Fish Food: Episode 637 - Being more intentional in the age of AI
Intentional and passive media diets, imagination in the era of AI, LLMs pass the three-party Turing Test, AI model collapse, agentic media buying and kairos
This week’s provocation:
Several years back the brilliant Faris Yakob wrote a triumvirate of posts making the case for a more intentional media diet (post one, two and three). He captured this in the pyramid visual above. The premise was that any content where ideology leads to falsehood (at the top of the pyramid) is simply not good for us and we should avoid it at all costs. Going down the pyramid, we should limit heavy social media use, ‘doomscrolling’ and rolling news since, in the age of the algorithm, this too can reduce our mood and make us angry or anxious. The bottom three layers, where we are making more intentional choices about which media we consume and what inputs we are seeking form a much healthier media diet and this is where we should prioritise the bulk of our time.
When we consume so much media (over 12 hours a day) it’s a good idea, wrote Faris, to be mindful about getting a balanced media diet in the same way that we are told to be mindful about balancing our food diet. We are what we eat, and we are the media that we consume. Attention is finite, time is finite, and yet content is everywhere, and we naturally hunger for ideas and information.
At the heart of what Faris is saying, is that active media is more satisfying than passive media, and intentional consumption is more fulfilling (and better for us) than algorithmically curated content streams which serve us a never ending feed of things that are designed to get our attention. As an example, TikTok users spend an average of almost an hour a day on the platform and open up the app an average of 14 times daily, which works out at an estimated range of between 140 to 224 videos viewed daily. The way that we consume intentional media is different from the way we engage with algorithmically curated feeds that are specifically designed to draw us in. This is not to say that we should all abandon TikTok and other feeds immediately, but simply that we should think more carefully about the balance of inputs we are choosing. Spending three hours reading a book that we’ve selected or programmes that we’re interested leaves us feeling better than spending three hours aimlessly scrolling social media.
Faris wrote a great book on attention, and I was reminded of his exceptional thinking about this topic through the wonderful conversation between him and Adam Morgan on a recent episode of Adam’s ‘Let Make This More Interesting’ podcast (Apple, Acast) - strong recommend.
There has been a number of broader things swirling around my head lately related to how technology is changing the way that we think, the hidden impact of changes in media consumption, and whether we are increasingly defined by what we pay attention to. There was the quite staggering data points in that FT piece that I shared last week about the growing struggle for people to concentrate, and the apparent broad decline we seem to be seeing in verbal and numerical reasoning capabilities (short video version here). There was the Kirby Ferguson video for the NY Times talking about how tech hegemony has produced cultural hegemony and the issues created by algorithms that make it harder to discover niche culture. And then there’s the steady rise in loneliness (particularly amongst the young), and the marked decline in people hanging out together (particularly amongst the young).
Image source
Many of these trends have accelerated since around 2010 leading many to believe that our phones and the way that we use them is having a significant impact on how we relate to each other, how we live, and how we feel. The reality is likely to be far more complex, and there’s no direct causal link that’s yet been proved. But there is something going on, and the importance of taking a more intentional approach in how we design systems of use and the application of technology is surely beyond dispute.
Faris’ pyramid was not about cutting out algorithmic feeds and passive media altogether (apart from the obviously toxic media of course). It was about establishing a balanced diet that involves more deliberate choices about how we give out our attention. As AI capability expands rapidly the impact of technology use on wellbeing becomes even more apposite and these intentional choices become more and more important. Research from OpenAI and MIT Media Lab released just a couple of weeks ago showed that both the AI model, and the way that users choose to use the model in the context of their personal circumstances can influence social and emotional outcomes.
As Piotr Bombol said recently, we are at something of a cognitive crossroads. He refers to the ‘jagged intelligence’ idea - AI’s uneven capabilities where it excels at some things but fails at others. Moravec’s paradox, a well known AI paradox that has been around since the 1980s, describes a counterintuitive aspect of AI development: tasks that humans find cognitively challenging (like playing chess or solving mathematical problems) are relatively easy for AI to replicate, while tasks that humans do effortlessly (navigating through space, walking, recognizing faces) are extremely difficult for AI to master. The acceleration of AI should be a call-to-action to dial up on the uniquely human attributes that AI will struggle to master for some time to come - empathy, human connection, humour, imagination.
We need to be more mindful of the impact of technology both on how we think, and how we feel. Whether it’s media or ideas, a more intentional approach can help. We need to design AI systems in ways that help us to be more, not less, human.
If you do one thing this week…
Not unrelated to the above, this week Margarat Heffernan wrote about how Andy Haldane, erstwhile Chief Economist at the Bank of England would host talks at the bank from a broad range of provocateurs including the artistic director of the English National Ballet, singer songwriters, astronomers and artists. Lots of people at the bank would ask what the speakers were going to talk about and Andy would happily say 'I don't know'. His intention was to spark new and different conversations... 'because in the midst of uncertainty, what’s needed are options: possibilities, backups, a proliferation of choices'. Says Margaret:
'Artists and scientists have more in common than is often expected. They...both move towards uncertainty, knowing that in it lies opportunity for discovery, invention and change.'
She echoes something that I keep coming back to - the ever greater need for imagination in leadership and business. The ROI on imagination is not always obvious. It's not the kind of thing that interviewers prioritise, or that leaders immediately value. Yet it has arguably never had more worth. For new ideas, says Margaret, you need new conversations. Imagination may not eliminate uncertainty but it can certainly illuminate possibilities.
The potential counterpoint to this is that AI is increasing the risk that we optimise ourselves into mediocrity. This week I serendipitously watched Rory Sutherland’s talk at Nudgestock 2024, titled ‘Are we now too impatient to be intelligent?’. Rory gives some good examples of how and why we prioritise the answers that we can get quickly, and which appear to be right, over knowledge that really matters:
‘If you can reduce decision making to an algorithm, formula, process or a procedure you avoid the risk of blame…instinctively people love to codify things and make them numerical and make them what you might call optimization problems with a single right answer…because the second you acknowledge any ambiguity…you now have to exercise choice. Whereas if you can pretend there's no ambiguity…you haven't made a decision, you can't be blamed, you can't be held responsible.’
Not everything is an optimisation problem. Which is why we need to design systems and habits that help us to think differently, forge new paths and to see things from different perspectives.
Photo by Marija Zaric on Unsplash
Links of the week
LLMs have just passed the standard three-party Turing test (where a machine and a human try to convince a third party that they are human) for the first time ever - summary here.
Google have just launched Gemini 2.5 Pro, and it’s very good - it has a comparatively huge context window, or data input capability, but also an excellent ability to pinpoint small details within that, and is reportedly very good at analysing YouTube videos. Meanwhile OpenAI have just launched a new image generation model, and it’s also good - although the hype around all the Studio Ghibli inspired images flying around completely ignores the sensitivity of ripping off Studio Ghibli content to train it to do this
Speaking of copyright issues, Library Genesis is a so-called ‘shadow library’ project which hosts a huge amount of copyrighted written material without the permission of the copyright holder. There’s millions of books, academic papers and other written material on there which have been illegally scraped. The Atlantic have created a search tool where authors and academics can see if any of their material has been pirated illegally, and yes, my three books are all on there. Meanwhile Meta actively decided that it was a good idea to use LibGen to train their latest AI model Llama 3. Nice one Meta (not).
This week I learned about AI model collapse, via Springboards.ai founder Pip Bingemann’s guest curation of the Genius Steals newsletter: ‘If future models train on AI-generated content instead of real human data, they start to “forget” reality. This phenomenon, called model collapse, happens when models lose the nuances of real-world data over generations, leading to dull, repetitive, and inaccurate outputs’. It’s like making copies of copies, and everything getting less accurate and more blurry - the solution to maintaining good fidelity is to keep people in the creative process and use AI to generate the new and novel. Amen to that.
This week I also had a fascinating conversation with Flynn Devine who is building Thoughtly, an AI tool designed to help writers throughout the writing process. Flynn got in touch to interview me about all the different ways in which I use AI in the writing process - something I’d not really thought through before (and maybe something that’s worth writing up at some point). I often use AI for research, extending my ideas, sourcing examples, understanding counterpoints, but I never use it to create finished drafts. This is mainly because writing is how I think (and so getting the AI to write for me would simply remove the value of writing in the first place) but also because whilst AI may understand the style in which I write, it still doesn’t understand the way I think (what I find interesting, the angle or the connections). This is something which I learned when I created a custom GPT trained on my three books and 16 years of blog posts.
This is how Jellyfish deploy AI agents to automate media buying and reduce campaign launch timescales by 65%
Dr Philippa Hardman is the best person to follow right now on the intersection between AI, learning and development. Her recent post on how agentic AI is going to upend online learning for the better is excellent.
In last week’s post on changing the game I included the example of the Mercedes-AMG Petronas Formula One team who (in 2014) flipped the script using a combination of strategic foresight and reinventing how F1 teams work, leading to a seven year run of dominance. A number of people said that they liked the example so I wrote a post on it
Clever little ad this, by Perplexity
And finally…
I rather liked the meaning behind the Ancient Greek word ‘Kairos’. HT Jonny Thomson
Weeknotes
This week I was mostly in Dublin working with my Diageo client, and a bunch of their European leaders, on digital and data transformation stuff. Fun fact - they sell 180 million pints of the Guinness in the US every year. Next week I’ll be travelling out to Dubai where I’m going to be for a week or so, working with PwC and a group of leaders from the media sector.
Thanks for subscribing to and reading Only Dead Fish. It means a lot. This newsletter is 100% free to read so if you liked this episode please do like, share and pass it on.
If you’d like more from me my blog is over here and my personal site is here, and do get in touch if you’d like me to give a talk to your team or talk about working together.
My favourite quote captures what I try to do every day, and it’s from renowned Creative Director Paul Arden: ‘Do not covet your ideas. Give away all you know, and more will come back to you’.
And remember - only dead fish go with the flow.