Nabila Cruz de Carvalho Artificial Intelligence: What are the threats and opportunities for the media?

Nabila Cruz de Carvalho is a PhD student at the University of Sheffield, whose research focuses on trust and inequalities in the context of digital news media in the age of Artificial Intelligence (AI).

In advance of her appearance at the Festival of Debate this Thursday at a free online event, she talks to Elim Lau and shares her thoughts on the threats and opportunities AI represents for the media.

Hi Nabila. You specialise in how generative artificial intelligence may affect the trust of young audiences in digital news media. How would you define AI?

Artificial intelligence is a system that uses machine learning from human interaction to create more algorithms and systems that can work in the background.

AI is a catch-all name for machine learning and systems that can think more, or be more intelligent than humans, in the sense that they have the computational capability to continue creating or working on algorithms in the background.

What is the current role of AI in the media industry?

At the moment, AI is a magical tool that people think can be a great opportunity for the media.

If you ask a publisher, editor, or someone with power, they think it’s a threat to the media because AI is helping work be more efficient by making the process much faster. If you are an artist or actor, it is going to take away your job.

The truth is that the role of AI and the media right now is to figure out how to go forward by understanding that it’s a new technology. The media is an industry that deals with technology change quite a lot, so the challenge is not actually new. The use of AI in the media right now is leading people to think about what their own role in the media might be in the future.

You recently wrote: “AI is not going to save or destroy us because the question is not whether we can trust AI, but whether we can trust the people and structures behind it.” But most people don't know the people and structures behind it. What should we consider before using AI in the media?

It’s difficult to consider if you don’t want to understand how it works. We call any system in technology that we don’t really understand how it works a “black box”. An AI is a black box because the big, famous AI companies, such as OpenAI or Google, don’t tell us how it works.

We need to consider the power relationships – basically, who benefits from this system when we use it, in terms of who’s profiting from us and interacting with the systems. The truth is that one of the main things about generative AI specifically is what we tell it. So when we give something to it, even if it’s just a prompt, it’s learning from that.

Let’s look at image-making AI, like Midjourney, Stable Diffusion, and even the video ones coming out now, like Sora. These systems look at these image libraries, where they’ve had access to millions of bits of data, so they learn statistical patterns to come out with an average image. So the image that comes out of the system is often stereotypical.

The New York Times has sued OpenAI and Microsoft over alleged copyright issues related to its written works. Can you tell us more about intellectual property and AI?

It’s very concerning. A lot of data from the internet, and things from artists that are public people, have not given permission to be used to train AI. So that’s the issue with data usage because AI looks at it and tries to find an average, which means that this artist’s art is picked up in bits, and has been regurgitated by AI.

Is AI creating something new, or is it just based on the things that are public?

It is both in a sense. It is creating new things if you think about how unique the images can be. Even though some images are very hyper-realistic, you may say: “That’s like a photograph.” It is new in the sense that perhaps we haven’t seen a specific image before.

I don’t think AI has its own style. I hate the kind of anthropomorphising. These are unique images but also generic — visuals that you could use as a stock photo.

Regarding the copyright issue, how can journalists use AI wisely? What can AI do for us?

Journalists have to be mindful of the sources. We were talking about what kind of data goes into it. You might not know how it went into the AI system and how it came out.

Hallucinations are errors that AI systems have. If you ask ChatGPT to give you the top cited articles in a specific subject, they might come up with fake articles with completely crazy bibliographies with people that don’t exist and links that you can’t even click on. That’s a hallucination, an error. Journalists have to be mindful of that when using AI, which gives them extra fact-checking work.

Some recent research has shown that when articles written by AI reporters are signalled to audiences, they think they trust the news less. Because of the bias against machines, it sounds very reductive, people ultimately don’t trust those articles.

Disclosure affects trust, but at the same time, audiences find it very important that journalists continue to uphold values of truth and honesty. So, to be able to use AI is not a problem; not being open about using it is the issue.

I’m a trainee journalist – what might AI mean for my future career?

An article I read recently gave a good example of when photography was developed as a technology in the 19th century, everyone thought it was the end of art. The art didn’t end there. It just changed, and new types of art came up, like photography and visual art. So it is difficult to predict exactly how AI will affect journalism. I don’t think journalists are going to be replaced, but the role of journalists might probably change.

What advice would you give to journalists?

Unfortunately, we have to adapt and learn the systems to be part of the conversation and not let them be the only solution. The top tip is to learn about them: becoming aware of them and understanding how they work and their effects is important to journalists. It can be a good starting point.