Dr Alex Connock is a Senior Fellow at Saïd Business School, University of Oxford and recently wrote a textbook on media and AI. He discussed AI through in the context of media perception with John Maxwell Hobbs.
Connock’s broadcasting background is practical as well as academic, having co-founded the indie Ten Alps with Bob Geldof in the late 90s and was the Managing Director of Endemol Shine North through 2017.
An explosion of interest in AI
Connock felt that AI’s major impact on the media landscape was getting short shrift in academia and set out to correct that. “The textbook is guide to media, reframing how you might see it in the 2020s. I was struck when teaching the subject that there really wasn’t a textbook that described the media that people actually live and work in today,” he said. “Most of the textbooks were framed around a rather 1980s vision, with a series of articles about things like exhibition, publishing, and agencies with no relation to, firstly, the way it is structured now with social networks, and streamers, and podcasters. Secondly, they didn’t really have any recognition of the technology that was defining who the biggest media players actually are: AI - in particular recommendation algorithms.”
“What was interesting when I pitched it in 2020,” he continued, “Is that Rutledge sent it off to a bunch of anonymised professors of media and business communications around the world, who all came back really cynically saying, ‘I don’t see why you would do a textbook framed around those misconceptions in the media. What do those two things have to do with each other?’ Then fast forward a year and a half, when it was published, and that subject is on the front page of every newspaper, every week, every day. It was a really interesting time when AI was sort of raising its head above the parapet in every other industry, and I just happened to write a book at the perfect time.”
Working backwards
Connock identified the path that AI has taken to become tightly integrated into media production from start to finish and came to the realisation that its introduction began at the end of the process.
“What I essentially landed upon was the idea that AI has come into the media business in four phases. But interestingly, those phases track the production lifespan of any given media project from the reverse end,” said Connock. “There is the paradigm of media projects getting produced and delivered - they start with ideation and development; then move into the production phase; then to distribution; and ultimately, they get monitored. Actually, AI came into the last part of that that sequence first. Reinforcement learning came into recommendation algorithms in about 2009. Netflix ran a competition for people to come up with the winning algorithm to determine their recommendations, and, if you ask now, what’s the singular capacity of Netflix, it’s still their recommendation algorithm.”
Postproduction was the next step in the introduction of AI technologies into the chain, according to Connock. “The second phase was around 2017 onwards, where you started to see Machine Learning integrated into lots of production and postproduction,” he said. “And you see it now integrated into pretty much every tool that people use for editing, effects, graphics, audio, you name it - Unity, Unreal, Premiere, Photoshop.”
Read more Artificial intelligence in broadcasting
Generative AI in the form of early image creation applications like Dall-E 1 began to attract the attention of early adopters at the start of the current decade. “But only really in 2022 when people started toying with Dall-E 2, did it finally make its way into the creation phase of media projects,” said Connock. And the rest is history. And later in the year, you had Stable Diffusion and Midjourney on the image generation side. And then the release of ChatGPT and GPT 4 on the text generation side. And now it’s completely native.”
The biggest AI myth
There is significant concern in the media world about the AI eliminating the human touch from the creation process. Connock feels that this fear comes from the fact that AI taking us into unfamiliar territory and is somewhat unfounded.
“I don’t think anyone in the AI world has advertised the idea that they’re going to create machines that will create fully synthetic content from scratch.”
“There was a really interesting quote from Kevin Scott, the CTO of Microsoft. He said the interesting thing about ‘The Queen’s Gambit’ is there’s no training data to say a female chess champion show would constitute a good bet. That was a human insight that gave them the confidence to take that leap. And Kevin Scott’s point is, tomorrow we could create a show about two AIs brilliantly playing chess against each other and no one’s going to watch it. There’s something about human created content and content about humans that makes it much more compelling.”
Connock is optimistic about the creative potential of AI in broadcasting. “I personally think that ‘Television Inc,’ particularly on the factual and entertainment side, has been in a period of not bankruptcy, but a kind of latency. The really big format insights are really 20 years old, and most format iterations are some kind of reinstatement of a bit of an idea from 20 years ago,” he said.
“I think what’s interesting is that the arrival of synthetic humans is probably going to create the most extraordinary opportunities for reinvention. The idea of generative AI and AI agents you can interact with in a kind of reality show way are really going turn upside down the paradigms of how we make TV and think about TV entertainment.”
Connock believes AI innovation requires a new way of thinking about the creative process and that needs people who are not tied to old ways of working. “One of the fascinating things I’ve noticed recently is that a number of quite creative and very visionary TV production companies have started hiring young people to be kind of Red Team units who sit outside, rather like the SAS, just doing AI ideas. ‘They’ve worked out that they’re much less likely to be stuck in the tropes of the past and much more likely to work from scratch,” he explained.
“They’re creating little guerrilla warfare teams consisting of a creative or two and some sort of programmer or coder who can actually fire up these tools and make it work. It shows you that the creative companies are starting to realise that there is something really interesting here on the creative development level, but they might need to go to a completely new generation to deliver it.”
Countering bias
There have been numerous examples AI systems showing bias, with searches for “criminal” resulting in a high proportion of images of people of colour, or a query for “doctor” producing only images of men.
Connock believes this is a significant issue and needs to be addressed with great care. “The British media and the American media, with other media around the world have been engaged in a very important and laudable quest to detoxify themselves with respect to gender and ethnic bias,” he commented.
“And AI is not healthy because the datasets that are driving these AI tools are incredibly biased. So, we have to be very, very careful that we are aware of the fact that training data is biased and that’s going to just prolong the historic biases about society. It’s very important that if we are going to embrace AI systems, we have to intervene in the data to make sure that the outputs that we get are not replicating the sins of the past. We need to make sure that all the good work that’s been done to make changes in what TV producers make isn’t done using systems that are trawling databases that are quite old and don’t necessarily reflect where society is now.”
Potential job loss and the Luddite
Almost every day we are confronted with alarmist headlines predicting that AI will put most of the population out of a job. Connock pointed to similar inflection points in history. “Strong parallels go right back to printing in 1460 and photography in the 1820s. I think the generative era is that big a zeitgeist, and we’re seeing the same kind of Luddite concerns about it,” he said.
“People are forgetting that what’s probably more likely to happen is not fewer jobs but different jobs. And in the same way that there wasn’t mass unemployment resulting from all the previous technological changes that we’ve had, quite contrary, the media is continuing to expand. The invention of the internet, which challenged pretty much the entire hierarchy of media production, has not ended up in fewer people working overall in media.
“In fact [the internet] has exploded the number of people working in media because it turns every local council and every Plumbing Supplies company into a media producer in an industry which had previously been dominated by five big American corporations and some press barons. I think this could be the same thing. This could actually create the Colombian jungle of new media species. That’s what we’re really seeing.
“We’re already seeing an explosion of creativity.”
Dr. Alex Connock is a Senior Fellow at Saïd Business School, University of Oxford and author of the recent textbook Media Management and Artificial Intelligence’ (Routledge). Connock’s broadcasting background is practical as well as academic. He co-founded the indie Ten Alps with Bob Geldof in the late 90s and was the Managing Director of Endemol Shine North through 2017.
No comments yet