Nordic AI in Media Summit

The one with the agents

Put hundreds of journalism+AI practitioners in a big room together with some of the brightest and best sharing their wisdom from the stage and you’re always going to get a dynamic conversation. The Nordic AI in Media Summit (NAMS) has developed a reputation all of its own for pithy provocations but according to veterans, the thing that marked out this year was the shift to more confident insights from those already deploying AI extensively. Beyond this, it was, very clearly ’the one with the agents’

It seemed there were three key ideas that shaped the debate:

Re-stock (what we offer)
Shuwei Fang’s assertion that “The article is not the asset” was a persistent theme - the notion that audiences will consume news less via browsing a selection of material created to attract their attention but rather via their direct intentions around information. So, what is the asset? For now, let’s call it information or facts.

When I worked in local news, building short form ‘live’ pages, we tried at one point to categorise stories using impact measures. Did a story affect many people or just a few in a locality? Was it highly time limited or were there long term repercussions?

• · If a lorry damages a building in the road next to us and shuts the road - those very close by will care for the short amount of time that there’s an issue

• · If the lorry damages a substation and takes a city off the grid for 3 days, a bigger radius of people will care for longer

Both these scenarios can be conveyed in a reasonably straightforward way via atomised facts. However, if the lorry is driven by an extremist with an agenda or the substation collapses because of safety violations or corruption, we start to get into territory in which witnessing (likely, a human who has observed things and can convey them to others) or investigation (likely guided by a human hand) is invaluable. Journalism, in other words.

While many of these outcomes might then become atomised facts over time, the more complex and potentially polarised a story is, the more care we need and the more the ecosystem is likely to be polluted by bad actors feeding in false information. In the chopping and filleting, do we risk the loss or contamination of those facts? We know the answer is yes. We have seen, see via work done by the BBC and the EBU, what AI has done to our news content through often clumsy summarisation.

For some information needs, deploying a fleet of winged monkeys to winnow the prairies of information and collect exactly the kernels you need, is a great outcome. But let’s not forget that to consider journalism as a simple sourcing or conveyance of facts completely understates it. Scoff at my romantic attachment to the beauty of writing if you will, but journalism layers on so much more - experience, expertise, context, understanding.

Which brings us to trust. I’ve invested a lot of time working on media provenance - ascertaining the source and integrity of a piece of content. This is vital for both trust and accountability. Where did that fact come from, who claimed it, how do we know it’s true and what more might we need to know?

It also leads us to the thorny issue of payment. If our website is dead and our content is out in the wild being encountered and reassembled by agents, who pays? Are there ways in which content carries not only trust signals but also a mechanism which triggers some kind of micropayment? This was something addressed in their sessions in thoughtful work by Florent Daudens and David Buttle amongst others.

Having embodied intelligence (people) deployed costs money - when they’re deployed to difficult or dangerous environments that most of us can’t see first-hand it costs LOTS of money. And as Politiken’s Amalie Kestler noted: “You can’t send a machine into the world.”

To those who believe AI oversight leads to the production of purer facts, somehow untainted by human perspectives or privileges (and one delegate did say this to me) I would remind them that AI is well known for carrying its very own biases. Mecha Hitler, anyone?

Re-frame (how we offer it)
There was a strong sense throughout the event that simply applying AI to our current business models isn’t what’s needed - rather, we need a complete re-architecting. In the wise words of Ezra Eeman it’s not about putting AI into your media but about putting your media into AI - accepting that we’re read now not just by humans but assimilated by machine. As Madhav Chinnappa warns, if this is to work, the data around our content must get better.

The article may survive, whole and as fodder — but increasingly, it looks like it’ll be just one offering among the many we are starting to engage more frequently with - podcasts, blogs, newsletters, highly personalised ‘streams’ and more. AI creates abundance in all these areas but also, if we’re clever, it helps audiences find a way through this abundance to things that enrich their lives. There was much talk of the ‘bar bell’ - commoditised journalism at one end; ‘luxury’ or premium content with a strong human factor at the other. As a strong advocate of public service journalism it’s important to me that luxury doesn’t become unaffordable.

And if we’re all plugged into IVs of news serving our very specific needs, even if they’re in a variety of formats, where is the shared experience - the things we gather around? Are we, if we are not careful, providing for a narrowing? With a diet of liquid news, is there enough to chew on? And what are our routes to this content, beyond the chatbot?

What we also need to be careful not to miss is that in some cases we don’t know what we ’need’ until we see it. Serendipity always feels like a faintly ethereal word, but without being out in the world scrolling, exploring, discovering, mining for content, I would never have found some of the most enlightening and life-enhancing things I have read. I always think of Dmitry Shishkin - how will I find the things I have no idea I’m seeking that might inspire, surprise or delight? What are the mechanisms for this to happen in an economy that’s gone from push to pull?


Re-create journalism for a new era
I’m aware that some will be impatient with criticism of how AI and agents might re-make our world without any attempt to come up with answers so let me have a go.

First of all, as the organisers so elegantly expressed in their conclusions we need to find better ways of engaging news audiences in the debate. What do they want and how do they want it - but crucially, asked at the same time as offering a glimpse of the possibilities that might unfold.

I was inspired by Nikita Roy’s challenge about what journalism might become, released from the containers we’ve traditionally packaged it in (the web page, the article). The idea of journalism developing new powers, new capabilities, and new audiences perhaps becoming more accessible and woven into our lives is an exciting prospect. As is the way we can derive more value from our existing news (like we have been doing in the BBC) via the multi modal capabilities of AI. Norway’s Verdens Gang newspaper has an AI driven app and market testing lab and Schibsted is using Videofy to create multi modal versions of text stories.

Newsrooms, it was agreed, urgently need to be empowered to experiment to diversify, and they need to learn. In the words of Nikita (much quoted in the breaks) ‘awareness is not immunity’. And then of course there’s the role of regulation and governance which I won’t dwell on here except to say I can’t see us building this new world without effectively engaging with it.

What I’d really like NOT to see is two schools of thought becoming (as everything else is in society) polarised. The ‘burn it to the ground and build from scratch’ and the ’stick a preservation order on it while we work it out’. Neither will work.

What’s clear is that we don’t, can’t have a full picture yet of what will come but thanks to NAMS and the many other events that work so hard to generate this kind of debate, we’re getting closer.

I’ll end with a provocation offered to me - I won’t say by whom to respect the Chatham House roles of the breakouts. “You’re either building the bomb Laura, or you’re watching the bomb building built. The bomb is getting built. Where are you?’ Having given this some consideration, my intemperate first answer ‘on the team, messing with the plutonium’ stands, but does it have to be a bomb? What’s the version of this that is the ploughshare that we can conspire to build so that the precious information that sustains our democracies and our human rights is used to nourish our society - and without exploiting those who create it?