Forget humankind,think about media

Scientists worry that Artificial Intelligence could destroy us all. Until it does, what can it do for journalism?

In October, the Nobel Prize was awarded to Geoff Hinton, the British "Godfather of Artificial Intelligence" who, a few months earlier, had been the highlight of my Radio 4 Analysis programme How Real is the Existential Threat from AI? I met him with the programme’s presenter, the UCL professor Jack Stilgoe, in a kitchen in a north London house — a very ordinary setting to talk about "existential threats" from Artificial General Intelligence (AGI), machines that can learn and perform any cognitive task that a human can. Hinton is emeritus professor at the University of Toronto and explained how he has spent the past 50 years working on neural networks. When he left Google last year, he went public with his fears of what might happen.

Sharing a cup of tea with Geoff, Jack asked whether he was really talking about the extinction of human life? "Yes," said Geoff, "the existential threat is they’ll just wipe us out. Right? We won’t be necessary anymore." What’s more, Geoff’s estimate of when that will happen has changed. "My latest guess is they’ll get more intelligent than us sometime between the next five and the next 20 years. It’s kind of urgent." If he’s right, then between five and 20 years from now, AGI machines will be more intelligent than us. So, if it looks like we’re doomed, where does that leave journalism and the media?

The striking thing about the interview was how Geoff Hinton has adjusted the timescale for when he thinks machines will overtake human intelligence. "I always thought that in the end systems running on computers will get better than us. I just thought it was 50 to 100 years off and we had plenty of time to worry about the threat that these things would take over. Now I think we’ve already got things like GPT-4, which know much more than us."

Chat GPT-4 is the AI chatbot that can mimic human speech and reasoning, launched by OpenAI in 2023. It’s capable of creating text, images, speech, music, video and code in response to prompts from a user. It has spurred on Big Tech to invest massively At the start of October, OpenAI raised $6.6billion in funding, taking its valuation to $157billion. To put that into a local context: the BBC’s total income from the licence fee in 2023 was £3.74billion, which accounted for about 65 per cent of the BBC’s total income of £5.7billion.

As well as OpenAI, other firms — Microsoft, Google, Meta, Amazon -have joined in. According to a Wall Street Journal article in October, the combined capital bill for Alphabet (Google’s owner), Microsoft, Metaplatforms (Facebook and Instagram) and Amazon this year is predicted to be $231billion. That’s almost 50 per cent higher than 2023. Most of it will "go on chips from Nvidia and other infrastructure to power generative artificial intelligence services".

Here’s how Geoff Hinton described the process and why he sees it as a threat: "These things will be much smarter than us and will want to take control. Because they’re much smarter than us, they’ll find it relatively easy We’re not going to understand what’s happening. Probably what’s going to happen is we’re going to become less and less crucial to the decisions that are determining things".

What if we apply that thought to media: to what extent is AI making journalists less crucial, and how far might that go? As with any journalism, the best way to find an answer is to ask an expert. So I asked Chat GPT: "What impact will Gen AI have on journalism?" Here’s what it came up with:

Generative AI is poised to significantly impact journalism in several ways:

1. Content Creation: AI can automate the generation of news articles, summaries, and reports, particularly for routine stories like financial earnings or sports results. This can free up journalists to focus on more complex investigations and storytelling.

2. Research and Fact-Checking: AI tools can assist journalists in quickly gathering information, verifying facts, and identifying trends in data. This can enhance the accuracy and efficiency of reporting.

3. Personalization: AI can help tailor news delivery to individual preferences, curating content based on reader interests, which may increase engagement and broaden audience reach.

4. Enhanced Storytelling: With AI, journalists can create interactive content, such as data visualizations and immersive storytelling experiences, making complex topics more accessible to readers.

5. Ethical Concerns: The use of AI raises questions about accuracy, bias, and the potential for misinformation. Journalists will need to navigate these challenges to maintain credibility and public trust.

6. Job Dynamics: While AI can enhance productivity, it may also lead to shifts in job roles within journalism, requiring professionals to adapt their skills and possibly reducing certain positions.

7. New Forms of Journalism: AI can enable innovative formats and approaches to storytelling, such as real-time reporting and automated updates, changing how news is produced and consumed.

Overall, while Gen AI offers tools to enhance journalism, it also presents challenges that the industry will need to address proactively.

The industry is addressing some of them. For instance, the annual Reuters Institute Digital Report noted: "As publishers rapidly adopt AI, to make their businesses more efficient and to personalise content, our research suggests they need to proceed with caution, as the public generally wants humans in the driving seat at all times. Our survey data show that, across all countries, only a minority currently feels comfortable using news made by humans with the help of AI (36 per cent), and an even smaller proportion is comfortable using news made mostly by AI with human oversight (19 per cent)."

As you might expect, the BBC has been giving a lot of thought to AI. In its editorial guidelines on AI, it warns: "Any use of AI by the BBC in the creation, presentation or distribution of content must be transparent and clear to the audience. The audience should be informed in a manner appropriate to the context and it may be helpful to explain not just that AI has been used but how and why it has been used."

The guidelines helpfully provide some case uses, some acceptable (creating a synthesised voice to deliver text based content, where it does not seek to replicate the voice of another individual), while others come with a warning light (for instance, any use of transcription technologies would require careful human editorial oversight).

Where does all this leave the human editorial factor — you and me? Under the sub-heading of Job Dynamics, Chat GPT offered three notes:

1. Enhance productivity

2. It will need adapting to new skills

3. It could lead to possibly reducing positions.

I can illustrate this with a hypothetical case study. Let’s say I make a radio programme or podcast interviewing a high-profile expert. After recording the audio, we’ll transcribe it. What used to be a thankless task can now be done instantly with an online subscription transcription service. There are a number around, but in this instance I’ll use one called Descript, which is often used for podcasting. It offers "AI-powered" end-to-end editing of audio and video. Simply upload the audio file to Descript and it generates a complete transcription.

But Descript’s transcript service goes much further, with a text-to-audio editing system. Cut the transcription and the audio is cut from a timeline on the screen below the text. Highlighted text exported as an edited audio interview. Descript has another AI-powered option it calls Underlord ("who wants an Overlord?" it asks) that can automatically clean up background noise and enhance voices. Another function, "edit for clarity", takes out all the filler words and digressions. It will generate a title, a summary, show notes for the podcast, a social media post and a blog post.

I also used Descript’s AI-powered "brainstorm" function to ask it about AI and the media — and had to remind myself I’m "brainstorming" not with colleagues, but with an AI-powered transcription service. Of its five suggestions I particularly like:

Conduct interviews with AI researchers, media professionals, and technologists to get their insights into how AI is changing the landscape of the media industry.

That’s actually what I do when I make a programme. Underlord will also write a script to my brief. When I typed in "AI and media" and specified an audience of journalism experts, this is what I got:

Welcome, everyone, to today’s discussion on AI and its impact on the media industry. As journalism experts, your understanding of traditional and modern media practices gives you a unique perspective on this rapidly evolving landscape.

— Artificial Intelligence is revolutionizing the way media content is created, curated, and consumed. From automated journalism to personalized news recommendations, AI is becoming an indispensable tool in the newsroom.

— Let’s explore some practical applications of AI in journalism:

1. Automated news writing and reporting.

2. Data mining and analysis for investigative journalism.

3. Enhancing audience engagement through personalized content.

4. Using AI for fact-checking and fighting misinformation.

— With great power comes great responsibility. As we incorporate AI into journalism, we must address several ethical issues such as data privacy, algorithmic bias, and the impact on employment for journalists.

— We’ll delve into some real-world case studies where AI has been successfully implemented in media organizations, highlighting both the benefits and challenges faced.

— Looking ahead, we will discuss the potential future developments in AI and their implications for the media industry, including emerging technologies and new business models.

— Thank you for joining this discussion on AI and Media. As journalism experts, your role in shaping the future of media is crucial. Let’s embrace AI with a critical yet optimistic mindset to drive innovation while upholding journalistic integrity.

I have to say, I’ve no idea where that last bit came from: to embrace AI with a critical yet optimistic mindset? Could do with a decent editor. Incidentally, would you prefer to listen to that script — or this article — rather than read it? Descript’s selection of stock voices, generated by AI, can do it with one click. So my only intervention, from the title — AI and Media and a target audience — to downloading a finished voice piece has been to choose a voice. (If you’d like to listen to this article, I’ve used the same voice and it is at philreevell.com.) Prefer to watch it as a video? If I paste the AI-generated text into a AI-powered video service called Synthesia, while I make myself a cup of tea, it will generate a short video, with an entirely realistic AI avatar. No talent required — literally

AI-powered text-based editing particularly appeals to podcasting. I like podcasting — mostly I’ve been recently working on the launch of The Times tech pod with Danny Fortson and Katie Prescott, which in its second episode discussed the Nobel Prize award to Geoff Hinton and to Google Deepmind’s Demis Hassabis. Podcasting offers reactive speed and ease of production to deliver innovative new format. And that lends itself to AI-powered tools, which offer end-to-end production. Remember Chat GPT’s AI job dynamics: enhanced productivity, new skills, fewer roles? That’s how podcasting operates in a casualised production sector where AI has hugely assisted the massive expansion of podcast content. What we might call AI’s podcastisation of audio production continues to reach into traditional speech formats in broadcast news, current affairs and documentary production. It’s just one example of AI’s transformative impact on journalism.

Not surprisingly, for those who are entering the industry, where podcasting might offer a more welcoming gateway than broadcasting, AI can be seen as an opportunity. The post-Chat GPT journalist will have a very different perspective on using AI compared to those with established careers, who might see it as a threat — what Chat GPT would call "possibly reducing positions".

Take the podcasting students I meet each year as an honorary professor at Stirling University, I talk to them about their projects. This month, I asked for a show of hands on AI as an opportunity or threat as they prepare to join the industry. The majority saw it as an opportunity… although concern was raised about how to check the reliability of information generated through AI-powered research.

That challenge (can we trust AI?) is at the heart of fears during this year of global elections. In particular, concern about AI-generated fake news and deep fakes. Geoff Hinton, in his interview, expressed his fears about the implications of deep fakes in the democratic process.

"In the US, they can’t even agree who won the last election. So that doesn’t make you very hopeful, particularly for things like fake videos. If one of the US parties is committed to perpetrating what they know to be false, namely the claim that Trump won the last election, then there’s no hope of them agreeing to legislation against fakes."

Last month, I produced a World Service radio documentary with the BBC’s Gary O’Donoghue about the challenges facing American small-town newspapers during the US presidential election We heard how deep fake detection tools are part of a subsidised offering funded by the Knight Foundation for small media operations in swing states. It seemed like an interesting metaphor for the dynamics of AI and the media: investing in digital journalism to check the spread of disinformation.

Another check on AI is the legal challenge to the way AI’s Large Learning Models (LLMs) are "trained", scraping massive amounts of information. The New York Times has sued OpenAI and Microsoft for unauthorised use of Times articles to train the AI models. In our BBC Analysis programme, we heard from AI academic expert and author Kate Crawford, who said we are witnessing "the largest ever experiment in creativity in a post-copyright environment". She added:

"All of the data that’s been collected online, all of our writings, images, films, everything has really been gathered up into these enormous vats of content that are driving generative AI. And, of course, for many writers, journalists, artists, and designers, this feels to them like another existential threat."

Earlier this year, the Financial Times announced "a strategic partnership and licensing agreement with OpenAI, a leader in artificial intelligence research and deployment, to enhance ChatGPT with attributed content, help improve its models’ usefulness by incorporating FT journalism, and collaborate on developing new AI products and features for FT readers." The deal enables ChatGPT users to see select attributed summaries, quotes and links to FT journalism in response to relevant queries. According to FT group CEO John Ridding: "Apart from the benefits to the FT, there are broader implications for the industry. It’s right, of course, that AI platforms pay publishers for the use of their material. OpenAI understands the importance of transparency, attribution and compensation — all essential for us. At the same time, it’s clearly in the interests of users that these products contain reliable sources." Other publishing groups have also signed AI licensing deals with OpenAI.

So, does this feel like an existential threat to journalism? In search of an answer, I asked ChatGPT. Here’s what it came up with:

The rise of artificial general intelligence (AGI) certainly has implications for journalism, but whether it poses an existential threat is more nuanced. In summary, while AGI could disrupt journalism, it also offers opportunities for transformation. The key will be how the industry adapts to these changes.

It would say that, wouldn’t it?