AI-generated news broadcasts – just pick your language!

Fully AI-generated media was bound to happen eventually, and with a proof-of-concept video released at the end of 2023, it could be closer than we think. Channel1.ai proposes to give users “a more personal way to watch the news” with a global news network powered by generative AI, fully launching in 2024.

Translation: these aren’t real news anchors talking on-screen.

Take a look at the introduction video for how AI will soon be used in news:

When Emmanuel Methivier, Business Program Director and Axway Catalyst, first shared this video with me, I was stunned. As a former journalist and news anchor myself, I had a lot of thoughts – and more questions than answers.

No, I don’t think AI anchor-bots will replace journalists. But there is still quite a lot to unpack here, and as Emmanuel points out, these technologies are going to change the way we do business. Read on as we share a few thoughts in a freestyle interview format.

AI production quality is getting very, very good

Emmanuel: As you can see, the first striking element is the quality of the production. Lip synchronizations have made further progress, transitions are perfectly executed, and the only remaining artificial aspect of this interface is the somewhat repetitive postures, especially noticeable in the hands.

Lydia: Or, as one ARS Technica reporter put it, “these AI-generated news anchors are freaking me out.”

Sure, there are some quirks, but the voices are remarkably smooth, the faces friendly and welcoming, and I’m amazed at how good this looks. Considering how few people these days actually WATCH television – vs scrolling through their phones while putting dinner on the table and listening from one ear – these anchors could pass muster if you’re not paying close attention.

Frankly, that’s terrifying. Channel 1 may have the best of intentions, but in an age where misinformation already runs rampant and deepfakes are a reality, I can’t imagine the damage true fake news could do.

Are news anchors out of a job?

Lydia: Forgive my soapbox moment here, but the most immediate question that sprang to mind for me was, “what is the source material?”

It’s a popular misconception that news anchors are talking heads, people who sit in a chair with makeup on and read a teleprompter. If that were the case, why not use AI-generated anchors? It would probably be a lot cheaper than prima donna humans!

But barring the Ron Burgundys of the world, the reality is that the news anchor, well, anchors the entire show – they drive the content, they are likely writing a good portion of the scripts, and they conduct live interviews of subjects and reporters.

When I anchored the nightly news in Casper, WY, I arrived hours earlier to write and order the newscast, decide on which stories to feature and promote… On days where I was also reporting, I would do advance research (like requesting city records and sifting through piles of printouts), make phone calls, then grab a camera and head out the door in a news van to report, shoot, and edit a story for the nightly news.

The point is that all of this requires two things:

• Real people who can talk to other real humans, and
• People and cameras on-location, where the news is happening.

So, will reporters be losing their jobs? They might suffer layoffs due to continued M&As and the general decline in journalism, but I doubt AI is quite there yet.

Emmanuel: Let us not forget that the sourcing of information has also been largely disrupted and uberized by digital technology, thanks to the proliferation of smartphones and video-surveillance cameras, whose images are already much more widely used than the videos of war reporters, for example. The “dirty work” side of Robert Capa, Henri Cartier-Bresson, and Jean-Pax Mefret is now replaced by the iPhones of local actors, ordinary people, whose images flood social networks!

AI-assisted translation is breaking down barriers

Emmanuel: A second observation when watching this demo is the great disruption of languages. Take a look at this clip, where an interview of a French man surveying the damage from a storm at his property is translated directly to English, using his voice and merely adjusting the video slightly to synchronize his lips.

Those who know me know I’m more of a German speaker than English, after my native French. But with an AI tool called HeyGen, I was able to record myself speaking in French, then generate a native-sounding English translation. Lydia did the same with some of her old newsreel clips.

Lydia: It’s scary how good this is – especially with the lip-syncing! If I didn’t know your real-life voice (I think the new video gives you a bit of an Australian Arnold Schwarzenegger vibe), I probably wouldn’t know this isn’t real at first glance. And it is quite uncanny to see myself speaking Mandarin when I don’t understand a word that’s coming out of my mouth – but it sure sounds like me!

In all seriousness, the synchronized translation element felt like the biggest breakthrough in this whole demo to me. The current practice of adding a monotone voiceover from an interpreter really takes us out of the “action” in a news clip, and the idea of being able to use a person’s own voice to convey their own story and emotions to any audience is really compelling.

Then again, I’m reminded of when polyglot journalist Philip Crowther went viral a little while ago for covering the news in no fewer than six different languages. It wasn’t just a parlor trick; he spoke those languages and understood their cultures.

Does Charlemagne’s belief that “to have another language is to possess a second soul” apply to ChatGPT?

Emmanuel: The Tower of Babel is about to fall, restoring to humanity a demiurgic power that nothing can stop anymore. We can start thinking of genetic creations and the man-machine symbiosis, which has also taken a big leap forward by surpassing neuromorphic computing to propose the integration of human brain tissue directly into the computer (yes, really: Brain organoid reservoir computing for artificial intelligence).

Bringing this back to our own circle of influence, at Axway, we will need to rethink integration tools and interfaces, my friends.

I shared some insights on this in our Looking Forward 2024 predictions guide: as we entered the 2020s, many understood that we had to move from a vision of technical APIs to business-oriented digital products. The coming year will force us to reassess these products and their marketing to adapt them to the new consumers of services: the AI generatives!

Download these 8 important digital integration predictions for 2024 today.

AI-generated journalism you can trust?

Emmanuel: My third observation: we can no longer truly distinguish the truth.

Admittedly, this is a philosophical question that dates back ages, and historical falsification can even be observed on the facades of Egyptian temples (erasing hieroglyphs to rewrite history), through the famous doctored photos of Stalin, or the videos of Kuwaiti incubators.

But the grand promise of internet’s media plurality, which was supposed to provide an answer to these centralized lies of the Pravda, now seems dissolved not only by the concentration of information sources (dominated by Big Tech, the GAFAM). The truth-faking machine has taken a further step with these new generative AI tools that can create an entirely unreal image and accompany it with an equally fictional oral testimony.

Watch: Putin confronts his AI ‘double’

Lydia: And even with well-meaning actors, what else would we lose out on? AI might get very good at scanning court documents to summarize criminal charges, but can you imagine an AI-generated email requesting that a fraudulent investor self-record an interview? Who exactly will be asking the proverbial tough questions?

All the footage in the AI-generated newscast we watched above was pulled from other news sources where real reporters and videographers were on the ground doing the “dirty work.” There’s been deserved criticism of personality cults in the media, but if we don’t even know who is doing the reporting, how can we trust the information?

Much has been written about the fact that AI is only as good as the data that trains it. Almost a decade ago now, Cathy O’Neil warned that some big data algorithms are increasingly used in ways that reinforce preexisting inequality, creating a “toxic cocktail for democracy.”

The Harvard Business Review has highlighted the urgency of remediating human biases that make their way into artificial intelligence systems with harmful results.

Combine that with the well-documented decline of local news, and any AI-generated news media – even if it does have human editors to verify content – would need to be especially careful and selective of its information sources.

See also: How AI will reshape journalism’s narrative in 2024

Emmanuel: Distrust of information and therefore truth will challenge the foundations of liberal democracies and could lead to a radical societal transformation. State sovereignty in these areas seems increasingly urgent, but that’s another debate…

In conclusion: These have been some relatively heavy topics, and we certainly won’t be able to resolve all these questions in one blog post, but we look forward to continuing the discussion. One thing is clear though: the AI revolution has very real and immediate applications in the world of APIs and integration, as Emmanuel has noted throughout this past year. At Axway, we’ll be here to advise and support enterprises as they seek the right balance on the API front.