Artificial Intelligence will be transformative across all elements of the media business. It may well kill off some jobs but will inevitably create others, as well as raise many interesting questions, not least about creativity, data and trust. George Jarrett talks to EBU Director of Technology and Innovation Antonio Arcidiacono about the good, the bad and the ugly.
The AI front line for all EBU members has been the AI and Data Initiative (AIDI), a huge discussion topic during IBC2023.
“It is very fashionable these days, but it has been there for the last four years. We were anticipating the fresh phases of AI applications for media, and there are plenty of possibilities with people interested in so many areas,” said Arcidiacono. “AI is seen as an accelerator in terms of automating many functions that have been manual historically.”
In most cases the human is combined with the automatic. “A typical example is language management, for which you can leverage AI to do special text conversion, text to text translation, and text to speech re-working,” he added. “All the media companies want to see what they can use AI for, but the main thing is that the evolution of AI is quite exponential currently. The new versions with revolutionary features are happening every 6-12 months.”
People need to follow this and anticipate the possibilities. “To do this you need to go from a linear product to a process that can digest an exponential evolution. This is an interesting challenge that we are confronted with,” said Arcidiacono. “It is also an opportunity for us to revolutionise the way we develop content.
“In order to stay at speed, the secret is in combining the work of creative editorial people, engineers, programmers, and data scientists. You are not going in a series of steps, but combining the expertise of your people,” he added. “You can develop new products and services using the tools put at your disposal by the various algorithms, and at high speed.”
New jobs will be created
The impact of AI will last for decades but is has a dark side to beware of. “You can use it for good or bad, so I am not surprised that so many people are concentrating on the dangers that exist. One danger is the number of jobs that could be supressed, but new jobs will be created,” said Arcidiacono.
“But the dangers are the more distorted way in which you can use these technologies in terms of disinformation – generating dangerous content that can create confusion in people, cities and nations,” he added. “But at the same time, it offers a lot of opportunities, and you can create new ways of communicating, and involve the end customers in what you do. But what is key is that we must consider helping people to get the education for using this new technology.”
Arcidiacono believes that younger generations must be shown how these AI tools impact on society, so that they understand the importance of leveraging them, but also how to defend themselves.
“It is like the story of fake news detectors. You can create an algorithm, but the fake news generators will discover it, and fool it. You could try to defend using detectors and filters, but most of all what we do at the EBU is to use trusted content that can be traced and identified easily, so it is like signing content you can read, and be certain that the content has been certified as trusted,” he said.
“Other, more dangerous things are in taking the identity of people and creating problems, particularly for the young generation, which is why young people should start learning from the age of curiosity, 7-10 years of age,” he added. “Through educating the young you will indirectly educate their parents, who will then discover something they had previously not completely understood.
“It is a long-term investment because it takes 15-20 years to a time when they are fully productive. We are public service, there to inform but also to educate. Today there is an enormous competition for the existing talent between the media people and all other industries like the banks, pharmaceutical giants, health, and automotive giants,” he continued.
We are condemned to be smarter
The EBU has melded data with AI, giving it more time to achieve what it wants with its plan to put both at the core of its organisation. It focusses on four key areas – PSM strategy, values and ethics, regulation, policy and advocacy, and content creation/delivery.
“You can take more time to be better off, and to increase quality. In a way we are condemned to be smarter. We need to be smarter in outperforming AI because if we leave it generating content at a certain moment the content it will generate is a simple re-shuffling of what it has been producing earlier on, so there is a sort of convergence to that quality,” said Arcidiacono.
“But using the tools that are accelerating your ability to produce content wisely, you can add both much more value and added quality. That is why I speak about improving our future rather than becoming slaves,” he added. “The other element which is important is that it is another step in the direction of getting a sort of pre-cooked answer. In this AI revolution and in generative AI in particular the information is pre-cooked for you, and you are just given the final version which is already summarised, so it is very important that the source is trustworthy, otherwise it is a typical disease of garbage in, garbage out. In the end the AI mechanism is just machinery, so it must be fed with interesting, smart new information.”
News has always led in the adoption of new technology, so how will AI gift news agencies efficiencies and savings?
“It is already used in language management. With our News Pilot for example we operate a network with 30 PSM sources of news. There are 4-5,000 stories generated every day, translated in all languages for each country,” said Arcidiacono. “The other one is the recommendation engine, where we use AI to find content that is semantically relevant. It is an amplification of all the tools that news people have already, and they now have more possibilities of generating consumer interest in news.”
“At the same time, it is an automatic mechanism that people can use to generate biased news, and we need to defend ourselves with ways of identifying where it comes from. We could be inundated by fake and biased news, so trusted sources are key,” he added. “We developed for the News Pilot a thing called News Relator. What we do is take any text that is on the Internet, we give it to this tool, which is AI-based, and this will extract from the two million news stories in our data lake and we can extract news elements that are relevant and equate to the text taken from the Internet. It gives you a reliability judgment.”
Arcidiacono was preparing for the EBU’s first AI summit, on the second day of the EBU general assembly (December 1). He said: “AI is in every area from legal and policy to technique, from content to communication so we have created a co-ordinated exchange offering shared information and documents.”
Should not hide behind a finger
Looking at other areas – people are looking at object-oriented encoding for 5G Broadcast that would use AI for further compression, and cyber security has been acknowledged as centrally important.
“We have a specific group that concentrates on cyber security within my team, and the strategic problem of the technical department and the technical committee. We have been working on this for 3-4 years and we have given it a lot of attention,” said Arcidiacono. “But the official element, which is at the crossroad of between production, and distribution, and cyber security is what we call business continuity management (BCM). This means having multiple infrastructures. I often refer to the MARS strategy, which is multi-layer, anywhere, resilient, sustainable.”
The AIDI action plan is always in motion, with the EBU working with its members to identify which areas it should concentrate more on, beyond what it is already doing.
“And we are involving the big players in what we do. We should not hide behind a finger because there are companies in the world that have spent tens of billions to develop algorithms and AI technologies. Our idea is to always work together with them, leaving the choice to each media company to select the one, or switch from one to another,” said Arcidiacono. “We give them inputs so they know what we need them to develop.
“We are using Amazon, Google, Speechmatics, and DeepL, and we are also using algorithms that come from universities. And the beauty of that is we have created is a meta layer that is able to switch from one to the other. You have the workflow for your production and move from one to the other and use the best in terms of quality or in terms of price,” he added.
The AIDI report on journalism was couched in ethical terms with provision of security guidelines and the protection of sources.
“The EBU has spent 5.2 billion Euros on new content, so it was giving journalists the tools and strategies to defend themselves, and the ways to help do their jobs,” said Arcidiacono. “The power of people trying to leverage AI to their individual personal advantage could be extremely dangerous.”