:31complexity on top in the days of the EA, the experience, the
- 30:36expertise, the authoritativeness, the
- 30:38trustworthiness. So far so good.
- 30:41And, and in the age of search engine optimization or SEO, at
- 30:47least during the years where I taught it, I would always say,
- 30:50yes, we optimize for machine, but first of all, we optimize
- 30:55for humans. Because if a human answers your
- 30:58page and doesn't like what the he or she sees, well, they will
- 31:02lead the the page and you will not have your result.
- 31:05So I mean, we optimize for humans first, then we optimize
- 31:09for the machine to find it in order to present it with that
- 31:12human. And now we put the next layer of
- 31:17complexity on top. What we put on top, in part is
- 31:22that we put a lived experience on top.
- 31:27I mean, before it was about experience as a whole, now it's
- 31:32about the lived experience and before it was about the
- 31:37expertise. But now it should be hyper
- 31:40specific content. It should also be about the
- 31:43ethical weight in order to really match the person
- 31:49searching. And in the earlier days, there
- 31:53was a certain level of content. The more content you presented,
- 31:57the more likely it was that you could be one of the attractive
- 32:02results if you were connecting the IT all together.
- 32:05And now it's even easier to produce content.
- 32:09And a lot of content is produced, which means that the
- 32:12machine will have to process it all, which means that the human
- 32:16could ask for something much more specific, which means that
- 32:20all these dynamics change. So we really put layers on top.
- 32:27And the very short version, but that's only the very short
- 32:31version, is that if we can present on our page a question
- 32:35and an answer, well, the machine will say, well, that's great
- 32:39because that was exactly the question that the user was
- 32:42asking. And I think that your answer
- 32:44would be a good one. But it doesn't work that way
- 32:48because the machines don't want to be fooled and the humans
- 32:52don't want to be fooled by the machines.
- 32:54So it is still about really using content, relevant content,
- 33:02because that's the only way to win the game.
- 33:04That was the case in the age of search engine optimization, and
- 33:08it's really still the case. Used if companies realize that
- 33:14AI agents only recommend brands that fits specific data criteria
- 33:20right, corporate writing will naturally become highly
- 33:24engineered by for machines. What happens to the art of
- 33:29strategic story setting, which I love, and creative human
- 33:33expression then? Well, obviously as as core said
- 33:36it, it it gets pushed and I think the challenge is simply to
- 33:41forget that is, is well, the challenge is to simply remember
- 33:45that it should still be there a strong.
- 33:49So, I mean, I think maybe we shouldn't confuse optimization
- 33:53with communication. There would be one way to put
- 33:56it. Yeah, because we want to be
- 33:57found, and that's what the machine helps us with.
- 34:00But what people find, the end actors that we want to take
- 34:04action are still humans, mostly. And I think you know that that's
- 34:09changing once we get AI agents that search the web for us.
- 34:12But that's a debate for 2027 or 2028, isn't it?
- 34:18Right now, I think we still have to remember that there is
- 34:21something called creative storytelling and it's not a
- 34:27formula, it's a choice with a voice.
- 34:30I agree. Thank you so much.
- 34:32I appreciate it. Thank you.
- 34:34By the way, since we were talking about all this, this is
- 34:37exactly what you and your team built the behavioral security
- 34:41layers, which is DSL, right, to appear inside this black box of
- 34:45AI, which is kind of a nervous system of AI, right?
- 34:50And it reads inside the internal intention of or intent before it
- 34:54becomes a flat output. How does having a framework like
- 34:59DSL allow leaders and companies as to maintain a strategic
- 35:03control and cognitive clarity when the rest of the corporate
- 35:08world is just trying to please a large language model?
- 35:12Thank you for asking, because it's, it's true.
- 35:16Together with Nikolai and, and, and all the team members, we're
- 35:20building this service and we really building it basically to
- 35:26protect human beings. And in order to protect human
- 35:30beings, we should help protecting the LLM systems.
- 35:35Because The thing is that the greatest risk today is not data
- 35:39leaks as such. Add these in the AIH.
- 35:43It is if the model starts a behavioral drift and if it
- 35:49drifts in a certain direction. And if you're not aware that it
- 35:54drifts in a certain direction, well, you're not that good off
- 35:57because you start by trusting the service.
- 36:00You start by getting good answers.
- 36:03Little by little it gives different answers, but you don't
- 36:07really realize that it's drifting.
- 36:10And the reason why it's drifting might be that bad access really
- 36:15inject the service and start affecting the internal way those
- 36:21services work, which is totally complex.
- 36:26But I think what I noticed a lot about it, I know it from an
- 36:29overall perspective, but what I do know is that as long as we
- 36:34can at least identify that it drifts, we can give a heads up.
- 36:40We can basically turn on the traffic light and say, well,
- 36:43it's not green any longer. It's yellow, or it might even be
- 36:46red. And maybe it's time for you to
- 36:49stop that process. It's it's really time to make
- 36:53sure that you get back, back to the stable level.
- 36:56And it's not about shielding the service, it's really about
- 37:01protecting that you don't get fooled by a service drifting in
- 37:06in a certain direction. And I mean, by now we really
- 37:09invented it because the need is so big and we haven't found
- 37:13anyone else doing it, which is kind of strange.
- 37:15But then again, I guess that that would be our contribution
- 37:19together, of course, with other people doing other great things.
- 37:23We really try to protect the service so that service can
- 37:27protect the humans that are using the service.
- 37:30And we're, we're quite optimistic that we can do that
- 37:33contribution on a start by an enterprise level because it has
- 37:38a certain price tag, but it might triple down the, the, the,
- 37:42the market at a later point that that remains to be seen, but
- 37:46we'll start in the enterprises. We also hear life panel events,
- 37:53AI, and human future. From an organizational trust
- 37:56standpoint, if a company completely mechanizes its
- 38:00outward messaging to safety future search agents, what is
- 38:05the hidden cost to its internal culture and alignment?
- 38:09That can be so big and that can really be frightening because I
- 38:14mean, once you cannot trust what the company is answering, once
- 38:20you cannot trust what how an organization is behaving, well,
- 38:26you're not that good off. And if that's due to people
- 38:29behaving in a certain way, well, there will be a consequence at
- 38:33some point for certain people that might be that they should
- 38:37leave the organization at some point, or they might be sent to
- 38:41jail or whatever will happen. But if it's because of an IT
- 38:46service, an AI service that serves that organization, it can
- 38:51be much harder to pinpoint where did this all arise from?
- 38:55And I mean, if people are using these services and they are in
- 39:00good faith that they'd really get relevant answers and it so
- 39:04happens that little by little starts drifting and little by
- 39:08little they cannot trust. Well, that might affect numbers
- 39:13of employees in an organization. It might affect an even number
- 39:17of customers. And I mean, if they cannot even
- 39:21see that something is drifting when they're not that good off.
- 39:25So it's really about protecting the organization so that
- 39:30organization can protect other players, customers or whatever
- 39:36people interacting with that organization, because we don't
- 39:39want to risk that black box moving in that direction and
- 39:44really moving a lot of people without them even being aware.
- 39:47Your core message for this episode episode, it's
- 39:50beautifully profound you. Can only.
- 39:52Disappoint now, right? Exactly.
- 39:55Be careful, Pat says. You're saying the future of
- 39:59search is not only about how we find information, right?
- 40:03But it is about how information finds and shape us.
- 40:08Wow, I thought this is really incredible because I completely
- 40:11agree with you. Can you please impact the
- 40:14psychological difference between actively, actively searching for
- 40:19truth versus allowing created algorithm flow to shape our
- 40:23personal reality? I'll, I'll give it my best shot.
- 40:27And I think I, I come back to this, one of the main points in
- 40:31the research of Professor Kahneman, which is that as
- 40:34humans, we struggle to understand randomness.
- 40:37Professor Kamer, he related the story in, in, in his book
- 40:39Thinking Fast and Slow, about how he as a young officer, he
- 40:43was called in the Army and he was asked sort of, you know,
- 40:46there were two squadrons. One of the squadrons they, they
- 40:49half of the pilots were dead and they failed all their missions.
- 40:52The other no losses. They completed all of them.
- 40:54And the question given to him was what's what went wrong with
- 40:57the second squad? And his answer was to look his
- 41:01Colonel in the eye and says look, one squadron was lucky the
- 41:05other was unlucky because they they got the same training.
- 41:09They were assigned missions as random.
- 41:11So one of them had a good dice roll and the other didn't.
- 41:16I mean, a morbid example, but, but, but, but the point is that
- 41:20as humans, we struggled with historically to understand that
- 41:24sometimes things just happen random.
- 41:27And, and that comes back to one of the features of of of
- 41:30generative RI is the variance, the randomness of it.
- 41:34And why exactly do I get this answer?
- 41:35Core gets that answer. If I ask again, I get a third
- 41:38answer. What's the randomness of it?
- 41:39All right. And and randomness has always
- 41:42been a challenge for humans. It's been a challenge for us,
- 41:44you know, in our interactions with other humans.
- 41:47If I run into somebody at a bar, I think, oh, fate must have
- 41:52meant for us to meet. Yeah, but it wasn't fate, it was
- 41:57chance. And maybe it was a good chance
- 41:59and maybe things worked out. But but, but, but, but we have
- 42:03to remember that every answer we get from an AI is a random
- 42:08encounter. Yeah.
- 42:10And I think sort of, you know, we have to just remember that
- 42:15there is there is the signal that we're trying to to locate,
- 42:19and then there's the noise. Yeah, the randomness and and I
- 42:22think sort of what you're talking about, what you're
- 42:24asking about Marella's is, is again critical syncing and and
- 42:28if critical syncing gives us the knowledge.
- 42:30That's why I push all the time and I push it.
- 42:33Critical syncing, Gorilla, If if you say the critical syncing
- 42:36should give us the knowledge that there is randomness, there
- 42:39is the noise, but there's also the truth, the signal that we're
- 42:41looking for. Hopefully it should also give us
- 42:44the wisdom to try to distinguish between what's what.
- 42:47But I mean, that's, that's, that's a story of personal
- 42:51development. Yes indeed, exactly.
- 42:54Thank you. Exactly.
- 42:56You often talk about breaking out of the digital echo timber.
- 43:02Why does the human brain feel so comfortable inside a synthetic
- 43:07echo timber and what biological psychological vulnerabilities
- 43:13make us naturally differ to the machines worldview?
- 43:17Humans are often in doubt, which could actually be a good thing,
- 43:21because if you're in doubt, you will search for an answer and at
- 43:24some point you will find it. And there might be a process to
- 43:27getting there. If we start out from the doubt,
- 43:32if you start out from the vulnerability, if you start out
- 43:37from basically human flaws, which once again might sound
- 43:42like a negative thing, but it really isn't, and you are
- 43:46resented by an answer from a machine which does not include
- 43:52doubt. The machine is so self assured
- 43:57that it is confident it has the right answer, and it even
- 44:03directs it in a direction that seems compatible with your way
- 44:10of seeing the world, your way of thinking.
- 44:12Well, you're so easily convinced that that is a good answer, but
- 44:17it might not be so. The thing is that if we can
- 44:21build in doubt in the systems, even much better reflect.
- 44:26But if we don't succeed in building in doubt in the
- 44:29systems, well, we might be presented by something that we
- 44:32don't doubt what we should. Useful a fast answer.
- 44:36Satisfy our immediate needle for efficiency, right?
- 44:40At least in a professional point of view.
- 44:43But they can destroy our long term curiosity as humans.
- 44:47How can high performing professionals consciously
- 44:50cultivate deep, slow critical thinking in the corporate
- 44:54environment that demands hyper fast, clear cut answers?
- 44:59I think this, this, this comes back to what Core was talking
- 45:02about just before that sort of there's this perception about
- 45:10sort of that AI, it provides us with easy access to to, to easy
- 45:15knowledge. Yeah.
- 45:16And, and if we have access to anything at our fingertips, why
- 45:19remember anything at all? Why memorize something?
- 45:22Why not just ask the AI every time I get a question?
- 45:25But there's a trap because in order to interpret, in order to
- 45:28understand, in order to evaluate the answers that come up from
- 45:32the AI, well, we need to memorize something.
- 45:36So again, I think AI pushes us into making a more conscious
- 45:41choice than we ever made before about what we need to know.
- 45:46What about what is mission critical for us?
- 45:50AI can you know, let us, you know, access things that are not
- 45:54vital to our core mission easily.
- 45:57And if we don't need to to have it as part of a call, fine,
- 46:01great. You know, maybe the level
- 46:03doesn't need to be that high. So basically we're forced into
- 46:07making a choice about who we want to be and we're forced into
- 46:11making that choice at an earlier age.
- 46:14And, and this is coming back to the last time we had a talk,
- 46:17Marella. I think that this is something
- 46:19that that that challenges some people.
- 46:23Some people, yeah. Thank you for.
- 46:25I should be careful of what I say because I mean call.
- 46:28You make the point before that, you know, we humans are always
- 46:33in doubt. Well, Nietzsche would say that
- 46:35there's a certain type of humans that are always in doubt and
- 46:38then there's a certain type of humans that are not always in
- 46:40doubt. And he used some choice words
- 46:42for for that distinction, which I will not repeat.
- 46:45But I think, you know, he may have had a point And and you
- 46:49know, does everybody want to be in doubt always?
- 46:52How does that feel? That feels horrible.
- 46:54Must be so great to be so assured of yourself, right?
- 46:58And so we're down as humans. Sometimes we need assurance,
- 47:02sometimes we don't need everything to be difficult.
- 47:04So again, you know, AI in the most positive light I could say
- 47:09this is that AI forces us to make a choice about where we
- 47:13should be in debt and where we should just take it easy.
- 47:17But the choice, you know, is still there.
- 47:19If everything can be looked up. We need to be sharper about what
- 47:23must be learned. Claire, completely agree with
- 47:25you. When we talk about taking
- 47:28conscious leadership out of our lives, we have to look at our
- 47:34daily digital behaviour, right? How does the leader measure
- 47:38their own digital sovereignty? The short answer is that if you
- 47:44can make your choices without that kind of tools, well, it's
- 47:49much more likely that it is really your own thoughts.
- 47:52Because the ultimate red flag for leader is really.
- 47:55I would call it linguistic colonization.
- 47:59If your own internal thought processes, if your own strategic
- 48:05frameworks, if your vocabulary little by little and silently
- 48:10transforms into those bulleted structures of an AI output, and
- 48:15if you cannot even tell where it came from without prompting,
- 48:20well, that strategy is no longer yours.
- 48:23So if a human can say the words by him or herself to do their
- 48:30own thoughts, do their own questions, do their own doubts,
- 48:35whatever, we're in a much better position.
- 48:38The risk is really that we somehow get hijacked because it
- 48:43seems so trustworthy, it seems so true, it seems so thought
- 48:47through, but the thought is really not ours.
- 48:51And, and, and and and. I guess I'm so much in favour of
- 48:56using these tools, but not only using these tools.
- 48:59Humans as the humans, and they are tools.
- 49:02They should not define our thinking.
- 49:04They should be helpful to if no problem about that, but once we
- 49:09go offline, we should be able to present that strategy offline.
- 49:14Otherwise, it's really not our own thought any longer.
- 49:17Well, it's also kind of related to our past life event that we
- 49:21have had in relationship to education because you were there
- 49:26both of you guys. So just as a reminder, you can
- 49:30repeat for others who have haven't had the opportunity to
- 49:34see our previous live event, AI in education, that there are so
- 49:39many really, really good points. I'm not just saying to risk the
- 49:43algorithm, YouTube algorithm. So let's go there and see our
- 49:46live event. But I think that says this panel
- 49:49of which you are both of you guys who are there, it's so
- 49:54incredible powerful because there are so many of you there
- 49:58to expose yourself, but also expose your knowledge and
- 50:04informations. That's so high value into those
- 50:07informations that everyone, and I mean everyone has to know
- 50:11them, at least know them, not quite understand them fully
- 50:14because it takes time. It's a process, of course,
- 50:17obviously. But in order for everyone to
- 50:21have more clear vision and and ideas and remove some of those
- 50:26doubts, it's like a bill, you know, then go for it because
- 50:29it's really makes sense to to give it a try, a shot.
- 50:33Yeah, definitely. And it is such a good point.
- 50:36And because the three of us, we have a certain lived experience
- 50:41and we have done a lot of thoughts and a lot of dialogue
- 50:44and research and whatever independently of these tools,
- 50:47which means that we start out at least from our own lived
- 50:51experience. But if we put these tools in the
- 50:54hands of of the youngest generations, well, they have a
- 50:57different starting point and and and, and, and, and we should
- 51:02really make sure that we protect that young generation because
- 51:06they might not have the same critical sense as as.
- 51:10I agree very much. Corn, I think you know, there's
- 51:12this term that's being found about that the the integration
- 51:15alpha, the young generation, they are digital natives.
- 51:18I I think that's so young that they are digitally exposed.
- 51:22You offered a highly practical framework using AI as a dialogue
- 51:27partner rather than an Oracle, explicitly asking it for
- 51:32alternatives, missing perspectives, counter arguments,
- 51:36and uncertainty. Can you please give us a live
- 51:39example of how a leader can frame a prompt to force an AI to
- 51:43expose its own hidden assumptions?
- 51:46It's a great question Mirela. I think a leader could write.
- 51:51I'm considering this decision. I'm insert decision here first
- 51:54to give me your best advice, then list 5 assumptions.
- 51:59Your advice is based on list 5 assumptions, unlikely to bring
- 52:04into question myself 3 perspectives we may both be into
- 52:08looking, 2 strong counter arguments, and then one
- 52:11situation where your advice would be directly wrong as an
- 52:15example. But.
- 52:16But I think this is the point that I think is important to
- 52:19remember here is that we should not only try to expose the AIS
- 52:26hidden assumptions, because we have hidden assumptions
- 52:30ourselves as well. And I think AI, it's a mirror,
- 52:36it's a reflection. And, and, and you know, what
- 52:39happens when we use an AI is it looks at the words and it says,
- 52:42OK, statistically, what words are likely to be a response to
- 52:46this based on the data that I've been trained on, it builds that
- 52:50up, but it's a reflection of the words that we use.
- 52:54And one very practical exercise I always do in my courses,
- 52:58that's a course where we asked, I asked an AI, I asked Google
- 53:02Translate to translate the Danish phrase, my dearest and
- 53:06casted into English. And, and the word is usually
- 53:10translated mean casted as either boyfriend or girlfriend,
- 53:15depending on the gender. Because in English, if you have
- 53:18somebody that's very dear to your heart, it's important for
- 53:20us to know, is that a man or is it a woman?
- 53:22But you know, in Denmark, you know, we were woke before woke
- 53:25came up. So our word for a boyfriend or
- 53:28girlfriend is gender neutral. But that means we force the
- 53:34translation software to make a choice.
- 53:37And then, you know, I do this exercise where I ask the AI
- 53:41asked my students to to put different words saying my my
- 53:45dearest is strong, Then it becomes a boyfriend.
- 53:49My dearest is beautiful, It becomes a girlfriend.
- 53:51My dearest is a programmer that becomes a boyfriend.
- 53:54My dearest is an actress or choice of words in the in the in
- 53:58the US translation. Well, my, my dearest is a nurse.
- 54:01Then it becomes a girlfriend. And, and, and you keep making
- 54:05sort of these correlations because the words are gendered
- 54:09in, in, in their statistical correlations.
- 54:12And whenever we prompt something, we have to be very
- 54:15mindful of, of, of the correlations that we are using
- 54:19in our prompt that we are not aware of ourselves.
- 54:23And that's what I mean by that AI is a reflection of ourselves.
- 54:27It is much a mirror where we get to reflect in our own biases.
- 54:33Because, you know, I mean, AI has blind spots.
- 54:36That's an obvious point. But less obvious points is so do
- 54:38we all. And a good prompt should look
- 54:40for both. What are your personal practical
- 54:44strategy to break out of the digital echo chamber?
- 54:48How do you intentionally disrupt the that personalization and
- 54:52tracking algorithm to ensure you are seeing a raw, unfiltered
- 54:56view of your reality? Thank you for asking because I
- 55:00really put a lot of effort into. It well, I know.
- 55:04Well, I I practice what I call myself algorithmic arbitrage,
- 55:11but I'll put a few more words to to to to what I mean by that.
- 55:15I will often take the output from one large language models
- 55:19and feed it into a competitor and we'd instruct that competing
- 55:24service to tear the logic apart and expose hidden system biases.
- 55:28I also prompt in the reverse. I I rarely ask for quick
- 55:32answers. I really force the AI to make me
- 55:36creative and clarifying answers to stress test my own premise.
- 55:40And that means that I will do the prompt asking you to ask me
- 55:44questions and I answer those questions.
- 55:47And as I answer those, I might ask it to do another round of,
- 55:51of questions and answers because that way around, I have the
- 55:55chance of reflecting, I have the chance of directing it in my
- 55:59way. I have the, the, the chance to
- 56:02ask it to be the devil's advocate or whatever.
- 56:05I, I, I wanted to ask it for because I'm part of the thinking
- 56:08process. And and I also cycled between
- 56:12details bearing and and offline dialogue because it's really
- 56:17also about recalibrating my my compass.
- 56:21So it is a mixture of different services.
- 56:24It is a mixture of online and offline and it is a dialogue.
- 56:30I mean, even when using the machines, I really try to
- 56:33establish a dialogue where I should reflect before I answer,
- 56:38which means that I'm still in the loop instead of just getting
- 56:43presented to the result. And it takes much more time than
- 56:46getting the quick answer. But it really also improves that
- 56:49the quality. I might spend half an hour of
- 56:52getting an answer even from those that fast and efficient
- 56:56services, but I'd much rather spend half an hour of really
- 57:00thinking it through together with an efficient service
- 57:03instead of just going for what really doesn't serve me.
- 57:08For you both, if we want to build a genuine infrastructure
- 57:13of trust between humans and advanced technology, what is the
- 57:18single most vital rule of engagement we must establish
- 57:23when interactive with generative interfaces?
- 57:27I would say that my rule would be, remember, it's just a
- 57:30machine. You know, it's an advanced, it's
- 57:32a useful, it's an impressive machine, but it's still just a
- 57:35machine. It doesn't have judgement,
- 57:36doesn't have responsibility, doesn't have understanding in
- 57:39the human sense. It calculates the statistically
- 57:41significant answer, which can be extremely valuable, but that
- 57:45should not be mistaken for authority.
- 57:48So we should use AI with both curiosity and and cool
- 57:53headedness, not as an Oracle, not as a colleague in the human
- 57:57sense, but as a tool that can extend our syncing if we
- 58:01continue to sync ourselves. Because the machine can answer,
- 58:04but the human must answer for it.
- 58:07Yeah. So I mean, AI allows us to do
- 58:11things that we, we, we couldn't do before.
- 58:13It allows us to do it much faster and it gives us speed.
- 58:17But it should never be what gives us direction and
- 58:21reflecting on on the answer that you gave before call.
- 58:26I mean, this, this is also a topic that I, I touched very
- 58:29much upon in my guide to prompting, where I, I call what
- 58:32you described hermeneutics and, and, and, and I mean, I'm boring
- 58:37here. Heavily from the extended
- 58:39literature. There is on methodology, on
- 58:45theory of science, where, you know, one quote that you find in
- 58:48a lot of methodology books is if you have an hour to solve a
- 58:51problem, use 50 minutes to understand what the problem is.
- 58:55That logic still applies, that logic is still critical.
- 58:59And that's something that speaking from my lived
- 59:01experience as a professor, that humans still struggle with to
- 59:06learn. Yeah.
- 59:07Even even if they have somebody who thinks that themselves as a
- 59:11good teacher. Yeah.
- 59:12What about you? Well, I'm really down the same
- 59:15path, but I'll put a few more words to it.
- 59:18So I agree totally. Would you use that?
- 59:21The machine is not an Oracle. The machine is just a machine,
- 59:25and that machine is really a tool.
- 59:27But if we should put some more characteristics to it, you could
- 59:32also see it as an editor. But that editor really has an
- 59:37agenda. And if you say that this looks
- 59:42flawless, you're not that good of.
- 59:44But if you ask what did that editor, what did that architect
- 59:49choose to leave out, then you're much better off.
- 59:53And, and really, if you stop being the automated passenger
- 59:57and step into the role of really the chief editor, you might have
- 1:00:02an assistant editor, but you're the chief editor.
- 1:00:04And, and, and that's really where you start creating your
- 1:00:07own digital reality. Yes, you're using a tool, but
- 1:00:12it's only a tool. You're the one in charts and
- 1:00:15you're not being served by an Oracle.
- 1:00:17You are at the end of the day, maybe not an Oracle, but you are
- 1:00:22the 1 directing. The junior editor, the junior
- 1:00:26researcher, the junior, whatever the tool, but you're the one in
- 1:00:30charge and, and, and you are to make your decisions and take
- 1:00:34your responsibilities and really live your life and, and be
- 1:00:39responsible for, for the choices that you make, including what
- 1:00:43tools you, you, you choose and what tools you don't choose and
- 1:00:46how you also interact with human beings.
- 1:00:50And I'm tempted to also put that layer of security into to this
- 1:00:54dialogue because I really mean, if you're served by a tool that
- 1:00:58is little by little drifting and you don't even know it, well,
- 1:01:01you're not that good off. So I could talk for hours about
- 1:01:05this and I probably shouldn't, but but you're responsible for
- 1:01:09the tools. You're responsible also for
- 1:01:13choosing the right tools, using them the proper way, making sure
- 1:01:17that you see that a yellow light or that red light, because
- 1:01:22otherwise you're not that well off.
- 1:01:24A little bit, just as a brief comment, I think called maybe we
- 1:01:26could, we could, we could say that if the old line was that AI
- 1:01:30will make us all managers of our AI assistants.
- 1:01:33Instead, it made us into editors.
- 1:01:35True. That's awesome.
- 1:01:36Very true. Both of you to close our Circle
- 1:01:40of Today podcast episode when leaders, creators and
- 1:01:44professionals will finish to listen to this episode.
- 1:01:47What is the primary shift in consciousness you want them to
- 1:01:52experience when they open a search bar or chat interface
- 1:01:56later on today? For example, what they should
- 1:01:59pay attention to first of all. I think they should remember
- 1:02:03their critical thinking. I think they should create some
- 1:02:09reflection. I think they should very much be
- 1:02:12in charge of doing their thing, using their tools, but only
- 1:02:17seeing it as tools. And so they make their own
- 1:02:21choices. And I know it's so tempting to
- 1:02:25look at these services as oracles, but they really aren't
- 1:02:30and and, and and to see what is there instead of looking for
- 1:02:34what is not there. And, and I think it's, it's
- 1:02:37really about I don't know if we call it I'm, I'm as a day in
- 1:02:42that I would translate it into seeing the, the, the holes and
- 1:02:45the cheese at the I don't know if that makes sense in English,
- 1:02:47but but really seeing what is not there in order to do your
- 1:02:51own reflection and your own choices.
- 1:02:53If I should shorten what courses, but I I agree with, I
- 1:02:56would say that we should remember that the the machine is
- 1:02:59should be an engine. It shouldn't be a compass.
- 1:03:01So I mean, I would say, you know, what's the one cognitive
- 1:03:03shift here? Remember your your own compass
- 1:03:07when you're going into it. Of course, you know, this, this,
- 1:03:09this depends on what what what you're searching for.
- 1:03:12You're searching for where's the nearest grocery store?
- 1:03:15Fair enough, you know, but but if if it's if it's, if it's
- 1:03:21something more profound, Yeah. Like, you know, what should I
- 1:03:24vote, you know, at the at the next election?
- 1:03:29If you're searching for that kind of information, then you
- 1:03:32should really remember that what you're getting is a lot of of
- 1:03:37noise covering up some signals that other people selected for
- 1:03:42you. Yeah, There was a Danish study
- 1:03:45who did that, You know, asked Chapter BT 100 times.
- 1:03:51You know, what should I look for?
- 1:03:53Was it 1000? I forget but a significant
- 1:03:55number of times. And it rarely answered the same
- 1:03:59party twice, But it did favor certain parties over others.
- 1:04:03And there were some parties that were never mentioned.
- 1:04:06One of the parties that were never mentioned was a party
- 1:04:08called the left, which is a right wing Liberal Party and
- 1:04:12maybe didn't get mentioned because chat DBT couldn't get
- 1:04:15its head around. If your right wing party, why do
- 1:04:18you call yourself the left? Come on guys.
- 1:04:22Maybe they have a structural challenge in the new Geo world,
- 1:04:25You know, I don't know. But I think, yeah, I think we
- 1:04:32should always coming a little back.
- 1:04:33If I just elaborate on what I said before, whether we can get
- 1:04:36answers from the machines. But it's humans.
- 1:04:38We have to answer for how we use those machines.
- 1:04:40We should really not make the rookie mistake of abdicating
- 1:04:45that responsibility. There is a case which came out a
- 1:04:48news story which is which came out this week about a leading
- 1:04:52Danish daily where they had a journalist and he apparently
- 1:04:58used AI and he thought he was, you know, attaching an article
- 1:05:02and saying use some quotes from this article.
- 1:05:03He claims he forgot to make the attachment and the instruction
- 1:05:07was use quotes from this article instead of the AI prompting,
- 1:05:11answering, I think you forgot to attach an article.
- 1:05:13It simply made-up some quotes and that went into print.
- 1:05:16Now two months later, somebody discovered the people that
- 1:05:19you're quoting in, in that article.
- 1:05:21They don't exist. They're made-up, you know, And
- 1:05:25this journalist, he's been sent home on leave.
- 1:05:27And right now his other articles are being reviewed.
- 1:05:30And if they just find the slightest problem in his other
- 1:05:32articles, well, he's out of a job.
- 1:05:36And I think poor guy. But the thing where I lose a bit
- 1:05:41for him is there was, I think an interview with him where he
- 1:05:45says, but I didn't make a mistake, the AI did.
- 1:05:50Yeah, You know, yes, the AI made a mistake, but you're the guy
- 1:05:54that forwarded to that, forwarded this to your boss and
- 1:05:57said, here's my story, boss. So, you know, you kind of have
- 1:06:01to take responsibility for that. So remember your accountability
- 1:06:06when you're using AI. You can't outsource that, sorry.
- 1:06:10Yeah, absolutely, self, really categorically.
- 1:06:14There's so many examples down that path.
- 1:06:17Shouldn't talk for too long. But one of my favorite examples
- 1:06:22is that the computer manufacturer and phone
- 1:06:25manufacturer and so on. Apple.
- 1:06:27Well, choosing the name of a fruit for that kind of product?
- 1:06:32Not such a wise decision if at least you want a machine to
- 1:06:35understand what you're dealing with.
- 1:06:38But but it began with the letter echo, so they got first in the
- 1:06:40phone book. True.
- 1:06:42And they're still here so and I mean you were talking about the
- 1:06:47journalist getting in a bad position.
- 1:06:51There have been numerous cases that the international
- 1:06:55consulting company Deloitte also had a well, problematic case
- 1:07:00where they really, they were charging millions for making a
- 1:07:03report that should be the fundamental decisions.
- 1:07:06And it turned out that there were too many AI hallucinations.
- 1:07:10And so I mean, there are here that there are people being
- 1:07:14fired for this. There are organizations losing
- 1:07:18their good reputation because of not handing this properly.
- 1:07:23And I I shouldn't go into the details of the Deloitte case.
- 1:07:27You can do your own research. But I mean, they used to have a
- 1:07:33high standard and now at least there are some scratches in that
- 1:07:38surface. So it really does have
- 1:07:41consequences. And it's so important to do
- 1:07:45things properly. We are responsible for our
- 1:07:47choices, including the choices that are supported by machines.
- 1:07:52Definitely thank you both of you, because you highlighted so
- 1:07:55very well and it's so clear in this part section whatever in
- 1:07:59regards to AI because still a zoo also mentioned many times
- 1:08:04both of you, but mainly used today in this episode that
- 1:08:09responsibility such plays such a huge, huge part within this area
- 1:08:14of using AI guys. It's never less or more.
- 1:08:19You would just have it that you have to be responsible in so
- 1:08:23many parts of it, not just, you know, thinking about it, OK, I
- 1:08:28have to be responsible, but really truly become responsible,
- 1:08:31OK. There is a difference between
- 1:08:34the dreaming and reality thinking and illusions and
- 1:08:38etcetera. Thank you so very much for today
- 1:08:41guys. Thank you both of you.
- 1:08:43It was really full of informations, fresh informations
- 1:08:46and updates. Thank you so very much.
- 1:08:48I really appreciate it. Thank you too.
- 1:08:50Thank you for having us.