Lauren Celenza

As AI reshapes work and creativity, the question that haunts us isn’t what AI can do, but what it’s leaving us to figure out about ourselves.

In this talk, Lauren makes the case that the designers, engineers, and leaders who will stand apart in the age of AI are not the ones who prompt the best, but those who bring taste, intention, community, and genuine humanity to everything they touch.

Through real-world case studies, sharp provocations, and an honest reckoning with how AI is reshaping our industry, economy, and environment, this talk offers a practical framework for reclaiming your creative soul and your strategic value.

Transcription

(Audience Applauds)

Thank you, Marc.

Hey, everyone. Welcome to the final talk at beyond tellerrand. Hello to everyone who’s watching on the live stream as well. Hi.

I’m very excited to be here with you all. So thank you so much for being here, for sticking it out to the final talk.

Now I have a confession to make with you all. And that is over the last few years, I’ve been feeling quite tired hearing about AI.

Does anyone else feel the same?

Yeah, a lot of hands are raised in this room. Well, unfortunately, this is a talk about AI.

But I wanna crack open the conversation a little bit and talk about what it means to live in this time of an AI takeover without losing your soul.

I wanna go beyond the hype, beyond the doom, beyond the sparkles,

beyond the gradients, although I love your gradients, James.

And examine how we in this room, no matter our job title or discipline or experience level, can elevate our craft in this time, sharpen our value.

And reclaim our creative soul.

Now I’m Lauren Celenza. This is actually my first time in Germany, and my first time at beyond tellerrand. And I’m based in the US, in Seattle, and I often travel to San Francisco to Silicon Valley.

And over the last decade or so, I’ve been working as a software designer and writer at Google, Adobe, Code for America, the World Resources Institute, lots of different big tech, non-profit, startup spaces. And throughout every sort of workplace or client that I’ve worked with, I’ve always really asked this question of, is it possible to build technology while still preserving human agency?

And so I started a newsletter five years ago called "Tech Without Losing Your Soul" on Substack. That kinda seeks out to figure out this very question.

And I’ve done a lot of personal narratives about my own experience working in the crazy tech industry. I’ve conducted a lot of interviews with experts across the world, and throughout every story I’ve written or interview I’ve conducted, I noticed in my research that there’s one common thread, which is that every technological breakthrough throughout human history has arrived with capabilities and consequences.

This is a picture of the birth of the internet.

In 1858, the first transatlantic cable was laid down on the ocean floor, creating near instantaneous communication through the telegraph. Then in 1988, serendipitously the year that I was born,

we created high voltage, high speed cables for longer distance, effectively establishing the internet as we know it today. I wouldn’t be standing here on this stage if it weren’t for a cable between North America and Europe to connect me to Marc over Zoom and email.

But of course, the cables arrived with a consequence.

The high voltage attracted a feeding frenzium on sharks, and they bit down on the cables, and it electrocuted the sharks, and it severed a lot of the cables. Protective sheathing was eventually added to save the shark, but they still bite down on these cables today as this video shows.

You see, we were so busy on creating high speed, you know, fast communication that we forgot that we were laying down these cables in someone else’s home.

Then came the breakthrough of the smartphone, the iPhone in 2007, which promised convenience. I just take my finger and I scroll.

That simple gesture didn’t just change the way that we interacted with the internet, but it catalyzed an entirely new economy of apps and startups, and suddenly we can connect with anyone, anytime, anywhere.

But soon, tech addiction started to grow, and we started to ask ourselves, why are we scrolling? What are we searching for?

And of course, the more powerful capability like large language models, generative AI and agentic AI,

the wider and deeper the consequences.

Just like the moon, there’s capability and consequence to everything that we create. There’s a light side and a shadow side. Yesterday, Marjan talked about the sun, and I’m gonna talk a little bit about the moon.

But you know, as tech makers, we often fixate and focus on the bright side, the capabilities. Well, the consequences, whether they’re good or bad, we often don’t see them. They’re far from our view.

But I believe that when we create new technology with both capability and consequence in mind, to the best of our ability, we can actually create something quite compelling. This is a real photograph, by the way, taken from the Artemis 2 mission from NASA just this month.

You see, when the moon eclipses the sun, it reveals more of the corona, the dimension, that the glare or the hype wouldn’t allow us to see.

Technology built with consequence in mind is ultimately more nuanced, more trusted, more inclusive. But of course, it’s difficult to know in advance what every consequence will be. Sometimes technology needs to meet the real world, right, to understand what it is. That’s all the more reason to keep paying attention,

to test at smaller scales, and to ask whether bigger and faster is always better.

But that’s quite difficult to do right now because the pressure and incentives to use AI for everything and put AI into everything is relentless.

And the capabilities of AI are genuinely quite compelling, so compelling that so many of us are asking now, where does AI end? And my own thinking begin.

Do my choices

still matter?

Well, to understand the path forward, I’d like to look at the path behind us and unpack a brief, non-exhaustive history of AI that many of us in this room may not actually know about because so much of a conversation of AI today is focused on the future. But really, there’s a lot we can learn from the past. Starting in 1950, 76 years ago, and a man named Alan Turing.

Alan Turing was a British mathematician who cracked the Nazi Enigma code during the Second World War, and after the war, he wrote a paper that posed a deceptively simple question.

Can machines think?

At this point in time, he had already designed a theoretical modern blueprint for the computer, and he had seen these machines come into existence in real time. And in this paper, he was starting to ask a lot of familiar questions, like,

where does computation end? And thinking begin.

Six years later, a group of scientists in 1956, all white men, came to Dartmouth College in the US, and they convened a workshop. They wanted to create a new discipline around Turing’s question, but they needed a name for this discipline. So John McCarthy here at the top left decided to name it Automata Studies to describe the pursuit of machines capable of automatic behavior.

But the name didn’t attract much attention.

So we decided to change it to a more evocative phrase, artificial intelligence.

It sounds inherently good, beneficial, sexy, sophisticated, impressive, and it did the trick. It gathered attention, not just from funders, but from scientists who wanted to be a part of this new discipline.

You see, the term artificial intelligence began as a marketing trick from the very beginning.

And as a result of this name change, the field of AI has gravitated towards measuring its progress against human capabilities. This is a quote from Karen Howe, journalist and author of the book Empire of AI, where much of this history is tracked.

And I think that how we name technology determines how we measure it. And what we measure is what we build.

Onward to 1964, and German professor and computer scientist, Joseph Weizenbaum, I’m so excited to talk about Weizenbaum right here in Germany, at MIT, he created the first chatbot called Eliza.

And people started to become obsessively reliant on Eliza.

In a 2006 interview, he had described how he had walked into a room with his colleague that was chatting with Eliza.

And he said it was as if I disturb an intimate moment. She had witnessed the development of this technology up close. Hardly anyone knew better than she, that it was nothing more than a computer program. The effect was absolutely astounding.

And this was back in 1964.

Later on Weizenbaum grew a bit critical about AI, and he wrote the 1976 popular critique, "Computer Power and Human Reason." And in this book he argues that while computers can provide great assistance and,

you know, efficiency in a person’s life, it can also restrict human to human relationships and decision making. And he advised to never allow computers to make important decisions because they lack human experience.

No other organism, and certainly no computer, can be made to confront genuine human problems in human terms.

But now let’s fast forward to 2012.

And the breakthrough of ImageNet.

ImageNet was a data set of 14 million images where a machine could detect and label what was inside the images. It was really the first of its kind. It was developed by Stanford computer scientist, Dr. Fei-Fei Li. And at the time it had about a 25% error rate in identifying the images. But then came AlexNet conducted by a group of scientists at the University of Toronto, including Ilya Sutskover and Jeffrey Hinton. And they were able to cut the image error rate in half. And they created a deep neural network here that expanded to speech and translation, and in large language models that are behind many of the AI tools that we use today. Which brings us to 2015. And the beginning of a small nonprofit called OpenAI.

In Silicon Valley, over dinner, two men, Sam Altman and Elon Musk,

shared a fear together.

Their fear was that AGI, artificial general intelligence, or the ability for AI to do any task that a human can do across a wide set of domains,

would be catastrophic if it was developed but just a handful of corporations. So they decided to create OpenAI as a nonprofit that would prioritize principles over profit and benefit humanity as a whole.

And a billion euros was pledged by Elon Musk. However, by 2018, Elon Musk believed that OpenAI had fallen fatally behind Google.

So he said, "I’m gonna take over." Majority equity, board control, CEO.

Altman said no.

Musk came back and said, "Okay, well, why don’t we just merge OpenAI into Tesla?" Altman said no again.

And then eventually Musk walked away and his funding walked away with him, leaving OpenAI without the funding that they wanted to build the models that they wanted to build. So they created a for-profit arm with an OpenAI, took an 11 billion dollar investment, 11 billion euro investment from Microsoft, and then Sam Altman became CEO. In 2020, Enthropic was founded by Dario Amade. He used to be a VP of research at OpenAI and he ended up leaving because he was concerned that OpenAI was scaling AI too fast without adequate safety measures.

Which all brings us to this moment, November 2022.

At this point in time, rumors were swirling that Google was gonna create their own large language model. And Enthropic was testing their own chatbot. OpenAI launched chatgbt 3.5 and they called it a low-key research preview. They didn’t think it warranted much testing or research because they were really focused on building GPT-4, the next model that will come out in just a few months. They thought maybe a few thousand people would use it over the weekend. But in just five days, chatgbt surpassed one million users.

In two months, it reached 100 million, becoming the fastest growing consumer application in history.

And for context, it took Google 11 years to reach 365 billion annual searches. Chatgbt achieved that number in just two.

At this point in time, it would seem that GenAI would erupt out of nowhere across the world, across industries with the launch of chatgbt. But really, it was all of these moments throughout history that laid the groundwork for it, as well as the corporate investments that had been skyrocketing in the background. From 12 billion euros in 2012 to 200 billion euros in 2022, to now 2.1 trillion euros is forecasted to be invested on AI just this year alone, a 44% year-over-year increase.

Today, over one billion people engage with AI tools.

And what I find fascinating, actually, about this history

is that AI wasn’t made purely out of a desire to create or innovate, but really, it emerged from power struggles, fear,

and the desire to get attention.

These are deeply human impulses that we all have.

And these impulses are shaping what’s getting built right now,

and what gets built

shapes us.

But how exactly is AI shaping us right now? Well,

many of the AI tools that we use promise to make us more decisive, innovative, faster, efficient, and productive. But of course, for each of these promises, there’s capabilities and there’s consequences.

Large language models can certainly help to make us more decisive. We can help to work with Claude, for example, to get different options to a problem, and it can recommend a path forward to us, and also we’re outsourcing a lot of our decision-making to just a handful of sparkles.

Agentic AI is really innovative. It can now automate a lot of complex and tedious tasks that have never been automated before, and also there’s real fear right now that it could eat software, meaning that it’s gonna reduce reliance on traditional software-as-a-service platforms, which is a 217 billion-year-old market right now.

And due to this fear, a lot of companies are rushing to bolt AI onto their standalone products, but the irony is striking. In the rush to stay ahead, products are starting to look alike.

The same sparkles, the same chatbots.

When everyone is feeling the same existential threat, design starts to conform.

And what makes the conversational UI in your product any different or better from Claude or chat GPT, especially as people become more and more comfortable and familiar with using it, and especially if they can just connect your product into theirs.

But of course, the simplicity of a chat interface

feels like a reprieve from years of doom-scrolling.

And it feels familiar, right? It feels intuitive. It feels something that we would do as humans. We converse with one another. So familiarity breeds accessibility. But at what point does familiarity breed contempt? You know, scrolling used to felt easy and fun once too, until the ease became the trap.

And prompting is not natural for every task, for every person, for every context.

And how long before zero AI features maybe becomes a selling point. I think people are craving some distinction now.

And it’s important to listen to that distinction because that’s where true innovation can live.

But of course,

we can generate content faster than ever.

No matter whether we have coding skills or writing skills or design skills, and this is causing a lot of confusion in the workplace. As product managers are becoming designers, designers are becoming engineers, engineers are becoming sort of managers of the chaos, right? It’s exciting, but it’s also a little confusing right now. And, you know, faster generation can create a lot of slop too. And slop dilutes meaning.

Last year, OpenAI launched their image tool, their GPT-40 image tool, that allowed people to prompt for GBLI style images, mimicking the highly acclaimed Japanese animation studio.

And it received a lot of copyright backlash. And they quickly

banned anyone from prompting GBLI style.

But

is GBLI merely a style?

Or is it a specific distinct body of work created by living human artists?

You know, how we build and use AI depends on whether we can slow down enough to ask these kinds of questions, and whether we still value art by living human artists.

It’s becoming more difficult than ever though to discern what’s human made versus AI made. And there’s more and more studies emerging this year showing that people can no longer reliably tell the difference.

It’s really cool James, to see you use AI in your work, but I also wanna know what part of that process is what you created.

But moving faster can make us more efficient, right? And for every task that agentic AI speeds up for us

is a cost somewhere else.

Germany actually has the largest data center market in Europe with over 500 facilities. And in Frankfurt, data centers are the leading source of electricity consumption, accounting for 40% of the city’s total power demand.

And the local energy supply is being pushed to its limits. Across

the world, the buzzing and droning and crackling of data centers can be heard for kilometers, 24 hours a day, creating relentless noise pollution.

And it affects underdeveloped communities the most where they’re seeing a lot of their energy bills skyrocket.

Researchers estimate that each chat GPT query consumes five times more electricity than a basic web search. And chat GPT users are sending 2.5 billion prompts each day. It’s quite astounding math.

But even as technology maybe helps to make these systems better and less strained on our resources, the reality is that our volume and rate of consumption increases. We just find more use cases for AI. This is called Jevons Paradox. And we’ve seen this before with air travel.

Planes became cheaper to fly so more people flew more often to more places.

According to the International Energy Agency, emissions have steadily grown in recent years, even though we’ve gotten more efficient per passenger kilometer, right?

So it’s not just about making these systems more efficient, it’s really about being mindful and intentional about our consumption as well.

But of course, being efficient allows us to do more, to be more productive. Software engineers like GitHub right now, they’re doing a lot of studies around this, and

the more they use AI, they’re able to complete their tasks 55% faster. But that doesn’t mean that it’s making the workplace any easier, right? Tasks are expanding, deadlines are accelerating, boundaries are blurring, miscommunication is happening more often.

The productivity gains are real, but so is the pressure. And underneath that pressure is the question, will AI include us in the coming years?

Some CEOs say that new jobs will emerge, but it’s difficult to tell what those jobs will be, how long they’ll last, and whether the gains will be shared. When the richest 10% of the global population now owns 76%

of all wealth.

So for each of these promises, there’s capability and there’s consequence. Which brings me to the question,

can we really call it intelligence? If it’s built to ignore its own consequences.

I love seeing this question on a big slide like this. I think we should take a lot more photos of it and share it. It’s really the crux of the soul, right? To ask these kinds of questions, that’s where our human value lies. It’s in the questions we ask, the bold questions that we ask.

But right now, we’re not really asking a lot of questions to ourselves or to each other.

We’re picking sides.

On one side, AI will save us. On the other side, AI will destroy us.

But what each of these sides do is actually the same thing. They collapse our agency into inevitability.

Both of these sides say the outcome is already decided, so your choices don’t matter.

And that can breed a sort of nihilism within us. I mean, how could it not? I certainly feel it from time to time.

But what both of these perspectives are missing

is the here and now.

And the here and now, no matter how much money or power we have, the here and now is actually all that we have as human beings.

To be clear, this is not a dismissal of long-term thinking, and of course, we cannot predict every consequence in advance. But living in the present does something really interesting.

It allows us to stay present with what’s unfolding so that we can make changes quickly when we need to.

And it allows us to focus not just on the hypothetical AGI,

but rather our everyday decisions. The dataset or model that we’re choosing when we’re using these tools, the automated workflows that we’re accepting or pushing back on, the questions that we ask before we open up, Claude.

It’s that everyday decision-making that matters.

You know, if AI could do most of the tasks that you do right now in your job, or maybe it will do most of them, what’s left?

Not of your title, of your tools, but really like what’s left of you? No.

I’ve been wondering this question myself a lot,

and the good news is,

I think there’s a lot left in us than we think.

So I’d like to introduce a little bit of a playful, practical framework

of five roles that we can experiment with.

And these roles aren’t skills. I’m not here to teach you skills.

It’s more of ways of being human that no model can condense into a prompt.

So what are the five roles?

Well, the first one is the catalyst.

The catalyst cuts through ambiguity to make the unclear visible because AI can generate endless options and prototypes, but it cannot discern

when a moment needs direction.

Next is the bridge. The bridge connects what’s been broken amid a technological breakthrough, right? Divergent systems, misaligned goals, competing teams. There’s a lot of miscommunication happening in our collaborations

because AI can synthesize both sides, but it can’t sit in the tension between them to build trust.

Next is the listener.

The listener brings the outside world in,

lived experience, cultural signals, industry signals, unspoken needs

because AI can analyze what’s already happened,

but it can’t sense what’s about to.

Then there’s the curator. The curator cuts through slop to protect quality

and empower human choice.

Because AI can recommend and rank and remix things, but it doesn’t have taste.

And then there’s the subverter.

Do you see the little fish here that’s swimming in the opposite direction?

The subverter challenges conformity when AI-driven work converges towards sameness

because AI can only recombine and reinforce things in its own patterns. It cannot step outside of them. That instinct is yours.

Each of these five roles aren’t necessarily new, right? But they are at risk of being devalued, right at the very moment that agentic AI is really popping off. I was in Las Vegas last week at an AI conference, and let me tell you, agentic AI is really, really accelerating and it needs humans. It needs humans to direct it.

So let’s dig in to each of these five roles more deeply.

The catalyst. The catalyst cuts through ambiguity to make the unclear visible, often through the art of prototyping. These are the initial prototypes that Johnny Ive and his team created at Apple, right? And before, without these prototypes, the iPhone wouldn’t exist today. And of course, anyone can generate prototypes, no matter your role, right? Faster than ever. But the reality is that speed is not the soul of this role.

I think the real risk isn’t that AI replaces the catalyst or replaces the designer, but rather that we outsource so much of our thinking to AI that we forget what we actually believe about the problems we are solving.

Our instincts, our values, what we know to be the right solution or the wrong solution to speak up for that.

You know, we become

prompt operators, skilled at extraction, but disconnected from conviction.

So what does conviction actually look like? Well, I think it starts by asking questions like, what problem are we actually solving here? And for whom?

Or is this a solution that’s in search of a problem?

And how would we solve this differently if speed weren’t the priority? Of course, speed is always the priority. It will remain the priority in business, right? But what if we just, just for a moment did an exercise together,

where we asked ourselves, how would we solve this differently if speed weren’t the priority? I do this with teams all the time. It’s actually quite astounding, the new conversations and questions and ideas that emerge. It sort of stirs the room up a bit.

So really critical thinking is the soul of this role.

There was a study last year from the Harvard Business School where they

brought together 640 entrepreneurs from Kenya and they gave half the group this AI tool through WhatsApp to help them make business decisions and it gave the other half human written guides, no AI at all. And the result was that there was no statistical difference in business performance between the two groups. What mattered from this study was the human judgment that they enacted with each decision.

And so know what is actually yours, your values, your conviction, your instincts, your beliefs about what you’re looking at day in and day out.

Because again, AI will generate prototypes and it will get better and better at that over time, but it cannot generate purpose.

So before you write your next prompt, ask yourself, do I know what I actually believe about the problem I am solving?

Next is the bridge. The bridge connects what’s been fragmented amid a technological breakthrough and they tie the thread between hype and doom by asking questions like what would happen if we actually put these two groups in a room together?

What’s the disconnect that maybe we all feel but no one is saying out loud?

And what trust have we been assuming that we haven’t actually built yet with our users, with our customers, with the people engaging with our products? What trust have we been assuming that we actually haven’t built yet?

A great product example of the bridge is this little app called Be My Eyes. It’s actually built for people experiencing blindness or low vision and it connects them to sighted human volunteers to help them with day-to-day tasks. And in 2023, it introduced GPT-4, so it gave people choice. They could use AI for quick tasks or they could still connect with a human volunteer for more nuanced ones. What was really cool here is like, the bridge is in action at saying, you know what? It’s not just about making life easier, although that’s great, but it’s also about offering you choice so that you can still connect human to human and make life shared. That’s where the meaning comes in.

It’s now the largest network of its kind and last year it saw a 76% increase in monthly active users.

And with agentic AI, you know, we’re really, the relationship between human judgment and machine execution is really being redefined. And we talk a lot about this phrase called human in the loop.

But I think it’s really about humans on the loop, not in it, right? Like how do we give ourselves a little bit more agency to reshape that automated loop, to change it, right? When we’re in the loop, then we’re being affected by it. And it kind of assumes that we can’t really get out of it.

See, the way that we define technology determines what we measure and what we measure is what gets built.

And we’re at risk of seeing fragmentation at a scale we’ve never really seen before, right? Where it is possible that agents could start doing things, you know, without us really understanding it.

It could do things that we don’t want it to do. We could lose visibility into what it’s doing and why. Because no one really thought about the relationship between human judgment and machine execution. And that’s the consequence of what happens when we design for speed and efficiency without designing for trust.

But the design for trust, we have to practice it, right, with each other, to know what trust actually means with one another.

So the bridge has something to reclaim in this moment, which is to show up for each other, right? Build your relationships. I’m assuming it’s part of the reason why you’re here at beyond tellerrand to have an in-person live experience and to be in community. You know, the real temptation right now is to take the AI summary, to read the transcript of what someone said versus actually hearing them.

You know, we’re optimizing our relationships rather than inhabiting them.

But let yourself, you know, be with people in ways that are inconvenient or unscripted and see what happens. Let yourself be seen and see others. That mutual recognition is not a feature that AI can ship.

So the next time you see a miscommunication, what if that misalignment isn’t a problem to solve, but rather a relationship to repair?

Next is the listener. Listener brings the outside world in because we cannot vibe code in a vacuum, right? We have to go outside and listen to people. And I think the faster the AI evolves, the more at risk we are of creating things that are technically impressive, but humanly disconnected.

So the listener asks, whose reality are we actually designing for? And whose reality are we ignoring?

What’s changing in people’s lives that our product or service isn’t addressing yet?

And what will we learn if we spent a day in the life of a person using this tool?

An example I want to give is when I was a designer for Google Maps, I was working in Seattle with this engineering team. And at the time, Google Maps had been primarily designed for the US, in the US, for car drivers. And of course, the majority of the world does not drive a car.

Yet we were asked to make Google Maps more inclusive for motorbikes in Southeast Asia, across Africa and South America, where motorbikes outnumber cars four to one.

I thought to myself, what a ridiculous ask to ask people in Seattle who have maybe a lot of technical expertise, of course, but no lived experience. And this happens all the time in the tech industry.

So we decided to go outside. We decided to connect with local teams that lived in these areas, with local communities. We got on the back of a motorbike. This is a picture I took in Jakarta, Indonesia on a Friday evening, stuck in traffic. We got lost for several hours to feel the pain

and to work with communities to actually put motorbike mode into Google Maps. So we were able to track the different routes that were different from cars that allowed for shortcuts and faster routes. We added more landmarks to the map to help people with voice navigations. They could listen versus looking at their phone, of course. And more landmarks, though, on the map in case they wanted to look at it before they head out on their bike.

And after we did this, we saw a lot of people get really excited for the first time that something was actually maybe built for their transportation mode. And Google Maps went from one billion monthly active users after this to two billion.

Being the listener is often underestimated.

But I think it’s the very thing that we need right now.

So protect your own embodied experience.

You know, screens and AI operate in abstraction.

It’s not until we go outside and we listen we hear what people say and don’t say that we really kind of start to understand things that maybe question or crack open our own assumptions.

One way that I protect embodied experience within myself is I do this little practice called motion sketches

where when I’m on a bus or in a car or in a van or on a paddleboard in the ocean, which was very wet,

I put my pen to paper and I record the motion that I’m feeling and I put my phone away just so that I can feel the environment around me. Sometimes I can strike up a conversation with a stranger as well when I do this.

This is when I was quarantined

in 2020 during COVID. I was stuck in the same room like a lot of us. Wasn’t feeling a lot of motion at that time. But I still wanted to feel it even though it was painful. I still wanted to feel it and record it.

The more the digital life expands, the more we need to seek out a world beyond screens

because the meaning that we bring to our work, the value that we bring to our work is only as rich as the realities that we’re willing to inhabit.

So when was the last time that you let a real environment change your mind about a design decision?

(Sighs)

One thing I wanna say here too is that AI can,

it can analyze, right, data that’s already happened. It can analyze a lot of data, which is great.

But it can’t truly listen.

It doesn’t feel empathy, context, or silence.

(Sighs)

That’s where the human, the listener, provides meaning.

Let’s move on to the curator. The curator cuts through SLOP to empower human choice and protect quality.

Last year, Elon Musk launched Grokipedia as a competitor to Wikipedia.

And in just a matter of days, it generated 900,000 AI articles. And he claimed that it would exceed Wikipedia in accuracy, breadth, and depth. But actually, it did the opposite. There were a lot of inaccuracies. There was a lot of bias, lack of nuance. And Wikipedia still kinda came out on top as the leading platform, largely in part because it has community verified content. You know, sustainable quality really still relies on human editorial judgment.

So the curator asks, "Where is quantity masquerading as quality?"

And who’s deciding what quality means? Is it a human? Is it a large language model? Or is it no one?

ChattyBT launched ChattyBT Pulse last year. And it’s this kinda cool, highly curated experience that gives you different tips. This is giving travel tips to you based on your preferences. But I think even in highly curated experiences like this one, the curator must stay vigilant and attentive and ask questions like, "Well, what if ads kinda creep into this? "How is that gonna diminish quality or trust?" Or what if, instead of ChattyBT recommending to you travel tips, what if it reminded you that you have a friend that’s actually been to that place before and has a lived experience, and maybe they can actually connect you to things that no computer would ever know?

Right, this is how the curator still thinks about empowering human choice and human to human connection. Even in highly curated experiences like this one.

So choose what you let in.

To be a good curator, you have to be aware of the fact that your mind becomes what it consumes. If you’re endlessly scrolling through AI-generated content or algorithmically optimized feeds all day, your taste will flatten.

And the curator with the flattened taste has nothing original left to offer.

So choose what you let in, but also let your mind wander.

You know, go to a new city without any agenda. Read a book with no application. Let yourself get a little lost. Let yourself be bad at things, right? Just wander for no reason.

So how might you remind people that discovery is one of the greatest pleasures of being alive?

And finally, the last role, the subverter. The subverter challenges conformity when AI-driven work converges towards sameness. And there’s a lot of really great examples of this. So there’s a product called U-Tune, not YouTube, but U-Tune that actually escapes or subverts the algorithm. And it shows you videos that have very few likes or views. Videos that maybe you’ve never seen before, right? The subverter in action. U-Tune, check it out.

There’s also this campaign from Polaroid last year, AI can’t generate sand between your toes.

I think about Annie Adkins’ talk last night, right? Where she was really kind of talking about how analog experiences can actually be more effective. And I think this is a great example of that. They leaned in to their analog nature to provide distinction.

And then there’s websitecarbon.com. If you go to websitecarbon.com, you can type in, I typed in chat GPT here. You can see

the carbon rating that any website has based on the energy that they’re consuming. And I think what’s really cool is that the subverter kind of pokes fun capability and consequence, right? They’re really looking for transparency to change things.

So they ask what default solution deserves to be questioned?

What’s the boldest version of this and why aren’t we building that?

And what if people don’t want what we think they want?

What if people don’t actually want what we think they want?

Now, of course, to ask these questions though, provocation demands agency. You have to hold your own position before you can challenge someone else’s. And right now, being a subverter sounds pretty risky, I know, because there’s a lot of layoffs going on and probably more to come.

However,

agency scales, right? And it shows up in different ways. A senior leader at a company can shape the policy, but an individual contributor can ask one question in a meeting that can change the whole trajectory of a project. And we as tech consumers can choose what models we’re using, whether or not we use AI, depending on the task at hand.

We often have someone too in the room that maybe has less agency than we do.

And we can lift them up through the process of this. But to play the role of the subverter authentically,

it requires us to do something that is quite difficult, that our current culture is eroding, which is to stay in contact with discomfort.

AI aims to be frictionless, to smooth things, to give us quick answers.

And the more we lean into that, the more that struggle feels like

a failure. Boredom starts to feel like a problem and uncertainty feels like inefficiency. Resist that, because struggle is where breakthroughs happen.

When we’re feeling bored, that’s where unexpected ideas emerge.

And when we’re frustrated, that’s when we ask ourselves questions that can change the trajectory of our lives or the work at hand or the way that we do art.

AI can only reinforce its own patterns. It cannot innovate outside of them.

That instinct is yours, if you’re willing to sit in the discomfort of uncertainty.

So if AI can only give you what already exists, what are you willing to create that doesn’t yet?

These are the five roles that we can step into, that we can play with, that we can experiment with. There’s probably a lot more than five roles, right? You don’t have to be all five at once, maybe just pick one or two. Whatever you think is most needed with the client that you’re working with or the organization you’re working within, this is how we enact our human agency.

And know what is actually yours.

Show up for each other.

Protect embodied experience.

Choose what you let in

and stay in contact with discomfort. 20 years from now, what world will we see based on the decisions and the choices that we make today?

Will we see technology that, you know, considers capability and consequence?

Or will we see an array of conflicts without the attempt of a resolution?

The stakes are high,

but they’re not inevitable.

It is our everyday choices

that ultimately shape who we are becoming.

Every choice you make shapes who we are becoming.

So what will your choice be?

Thank you.

(Audience Applauding)