Monki Gras 2024: Prompting Craft
Monki Gras is the RedMonk conference that celebrates software, craft and tech culture. This year’s theme was theoretically about prompt engineering, but as always, the talks went so much more meta than that.
It’s hard to believe five years have passed since the last Monki Gras. Of course the pandemic stole a few years from all of us, but wow it went by fast. This year’s theme couldn’t not be about AI, because everyone is currently obsessed with the impact of Large Language Models. That said, most people I spoke to actually came despite the topic: in a “yeah, I wouldn’t usually go to a conference about prompt engineering, but it’s Monki Gras, so you know it’ll be good” kind of way. Naturally all the talks were incredible, but I’m going to pick out just three that really made me think.
Cat Hicks on AI skill threat
Everyone talks about what developers are doing, but few people really delve into how developers are feeling, as a class. Working in tech (and being married to a developer), I’m acutely aware of the existential anxieties AI is fuelling. Long before people are actually replaced by AI, the fear of being replaced is damaging the profession, way ahead of the change.
As a researcher, Dr Cat Hicks brought science to the problem: what if we measured developers’ attitudes to AI and its perceived threats, and observed if there were any patterns in the kinds of workplaces where they reported feeling more confident or resilient? What might we learn about the best environments for developer thriving?
The “brilliance trap” has been observed in academic fields, where there’s a culture of hero-worshipping innate genius, and under-emphasising the value of hard graft and team work. Fields that over-stress the importance of naturally gifted individuals tend to undervalue collaboration. Compare against the classic XKCD strip about purity: academics love to argue about whose subject is harder, whose work is more abstract and true. Studies have shown that women (and minorities, perhaps?) are under-represented in fields that skew towards valuing brilliance. The brilliance narrative positively correlates with “masculinity-contest culture”, with the two reinforcing each other.
Cat defined contest culture as fixating on individual performance and aggressive competition between colleagues.
Brilliance: A good developer is brilliant, brilliance is a trait
Contest: On this team, we expect you to prove yourself in technical contests
In contrast, a thriving culture is one based around belonging and learning.
Learning: A good developer learns, learning is an activity
Belonging: On this team, we care about whether you feel like you belong
Which one sounds more like most tech companies to you? Or equally, most consultancies? Stack ranking, anyone?
If you promote the idea that “10x developers” are magical unicorns who could code before they could walk, you’re going to put a lot of people off careers in tech. Young people might think: “Oh, I had to try really hard to learn to code, so I guess I’m just not gifted enough, I’ll never succeed in this sector.” If you’re constantly forced to demonstrate your technical prowess in competition against your colleagues, you’ll constantly be questioning your own ability. And if all your self-worth is tied up in your technical competence, the idea that LLMs will eat your lunch is potentially devastating to your ego and perceived career prospects. If you believe you are primarily valued for an innate trait, and that trait is no longer valuable, you’re going to be worried about the future.
(Bonus points to Cat for linking to this adorable Brilliant Blobs comic from the Developer Success Lab.)
On the other hand, if you believe that life long learning is natural and healthy, you can be much more sanguine. If your teams are caring and supportive, lifting each other up, believing in learning new skills as a response to change, then adoption of LLMs is just one more change for us to tackle together. There will be new skills, there will be new opportunities. We got this 💪
We could perhaps extrapolate to a broader conclusion: people who work in toxic environments are more likely to be afraid of change (and for good reason).
Jim Boulton’s Unsung Heroes
Jim calls himself a Digital Archaeologist, which is the best job title ever. I remember his excellent exhibition 64 BITS at Here East in 2017: a collection of historic websites and early computer art. Technology becomes obsolete very quickly, with digital decay making it difficult to relive the formative internet experiences of our youth. Similarly, the stories we tell about the internet’s history are constantly being written and rewritten, with some figures fading from view faster than others.
Unsung Heroes is Jim’s effort to spotlight some influential characters in computing history whose contributions have not been so widely recognised. The first two issues feature Alan Emtage and Lynn Conway: short, easily readable and very affordable, they aim to introduce a wide audience to a more diverse cast of tech pioneers.
While Jim worked with a human illustrator for the first edition, he experimented with Midjourney to create the second, running up against the challenges of using a nascent technology for a novel use case. While Midjourney can create quite compelling individual images, it doesn’t yet have a foolproof mechanism for character continuity: the same prompt can generate similar style faces, but they might not always look like the same person.
(Photo credit to Dormain Drewitz 🙏)
In one panel she looks like the actress from The Queen’s Gambit, in another panel she looks more like a 1950s housewife from an advert, on another page she might look like Barbarella! It took a lot of patience and prompt tweaking to get acceptably similar faces for this project. (Apparently Midjourney just launched a new character consistency feature last week, so maybe this is already fixed? 🤷♀️)
Another issue Jim encountered was bias in the model. When he wanted to generate a male and female character walking together towards something, the model rendered them holding hands! Perhaps there is so much more romantic source material in the model that it has a hard time imagining platonic co-worker relationships?
Finally, we can’t ignore the ethical problem of attribution and fair compensation. Jim referenced a particular artist’s style within his image generation prompt, aiming to get a specific look and feel for the comic. Is it fair to use an AI model to generate illustrations for a comic, without crediting/compensating the artists whose work fed the model? I expect he would argue it’s fair use for a good cause.
Matt Webb and his AI clock
Introduced by James Governor as the original “Granddad Shoreditch”, Matt Webb has launched more products than most people have had hot dinners. BERG’s Little Printer was the darling of the IOT scene a decade ago, and now he’s done it again with the Poem/1. The internet hype around this concept clock is unreal: it has been insanely viral on social media, featured in the New York Times and mainstream news outlets across the world. Now there’s a fully funded Kickstarter for a real, industrially designed product.
Why do people love this clock? I think it’s popular because it gives us a whimsical avatar of AI to interact with. When a computer writes poetry (however good or bad the quality of the poetry), you can’t help but think: “Aww, it thinks it’s people…” Even when it lies about the time to make a rhyme work, it feels cute rather than sinister. A playful product neutralises the poisonous hyperbole of the whole AI debate, and reminds us that what we choose to do with technology is our choice.
AI == Magic obscures invention. Use play to demystify and find the new opportunities.
On the “serious slide”, Matt cut to the chase: when you invite AI into your home in a physical vehicle, what *vibes* are you welcoming in with it? What power are you giving an unknown entity to influence your mood? As a thought experiment, it’s not quite as extreme as “What if AI takes over the world and turns everything into paperclips?”, but “What if AI clocks lied about time and we didn’t even care?” is still pretty dystopian, in my view.
I have resisted the temptation to use ChatGPT to write this summary for me, because I personally find it useful to do my own reflection on what I have learnt from my experiences. Laboriously writing my thoughts down is how I process the world with my squishy, imperfect brain 😅
Thanks to James and the team for another fantastic Monki Gras 💕
If you’d like to read another Monki Gras round up, French has posted his reflections over on his website.
