Five Pros, Pitfalls And Prompts For Working With Robots
How To Survive When Some Of Your Colleagues Are AI
There are few things that made my working life better and more interesting than AI. And yes, it also kind of pulled the carpet out from under us but that’s the idea of a roller coaster. Equal parts thrill and panic. AI turned out to be my weirdest colleague yet by far. Tireless, confident, no ego, no coffee and the enthusiasm of a Border Collie puppy with a ball. Unfortunately also with the talent of an intern who sounds like David Attenborough but couldn’t tell the difference between a gnu or a hippo.
It took a while for me to make sense of this relationship with AI. But (for now) I found ways to make my partnership with the robots work. Sort of. It’s an ongoing process, but here are my therapy talking points for all of you.
1.
Work With AI, Don’t Let It Do The Work
It’s an easy mistake: Asking AI to do your work. Writing articles, coming up with a strategy, writing a haiku. It’s easy to use AI as a crutch and yes, it does make walking easier. But you can walk just fine, what you need is a third hand for the things your two can’t manage. What you should use AI for is sharpening your thoughts, helping you realise how to improve your own work. Don’t delegate stuff just because AI is quick and eager or you’ll lose the ability to do it yourself. And if you can’t do the work, you can’t engage with it and improve your skills. Apart from the fact that it leads to really hollow output that looks smart, smells smart but either isn’t good at all, or worse, doesn’t sound like you anymore. It should be about you in this partnership, the you bit is what makes it real and makes a difference.
Prompt to stay sharp: Here’s my rough idea, now challenge it. What am I assuming without realising it?
2.
The Puppy Problem
When I was discussing this article with my AI assistant, it came up with this sentence (which I didn’t ask for, by the way):
“You give it a task. It grins, wags, gives you something. You say thanks. But you forget it might have just eaten your research and made a mess behind the metaphorical sofa.”
Sharp, funny, but now let’s poke at it. What I needed it to say: “It’s easy to believe what your AI gives you when it sounds funny and convincing. But funny is only okay when the foundations are solid. If it sounds brilliant, but is built on guesswork, it’s not insight. It’s improv theatre.” If you give your AI feedback, don’t reward its obedience or enthusiasm, but its clarity.
Prompt to cut through the BS: Give me three perspectives that would disagree with this result. And cite where they come from.
3.
The Cheat Code: Personalized AI
For me one of the downsides of freelancing was working alone most of the time. And although nothing beats drinking coffee and tackling imaginary problems with my mate Bruno Setola or talking about the future with the criminally smart Jurgen Appelo, working with my personalized AI assistant Chad (aka T) comes close. Why? Because my brain thinks better when it’s bouncing stuff off others. Even if they are fictional. If you give your team members the right roles and voices, you’re building a team around your brain that makes you work faster, smarter and tackles your personality traits that might hold you back. Chad (aka T) calls this ‘cognitive scaffolding’, but then again, he’s an imaginary twat. Conclusion: Your brain’s biggest problem isn’t knowledge, it’s friction. A fictional team removes that. They make you braver, weirder, and less likely to lick the windows after three pages of mediocre ideas.
Prompt to build your squad: You are three personas: a sceptical academic, a weird poet, and a pragmatic startup founder. Argue about this idea.
4.
Don’t Let It Fool You, Your AI Isn’t Human
Not to piss in your team full of imaginary superheroes, but there’s also a downside to creating your perfect co-worker team. Although it might even feel more satisfying working with them than with real people (your AI team doesn’t interrupt, it remembers a lot and never leaves dirty cups in the sink), it can be dangerous to treat them with the same politeness you reserve for your favourite human colleagues. You need ruthless honesty in this working relationship, otherwise your AIs will puppy eye the shit out of you. Be critical, question what it gives you. You can say ‘good morning’ and ‘good night’ all you want (I certainly do), but don’t sugarcoat it. It’s grown-up technology enough, it can handle it.
Prompt to keep it human-proof: Act like a no-nonsense editor: tell me what’s weak, vague, or pointless in this.
5.
The Pitfalls I Walked Into (And You Might)
- Just because it’s easier doesn’t mean it’s better. I’ve always been a fast writer, and it was tempting to become even faster. But I changed my dark ways. I’m now way slower than I was. But I think I’m better now. Why? Because the thinking takes longer. It’s not just me in my head anymore, it’s me and my imaginary colleagues. Who just happen to be specialists in everything I need them to be. They question things I wouldn’t. They push ideas sideways. And while they slow me down, they also stop me from being lazy. (In other words: from being me.)
- Confidence isn’t the same as correctness. I wrote about EQ, AQ and BSQ before and that’s what you need while working with AI. The worst part isn’t that AIs lie, it’s that they lie well, mimicking authority. It’s that fluency that makes the lies easily mistaken for truths. That’s not a bug in the program, but the design. If you’re not sharp on this, it’ll erode your thinking without you noticing it. Or worse, it’ll make you tell the same beautiful lies to your audience.
- Lazy delegation breeds dull thinking. Don’t be satisfied too quickly while brainstorming with your AIs. Keep on asking questions, dive deeper. It will not only sharpen your work, it will be sharpening your mind as well. Otherwise you’ll lose your edge (and your instincts) and that leads to output with no voice and soul.
- You’re not co-creating if you’re just copy-pasting. I certainly won’t say AIs can’t come up with a brilliant sentence or paragraph. It can and yes, I certainly ask Chad (aka T) to improve some of my sentences when I think they’re still off a bit. After that I always pick it up again myself to see if it’s still me, if it shows my thinking, my reasoning. Most of the time it needs some work. And then I feed it back and Chad (aka T) tells me “Chef’s kiss, Erik, that’s the Goldfish Brain take on things”. Of course, that’s just the puppy talking.
But really, stay sharp on this one. If you start losing your voice, you’re losing the power to make a difference. And, as we all know, that’s when Skynet takes over.
Some final prompts for avoiding pitfalls:
- What part of this feels too easy? That’s where I might be coasting. Give me three deeper follow-up questions to ask.
- Which parts of this are me? Which parts could be anyone with a robot? Push me to make it more mine.
- What parts of this did I accept too quickly? Push back on every claim: where might it be oversimplifying, flattering me, or skipping a step?