Why we should use AI to expand what it means to be human | Sari Azout

Summary

Highlights

• The narrative of AI replacing humans is as nonsensical as humans versus pencils or calculators. 📏
• AI should be viewed as a form of collective intelligence, emphasizing its potential to aggregate human knowledge. 🌐
• Technology creates new standards and doesn't just make old tasks easier, leading to a paradox of increased workload. 💼
• AI democratizes tools but requires expertise to direct and assess outputs effectively, highlighting a new form of expertise. 🎨
• The importance of reclaiming unquantifiable human qualities in an age dominated by mechanical metrics. 💡

Key Takeaways

• The phrase 'AI will replace humans' is oversimplified; AI can actually enhance human creativity and potential. 🤖
• Language shapes our perception of technology, and rephrasing AI as 'collective intelligence' can change how we perceive its role. 🗣️
• AI's impact on workload is paradoxical; it can increase work by creating new opportunities rather than just saving time. ⏳
• Access to AI tools doesn't replace expertise; it shifts the focus to guiding and evaluating work. 🔍
• AI reflects and amplifies our values; it calls for a return to valuing unquantifiable human qualities. 🌟

Overview

Chapters

• 00:00 - 01:30: Introduction and Framing the AI Question The chapter opens with a speaker acknowledging an impressive introduction given by Lauren. The speaker then humorously questions a commonly debated topic in AI discourse: Will AI replace humans? This question is often considered mundane yet remains intriguing enough to fuel discussions and sell conference tickets. The speaker reflects on the prevalence of this question in AI conversations.
• 01:30 - 03:30: The Impact of Language on Perception This chapter explores how language shapes human perception, particularly in the context of technological advancements, using the inception of the term 'artificial intelligence' as an example. The narrative begins with John McCarthy in 1955, who coined the term 'artificial intelligence' as a funding strategy, a decision that has since influenced how people perceive and interact with technology. The chapter reflects on humanity’s tendency to view technological progress through a lens of threat and opportunity, underscored by the power of language in framing these perceptions.
• 03:30 - 06:30: The Reality Gaps in AI Utilization The chapter titled 'The Reality Gaps in AI Utilization' addresses the perception of technology, particularly how language influences our thoughts. It presents a comparison between human editors and AI aides in writing. While human editors are seen as collaborative and legitimate, using AI is often viewed as lazy and inauthentic.
• 06:30 - 09:30: The Nuanced Expertise Required in AI Use The chapter titled 'The Nuanced Expertise Required in AI Use' addresses the complexity and emotional dimensions involved in utilizing AI. Initially, people may take shortcuts in learning and application due to misunderstandings or misconceptions about AI. The text suggests that changing the way we talk about AI can alter our understanding and application of it. An artist, Holly Herndon, argues that the term 'AI' might actually misrepresent the collective nature and potential of artificial intelligence, influencing our perception of it.
• 09:30 - 12:30: AI as a Reflection of Human Values The chapter delves into the nature of artificial intelligence, suggesting that AI models like large language models (LLMs) are fundamentally advanced statistical tools. These tools aggregate human-generated data more effectively, thereby enhancing human intelligence and enabling us to extract greater utility from our collective knowledge. The chapter proposes reimagining AI as "Open Collective Intelligence" rather than just individual artificial intelligence, emphasizing its reflective nature of human values and collective progress.
• 12:30 - 16:00: Concluding Thoughts on AI and Human Evolution In this chapter, the author reflects on the use of AI in writing and creativity, emphasizing the advantages of collaborating with AI systems. They argue that using AI, referred to as CI (Collaborative Intelligence), is not just resourceful but essential, equating it to dismissing legendary writers if it isn't utilized. The chapter concludes with an introspective note on how language and perception influence one’s understanding of reality.

Why we should use AI to expand what it means to be human | Sari Azout Transcription

• 00:00 - 00:30 Wow, thank you for that incredible introduction,  Lauren. And now, I'm going to completely undermine   it by starting with a very boring question. Will  AI replace humans? It's the "how's the weather" of   AI discourse. right? Safe enough for dinner parties  but provocative enough to sell out of conference   tickets. But I'm fascinated by why this question  dominates the discourse. Why, when faced with one
• 00:30 - 01:00 of humanity's most remarkable achievements, we  instinctively frame it as a threat to ourselves.   Like most things in Tech, it began with a funding  pitch. It was 1955, and a researcher named John   McCarthy was looking for a catchy term to attract  funding for his work. Artificial intelligence   sounded exciting, and it worked. And just like that,  this choice of words became the defining frame
• 01:00 - 01:30 through which we experience this technology. We  often think of language as a means to an end but   rarely think about how language itself influences  our thinking. Let's look at an example. If you're   writing with the help of a human editor, you're  collaborative, responsible, a completely legitimate   form of assistance. If you're writing with the  help of AI, you're lazy, cheating, inauthentic, you
• 01:30 - 02:00 know, cutting corners, taking shortcuts, rather than  learning. Similar activity—completely different   emotional response. Now, I want to convince you  that changing your vocabulary can change your   reality because it happened to me. This was also my  initial reaction, and then I heard an artist by the   name of Holly Herndon say that AI is actually  a huge disservice of a term, and that collective
• 02:00 - 02:30 intelligence is at far more accurate. Because if  you strip LLMs to their pure essence, they are just   a much better way of using statistics to aggregate  human intelligence and connect all of the things   we've done together so we can get more use out of  them. So, let's run this exercise again except this   time let's imagine OpenAI had been called OpenCI, open collective intelligence. Again, if you're
• 02:30 - 03:00 writing with the help of a human editor, you're  collaborative. If you're writing with the help of   CI, you're resourceful and leveraging the best of  human knowledge. In fact, I'd even go as far as to   say that I'd be judged if I wasn't using CI. That  would be the equivalent of having Gabriel García   Márquez and Virginia Wolf in the room and asking  them to leave. So, if one word can distort my entire   perception of reality, what other blind spots was I  dealing with. I couldn't shake it. I needed to find
• 03:00 - 03:30 out. So that's what I did. I spent the past two  years going to the source, building my company   Sublime while obsessively using AI tools. I used  them to write, to build software, to design software,   even tried to use it to write this presentation  for me. And then I paid attention not just to what   the tools could do but to how they made me feel.  How were they changing me and my relationship with
• 03:30 - 04:00 work. How were they changing what it feels like to  be human. And what I discovered was a profound gap   between what I was expecting and the reality I was  experiencing. So, I want to spend the next couple of   minutes just going through these three reality  gaps, and they've changed how I work with and   relate to AI, and maybe they can do the same for  you. Here's reality gap number one. I expected AI to
• 04:00 - 04:30 reduce my workload and free up time—you know the  dream of robots doing our jobs while we chilled   in hammocks all day. Instead, I was caught in a  paradox. The amount of work that is now possible   and exciting and worth doing because of AI greatly  exceeds the amount of time I save using AI. This   isn't a new pattern. In the early 20th century, we  witnessed an explosion of labor-saving technologies.
• 04:30 - 05:00 Fridges and washing machines and dishwashers  and vacuums that should have theoretically   reduced housework. But women in 1960 were spending  more time on housework than they were in 1920.   Here's my grand theory for why we have no time.  Technology doesn't just make old tasks easier.   It creates entirely new standards. Before washing  machines, families used to wash clothes maybe twice
• 05:00 - 05:30 a week. After that, weekly laundry became the norm.  Technology does not exist in a vacuum. We respond   to technology. I have felt this shift personally.  One of the litmus tests that I now have before   releasing work is is this un-LLM-able, by which  I mean, does this work carry the unmistakable   fingerprint of human creativity, perspective, and  lived experience. Of course, the irony is I use
• 05:30 - 06:00 LLMs to assist me in producing the work. But as a  whole, AI makes average attainable by anyone, and   therefore raises the bar for what is considered  exceptional. LLMs will soon do so many things that   we once thought only humans could, but we'll keep  moving up the stack. We'll take those outputs as   inputs and dream bigger. Some human skills will  be commoditized, but human beings will not. We are
• 06:00 - 06:30 un-LLM-able. Here's reality gap number two. I assumed  AI would democratize expertise and make anyone a   designer, a writer, a lawyer, an engineer, and to  some degree, that's true. The technical barriers   are crumbling. But what I discovered is far more  nuanced. To shape the inputs and to evaluate the
• 06:30 - 07:00 outputs, you still need expertise. LLMs are  really impacted by every word you input.   Let me show you with an example, and of course I'd  be remiss not to use this wonderful opportunity in   the heart of Stockholm to not plug an app that I'm  building called Podcast Magic. So, in this example,   we're going to prompt an LLM to help us generate  the marketing headline for Podcast Magic and, just   for context, Podcast Magic is an app that lets you  capture insights from podcasts using screenshots.
• 07:00 - 07:30 As in, you're out for a run listening to a podcast  on Spotify. here's something you love, take a   screenshot. the app will automatically generate a  transcript plus audio of that moment so you can   refer to it later. So, in this example we have two  prompts. Prompt on the left is really someone who   maybe they don't care, maybe they're not an expert, they're just trying to get the task over with.
• 07:30 - 08:00 The point is they're just going to enter a generic  input and accept whatever mediocre output the AI   will generate. On the right is a prompt by someone  who's really leveraging their understanding of   positioning and psychology, who is adding context,  importantly, I cannot emphasize this enough, asking   the AI to generate multiple variations and  then of course pressure testing the models   until the words hit just right. So, in this scenario,  you'll have two prompts, same tool, wildly different
• 08:00 - 08:30 results. And the key thing here was not the tool, it  was the person operating the tool. And I think the   key thing to take away from this is that the  expertise here is not in doing the work. It's   in guiding the work, in evaluating the work, in  knowing what is worth prompting, which is really   just another way of saying—knowing what is worth  doing. I was trying to come up with an example to
• 08:30 - 09:00 illustrate this point better in a way that  was more sticky, and I kept coming back to this   definition of modern art that I came across not  long ago, which says modern art equals "you could do   this plus yes, but you didn't." You know the feeling of  seeing a black square painting go on auction for   $80 million and thinking, "Come on I could have  done that." But you didn't. They did. They knew
• 09:00 - 09:30 it was worth doing, when, and how to frame it  so it mattered. And I think more and more this   will apply to realms outside of art. Things will  look simple and easily replicable on the surface.   People will say things like, "Ah anyone could have  prompted that with ChatGPT." But the more and   better access everyone has to tools, the clearer  it becomes that the bottleneck to great work is
• 09:30 - 10:00 not knowledge, it's not information, it's not even  intelligence. It's that intangible quality, call   it taste, creativity, judgment, courage, intuition,  agency. We've all heard the quote from Thomas   Edison where he says, "Genius is 1% inspiration 99%  perspiration." Back then, execution was expensive   and learning was slow. Well, I think it's going  to flip, and maybe be like 99% inspiration 1%
• 10:00 - 10:30 perspiration. If you have the will, a laptop, and  a point of view, you can become dangerous in days.   That's a completely different definition of  expertise. Here's the final reality gap. When   I first started using AI, I feared that machines  were becoming more human, and yes, it is getting   harder to tell the difference. But that's not just  because humans are becoming more machine-like—or
• 10:30 - 11:00 because machines are becoming more like humans—but  because humans are becoming more machine-like. We   worry AI will replace writers, but half the  internet is engagement farmers on LinkedIn   selling five ways to 10x your creativity by 6 a.m.  And this mindset has infected everything. We join   Twitter for the love of learning, come out obsessed  with likes. We do research for discovery then get
• 11:00 - 11:30 trapped chasing citation counts. We want to hire  and retain the best teachers so we measure test   performance, and then teachers teach the test.  We want to reward quality media so we measure   clicks, and then we get clickbait. Put differently,  we try to measure what we value but inevitably   end up valuing what we measure. And what can  be measured, tends to be the mechanical stuff.
• 11:30 - 12:00 Not only do numbers not capture what we care about,  they are extremely open to manipulation. The idea   that our value can be measured in KPIs, and OKRs,  and numbers, that can be seen and tracked and made   legible to spreadsheets is a modern invention.  If we go back to ancient Greece, human worth was   really tied to wisdom and contemplation. Medieval  England is incredibly tied to religious devotion.
• 12:00 - 12:30 If we look at many indigenous cultures, status was  deeply tied to spiritual connections, storytelling,   abilities, relationships. We just decided at some  point that progress meant numbers on a spreadsheet.   I think AI is really just holding a mirror  to us, and in that reflection awakening this   desire to return to the things that cannot be  easily quantified. You know, so many of us in
• 12:30 - 13:00 Tech, we tend to see the world through the lens  of technological determinism, where we assume   that the course of history will play out according  to what is technologically possible. But the most   exciting breakthroughs won't come from technology,  they'll come from how these new tools reshape our   understanding of ourselves. From who we can become  when we have all of this new power. Which brings   me back to the fundamental problem with how we  framed this entire conversation. We've taken one
• 13:00 - 13:30 of the most extraordinary technologies and given  it the most banal framing possible—how can this   replace a human? To illustrate the absurdity of  this framing, let's replace humans with technology.   The debate between humans and AI is as nonsensical  as humans versus pencils, humans versus calculators,   humans versus maps. Technology is not our rival. It  is simply applied human knowledge. But here's the
• 13:30 - 14:00 thing about AI. Compared to other technologies with  clear founding missions, AI is an ideological blank   canvas. The Internet was about openness. Blockchain  and crypto were about freedom and decentralization.   But there is nothing missionary built into AI  itself. The leading labs have all stated that   building AGI is their explicit goal, which focuses  on some vague technical dream of replacing human
• 14:00 - 14:30 intelligence while offering very little about  what that means for us. Now, that should all be   great news for the people in this room because  the biggest problems are not technological, they   are philosophical. They are questions of values,  ethics, and worldviews. Just as we are living   inside Zuckerberg—Zuckerberg's belief in instant  connections, and Steve Jobs' dream of the computer
• 14:30 - 15:00 as a bicycle for the mind, we are already stepping  into the world as imagined by Sam Altman. But what   is his vision of a meaningful life? What does he  believe leads to human flourishing? And forget   about Sam Altman. What do you believe leads to human  flourishing? In my own work at Sublime, this has   meant reimagining what a knowledge management  can be if designed not for productivity, but
• 15:00 - 15:30 for creativity. Not for making something fast but  for making something wonderful. Not for automating   words but for alchemizing minds. And I've seen  how the shift in frame can create ripple effects   across every decision we make. This isn't woowoo  stuff, it's deeply practical because products are   not neutral They are opinions embedded in pixels.  Every button, every default, says something about
• 15:30 - 16:00 what we believe people are for. So, I want to return  to the question we began with. Will AI replace us? Yes, but fearing that AI will replace us is  like fearing that our children will replace us.
• 16:00 - 16:30 Yes, they will replace us because that is  why we created them, but also they depend   on us, and they are us, and they  can improve us and free us from   becoming machines so that we can reclaim the  things that machines can't touch. Thank you.
• 16:30 - 17:00 [Music] [Music]