From miraculous to taken-for-granted: How AI is already changing every aspect of the media and marketing world
Welcome to a midweek update from Unmade.
Today, Unmade’s Cat McGinn distills what she learned about the impact of AI on the media and marketing world during her months researching last week’s HumAIn conference.
Plus, we have news of an end to the Unmade Index’s five-day winning streak.
Only Unmade’s paying members get access to our archive, where all our content goes after two months.
Plus they have the satisfaction of knowing they are helping help to fund experiments in publishing, like the one set out below. Become a member today.
The arrival of AI is no time to be an ostrich
Last week Unmade ran our first humAIn | human creativity x AI conference. In the process of curating the event, Unmade’s Cat McGinn spent months considering the implications of the impacts that generative AI, and other emerging forms of artificial intelligence will have upon the marketing and media industry.
There is a not-insignificant subset of people in the creative industries who are choosing to take the ostrich approach, or perhaps just hoping to wait and see whether the early adopters and fast followers fail.
I do not believe we have the luxury of waiting.
I’m no AI expert - and I would urge you to check the credentials of anyone claiming to be - but I’ve been thinking pretty obsessively about AI now for months. It isn’t an exaggeration to say it haunts my dreams.
The context that we find ourselves in is one in which we’re hearing about global holding companies banning their staff from using, or even talking about generative AI.
Influential figures like Martin Sorrell are issuing pronouncements that AI will replace a quarter of a million jobs in media in the near future - but offering little in the way of solution.
Those people and companies willing to investigate and experiment, to engage in open dialogue about the benefits - and the very real problems - are amongst our industry’s more courageous and future-focused. We’re in uncharted territory, and we have the opportunity to shape the way AI will impact media and marketing.
I've been able to have wide-ranging and unusual conversations, about the nature of creativity; the future - the likelihood we’re living in a simulation, and even one’s personal P(doom); this is the estimate, expressed as a percentage, of your view of the probability that artificial intelligence will lead to the downfall of humanity).
I want to share a little of what I’ve learned. I cannot emphasise enough the unprecedented scale, breadth, and extent of AI's impact, and the consequences of AI on creativity, humanity...
I don't think people realise quite how surprising many machine learning and artificial intelligence experts found the capabilities of the recent explosion in Large Language Models.
In 2017 McKinsey’s research predicted that by 2037 AI would be capable of “median human-level performance,” or in other words, could perform at the level of an ‘average’ human for some tasks - particularly in terms of creativity, language, logic and reasoning.
That prediction was blown out of the water this year. And based on that trajectory, we might reasonably predict that AI will write as well or better than the “best” writers, generate images and carry out analysis to the highest standard… not in 2055: next year.
Is this the way the future’s meant to feel?
For a generation brought up on outlandish claims about what the future would hold, and the ensuing disappointment at our lack of jetpacks, it seems that at long last Gen X and the elder millennials have found renewed social purpose.
We’re the generation with a perspective that spans digital and analogue worlds, and we recall a time before the internet. We remember the days when if you wanted to meet someone, you just had to turn up at the time you said you’d be there. If you wanted to know something you had to find a book and read it.
Barbaric times.
But that knowledge of not only how much things have changed, but critically the way they changed us, is going to come in really handy as we navigate this shift.
This unique experience will help us as we adopt AI and adapt to the changes
Learning from history isn’t our strong suit as a species, but I think we can agree that the internet and social media have had implications we didn’t consider 20 years ago, and some of those consequences have been really harmful.
We have an opportunity to be a little more considered.
I cannot emphasise this enough - and honestly, I would be happy to be wrong - but the scale of the change and the speed of this transformation is something we’ve never seen before in human history, and it is going to change everything.
Not all of these changes will be positive or beneficial for individuals, for companies or for the planet.
Experience
Talking about the fundamental tenets of Buddhism in this context might seem like a stretch, but hear me out. Buddha used a word when describing the practice of meditation that means to “come and see for yourselves,” a Pali word: Ehipassiko.
AI shares something of this quality: it is experiential in nature. In the words of one of the advisory panel: "Stop reading, go do something with it, then we talk," - PJ Pereira, Creative Chairman, Pereira O'Dell.
Only by playing and experimenting with the tools can you start to understand the limitations and capabilities; get a sense of how to get the outcome you want.
The way in which you are in dialogue, of some sort, with a machine, an entity.
I’m not saying it’s conscious, but it’s qualitatively different from using a spreadsheet, Photoshop and so on.
The creative possibilities are vast. I’m seeing a really playful characteristic in the way people are using these tools. An example I love is this side project by Ant Keogh, Chief Creative Officer at the Monkeys.
The huge benefit is the democratising effect. Small brands, individuals will now have access to the kinds of tools, to create the kinds of outcomes that were once the province of global brands, large multinational companies and Hollywood.
Breaking free from constraint
One of the reasons gen-AI is so exciting is the ability to deliver creative ideas without the limitations of budget, time, or resources.
This is an AI-generated video created as a promo for PJ Peirera’s new book, the Girl from Wudan. Publishers would never have had the budget allocated to create promotional assets like this.
And this makes me really optimistic about the potential for art, content, writing, creativity. The potential for conceptualising ideas - ideas that wouldn’t have made it out of the brainstorm - are now within the realm of the possible.
Speed the Plow
NotContent is a new - they claim the world’s first - generative AI creative agency founded by one of our advisory panel, Jeremy Somers. NotContent uses generative AI to concept, ideate and create work, turning around campaigns from brief to execution in a few days, and in some cases only a matter of hours.
Everything in that clip, other than the brand assets, was created by AI.
On one hand, the possible benefit is that there will be greater freedom for creatives, more time and budget to invest in craft and storytelling.
On the other, the fact we can compress the process means jobs will change, some will disappear, and remuneration models will have to shift.
Charging hourly rates when you’re working with tools which create outputs in seconds ceases to have any meaning.
The value comes in the experience and skill to know what “good’ looks like.
The new skills include the ability to effectively prompt. Coding in English. It means articulating your vision, your ask, and then to curate and distil, refine and polish. It’s not the work of grinding out scamps and concepts, scouring spreadsheets, tweaking decks, nudging text and so on.
But it begs the question: What happens to the 10,000-hour rule when the machine does all our heavy lifting?
Beigification
Relying on AI comes with the risk of homogenising creativity - if every output is generated and finessed by AI as the norm, will it all start to look similar?
The limitations of synthetic creativity speak to the increased need for human ingenuity and creativity to bring new and original ideas.
Any large language model is trained on a large amount of data from the internet. I think it’s fair to say that most of the stuff on the Internet is not exquisitely creative or compelling. As Marshall McLuhan put it, “the content of each new medium is the old medium.”
Keyword-stuffed landing pages, banal social posts, toxic forums - not the pinnacle of human capability….Most of the content on the web is mediocre at best, and this is the issue when we are training large language models.
To my mind this makes the new AI initiative launched by clown-shoe wearing billionaire Elon Musk, who every single day resolves not to use his wealth to make the world a better place, even more ludicrous. Musk’s objective is to create an AI which will “understand the true nature of the universe,” but has spoken of plans to use Twitter’s data to train his LLM. I think we can agree that those seeking the true nature of the universe are unlikely to find it on Twitter.
When the training data is flawed, what do we expect the outcome to be?
Search is going to change beyond recognition. In the near future, when user queries are directed at an AI-powered platform, or my personal AI is chatting with a brand AI, the content will need to be optimised for AI x AI interaction. And without the commercial imperative to create content for human users, what is the role of digital content?
The ability to generate a large volume of content at speed is already flooding the internet with low-quality, fast produced AI content.
"Just as we’ve strewn the oceans with plastic trash, and filled the atmosphere with carbon dioxide, so we’re about to fill the Internet with blah,” - Professor Ross Anderson, Cambridge University/ University of Edinburgh
AI-ception
The additional complication is that as more and more of the content on the internet is AI-generated, future LLMs will be trained on synthetic data. Bear with me, because this is Inception-worthy. There are theories that AI trained on AI data will begin to degrade; the models will collapse. Greater errors, more hallucinations…
Another crucial consideration is the issue of bias. We’ve trained the LLMs on biased data.
Bias in, bias out.
This example really highlights it; the gen-AI prompt, repeated over and over was simply “mother and baby”.
Hard to think of a more universal concept than mother and child, and yet….What do you notice about the images?
Ethics
Another issue here is the erosion of already weakened consumer trust. In a post-AI world, trust, personalisation and human creativity will become the key point of difference for brands and media.
The provenance of assets created by genAI needs further, deeper exploration.
AI has been described as a plagiarism engine, and issues of copyright and fair use is going to become paramount. The lawsuits are only just beginning.
The fact is that the expertise and work of countless artists, illustrators, writers, comedians and so on have been used to train the very artificial intelligence tools that are beginning to cost those same creative professionals their livelihoods.
It’s inconvenient, but it’s inescapable.
Trust is in short supply. Media literacy is low, and we are facing a world of undetectable deepfakes.
What happens to the social contract when we can’t believe our eyes or ears?
AI and Knowledge Bias
We already have a tendency to take AI for granted.
Knowledge bias means that AI's transformative effect and impact quickly become mundane. The first time we experienced AI in the context of our lives - ChatGPT, Midjourney, DALL-E, Code Interpreter - it felt miraculous, unprecedented. As Arthur C Clarke wrote, “Any sufficiently advanced technology is indistinguishable from magic”. And yet we rapidly begin to feel entitled to things that were unimaginable mere moments earlier.
I want to share a little about the AI chatbot we created with Fear of Missing Art.
I’ve learned more from working on TimBot than from hundreds of hours of research.
We trained Timbot on my colleague Tim Burrowes’s book, and then on the content from the Unmade archive.
We use GPT-3 to pull the data in response to a user question, then GPT-4 to finesse the language of the output, so while it uses ChatGPT, the training data is all Burrowes.
Remarkably, users so far have spent around 20 hours chatting with TimBot. TimBot isn’t just firing out the first response, but rather it uses a series of reasoning and logic steps.
The new feature we’ve been working on is replicating the sound of Tim’s voice. It’s taken more effort, strangely, than getting the written language close to being right.
We think it’s pretty close, and you’re hearing it for the first time now:
As we announced back in June, the text version of TimBot is already live. The spoken version will go live soon.
There is something - we think - very exciting about this process. Imagine every non-fiction author giving their audience access to their ideas through chat and voice. The accessibility this offers could have a huge impact on a wide audience.
Of course, the security issues are enormous, and we’ve inadvertently destroyed Tim’s ability ever to use phone banking again.
Building Timbot was a collaboration, between Nic at FOMA, Unmade - and AI. Isolating hours of Tim’s voice from the Unmade Podcast was the work of Abe’s Audio.
I think it gives a tiny sense of what happens if we think differently about intelligence: Intelligence as a collective idea.
Individually humans are very smart. Crowds are famously… less so.
But if we connect and network our knowledge - which is what AI does - the potential for creativity, problem-solving, and ingenuity - is almost limitless. Think of this as a sort of planetary intelligence, an ecosystem: like forests, or like mycelium.
As we navigate these turbulent times, I want to stress my belief that we should not leave the inventors and a handful of tech giants to decide how their work will be used - that’s up to us, the users.
AI is a collective machine, and it requires a collective response.
So, knowing what we know, and seeing the glimmers of possibility, in either direction, I believe we find ourselves at a fork in the road moment.
Let’s come back to the human.
A crude metaphor is that AI is mind; not heart, soul… it doesn’t have empathy because empathy is grounded in being able to share the feelings of another. Until AI inhabits a meat suit and experiences joy or pain (and what a chilling phrase that is) it cannot have true empathy.
In a world dominated by AI tools in every facet of life, which is our near future, it is these qualities of empathy, of being able to imagine and dream up new ideas (rather than the juxtaposing of what already exists in new ways), of a desire to be creative, a need to create that will be most in demand.
Something our industry is perhaps often guilty of is an excess of confidence and certainty. We’re facing a scenario in which even the engineers and experts building AI don’t fully understand what they’re creating.
I’m not using hyperbole: AI engineers can’t fully explain the processes, the outputs or the interpretability of the way that an AI operates.
This presents an opportunity to lean into curiosity and uncertainty, to acknowledge that we don’t have all of the answers and indeed, maybe we have not been asking the right questions.
I hope we can ask some of them together.
This piece was based on Cat McGinn’s opening presentation at last week’s HumAIn conference
Unmade Index takes a downward turn
Seja Al Zaidi writes:
After five days of growth, the Unmade Index went backwards yesterday.
The Index, which measures the performance of ASX-listed media and marketing stocks, dropped 0.77% to land at 668.5 points.
The dip was largely thanks to falls in share price seen by the two largest stocks on the Index - Nine, which fell 1.43% and Domain, falling 0.52%.
ARN Media had a fall of 2.31% while printing and marketing company IVE Group declined 1.3%.
Enero Group saw a continued positive trend with a 3.69% lift in its share price, while Seven West Media rose 2.63%. Southern Cross Austereo lifted 2.08%.
Time to leave you to your Wednesday.
We’ll be back tomorrow with an audio-led edition of Unmade, in which I talk to Chris Janz, co-founder of Scire, about his plans for the most talked about media startup of the year.
Toodlepip…
Tim Burrowes
Publisher - Unmade
tim@unmade.media
A frighteningly accurate conjuring of AI in our future. Whilst I generally am a glass-half-full sort of guy (or even a, "Hey wow, I have a glass!, sort of guy), I'm apprehensive over the possibilities of how AI and humans will evolve over the next few years.
Because if we already can't trust humans to do the right thing...
Great speech Cat!