Data Insiders: Fake news and AI — The race against disinformation (2024)

Tiivistelmä

Tietoevryn Data Insiders -podcastissa Kaius Niemi — Helsingin Sanomien, Ilta-Sanomien ja Taloussanomien entinen päätoimittaja, Miltton Groupin varatoimitusjohtaja ja Finnish Reporters Without Borders -puheenjohtaja — ja Tietoevryn arkkitehtuurineuvonnan johtaja Thomas Rosqvist keskustelevat generatiivisen tekoälyn vauhdittamasta disinformaatiosta. Niemi erottaa disinformaation (tarkoituksellinen, strateginen) ja misinformaation (tahaton leviäminen) ja korostaa, että avoin demokratia ei toimi ilman faktoja: "there is no democracy without facts". Rosqvist avaa teknisiä ratkaisuja — C2PA-konsortion standardit, Metan Stable Signature -vesileimat, asymmetriseen kryptografiaan perustuva sisällön todentaminen ja hahmontunnistus — ja toteaa, että ihminen on informaatiotulvan pullonkaula. Jakso käsittelee deepfake-robocalleja (Biden), Makedonian Velesin Facebook-tehdasta 2016, Brexitiä, Federal Communications Commissionin sääntelyä, EU:n, USA:n ja Kiinan eriäviä AI-sääntelyfilosofioita (rights-driven vs. market-driven vs. state-driven), Open Society Instituten Media Literacy Indexiä (Suomi kärjessä, 75 % resilienssitaso), pohjoismaista destination-strategiaa mediabrändien kestävyytenä ja organisaatioiden vastuuta läpinäkyvyydestä, henkilöstön kouluttamisesta ja kriisivalmiudesta tekoälyaikakaudella. Kirjasuositukset: Anu Bradford — Digital Empires (Kaius Niemi suosittelee); Richard Feynman — The Pleasure of Finding Things Out (Thomas Rosqvist suosittelee).

Listen

Spotify — full episode · Tietoevry blog (show notes)

Main topics

• Definitions: fake news, disinformation, misinformation
• Generative AI and the acceleration of propaganda
• Deepfakes, AI-generated voices and robocalls (Biden case)
• Election interference: 2016 US elections, Veles (Macedonia), Brexit
• C2PA standards, Meta's Stable Signature, asymmetric cryptography
• Pattern recognition: Infopulse + ChangeTogether + Microsoft case
• Media literacy as democratic infrastructure (Open Society Institute index)
• Nordic destination strategy and trusted media brands
• EU rights-driven vs. US market-driven vs. China state-driven AI regulation
• Organisational resilience, internal culture and crisis preparedness
• Transparency in AI use: client disclosure, content tagging
• AI augmentation in software engineering and IPR considerations
• Federal Communications Commission ruling on AI robocalls
• Critical thinking in the school system
• Information warfare and the volatile global environment

Referenced people

• Kaius Niemi — former editor-in-chief of Helsingin Sanomat, Ilta-Sanomat and Taloussanomat; Deputy CEO of Miltton Group; chair of Finnish Reporters Without Borders.
• Thomas Rosqvist — Head of Architecture Advisory at Tietoevry Create.
• Oona Ylänkö — host, Data Insiders podcast.
• Joe Biden — referenced in the AI-generated robocall case in the US.
• Donald Trump — referenced in the 2016 US elections context.
• Vladimir Putin — implicit in the discussion of authoritarian influence and information operations.
• George Orwell — cited via the "two minutes hate" concept from 1984.
• Anu Bradford — Finnish-American professor of law at Columbia Law School, author of Digital Empires – The Global Battle to Regulate Technology.
• Richard Feynman — Nobel laureate physicist, author of The Pleasure of Finding Things Out.

Referenced organizations

• Helsingin Sanomat — largest subscription newspaper in the Nordics; Niemi's editor-in-chief tenure 2013–2022.
• Ilta-Sanomat and Taloussanomat — Niemi's editor-in-chief tenure 2010–2013.
• Miltton Group — Nordic independent consultancy where Niemi serves as Deputy CEO.
• Finnish Reporters Without Borders — Niemi as newly elected chair.
• Tietoevry and Tietoevry Create — podcast publisher and Rosqvist's employer.
• Infopulse — Tietoevry's Ukrainian subsidiary with around 2 000 employees.
• ChangeTogether — Ukrainian NGO partnering with Infopulse on disinformation pattern detection.
• Microsoft — technology partner in the pattern recognition system and in the C2PA consortium.
• Meta — developer of the Stable Signature watermarking technology.
• C2PA (Coalition for Content Provenance and Authenticity) — multi-vendor standards consortium.
• European Union — rights-driven AI regulation.
• NATO — implicit in the broader democratic resilience discussion.
• Columbia Law School — Anu Bradford's institution.
• Open Society Institute — author of the Media Literacy Index.
• Federal Communications Commission (FCC) — US regulator that ruled AI-generated robocalls illegal under federal telecoms law.
• US State Department — author of a blueprint on AI's potential threats and incident preparedness.

Referenced technologies

• Generative AI and large language models (LLMs)
• Deepfake technology and AI-generated voices
• Robocalls driven by synthetic speech
• Meta Stable Signature — invisible pixel-level watermarking
• C2PA content provenance standards
• Asymmetric cryptography — private/public key signing (SSL/TLS analogue)
• Pattern recognition for narrative analysis across media channels
• Audio transcription and diarisation for speaker identification
• AI augmentation in software engineering (coding assistants)
• Social media algorithms and echo chambers
• Content tagging and labelling for AI-generated media

Referenced geopolitical themes

• Democratic resilience and the open society
• Authoritarian influence and information operations
• Election interference — record number of elections in 2024
• EU rights-driven vs. US market-driven vs. China state-driven AI governance
• Brexit referendum and external influence campaigns
• 2016 US presidential election and the Veles (Macedonia) Facebook operation
• Ukraine — Infopulse, ChangeTogether and documentation of war narratives
• Nordic democratic and educational systems as models
• Weaponisation of social media
• The volatile global information environment

Key insights

• "There is no democracy without facts." Niemi grounds the discussion in democratic theory — an open society needs a shared factual base.
• Humans are the bottleneck. Rosqvist: information overload is the real constraint, not the supply of content.
• Nordic destination strategy. Trust in media brands — not in aggregators — is the long-term "vaccine" against AI-driven havoc.
• Media literacy as infrastructure. Finland's 75% resilience rate (Open Society Institute, 2022) is the product of the school system, not innate national character.
• Two technical levers: watermarks and signatures. Stable Signature and C2PA versus asymmetric cryptography — but both require coordinated adoption by creators and platforms.
• AI verification needs cross-vendor consensus. Standards exist in fragments; the next twelve months are pivotal.
• Robocalls are already operational. The Biden deepfake call triggered the FCC ruling — the future is already here.
• Organisational resilience starts with culture. Strong internal trust reduces susceptibility to external smear campaigns.
• Transparency is the guiding star. Audiences and clients expect organisations to disclose how AI is used.
• Journalism as societal defence. Trusted media brands are democratic infrastructure, not merely commercial products.

Key quotes

"There is no democracy without facts. And that's why it's so important, it's the basis of everything, as we know it, in the Nordic countries especially."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"I think we need to fight against disinformation, but we need to be resilient against misinformation."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"We are going to see interestingly some very imaginative different permutations of using these technologies, but how much do they then have impact, that's another case."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"Humans for sure are a bottleneck in the question here, so there's too much information coming towards us, and we are definitely the bottleneck in processing that information."

— Thomas Rosqvist, Data Insiders (Tietoevry), 16.5.2024

"There has been a strategy, which is called destination strategy. And it's very, in the Nordic countries, it has been used by the media companies, which means that the most of the news is being consumed through the media brands, not through aggregators."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"Even if there would be AI kind of a havoc happening in the future years, I think the media brands could be sustainable and people could hopefully trust them and there would be this kind of a vaccination against it."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"People do expect organizations, companies, politicians, state actors to be as transparent as possible. So this is something that needs to be sort of like underlined."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"We are entering a little bit more volatile world, not only from election wise or media wise, but also that the environment is changing, and people do have these rages and it might affect a lot."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"Building a strong culture in an organisation is something that generates trust. When you have a strong culture and a strong capacity to trust, it's much more difficult to influence internal employees from an external world point of view."

— Thomas Rosqvist, Data Insiders (Tietoevry), 16.5.2024

"It's about understanding what's human made and what's machine made, what's something synthetic."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"We can see that European Union is actually willing to regulate the AI more thoroughly in a kind of a rights driven approach compared to the US, which is market driven. And then for instance, China has got this kind of a state driven model."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

"The Federal Communication Commission is actually in the US ruled that the robocalls using AI generated voices are illegal under federal telecoms law, which is like really opening the door to find lawsuits against these people who might be conducting those."

— Kaius Niemi, Data Insiders (Tietoevry), 16.5.2024

Context: why this conversation matters

This 2024 Data Insiders episode sits at the intersection of three threads that run through Kaius Niemi's entire career. His journalistic identity — built as a foreign correspondent in Kosovo, Afghanistan and Iraq, and refined through nine years as editor-in-chief of Helsingin Sanomat — gives him a long view of how propaganda has always existed but only now scales industrially. His leadership of the HS digital transformation — which broke a twenty-five-year subscriber decline in 2017 and reached over 400 000 subscribers by 2021 — is the lived foundation for the "destination strategy" argument he advances here. And his current platform at Miltton Group, where he co-hosts the Tekoälyn rajamailla podcast on AI, democracy and geopolitics, places him directly in the AI governance conversation. The Tietoevry episode also anticipates the press freedom work he leads as chair of Finnish Reporters Without Borders — fact-based public discourse is the same battle whether it is fought against editorial intimidation or against synthetic media. The conversation belongs in the longer arc of his career timeline alongside the awards recognising work on misinformation resilience.

Full transcript

Source: Tietoevry — Data Insiders, "Fake news and AI – The race against disinformation", 16 May 2024. Transcript reproduced verbatim from the publisher's transcript with only light formatting for readability. Speakers: Oona Ylänkö (host), Kaius Niemi, Thomas Rosqvist.

Cold open

Kaius Niemi: We are going to see interestingly some very imaginative different permutations of using these technologies, but how much do they then have impact, that's another case.

Oona Ylänkö (host): Welcome to the Data Insiders podcast. I'm your host, Oona Ylänkö. Joining me are two guests with over thirty years of experience in journalism. Kaius Niemi is the former editor-in-chief of Helsingin Sanomat, Ilta-Sanomat and Taloussanomat, is currently serving as deputy CEO at Miltton Group, as well as the newly elected chair of the Finnish Reporters Without Borders.

Also joining us is Thomas Rosqvist, Head of Architecture Advisory at TietoEvry, leading the team of architects and guiding architecture topics, including adaptations of generative AI. Warm welcome to both of you.

Thank you. All right, so big topic. To get us started on it, let's first define some terminology. Kaius, what exactly do we mean when we talk about fake news and disinformation?

Defining fake news, disinformation and misinformation

Kaius Niemi: Well, that's a big question. Fake news is something which emulates sort of like traditional news. There's always a name behind doing fake news, a purpose. So that's kind of a disinformation, it's purposeful for some reason. Trying to make people understand or actually accept, you know, the news, which is not fact based. So basically, that's the element. Disinformation is something which is very purposeful, and there is — it's part of a strategic sphere where people are being targeted. I guess there's also then misinformation.

Oona Ylänkö: Is there a difference? What does that mean?

Kaius Niemi: Well, misinformation is something which is not that purposeful. It might be accelerated by disinformative ways and then misinformation follows. That means information can just appear without a kind of a purpose. So there is a distinction between those two. And it's actually important to understand the difference. I think we need to fight against disinformation, but we need to be resilient against misinformation.

Oona Ylänkö: Okay, can you expand a bit on why? Why is it important to fight both?

Kaius Niemi: Well, if we are thinking about the basic democratic values and open society, as we know it, it's based on facts. So that, you know, individuals and communities can do their thinking and understand and be actors in the society based on factual topics and issues. If you have distorted understanding of reality, you cannot be an active member of society — you are under authoritarian rule, or you are misguided by someone. So that's like the deep end of it — so there's no democracy without facts. And that's why it's so important, it's the basis of everything, as we know it, in the Nordic countries especially.

How the media landscape changed

Oona Ylänkö: With your very long career in journalism, you've been at the forefront and seen this phenomenon develop over the years. I feel like in order to really understand the challenge of disinformation, it's worth spending some time dissecting the developments within the media landscape in recent years, starting from the way that we consume news. That has radically shifted. So how have you observed this change?

Kaius Niemi: I would say when talking about disinformation, there's nothing new. Propaganda has been there always for thousands of years. We have had it. On the other hand, the digital sphere enables propaganda to be much more effective, and it really goes deep into our psyche. So that's the element. When we are discussing misinformation and the digital sphere, we are talking about leveraging the ways that propaganda has been there. So it's much more sophisticated, it's much more scalable.

So basically what we see in every part of the digitalization, both good and bad, it's also part of the misinformation and disinformation, and special disinformation campaigns. And then we can see the change in various aspects. The media landscape has changed a lot. There was formerly a monopoly of delivering the news, that has changed radically, in the beginning of the Internet. Then there's the social media sphere, which came like fifteen years ago, pretty much — that was the time when it evolved radically, that has changed relationships and created echo chambers, it has diminished the power of traditional media.

Oona Ylänkö: And then we have now AI, which is a third question — what's the next phase of it?

Kaius Niemi: I think that disinformation has been there always, it's the question how effectively it can be used, part of the strategic sphere with new ways of digitalizing everything around us.

The technical view: information overload and weighting

Oona Ylänkö: Absolutely. I think this is maybe a good time to turn to you, Thomas. ChatGPT was released nearly one and a half years ago, further changing our media landscape. And it is easier than ever to fabricate and spread misinformation. Can you tell a little bit about how you view this situation from a technical standpoint?

Thomas Rosqvist: Yeah, for sure. We used to be having institutions and talking with our families and friends — that's where we gained our information. Now we have all the automatically generated content and with social media and internet of course, and so on. It sort of feels like for humans it becomes a sense of an overload. Humans for sure are a bottleneck in the question here, so there's too much information coming towards us, and we are definitely the bottleneck in processing that information. I guess the issue is that when you have that much information coming towards you, it's very difficult to filter out and put weights on that information. So, if you have old monopolies or institutions of media channels that you have trusted in the past, and now you are getting all this information from various other channels like YouTube and all these others, it's very difficult for us humans to put a weight on this. And we start to treat all this information with a similar weight structure. And then it becomes very, very blurry for us.

Standards: C2PA, watermarks and asymmetric cryptography

Oona Ylänkö: Are there already some tools for us to kind of place this weight then on the different sources?

Thomas Rosqvist: There are some technology consortiums like C2PA that are working with Microsoft and other technology vendors on creating standards. The issue is that you need both parts of the table. You need technology partners who are part of creating the content, like images and text and so on. And you also need the publishing part of the table — those platforms that are publishing that information. Those two parts of the table need to play together and create standards. We are really on the brinks where it's important at this stage that technology vendors and others are playing together into getting these standards in place, so that it's possible to verify content that is being generated.

Oona Ylänkö: I've heard this mention of standards previously, but I have to say, I don't understand what it actually means in practice. So what would something like that look like?

Thomas Rosqvist: Yeah, I guess there are two ways to tackle the situation. One is identifying content that has been generated. The other is identifying patterns of content that is being generated on the internet and in media sites.

Regarding the first one — how to detect content that has been generated — there are basically two ways. One is what we call watermarks. There are different standards for this, but nothing to my knowledge, at least anything that has been agreed over all vendors. When you create text or images or video or whatever, you can assign a fingerprint into the media. That is something that is not visible to us humans. The fingerprint alters some pixels with a certain algorithm. Then you publish that media content and the receiving end — being Meta or LinkedIn or some other publishing platform — when they have that signature, they are able to verify when that content was created and by whom.

Oona Ylänkö: How effective are tools like this? I guess this is a question to the both of you.

Thomas Rosqvist: Meta has this Stable Signature, which is one example of creating these signatures. To my knowledge, it's very effective when the image hasn't been altered. But if you, for example, crop the image to something like 10% of the original image, it can still detect the accuracy with something like 84% or so. And even if you crop the image and you do some alterations to the coloring and other of the image, it's still 65 or so percent possible to identify that image.

But then we come into problems like if you, for example, take screenshots with low resolution of the image and then publish it, then you may not be able to identify that anymore. But that's one way to add watermarks. And then the other one is the usual signing of content. So using asymmetric cryptography that we know from SSL and TLS. When the browser establishes a secure connection to the server, it's encrypted; when you send encrypted emails — that's where asymmetric cryptography is being used. The same can be used here. When you generate content, you assign or encrypt it with a private key, then you publish the public key to the publisher platform. And that publisher platform is able to open up the message and provide it to user.

Kaius Niemi: It's complicated, isn't it? Yeah. I think the most relevant thing is that we have technological solutions in order to tackle this. And it definitely regards different kinds of regulations. We can see that the European Union is willing to regulate AI more thoroughly in a kind of a rights driven approach compared to the US, which is market driven. And then for instance China has got this kind of a state driven model. But even China is willing to regulate, at least at some point. Of course, like then there are different kinds of approaches here. But I would say that technical solutions to prove authenticity are something which is really, really needed in the very near future, even now, because we are ending up into situations when deep fakes or whatever — AI based emulations of the reality around us — going to create havoc, you know, and it's definitely going to be used. I'm counting on Thomas that you guys are going to find out the right technological solutions.

Thomas Rosqvist: To my knowledge, no one agreed standard across platforms. There's no single way to identify. Of the ways to identify content, one is really using human based knowledge — how we identify content, how we identify patterns, something suspicious — like if you delve into comment sections in discussion forums, it's quite easy in many times to detect bots when an account has subscribed to a lot of other accounts but does not necessarily have that many followers.

Election interference: Brexit, Trump 2016, robocalls

Kaius Niemi: Right. There was a blueprint being done by the State Department of the US, which tackled the issues of how to understand AI's potential threats. One thing was actually the possibility to have these incidents. So from the state level, we can see that there is a huge and urgent need to even prepare for incidents. This year is going to be very crucial, because there are so many elections around the world. There are more elections around the world than never before, just happening. I'm wishing that there wouldn't be any sort of real serious incidents, but we have seen those already, when it comes to Brexit or the first Trump elections and so forth.

Oona Ylänkö: Can you maybe expand on that a little bit? Like you said, the topic of disinformation — the urgency is very clear right now with nearly half of the world's adult population heading to the polls. What are some examples where we've seen?

Kaius Niemi: We've seen a lot. We can expect social media influencing, that's for sure — doing it more thoroughly and effectively, emulating traditional media. Deep fakes such as robocalls. And these robocalls are something — even that's not so new — but you can have Biden calling you and it really sounds like that. That has happened in the US. The Federal Communication Commission ruled in the US that robocalls using AI generated voices are illegal under federal telecoms law, which is really opening the door to find lawsuits against these people who might be conducting those.

There is certainly some sort of preparedness, but we are having elections all over, so this is only the US. We are going to see interestingly some very imaginative different permutations of using these technologies, but how much do they then have impact, that's another case. It was wild — in 2016 when there was the presidential elections, even there was this Macedonian city called Veles and the youngsters were doing all these kind of Facebook sites and really that was industrial, paid by entities that we don't know so much about, but they were very effective. Facebook sites were very efficient when it comes to the presidential elections.

Did Trump win because of those? I don't know. But there were actions conducted outside the US, and so there was a kind of perpetration against the free elections. And that's how it goes. There is a threat.

Pattern recognition in practice: Infopulse and ChangeTogether

Oona Ylänkö: Thomas, you spoke a little bit about pattern recognition. Can you explain what that means in practice?

Thomas Rosqvist: Well, in practice it's about — if you're not able to technologically verify content, but you see that the content is there, then of course there are many ways of being able to detect that, with specific events over the world. Are there specific topics that are immediately or suddenly being more discussed one week before the event, or at the same time, or afterwards? For example, in our case, one example of this — there are around 2,000 Ukrainians in Tietoevry, in a subsidiary organization called Infopulse. They have created a system together with a non-government organisation called ChangeTogether, and also together with Microsoft as a technology partner.

What they are able to do is — because they need to collect evidence for their own specific purposes, so that later on they can have evidence of what has been spoken, who has been speaking and so on — they have created a system that can connect to tens of media channels, hundreds of social media environments, scanned hundreds of thousands of media content, millions of articles. What you're then able to do is, you're able to analyse the videos and retrieve the audio from those, you're able to transcribe the audio, you're able to diarize the audio, meaning that you're actually able to identify speakers with various tools in those audio tracks. With that information, it's possible to see who has been speaking what in what information sources, and when. And you can see trends.

If you have specific events happening in the world, it's possible to source those channels and see who is talking about what in which sources. So these are just one example, I'm sure there's a lot more elsewhere.

Oona Ylänkö: What can you then do with this information? What's the outcome?

Thomas Rosqvist: Well, the outcome is that you can just recognise patterns. If there are specific speakers who have not typically been talking about specific things, specific channels have typically not been talking about specific things, then due to whatever reason, suddenly many similar channels all over start talking about the same subject suddenly. That way you're able to detect the pattern that is happening and maybe even forecast some kind of events or later on, analyze the situation better.

Resilience: media literacy and the school system

Oona Ylänkö: I know this is a data podcast and we're talking about technical matters, but when it comes to disinformation, it seems like we're dealing with two different kinds of solutions that we need to consider. There are these technical tools, but it has been said that even the most well-read news consumer might not be able to identify false information from correct ones. So what can we do to reinforce our resilience against this? If it's not a tool? I think you guys had in our earlier discussions mentioned that Finland had ranked fairly high in some of these resilience studies.

Kaius Niemi: So true. We are discussing the resilience part — it's about humankind and humanity. So we cannot count only on technology. I think it's very important that we get this data driven way of looking at it and predicting, as Thomas said. This prediction part is the very positive side of AI — understanding a lot more, tackling things.

But there's this kind of a media literacy index, which is a global one created by the Open Society Institute, which kind of evaluates the ability to withstand the impact of fake news. So it's kind of like being vaccinated against the fake news. Interestingly, Finns are less likely to fall for fake news in the world, which is great, because it tells like 75% of Finns — this is a study conducted in 2022 — says that 75% are pretty much resilient against fake news. If we are looking at the US for instance, it's only like 60%, or Italy, it's about 48%. And when going to Ukraine, it's 37% or in Turkey, it's staggering, it's 29%.

But even in Finland, if we think about 75% people are very media literate, you still have 25% of people who might be taken and carried away by the fake news. I'm afraid this number will not get better, it might deteriorate. When talking about AI, it's very important that we are sticking into what the media is holding a grip of being trusted at.

If we think about the internet, and we think about the digital world, there has been a strategy, which is called destination strategy. It's very — in the Nordic countries, it has been used by the media companies, which means that most of the news is consumed through the media brands, not through aggregators. That's a trust building thing. You're gravitating towards the media brands rather than going here and there, not actually knowing where you read your news. You might like one media, you are following it, but you trust that media.

Even if there would be AI kind of a havoc happening in the future years, I think the media brands could be sustainable and people could hopefully trust them and there would be this kind of a vaccination against it. But I think we need to stick to it. We don't want to go in the Turkish way.

Oona Ylänkö: What can we do to reinforce the media brands? Is there something that we as regular readers can do, or is it in the hands of journalists?

Kaius Niemi: I think it's in the hands of media houses, but it's very much in the hands of schools — that we are learning to understand and to be critical on where we consume what and be very mindful that there are so many different entities and states being vigilant and trying to make us believe something which is not true. Basically we need to stick that the school system is giving us the background and the resilience against the fake news. It's easier said than done, but I think there is a good track record in Finland and in the Nordic countries. Worldwide, so many countries are looking at us — how on earth are these guys able to do that? It's not because we are better than the others, but it's the system that has been taken very seriously. That's something to really, really — we need to take good care of it in the future as well.

Thomas Rosqvist: I wonder if these are skeptical by nature. Just a few days ago, I was discussing with my children over the breakfast table — how much do you have in school, how much are you taught in school about this critical thinking? Apparently they are, so.

Kaius Niemi: Good, yeah.

Thomas Rosqvist: Even if all children at that age wouldn't have access to YouTube, there's always someone who has, and they will of course discuss it at school. So it's good that the school system, not necessarily even saying what is a trusted source of information, but just teaching what does it mean to apply critical thinking in general.

Kaius Niemi: That's right, exactly.

The role of businesses

Oona Ylänkö: Let's discuss for a little bit, what's the role of businesses in all of this? What can businesses do? And should they, first of all, why should businesses care about misinformation?

Thomas Rosqvist: I think for businesses, it's actually quite important and not always enough thought about. For businesses there are a lot of possibilities to smearing campaigns in the external media. That can have, I don't have any specific examples, but it can have direct impact for shareholders as example. I think that's also one really important factor why organizations in general should think about organisational culture. Because building a strong culture in an organisation is something that generates trust. When an organisation has a strong culture, we are talking about brand, we are talking about tales around bonfire basically — these type of things and the stories. That is a part of organization culture. That generates a sense of belongingness and trust among its employees. When you have a strong culture and a strong capacity to trust, it's much more difficult to influence, for example, internal employees from an external world point of view.

Kaius Niemi: Yeah, George Orwell wrote about this two minute rage, which was part of the authoritarian system he described in 1984. But even in the democracies you get this rage of factual based things, and you might face it because of misinformation. I think there are really big risks nowadays for companies to face which might really make a serious hit on their brand reputation, loss of consumer trust and then lead to potential financial losses. You have to be very streetwise, you have to preplan if something goes wrong. There is nothing new in that. But it's really important that there are plan B, C and D if something goes wrong.

That's why we are entering a little bit more volatile world, not only from election wise or media wise, but also that the environment is changing, and people do have these rages and it might affect a lot. Is it serious? Might be serious, but if you're well prepared, you're better off.

Oona Ylänkö: In most cases when we talk about disinformation, our minds go to political disinformation and objectives and agendas, but then of course there's also commercial kind of objectives behind misinformation. What can an organization do to counter misinformation? You were talking about trust, but is there something else?

Kaius Niemi: I think it's important that companies are monitoring what is being said, and then developing a response strategy, educating their employees. There's an internal aspect as well, and maybe using technology, as Thomas said, like there's different means of doing that. And then engage the stakeholders as transparently as possible. The transparent thing is so important nowadays. People do expect organizations, companies, politicians, state actors to be as transparent as possible. So this is something that needs to be sort of underlined.

Transparency, AI labelling and ethical guidelines

Oona Ylänkö: And in the context of AI, what would transparency look like?

Thomas Rosqvist: Well, one that technology vendors are implementing, as one example, is just to label or tag images or any media content. When some content has been AI generated on a platform, it should be tagged, visibly tagged, so that users know that this has been generated. And that's a good question — when do you come into censorship? My view is that we probably shouldn't disband those content generated by AI, but at least people, humans should see that this has been generated by AI, so they can make informed decisions by themselves. And we come to the critical thinking part again.

Oona Ylänkö: Sure. Both TietoEvry and Miltton have some sort of ethical rules and guidelines around the use of AI. Can you tell a little bit about that?

Kaius Niemi: It's important that the clients know how much AI is being used, for instance, to a project. It's not like giving a full disclosure if you have used AI for instance just to make the grammar better. But if you are using AI in a much more thorough way, then the clients need to know that. The same applies to media, almost every industry, that the clients know how things are being generated. And again, it's transparency. It's about understanding what's human made and what's machine made, what's something synthetic.

Thomas Rosqvist: We have similar. I come from the engineering side. For example, if we plan to use in software engineering, we call these AI augmentation. Software developers use this tool, kind of like programming buddies, to have AI tools assist the software development. But we do need sign off from our customers to be allowed to use these, because some customers can be quite sensitive regarding the data that is being used. We come into IPR — is it used to train data? If you are a software product company, you are creating a product that you are then selling to your customers. It's quite important for them that you don't breach the IPR when generating code to those products. For other software products, it's not as important.

Kaius Niemi: Data security is really important and big issue when talking about AI and these LLMs, large language models. How much you put in and are these buckets safe? That's something which has been discussed a lot throughout the industries.

Closing book recommendations

Oona Ylänkö: We're reaching towards the end of our episode, so it's time for our last question. For our listeners, is there something that you've read or come across that might be interesting, if they're interested in digging into this topic deeper?

Kaius Niemi: My favorite is Anu Bradford, a Finnish-American legal scholar. She's a professor of law at Columbia Law School. She just wrote a book very recently, which is called Digital Empires – The Global Battle to Regulate Technology. That's a fascinating one, and that has really made a big change in the US and worldwide. Anu's book is really tackling the issues how to regulate AI and understanding the differences between the different continents, how things are being done. Very recommendable.

Oona Ylänkö: Thank you.

Thomas Rosqvist: My book is not related to technology at all. But I have some books that I usually recommend. For this one, I think I read a book many years ago, probably six years ago, by Richard Feynman. Richard got the Nobel Prize in quantum physics. He was into theoretical physics, and he has this book called The Pleasure of Finding Things Out. I think it's a wonderful book for this topic, because it goes into critical thinking and being curious about things, because all the topics that we have been discussing today are about curiosity and critical thinking.

He has this wonderful example where when he was a boy, he was in the woods with some of his friends, and the friends were asking him what is the name of this bird. He didn't know, and he was sort of ashamed of that, then discussed with his father. His father said the bird's name in all five languages, then said that the only thing that you have learned now is how we humans describe this bird. We humans have given a name for this bird in five different languages, but you haven't actually learned anything about the bird. The important thing is to actually observe and be curious about the bird and its behavior. I think that's sort of almost a guiding line for this discussion.

Oona Ylänkö: Great. Very much on theme. That brings us to the end of our discussion today. Kaius, Thomas, thank you both for sharing your thoughts today. To our dear listeners, thank you for tuning in. If you found today's discussion enlightening, don't forget to subscribe to the Data Insiders podcast by clicking on the bell sign. Until next time.