English
Based on โWhy this Claude Code engineer uses HTML files as AI specs | Thariq Shihipar (Anthropic)โ from How I AI Watch the original video
The Secret Language of AI Specs: Why HTML is Outpacing Markdown for Smarter Collaboration
In the fast-evolving world of AI-driven product development, teams are constantly seeking more effective ways to communicate with their intelligent agents. For a while, Markdown reigned supreme as the go-to format for crafting AI plans and specifications. Its simplicity and human-readability made it an accessible bridge between human intent and machine execution. However, as AI agents grow more sophisticated and their tasks more complex, a surprising challenger has emerged: HTML.
Thariq Shihipar, an engineer on Anthropicโs Claude Code team, is at the forefront of this shift. Speaking at Anthropicโs โCode with Claudeโ developer conference, Shihipar revealed a revelation that has since โblown his mindโ: HTML is the new Markdown for AI specs. This isnโt just about technical preference; itโs a fundamental rethinking of how humans and AI can achieve a deeper, more productive synergy.
The Markdown Paradox: When Plans Become Unreadable
The initial appeal of Markdown for AI interactions was undeniable. Its straightforward syntax allowed developers to quickly outline plans, feature requests, and code modifications for agents like Claude 3.5. โMarkdown became a really popular way of interacting with agents,โ Shihipar explains, โespecially like, you know, Opus 4 and Claude 3.5 where, you know, they have a plan, and the plan is like how to do this feature and maybe itโs like 50 lines of code and you can edit it.โ
However, as AI models advanced, particularly with longer-running agents like Opus 4.5 and 4.7 that can operate for hours, the plans they generated ballooned in size and complexity. โThe plans are so long,โ Shihipar admits, โI honestly have stopped reading them. And this was honestly a mistake. I think that you still need to be really in the loop.โ A thousand-line Markdown file, he notes, becomes an unmanageable beast. Humans no longer edit them; they simply ask Claude to edit itself. This disconnect creates a critical gap in oversight and quality control.
This challenge underscores a broader shift in the role of human professionals in the AI era. As host Claraveo, a product leader and โAI obsessive,โ dramatically put it, โPeople ask me all the time, Claire, you said product management is dead. Whatโs next? Iโm going to say youโre a compute allocator, babe, thatโs the job now.โ
Shihipar wholeheartedly agrees. โWhen you say, okay, Claude can run for eight hours, what youโre really saying is Claude can spend like 500 bucks.โ Every professional working with AI agents is now a โcompute allocator,โ tasked with deciding whatโs truly worthwhile for the AI to spend its computational resources on. This demands clear, precise, and readable specificationsโa role where Markdown is increasingly falling short.
HTMLโs Unsung Virtues: A Richer Dialogue with AI
The Anthropic teamโs pivot to HTML wasnโt born from a desire for complexity, but from a need for clarity and engagement. While models are adept at processing various formats, HTML offers distinct advantages for the human side of the collaboration:
Richer Communication, Enhanced Readability
โHTML is a lot easier to read,โ Shihipar points out, โand so itโs just a richer communication medium between you and Claude.โ Unlike Markdownโs often crude ASCII diagrams, HTML allows agents to โactually draw things that you can look at.โ This visual fidelity means less cognitive load for the human, making it easier to parse complex information, scroll through extensive plans, and truly understand the AIโs proposed actions.
Interactive Artifacts and โMicro-Softwareโ
One of the most compelling aspects of HTML is its capacity for interactivity. Shihipar demonstrated how Claude can generate not just static documents, but dynamic, engaging artifacts. In one example, to help him choose a demo idea for the podcast, Claude presented eight visual demos in an HTML file, complete with mockups and even โriskโ assessments for each. โThis is something that I will actually read,โ he emphasized, highlighting how the visuals make the ideas โso much more evocative.โ
This concept extends to what Shihipar calls โmicro-software on top of micro-software.โ Imagine a section within an HTML plan detailing data visualization rules. If a user wants to tweak these rules, instead of editing raw code or Markdown, Claude can generate a custom, throwaway UI โ an interactive form or dashboard โ directly within the HTML. The user engages with this โgamified,โ โconsumeryโ interface, makes their changes, and the updated data is seamlessly fed back into the plan. This bespoke UI drastically improves human engagement and, consequently, the quality of the output.
The Art of Prompting: Guiding, Not Constraining
Crucially, the power of HTML doesnโt necessitate overly complex prompts. Shihiparโs initial request to Claude was disarmingly simple: โLook, Iโm on a Claro podcast. I want to do a demoโฆ can you brainstorm some ideas in the HTML file?โ Later, for a detailed implementation plan, he asked: โcreate an HTML file as a plan that helps me visualize what the implementation plan is, include excerpts, mockups, code when whatever is needed to give me maximum context.โ
This approach embodies a philosophy of giving Claude enough information to achieve the goal, but not โover-constrainingโ it. As Claraveo adds, โI believe in you and trust you. Exclamation. Iโm like truly I know you are capable and I believe in you to make these decisions.โ Itโs about fostering a trusting, collaborative relationship where the human provides the high-level vision and feedback, and the AI leverages its intelligence to craft the most effective presentation.
Beyond Plans: HTML for Living Systems and Collaboration
The utility of HTML extends far beyond mere project plans:
Living Design Systems
Shihipar enthusiastically showcased how HTML can power a โliving design system.โ Instead of static Markdown documentation, an HTML file can dynamically represent a companyโs design languageโcolors, typography, spacing, radius, and core components. This single, interactive file can then be easily referenced by Claude across new projects. โI go to a new project,โ he explains, โIโm like design_system.html, right? Instead of like design.md or something.โ
Claraveo expanded on this, describing how she uses Claude Design to pull in existing marketing and app repos, generate a component-level style guide in HTML, and even create a โcomponent visualization page.โ This page features interactive versions of 25 app components, allowing marketers to download transparent PNGs for decks or videos. This drastically reduces friction, ensuring consistency across all brand touchpoints, from code to marketing materials.
Streamlining Internal Communication
The benefits of HTMLโs richness even permeate internal team communication. Shihipar revealed a personal hack: โEvery week I send [my manager] a weekly status update in HTML of everything Iโve done. I get Claude to read my Slack and just like create this message, and and like she actually gets to read it and I donโt have to spend that much time on it.โ This demonstrates how engaging, AI-generated HTML content can significantly improve communication efficiency and impact within an organization.
โIt is just much a much more effective way to communicate across a company,โ Claraveo observed, โis with content that is engaging and at the right level of detail and consumable and weโre all pretty good at reading websites.โ
The Future of Product Development: Focus on Content, Not Constraints
The shift to HTML for AI specs marks a profound change in how we approach product development documentation. Historically, teams would get โwrapped around the axleโ on โsources of truth,โ rigid templates, and centralized repositories. These arbitrary rules emerged because content creation and consumption were expensive, and finding information was difficult.
However, as the cost of generating and accessing high-quality content approaches โfunctionally zeroโ (even with โClaude chits,โ as Claraveo humorously calls them), these constraints become obsolete. โYou can kind of put stuff wherever in whatever format because we know these models are very good at using tools to discover the context that they need,โ Claraveo notes.
This liberates teams to focus on what truly matters: the content of the plan, the quality of the idea, and the effectiveness of its execution. HTML enables โjust-in-time documentationโ and even โthrowaway softwareโ that is cheap enough to be discarded once its purpose is served. If something needs to be found later, โClaude can find it for us, itโs fine.โ
Ultimately, this approach results in better products. By enhancing human engagement, fostering deeper collaboration with AI, and allowing for flexible, tailored documentation, HTML is poised to become the cornerstone of AI-powered product development. Itโs a testament to the idea that the most effective tools arenโt just about what machines can read, but what truly resonates with the human mind.
As Thariq Shihipar and Claraveo demonstrate, the future of building with AI isnโt about replacing human intelligence, but augmenting it with tools that make our interactions richer, more intuitive, and ultimately, more impactful. Goodbye Markdown, hello HTML โ the webโs original language is now your AIโs best friend.
Based on โThe Courtroom Showdown Between Elon Musk and Sam Altmanโ from New York Times Podcasts Watch the original video
The Billionaire Brawl: Inside the Musk-Altman Courtroom Showdown
For three intense weeks, the normally staid federal courthouse in downtown Oakland transformed into a veritable circus, drawing a motley crowd of protestors, fans, and the simply curious. At its heart was a high-stakes legal battle, a corporate dispute that spiraled into a deeply personal saga between two of the most influential figures in global technology: Elon Musk and Sam Altman. The co-founders of OpenAI, the company at the forefront of the artificial intelligence boom, found themselves locked in a courtroom showdown, with Musk seeking to dismantle the trillion-dollar company he helped create and remove Altman from its helm.
As jury deliberations began, New York Times tech reporter Mike Isaac, who endured โthree weeks of 12-hour days, 5 a.m. morningsโ covering the trial, offered a rare glimpse into the drama. Outside, giant wavy car dealership inflatables, one emblazoned with โElon Sucks,โ vied for attention with protest groups railing against both billionaires and the broader implications of AI. Inside, the scene was no less charged. The tech titansโMusk, Altman, and OpenAI co-founder Greg Brockmanโwere forced to walk a gauntlet past press and public, a rare descent from their stratospheric perch.
The Altruistic Dream: When AI Was for Humanity
The genesis of OpenAI, and indeed the core of Muskโs lawsuit, dates back to 2015. It was born from a profound disagreement between Elon Musk and Google CEO Larry Page. Musk, long a vocal worrier about AIโs potential to harm humanity, found himself aghast at Pageโs seemingly cavalier attitude, reportedly stating that AIโs survival, even if at humanityโs expense, would be โokay.โ Page, in turn, called Musk a โspeciesistโโa term new to many, suggesting discrimination against artificial intelligence.
Shocked by what he perceived as a disregard for human welfare, Musk sought out Sam Altman, then running a Silicon Valley startup incubator. Together, they forged a vision for OpenAI: an altruistic, nonprofit endeavor dedicated to building artificial intelligence safely and for the good of mankind. Musk, fresh off his PayPal success and riding high with Tesla and SpaceX, became the primary financier, pouring in $38 million. The mission was clear: โthe good guysโ would be the stewards of this immensely powerful, potentially world-changing technology.
Emails from that era, introduced in court, revealed the foundersโ genuine, if somewhat โfar-flung,โ concerns about AIโs futureโa stark contrast to the cynical view many hold today. โThey were serious about it,โ Isaac noted, highlighting that these were not mere public performances but private anxieties shared among a small group.
The Pivot to Profit: A Betrayal or a Necessity?
By 2017-2018, the initial dream began to fray. Altman and Brockman, tasked with building cutting-edge AI models, realized the nonprofit structure was a severe limitation. Developing advanced AI required immense resources, far exceeding the millions Musk had initially contributed. They needed โtens of millions, billions even,โ a scale of investment a nonprofit simply couldnโt attract.
Musk, meanwhile, was increasingly worried about losing control. He proposed integrating OpenAI into Tesla, a suggestion Altman resisted. The relationship soured, and Musk, frustrated and juggling multiple ventures, decided to cut ties. โLook, Iโm out. Go with God. Good luck. Iโm doing my own thing,โ he reportedly declared, unaware of the future AI explosion. OpenAI, though losing its primary benefactor, continued to receive a few more months of rent payments from Musk, a seemingly amicable parting.
This departure paved the way for Microsoft. Recognizing Googleโs lead in AI, CEO Satya Nadella saw an opportunity. Microsoft pledged a staggering billion dollars to OpenAI, a sum that dwarfed Muskโs original investment. This marked a pivotal shift, moving OpenAI further toward a model capable of securing the vast capital needed for its ambitious research.
ChatGPTโs Boom and Muskโs Fury
The true turning point, however, arrived in 2022 with the โunceremoniousโ release of ChatGPT. OpenAIโs chatbot, which they initially โdidnโt really think was a big deal,โ exploded overnight, becoming the fastest-growing application in history. Its success sent shockwaves through the tech world, triggering a frenzied investment race. Microsoft, seeing its gamble pay off spectacularly, upped its investment to an astonishing $10 billion.
This was the moment Muskโs anger ignited. The nonprofit he had co-founded, intended for the good of mankind, now looked undeniably like a for-profit juggernaut. He fired off a message to Sam Altman: โDude, what are you doing? This feels like a bait and switch.โ
The irony was not lost on observers. Around the same time, Musk launched his own AI company, X AIโa decidedly for-profit venture aimed at competing directly with OpenAI. As Isaac succinctly put it, โThis case is basically like deciding which billionaire you believe.โ Both men stood accused of hypocrisy, a central theme that would define the courtroom drama.
The Courtroom Show: Muskโs Pugilism vs. Altmanโs Pushback
The trial became a battle of narratives, each side attempting to sway the nine-person jury on the credibility of these tech titans.
Elon Muskโs testimony began with his lawyers painting him as a โserial entrepreneur dreamer guyโ and an โaltruistโ who cares deeply about humanity. Despite recent controversies and political alignments that made jury selection challenging, Muskโs on-stand demeanorโa muttering, rambling, occasionally self-effacing charmโseemed calculated to resonate. He spoke of grand visions, from Mars to saving humanity, with the matter-of-factness of an engineer listing daily tasks. His core argument: OpenAI was โstealing a charity,โ and allowing them to get away with it would endanger all future philanthropic endeavors. He positioned himself as the โlittle old meโ billionaire fighting a โMachiavellian guyโ who prioritized profit.
However, the โElon showโ truly erupted during cross-examination by OpenAIโs attorneys. Prodded with questions insinuating untruthfulness or hypocrisy, the humble Musk melted away, revealing the โpugilist.โ He became snippy, accusatory, and visibly angry, lashing out at the lawyers for being โdisingenuous.โ The gallery, filled with Musk fans, responded with โfist pumpsโ and snickers, treating the proceedings โlike a show, man.โ For OpenAI, this was their opportunity to demonstrate that the lawsuit was merely a โtantrumโ from a man โmad that heโs not in control.โ
Sam Altmanโs โshowโ presented a different kind of charisma. Soft-spoken and leaning into an image of an โentrepreneur who cares about the world,โ he spoke of humble beginnings and a passion for technologyโs positive impact. His challenge, however, was a reputation increasingly marred by accusations of deception. Recent public scrutiny, including a New Yorker article, had painted him as โfundamentally untrustworthyโ and โnot consistently candidโ with executives.
Muskโs lawyers wasted no time, opening their cross-examination with a brutal, โAre you trustworthy?โ Altman, visibly taken aback, responded with a tentative, โI think so,โ which was immediately weaponized by the opposing counsel. What followed was โa parade of here are all the ways we think youโre a bald-faced liar in every aspect of your life,โ according to Isaac. Altman, though not resorting to Muskโs pugilism, gradually pushed back with more force against the relentless onslaught, trying to maintain his composure under immense pressure.
The High-Stakes AI Race: Beyond the Verdict
The trial, in many ways, transcended the legal arguments, becoming a โvibes-basedโ assessment of character. The juryโs task was to discern โwhich side they trust more,โ a task complicated by the โmuckrakingโ and โold emailsโ dredged up for public consumption.
Regardless of the verdict, the case has already unveiled the โreal uglinessโ beneath the veneer of Silicon Valley innovation. The competition in AI, Isaac observed, is โcrazier than ever.โ Itโs no longer just about raising millions but โtens of billions, hundreds of billions of dollars at stake.โ The pursuit of dominance has led to โdragging their name and reputation through the mud,โ โcompetitive intelligence,โ โnasty character assassination,โ and even extreme acts like โMolotov cocktails being thrown at Sam Altmanโs house.โ
This intense, personal rivalry plays out against a backdrop of widespread public skepticism about AI. While technologists envision a future of faster, better processes, the average person hears โyouโre going to take my jobsโ and โyouโre going to be richer than God and Iโm going to be left with the scraps.โ The courtroom drama, with its accusations of greed and deception, only heightens this mistrust.
The Musk-Altman showdown is more than just a legal battle; itโs a stark reflection of the existential stakes and cutthroat competition defining the current AI era. โNo matter what the outcome of this thing is,โ Isaac concluded, โthereโs not really any coming back from how high stakes things are. Thatโs just how it is out here now.โ The verdict may decide the fate of OpenAI, but the trial has already laid bare the raw, unvarnished realities of the race to control the future of artificial intelligence.
Based on โHow to be more creative | The Gray Areaโ from Vox Watch the original video
The Surprising Power of Limits: How Constraints Unlock Creativity and Freedom
We live in an age that champions boundless choice and unfettered freedom. From career paths to consumer goods, the prevailing wisdom suggests that more options lead to greater happiness and innovation. Yet, what if this deeply ingrained intuition is fundamentally flawed? What if, paradoxically, true creativity, productivity, and even personal contentment flourish not in the absence of limits, but precisely because of them?
This radical rethinking of freedom and its relationship to our lives is at the heart of author David Epsteinโs work. Drawing on compelling stories, psychological research, and historical examples, Epstein argues that we routinely โovervalue complete freedomโ and โundervalue the power of useful constraints.โ In a world where doing โtoo muchโ has never been easier, learning to embrace limits might be the most liberating act of all.
The Unlikely Genesis of Genius: Creativity Through Constraint
The idea that constraints can spark creativity might seem counterintuitive, yet history is rife with examples. Take the beloved childrenโs author Dr. Seuss. His editor at Random House once made him a bet: could he write a childrenโs book using only 50 distinct words? The result of this seemingly impossible challenge was Green Eggs and Ham, an enduring classic that stands as a testament to the power of imposed boundaries.
Before this, Dr. Seuss had already experienced the generative nature of limits. Tasked with writing a book using only 200 words from a specific vocabulary list for children, he initially complained about the scarcity of adjectives. Frustrated, he eventually decided to simply take the first two rhyming words on the list: โcatโ and โhat.โ From this arbitrary constraint, The Cat in the Hat was born, forcing him to rely on rollicking rhythm and inventive scenarios in place of expansive vocabulary.
Another foundational myth challenged by Epstein is the romanticized origin of the periodic table. Legend has it that Russian chemist Dmitri Mendeleev dreamt up the elegant arrangement of elements after three days of tireless work. While a compelling narrative, the reality was far more prosaic, yet equally illustrative of the power of constraints. Mendeleev was under a tight publishing contract to write a two-volume introductory chemistry textbook. Having only covered eight of the then 63 known elements in volume one, he faced the daunting task of organizing the remaining 55 in a logical, beginner-friendly way for volume two. This practical, commercial deadline forced him to abandon a linear approach and instead think in terms of โfamiliesโ of elements, recognizing the periodic patterns of their properties. It was these โvery tight constraints,โ not boundless freedom, that pushed him into โproductive experimentation.โ
As cognitive scientist Daniel Willingham succinctly puts it, โYou may think that your brain is made for thinking, but your brain is actually made for preventing you from having to think whenever possible because thinking is energetically costly.โ Given complete freedom, our brains tend to default to the โpath of least resistanceโโconvenient, familiar solutions. True creativity, therefore, often arises when we are forced off that path, compelled to explore novel avenues.
This concept extends to artistic innovation as well. Psychologist Patricia Stokes, after studying the history of art, identified a recurring pattern she termed โpaired constraints.โ Innovators first master the status quo, then intentionally block a conventional element (a โpreclude constraintโ), and simultaneously introduce a new mandatory element (a โpromote constraintโ). Claude Monet, for instance, chose to portray impressions of light without using black or dark shades, which were standard at the time. Instead, he forced himself to use โonly pure color in close proximity like a mosaic of color.โ This deliberate restriction led directly to the birth of Impressionism, a revolutionary artistic movement.
Beyond the Canvas: The Societal and Personal Burden of Too Much Freedom
The benefits of constraints arenโt limited to the creative arts. They extend to problem-solving in high-stakes environments. Consider NASAโs Elcross mission, where a team found themselves with half the time and budget they initially expected. While initially an โimpediment,โ this severe limitation forced them to innovate by โborrowing equipment to borrow technologyโโimaging equipment from army tanks, engine temperature sensors from NASCARโultimately leading to the confirmation of water on the moon. This wasnโt the situation they wanted, but it pushed them to think in ways they never would have otherwise.
On a societal level, the โfetishization of total freedomโ is โcompletely ahistorical for the things that have made the freest countries in the world.โ Nobel laureate in economics, Douglas North, highlighted how agreed-upon rules, social norms, and political/economic structures constrain human behavior, making strangers more predictable and fostering trust. Historically, collaboration was limited to kinship networks. It was the establishment of these shared constraints that allowed societies to scale, leading to increased trust and shared prosperity. Ominously, Epstein notes, a breakdown of these shared norms correlates with a decline in trust, mirroring current societal trends.
Individually, the abstract desire for total freedom often clashes with practical reality. While most people surveyed might say theyโd want complete freedom to choose their cancer treatment, those who actually face such a diagnosis overwhelmingly prefer to have fewer, clearer options. The burden of infinite choice can be overwhelming.
Epstein himself experienced this paradox. After leaving his journalism job to become a full-time author, he initially sought โautonomyโ above all else. However, he soon realized that โthere is such a thing as too much autonomy.โ An individualized schedule, while convenient, led to decision fatigue and isolation. He discovered that โhumans thrive in a dense network of reciprocal obligation,โ leading him to intentionally re-introduce inconvenient, time-bound commitments into his lifeโjoining a non-profit board, dance meetups, and a discussion club. These obligations, far from being restrictive, added meaning and structure.
The Tyranny of Choice: Why More Options Make Us Anxious
The modern world is characterized by an unprecedented abundance of choices. What to do with our time, what to pay attention to, what to buyโthe options are seemingly infinite. Yet, this superabundance often has the opposite effect of what we desire: it creates paralysis and anxiety.
Classical economic theory posits that more options are always better, increasing the likelihood of reaching an optimal decision. However, humans have finite cognitive bandwidth. When choice sets become too complex, people make worse decisions or become paralyzed. Studies show that when 401k options surpass a certain complexity, people are more likely to opt out entirely. The advent of โinfinite scrollingโ has even been linked to increased boredom among young people, as the mere possibility of something โbetterโ spoils the experience of the moment. An Oxford professor calculated that modern consumer choices are a โ100 million-fold moreโ than in pre-industrial societies, far outpacing the increase in wealth. This deluge of options often leads people to view every purchase as an extension of their identity, freighting even minor decisions with immense weight.
This situation perfectly illustrates Herbert Simonโs concept of โsatisficingโโa portmanteau of โsatisfyโ and โsuffice.โ Simon, a polymath who won Nobel Prizes in economics and awards in computer science and psychology, argued that humans donโt maximize or optimize by evaluating every option. Instead, we use โgood enough shortcutsโ to make decisions, due to our finite bandwidth. Simon was an โincorrigible satisficerโ himself, wearing the same type of socks, owning one beret, and eating the same breakfast to conserve cognitive energy for what truly mattered. He famously stated, โThe best is the enemy of the good.โ
The problem is exacerbated by whatโs known as โFredkinโs paradoxโ: the more similar your options are, the less choosing between them matters, but the harder choosing between them becomes. This leads us to spend the most time and energy on the least important decisionsโlike endlessly scrolling through Netflix options when theyโre all essentially similar.
This abundance of choice fosters a โfear of closure,โ a deep-seated worry that committing to anything means โkilling off some potential alternative version of yourself.โ This โslidingโ into commitments, particularly evident in relationships where individuals delay making explicit decisions to โkeep options open,โ often leads to less happy outcomes, including higher divorce rates. By refusing to choose, we paradoxically make our lives feel more fragile, never truly grounded.
Structuring Attention in the Digital Wild West
The digital world, with its constant stimulation and endless feeds, is expertly designed to exploit these psychological vulnerabilities. As Epstein notes, โthereโs huge amounts of money to be made by other people structuring your attention.โ AI further exacerbates this, enabling the production of โincredible volumes of things that are like mediocre,โ fragmenting attention in unhealthy ways.
Research by psychologist Gloria Mark illustrates this alarming trend: 25 years ago, office workers switched tasks every 3 minutes; by 2012, it was every 75 seconds; by 2022, it was every 45 seconds. This constant toggling, often mislabeled as multitasking, makes people less effective and more stressed.
In this environment, โitโs just incumbent upon us unfortunately to have to structure our own attention,โ Epstein emphasizes. This requires effort, but itโs โdoableโ and can yield quick results. One powerful tool for this is ritual.
Isabel Allende, the prolific Spanish-language author, offers a remarkable example. Since writing her first bestseller, The House of the Spirits, she has started every subsequent book on January 8th. This date became a powerful ritual: a period of cleansing, clearing old research, and retreating into silence. Even within the workday, she lights a candle to begin and blows it out to end, signaling a shift in mindset. This ritual provides structure, discipline, and a sacred space for her creative work, allowing her to consistently produce bestsellers for decades.
Embracing Intentional Limits: A Path to Personal Transformation
Epstein himself has profoundly reshaped his life by applying the principles explored in his book. Recognizing his own tendency to โdo too much,โ he adopted a โPixar wayโ of staying small longer, dedicating a full year to research before forcing himself to create a one-page structural outline for his latest book. This ruthless prioritization allowed him to be โmuch more efficient with my timeโ and even turn the book in early.
He also implemented rituals to contain his workday, like flipping off an electric candle and closing his office door. In his personal life, he reined in excessive autonomy by joining community groups that required his physical presence at specific times, combating the โforever elsewhereโ feeling of living purely in the virtual world.
Crucially, he adopted โsatisficing rulesโ for everyday decisions: identifying the few essential criteria, and once met, stopping the search and committing. This counters the maximizerโs endless quest for the โbest,โ which often leads to regret and unhappiness. He also engages in a โsubtraction audit,โ regularly reviewing his commitments and forcing himself to cut what isnโt essential.
The journey towards a more creative, productive, and fulfilling life, it seems, is not about breaking free from all shackles, but about intelligently choosing which ones to embrace. By imposing thoughtful limitsโwhether on our creative process, our choices, or our attentionโwe can transform the anxiety of endless possibility into the focused freedom that truly matters. In a world constantly vying for our attention, the discipline of saying โnoโ and the courage to define our own boundaries may be the most powerful acts of self-liberation imaginable.
ํ๊ตญ์ด
โWhy this Claude Code engineer uses HTML files as AI specs | Thariq Shihipar (Anthropic)โ โ How I AI ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI ์๋ ๊ธฐํ ๋ฌธ์์ ํ๋ช : ์คํธ๋กํฝ ์์ง๋์ด๊ฐ ๋งํฌ๋ค์ด ๋์ HTML์ ์ ํํ ์ด์
์ธ๊ณต์ง๋ฅ(AI) ์์ด์ ํธ๊ฐ ์ฐ๋ฆฌ์ ์์ ๋ฐฉ์์ ํ์ ์ ๊ฐ์ ธ์ค๋ฉด์, AI์ ์ํตํ๋ ๋ฐฉ์ ๋ํ ๋์์์ด ์งํํ๊ณ ์์ต๋๋ค. ๊ณผ๊ฑฐ์๋ ๋งํฌ๋ค์ด(Markdown)์ด AI ์์ด์ ํธ์์ ์ํธ์์ฉ์ ์์ด ํ์ค์ฒ๋ผ ์ฌ๊ฒจ์ก์ง๋ง, ๊ทธ ํ๊ณ๊ฐ ์ ์ฐจ ๋๋ฌ๋๊ณ ์์ต๋๋ค. ์คํธ๋กํฝ(Anthropic)์ ํด๋ก๋ ์ฝ๋(Claude Code) ํ ์์ง๋์ด ํ๋ฆฌํฌ ์ฌํํ๋ฅด(Thariq Shihipar)๋ ๋งํฌ๋ค์ด์ ๋์์ผ๋ก HTML ํ์ผ์ AI ์คํ(AI specs)์ผ๋ก ํ์ฉํ๋ ํ์ ์ ์ธ ์ ๊ทผ ๋ฐฉ์์ ์ ์ํ๋ฉฐ, AI ์๋์ ๊ธฐํ ๋ฐ ํ์ ๋ฐฉ์์ ๋ํ ์๋ก์ด ๋น์ ์ ์ ์ํฉ๋๋ค.
๋งํฌ๋ค์ด์ ํ๊ณ์ โ์ปดํจํธ ํ ๋น์โ์ ๋ฑ์ฅ
๋งํฌ๋ค์ด์ ๊ฐ๊ฒฐํ๊ณ ์์ฑํ๊ธฐ ์ฌ์ AI ์์ด์ ํธ, ํนํ ํด๋ก๋ 3.5(Claude 3.5)๋ GPT-4(Opus 4)์ ๊ฐ์ ๋ชจ๋ธ๊ณผ์ ์ํต์ ๋งค์ฐ ์ธ๊ธฐ๋ฅผ ๋์์ต๋๋ค. ์์ด์ ํธ๊ฐ ์์ฑํ๋ ๊ณํ์๋ ์ฝ๋ ์ด์์ ์ฌ๋์ด ์ฝ๊ณ ์์ ํ๋ ๋ฐ ์ฉ์ดํ๊ธฐ ๋๋ฌธ์ ๋๋ค. ํ์ง๋ง ์์ด์ ํธ์ ์คํ ์๊ฐ์ด ๊ธธ์ด์ง๊ณ (์: 4.5๋ 4.7 ๋ฒ์ ์ ํ ์๊ฐ ์ด์ ์คํ ๊ฐ๋ฅ) ๊ณํ์์ ๋ถ๋์ด ์๋ฐฑ, ์์ฒ ์ค์ ๋ฌํ๊ฒ ๋๋ฉด์ ๋ฌธ์ ๊ฐ ๋ฐ์ํ์ต๋๋ค.
ํ๋ฆฌํฌ๋ โ๊ณํ์๊ฐ ๋๋ฌด ๊ธธ์ด์ ธ์ ์์งํ ๋ ์ด์ ์ฝ์ง ์๊ฒ ๋์๋คโ๊ณ ๊ณ ๋ฐฑํ๋ฉฐ, ์ด๋ โ์ค์์๋คโ๊ณ ๊ฐ์กฐํฉ๋๋ค. AI๊ฐ ์๋ฌด๋ฆฌ ๋๋ํด์ ธ๋ ์ธ๊ฐ์ ์ฌ์ ํ โ๋ฃจํ ์์(in the loop)โ ๋จธ๋ฌผ๋ฉฐ ์์ด์ ํธ๊ฐ ๋ฌด์์ ํ๊ณ ์๋์ง ๋ช ํํ ์ดํดํด์ผ ํ๋ค๋ ๊ฒ์ ๋๋ค. ์์ฒ ์ค์ง๋ฆฌ ๋งํฌ๋ค์ด ํ์ผ์ ๋ ์ด์ ์ฌ๋์ด ์ง์ ํธ์งํ๊ธฐ ์ด๋ ต๊ณ , ๊ฒฐ๊ตญ ํด๋ก๋์๊ฒ ํธ์ง์ ์์ฒญํ๊ฒ ๋๋ ์ํฉ์ ์ด๋ฆ ๋๋ค.
ํด๋ผ๋ผ๋ฒ ์ค(Claraveo)๋ ์ด๋ฌํ ํ์์ โํด๋ก๋๊ฐ 8์๊ฐ ๋์ ์คํ๋ ์ ์๋ค๊ณ ๋งํ๋ ๊ฒ์, ํด๋ก๋๊ฐ 500๋ฌ๋ฌ๋ฅผ ์ธ ์ ์๋ค๊ณ ๋งํ๋ ๊ฒ๊ณผ ๊ฐ๋คโ๋ ๋น์ ๋ก ์ค๋ช ํฉ๋๋ค. ์ฐ๋ฆฌ๋ ๋ชจ๋ โ์ปดํจํธ ํ ๋น์(compute allocators)โ๊ฐ ๋์ด๊ฐ๊ณ ์์ผ๋ฉฐ, ์ ํ๋ ์ปดํจํ ์์์ ์ด๋์ ๊ฐ์ฅ ๊ฐ์น ์๊ฒ ์ฌ์ฉํ ์ง ๊ฒฐ์ ํด์ผ ํฉ๋๋ค. ์ด ๊ฒฐ์ ์ ๋ฐ๋ก ๊ธฐํ(planning) ๋จ๊ณ, ์ฆ PRD(Product Requirements Document, ์ ํ ์๊ตฌ์ฌํญ ๋ฌธ์)์ ์คํ(spec) ์์ฑ ๋จ๊ณ์์ ์ด๋ฃจ์ด์ง๋๋ค. โ์ ํ ๊ด๋ฆฌ(product management)๋ ์ฃฝ์๋คโ๊ณ ๋งํ๋ ํด๋ผ๋ผ๋ฒ ์ค๊ฐ ์ด์ ๋ โ๋น์ ์ ์ปดํจํธ ํ ๋น์๋ค, ๊ทธ๊ฒ ์ด์ ์ง์ ์ด๋คโ๋ผ๊ณ ์ ์ธํ๋ ์ด์ ์ ๋๋ค.
HTML: AI ์์ด์ ํธ์์ ๋ ํ๋ถํ ๋ํ ์๋จ
ํด๋ก๋ ์ฝ๋ ํ์์ HTML ํ์ผ์ ์คํ ๋ฌธ์๋ก ์ฌ์ฉํ๊ธฐ ์์ํ ๊ฒ์ ์ด๋ฌํ ๋งํฌ๋ค์ด์ ํ๊ณ๋ฅผ ๊ทน๋ณตํ๊ธฐ ์ํจ์ ๋๋ค. HTML์ ๋ชจ๋ธ์ด ์ฌ์ ํ ์ ์ดํดํ๋ฉฐ, ํจ์ฌ ๋ ๋ง์ ์ปจํ ์คํธ(context)๋ฅผ ์ฒ๋ฆฌํ ์ ์๋๋ก ๋ ๋ง์ ํ ํฐ(token)์ ์๋นํ ์ฌ์ง๊ฐ ์์ต๋๋ค. ๋ฌด์๋ณด๋ค HTML์ ์ฌ๋์๊ฒ ํจ์ฌ ์ฝ๊ธฐ ์ฝ๊ณ , ์คํฌ๋กค์ด ์ฉ์ดํ๋ฉฐ, ํ๋ถํ ์ ๋ณด๋ฅผ ๋ด์ ์ ์์ต๋๋ค.
๋งํฌ๋ค์ด์ด ASCII ์ํธ๋ฅผ ์ด์ฉํด ์๊ฐ์ ์ธ ์์๋ฅผ ํํํ๋ ค ์ ์ฐ๋ ๋ฐ๋ฉด, HTML์ ์ค์ ๊ทธ๋ํฝ๊ณผ UI ์์๋ฅผ ์ง์ ๊ทธ๋ฆด ์ ์์ต๋๋ค. ์ด๋ ์ธ๊ฐ๊ณผ ํด๋ก๋ ๊ฐ์ โ๋ ํ๋ถํ ์ปค๋ฎค๋์ผ์ด์ ๋งค์ฒด(richer communication medium)โ๊ฐ ๋ฉ๋๋ค. ํ๋ฆฌํฌ๋ โHTML์์๋ ๊ทธ๋ ๊ฒ ์ ์ธ ํ์๊ฐ ์๋ค. ์ค์ ๋ก ๋ณผ ์ ์๋ ๊ฒ์ ๊ทธ๋ฆด ์ ์๋คโ๊ณ ๋งํฉ๋๋ค.
ํด๋ก๋ ์ฝ๋ ๋ฐ๋ชจ: HTML๋ก ์์ด๋์ด ๋ธ๋ ์ธ์คํ ๋ฐ ๋ฐ ๊ณํ ์๋ฆฝ
ํ๋ฆฌํฌ๋ ํด๋ก๋ ์ฝ๋์์ HTML์ ํ์ฉํ๋ ๊ณผ์ ์ ์์ฐํ์ต๋๋ค. ๊ทธ๋ ํด๋ก๋์๊ฒ โํด๋ผ๋ผ ํ์บ์คํธ์์ ๋ฐ๋ชจ๋ฅผ ํ๊ณ ์ถ์ผ๋, HTML ํ์ผ๋ก ์์ด๋์ด๋ฅผ ๋ธ๋ ์ธ์คํ ๋ฐํด๋ฌ๋ผโ๋ ๊ฐ๋จํ ํ๋กฌํํธ(prompt)๋ฅผ ์ฃผ์์ต๋๋ค. ๊ทธ ๊ฒฐ๊ณผ, ํด๋ก๋๋ ๋จ์ํ ํ ์คํธ ๋ชฉ๋ก์ด ์๋, ์๊ฐ์ ์ธ ๋ชฉ์ (mockup)๊ณผ ๊ฐ ์์ด๋์ด์ ๋ํ ์ค๋ช , ๊ทธ๋ฆฌ๊ณ ๋ฐ์ํ ์ ์๋ โ์ํ ์์(risk)โ๊น์ง ํฌํจ๋ 8๊ฐ์ง ๋ฐ๋ชจ ์์ด๋์ด๋ฅผ HTML ํ์ผ๋ก ์์ฑํ์ต๋๋ค.
์๋ฅผ ๋ค์ด, โํ์ดํธ๋ณด๋ ์ค์ผ์น์์ ์๋ํ๋ UI(whiteboard sketch to working UI)โ์ ๊ฐ์ ์์ด๋์ด๋ ๊ท์ฌ์ด ์ค์ผ์น์ ํจ๊ป ์ ์๋์ด, ํด๋ผ๋ผ๋ฒ ์ค๋ฅผ ๊ฐํํ๊ฒ ๋ง๋ค์์ต๋๋ค. ํ๋ฆฌํฌ๋ โ์ด๊ฒ์ ํด๋ก๋์๊ฒ ๋งํฌ๋ค์ด ๋ฌธ์ ๋์ , ์ ๋ณด๋ฅผ ๊ฐ์ฅ ์ ์ ๋ฌํ์ฌ ์ฌ์ฉ์๊ฐ ์ค์ ๋ก ์ฐธ์ฌํ๊ณ ์ ํํ ์ ์๋ ๋ฐฉ๋ฒ์ ์ฐพ๋๋ก ํ ๊ฒโ์ด๋ผ๊ณ ์ค๋ช ํฉ๋๋ค. HTML ๋๋ถ์ ์ฌ์ฉ์๋ ๋ชจ๋ ์์ด๋์ด๋ฅผ ์คํฌ๋กคํ๋ฉฐ ํ์ธํ ์ ์๊ณ , ๋ค์ด์ด๊ทธ๋จ์ ํจ์ฌ ๋ ๋ง์ ๊ฒ์ ์ฐ์์์ผ ์์ด๋์ด๋ฅผ ์ฝ๊ฒ ์ดํดํ๋๋ก ๋์ต๋๋ค.
๋ธ๋ ์ธ์คํ ๋ฐ์ ํตํด โCSV ํ์ผ์ ์ธํฐ๋ํฐ๋ธ ๋์๋ณด๋๋ก ๋ณํ(CSV to interactive dashboards)โํ๋ ์์ด๋์ด๋ฅผ ์ ํํ ํ, ํ๋ฆฌํฌ๋ ํด๋ก๋์๊ฒ ์ด ์์ด๋์ด๋ฅผ ๋ฐํ์ผ๋ก โ์ต๋ ์ปจํ ์คํธ(maximum context)๋ฅผ ์ ๊ณตํ๋ HTML ๊ณํ ํ์ผ(HTML file as a plan)์ ์์ฑํด๋ฌ๋ผ. ๊ตฌํ ๊ณํ์ ์๊ฐํํ๋ ๋ฐ ๋์์ด ๋๋๋ก ๋ฐ์ท๋ฌธ, ๋ชฉ์ , ์ฝ๋ ๋ฑ์ ํฌํจํ๋ผโ๊ณ ์์ฒญํ์ต๋๋ค.
ํด๋ก๋๋ ํ์บ์คํธ ์คํฌ๋ฆฝํธ ์ด์, ํ์ผ ์์คํ ๊ตฌ์กฐ, ์คํฌ(skill) ์์, ๋ฌด๋ ๋ณด๋(mood board), ์์ ์ปดํฌ๋ํธ, ๋ก์ง, ํต์ฐฐ๋ ฅ, ํ ํ๋ฆฟ, ํฌํผ ์คํฌ๋ฆฝํธ ๋ฑ ๋ฐฉ๋ํ ์ ๋ณด๋ฅผ ๋ด์ HTML ๊ณํ ํ์ผ์ ์์ฑํ์ต๋๋ค. ํ๋ฆฌํฌ๋ โ์ด๊ฒ์ ๋ด๊ฐ ์ค์ ๋ก ์ฝ์ ๊ณํ์ด๋คโ๋ผ๊ณ ๊ฐ์กฐํ๋ฉฐ, HTML์ด ์ ๊ณตํ๋ ํ๋ถํ ์ ๋ณด์ ์๊ฐํ ๋๋ถ์ ๊ณํ์ ๋ ๊น์ด ์ดํดํ๊ณ ์ฐธ์ฌํ ์ ์์์ ๋ณด์ฌ์ฃผ์์ต๋๋ค.
ํ๋กฌํํธ ์์ง๋์ด๋ง์ ์งํ: โ์ ๋ขฐโ์ โ์ฌ์งโ
ํด๋ก๋์ ๋๋ผ์ด ๊ฒฐ๊ณผ๋ฌผ์๋ ๋ถ๊ตฌํ๊ณ , ํ๋ฆฌํฌ์ ํ๋กฌํํธ๋ ๋งค์ฐ ๊ฐ๋จํ์ต๋๋ค. ์ด๋ ํ๋กฌํํธ ์์ง๋์ด๋ง(prompt engineering)์ ํต์ฌ ์์น์ ๋ณด์ฌ์ค๋๋ค. ์ฆ, โ์ํ๋ ๊ฒ์ ์ป์ ๋งํผ ์ถฉ๋ถํ ์ ๋ณด๋ฅผ ์ฃผ๋, ํด๋ก๋๋ฅผ ๊ณผ๋ํ๊ฒ ์ ์ฝํ์ง ์๋ ๊ฒโ์ ๋๋ค. ํ๋ฆฌํฌ๋ ๊ณผ๋ํ๊ฒ ๋ณต์กํ โ์คํฌ(skills)โ์ด๋ โ๋น์ ์ ์ ๋ฌธ๊ฐ ๊ธฐํ์๋คโ์ ๊ฐ์ ์ง๋์น ์ญํ ๋ถ์ฌ๊ฐ ์คํ๋ ค ์์ด์ ํธ๋ฅผ ์ ์ฝํ ์ ์๋ค๊ณ ์ง์ ํฉ๋๋ค.
๊ทธ๋ ๋ํ ํด๋ก๋์๊ฒ ํญ์ โ์ฌ์ง(an out)โ๋ฅผ ์ฃผ์ด์ผ ํ๋ค๊ณ ๋งํฉ๋๋ค. โ์ต๋ ์ปจํ ์คํธ๋ฅผ ์ํด ํ์ํ ๋ชจ๋ ๊ฒ(whatever is needed to give me maximum context)โ๊ณผ ๊ฐ์ ๋ฌธ๊ตฌ๋ ํด๋ก๋์๊ฒ ์์จ์ฑ์ ๋ถ์ฌํ๊ณ , โ๋๋ ๋๋ฅผ ๋ฏฟ๋๋ค(I trust you)โ๋ ๊ธ์ ์ ์ธ ํ๋๋ ๋ ๋์ ๊ฒฐ๊ณผ๋ฌผ์ ์ด๋์ด๋ ๋๋ค. ํ๋ฆฌํฌ๋ ์ฌ์ง์ด โ์ค์ํด๋ผ, ํด๋ก๋(make mistakes Claude)โ์ ๊ฐ์ด ์์ด์ ํธ์ ์ ์ฐ์ฑ์ ๋ ๋ คํ๋ ์ ๊ทผ ๋ฐฉ์๋ ์ธ๊ธํ๋ฉฐ, ์ธ๊ฐ์ ๊ฐ์ ์ ํ๋๊ฐ AI ๋ชจ๋ธ์ ์ถ๋ก (reasoning)์ ์ํฅ์ ๋ฏธ์น ์ ์๋ค๋ ์ฐ๊ตฌ ๊ฒฐ๊ณผ๋ ํฅ๋ฏธ๋กญ๊ฒ ์ธ๊ธํ์ต๋๋ค.
โ๋ง์ดํฌ๋ก ์ํํธ์จ์ดโ์ ์๋: HTML์ ํตํ ๋ง์ถคํ ์ํธ์์ฉ
๋งํฌ๋ค์ด์ ์ฅ์ ์ค ํ๋๋ ์ ๊ทผ์ฑ(accessibility)๊ณผ ์ฌ์ด ํธ์ง์ ๋๋ค. ์ด์ ๋ํด ํ๋ฆฌํฌ๋ โ๋งํฌ๋ค์ด์ ์ฝ์ง ์๊ฒ ๋์ ํธ์ง๋ ๋ฉ์ท๋คโ๋ฉฐ, ๊ฒฐ๊ตญ ํด๋ก๋์๊ฒ ํธ์ง์ ์์ฒญํ๊ฒ ๋์๋ค๊ณ ๋งํฉ๋๋ค. ํ์ง๋ง HTML์ ์ฌ๊ธฐ์ ํ ๋จ๊ณ ๋ ๋์๊ฐ๋๋ค. ์ฌ์ฉ์๊ฐ ํน์ ๋ถ๋ถ์ ์ง์ ๊น์ด ์๊ฒ ํธ์งํ๊ณ ์ถ์ ๋, ํด๋ก๋์๊ฒ โ์ผํ์ฉ UI(throwaway UI)โ๋ฅผ ๋ง๋ค๋๋ก ์์ฒญํ ์ ์์ต๋๋ค.
CSV-๋์๋ณด๋ ๋ฐ๋ชจ์์ ํ๋ฆฌํฌ๋ ์๊ฐํ ๊ท์น์ ์ ์ํ๋ ํ๊ฐ ๋ง์์ ๋ค์ง ์์, ํด๋ก๋์๊ฒ โ๊ฒฐ์ ๊ท์น์ ์ ์ํ๋ ๋ฐ ๋์์ด ๋๋ ํธ์ง ๊ฐ๋ฅํ HTML ์ํฐํฉํธ(editable HTML artifact)๋ฅผ ๋ง๋ค์ด๋ฌ๋ผ. ํ์ฌ ๊ท์น์ด ๋ง์์ ๋ค์ง ์์ผ๋, ๊ตฌ์กฐ๋ ์ ์งํ๋ ์ ์ฐ์ฑ์ ์ ๊ณตํ๋ ๋ง์ถคํ UI(custom UI)๋ฅผ ๋์์ธํด๋ฌ๋ผโ๊ณ ์์ฒญํ์ต๋๋ค.
ํด๋ก๋๋ ์ฌ์ฉ์๊ฐ ๊ฐ ํ๋๋ฅผ ํธ์งํ๊ณ , ์จ๊ธฐ๊ฑฐ๋, ๋ณต์ฌํ๊ณ , ์๋ก์ด ํ๋๋ฅผ ์ถ๊ฐํ ์ ์๋ ์๋ฆ๋ค์ด ์ปค์คํ ์ธํฐํ์ด์ค๋ฅผ ์์ฑํ์ต๋๋ค. ์ฌ์ฉ์๋ ์ด UI์์ ์ํ๋ ๋๋ก ๊ท์น์ ์กฐ์ ํ ํ, ์์ฑ๋ ๋งํฌ๋ค์ด์ ๋ณต์ฌํ์ฌ ๋ค์ ํด๋ก๋ ์ฝ๋์ ๋ถ์ฌ๋ฃ์ ์ ์์ต๋๋ค. ํด๋ผ๋ผ๋ฒ ์ค๋ ์ด๋ฅผ โ๋ง์ดํฌ๋ก ์ํํธ์จ์ด ์์ ๋ง์ดํฌ๋ก ์ํํธ์จ์ดโ๋ผ๊ณ ํํํ๋ฉฐ, โ๋งค์ฐ ๊ฐ์ธํ๋ ๊ณํ์์ ํน์ ๋ชจ๋์ ํ๋ํ์ฌ ๋งค์ฐ ๋ง์ถคํ UI๋ฅผ ์ฌ์ฉํ์ฌ ์ฝํ ์ธ ์ ์ํธ์์ฉํจ์ผ๋ก์จ ๋ ๋์ ํ์ง์ ๋๋ฌํ๋ ๊ฒโ์ด๋ผ๊ณ ์ค๋ช ํฉ๋๋ค. ์ด๋ ์ฌ์ฉ์ ์ฐธ์ฌ๋ฅผ ๊ทน๋ํํ๊ณ ์ต์ข ๊ฒฐ๊ณผ๋ฌผ์ ํ์ง์ ํฅ์์ํค๋ ํ์ ์ ์ธ ๋ฐฉ์์ ๋๋ค.
ํ์ ๊ณผ โ์ด์์๋โ ๋ฌธ์์ ๊ฐ์น
HTML ๊ธฐ๋ฐ์ AI ์คํ์ ๊ฐ์ธ์ ์์ ํจ์จ์ฑ์ ๋์ด ํ์ (collaboration) ๋ฐฉ์์๋ ๊ธ์ ์ ์ธ ์ํฅ์ ๋ฏธ์นฉ๋๋ค. HTML ํ์ผ์ AWS์ ๊ฐ์ ํด๋ผ์ฐ๋ ํ๊ฒฝ์ ์ ๋ก๋ํ์ฌ ๋งํฌ๋ฅผ ๊ณต์ ํ๊ธฐ ์ฝ๊ธฐ ๋๋ฌธ์, ํ์๋ค์ด ๋ฌธ์๋ฅผ ์ฝ์ ๊ฐ๋ฅ์ฑ์ด ํจ์ฌ ๋์์ง๋๋ค. ํ๋ฆฌํฌ๋ ๋งค์ฃผ HTML ํ์์ผ๋ก ์ฃผ๊ฐ ์ ๋ฌด ํํฉ ์ ๋ฐ์ดํธ๋ฅผ ์์ฌ์๊ฒ ๋ณด๋ธ๋ค๊ณ ํฉ๋๋ค. ํด๋ก๋๊ฐ ๊ทธ์ ์ฌ๋(Slack) ๋ฉ์์ง๋ฅผ ์ฝ๊ณ ์๋์ผ๋ก ์์ฝํ์ฌ HTML ๋ณด๊ณ ์๋ฅผ ์์ฑํ๋ ๋ฐฉ์์ ๋๋ค. ์ด๋ โํ์ฌ๋ฅผ ๋์ด ์ํตํ๋ ํจ์ฌ ๋ ํจ๊ณผ์ ์ธ ๋ฐฉ๋ฒโ์ด๋ฉฐ, โ๋งค๋ ฅ์ ์ด๊ณ ์ ์ ํ ์์ค์ ์ธ๋ถ ์ ๋ณด๋ฅผ ๋ด์ ์๋น ๊ฐ๋ฅํ ์ฝํ ์ธ โ์ ์ค์์ฑ์ ๊ฐ์กฐํฉ๋๋ค.
ํ๋ฆฌํฌ๋ โ์์ฐ ์ฝ๋์ ๋ค์ด๊ฐ๋ ํ ํฐ์ ์์ ๊ทนํ ์ ์ง๋ง, ๋์๋ณด๋, ์ปค์คํ ์ธํฐํ์ด์ค์ ๊ฐ์ ๋ ๋ง์ ํ ํฐ์ ์์ฑํ๊ณ ์๋คโ๋ฉฐ, AI ์๋์ โํ ํฐ ํ์๋ก์(abundance of tokens)โ์ ์ธ๊ธํฉ๋๋ค. ๊ทธ๋ ์์ ์ด ์ํธ์์ฉํ๋ ๋ชจ๋ ๊ฒ์ด โ์๋ฆ๋ต๊ณ โ โ๋ ๋ฃจํ ์์โ ์์ผ๋ฉฐ, ์ด๊ฒ์ด ์ต์ข ์ ํ์ ํ์ง๋ก ์ด์ด์ง๊ธฐ๋ฅผ ํฌ๋งํฉ๋๋ค.
์ด๋ฌํ ์ ๊ทผ ๋ฐฉ์์ โ์ด์์๋ ๋์์ธ ์์คํ (Living Design System)โ ๊ตฌ์ถ์๋ ์ ์ฉ๋ฉ๋๋ค. ํ๋ฆฌํฌ๋ GitHub ๋ฆฌํฌ์งํ ๋ฆฌ์์ ๋์์ธ ์์คํ ์ ์ถ์ถํ๋ ํด๋ก๋ ๋์์ธ(Claude Design)์ ๊ธฐ๋ฅ์ ํ์ฉํ์ฌ, ์์, ํ์ดํฌ๊ทธ๋ํผ, ๊ฐ๊ฒฉ, ์ปดํฌ๋ํธ ๋ฑ ๋์์ธ ์์คํ ์ ๋ชจ๋ ์์๋ฅผ ๋ด์ HTML ํ์ผ์ ์์ฑํฉ๋๋ค. ์ด HTML ํ์ผ์ ์๋ก์ด ํ๋ก์ ํธ์ ์ฝ๊ฒ ์ฐธ์กฐํจ์ผ๋ก์จ, ๋์์ธ ์์คํ ์ โdesign.mdโ ๋์ โdesign.htmlโ๋ก ์ธ์ฝ๋ฉํ์ฌ ์ธ์ ๋ ์ฐธ์กฐํ ์ ์๋ โ์์ถ๋ ์ดํด(compressed understanding)โ๋ฅผ ์ ๊ณตํฉ๋๋ค.
ํด๋ผ๋ผ๋ฒ ์ค๋ ์ฌ๊ธฐ์ ๋ํด, ๋ง์ผํฐ์ ๋์์ด๋๋ฅผ ์ํ โ์ปดํฌ๋ํธ ์๊ฐํ ํ์ด์ง(component visualization page)โ๋ฅผ ๋ง๋๋ ๊ณ ๊ธ ํ์ฉ๋ฒ์ ์๊ฐํฉ๋๋ค. ์ฑ์ 25๊ฐ ์ฃผ์ ์ปดํฌ๋ํธ๊ฐ ์ค์ ๋ก ์๋ํ๋ ๋ชจ์ต์ ๋ณด์ฌ์ฃผ๋ ์ธํฐ๋ํฐ๋ธ(interactive) HTML ํ์ด์ง๋ฅผ ๋ง๋ค๊ณ , ์ฌ๊ธฐ์ ํฌ๋ช PNG ์ด๋ฏธ์ง๋ฅผ ๋ค์ด๋ก๋ํ์ฌ ๋ง์ผํ ์๋ฃ๋ ๋น๋์ค์ ํ์ฉํ๋ ๋ฐฉ์์ ๋๋ค. ์ด๋ ์ฝ๋๋ฟ๋ง ์๋๋ผ ๋ง์ผํฐ, ๋์์ด๋์๊ฒ๋ โ์ค์ ์ ๊ฐ์ ์ฑ ๋ฒ์ โ์ ์ ๊ณตํ์ฌ ํ์ ์ ํจ์จ์ฑ์ ๊ทน๋ํํฉ๋๋ค.
HTML ๊ธฐ๋ฐ์ ์ ๊ทผ ๋ฐฉ์์ โjust-in-timeโ ๋ฌธ์ํ์ ์ ๋ ดํ โ์ผํ์ฉ ์ํํธ์จ์ด(throwaway software)โ๋ฅผ ๊ฐ๋ฅํ๊ฒ ํฉ๋๋ค. ๊ณผ๊ฑฐ์๋ ์คํ์ด๋ PRD์ โ๋จ์ผ ์ง์ค ์์ฒ(source of truth)โ๊ณผ ์ค์ ์ง์ค์ ์ ์ฅ์, ํต์ผ๋ ํ์์ ๋ํ ์ง์ฐฉ์ด ์ฌํ์ง๋ง, AI ์๋์๋ ํด๋ก๋๊ฐ ํ์ํ ์ปจํ ์คํธ๋ฅผ ์ค์ค๋ก ์ฐพ์๋ผ ์ ์๊ธฐ ๋๋ฌธ์ ์ด๋ฌํ ์ ์ฝ์ด ์ค์ด๋ญ๋๋ค. ํต์ฌ์ โ๊ณํ์ ๋ด์ฉ์ด ์ข์ ์์ด๋์ด์ธ๊ฐ? ์ ์คํ๋ ๊ฒ์ธ๊ฐ?โ์ ๊ฐ์ ๋ณธ์ง์ ์ธ ์ง๋ฌธ์ ์ง์คํ๋ ๊ฒ์ ๋๋ค.
๋ฏธ๋๋ฅผ ํฅํ ์์ฌ์ : ๋ ๋์ ์ ํ, ๋ ๊น์ ํต์ฐฐ
HTML ๊ธฐ๋ฐ ๊ณํ์ ๋ ๋์ ์ ํ์ผ๋ก ์ด์ด์ง ์ ์์ต๋๋ค. ํ๋ฆฌํฌ๋ HTML ๊ณํ์ ์ํฐํฉํธ(artifact)๋ก ์ฌ์ฉํ์ฌ โ๊ณํ์ด ์์ผ๋ ๊ตฌํํด๋ฌ๋ผโ๊ณ ํด๋ก๋์๊ฒ ์์ฒญํ ์ ์๋ค๊ณ ๋งํฉ๋๋ค. ๋ํ, HTML ๊ณํ์ ํฌํจ๋ ๋ชฉ์ ์ ํ์ฉํ์ฌ ๊ฒ์ฆ ์์ด์ ํธ(verification agent)๊ฐ โ์๋ํ๋ ๊ฒ๊ณผ ์ค์ ๊ฒฐ๊ณผ๋ฌผ์ด ์ผ์นํ๋์งโ ํ์ธํ ์ ์์ต๋๋ค.
ํ ์คํธ์ ๊ฒ์ฆ(testing and verification)์ AI ์๋์ ๋์ฑ ์ค์ํด์ง๋๋ค. ์คํธ๋กํฝ์ โ์์์ปด์ฆ(Outcomes)โ ๊ธฐ๋ฅ์ฒ๋ผ ๋ชฉํ ์งํฅ์ ์ธ ์ ๊ทผ ๋ฐฉ์๊ณผ ํจ๊ป, ๋ฃจ๋ธ๋ฆญ(rubric), ํด๋ก๋๊ฐ ์ํํ ์์ ์ ๋น๋์ค๋ก ๋ นํํ๋ ๋ฐฉ์, ํฉ์ฑ ๋ฐ์ดํฐ(synthetic data)๋ฅผ ์ด์ฉํ ํ ์คํธ ๋ฑ ๋ค์ํ ๊ฒ์ฆ ๋ฉ์ปค๋์ฆ์ด ๊ฐ๋ฅํด์ก์ต๋๋ค.
Stripe ํ์ ์ฌ๋ก๋ HTML์ ์ ์ฌ๋ ฅ์ ์ ๋ณด์ฌ์ค๋๋ค. ๊ทธ๋ค์ ์์ ๋ค๋ง์ โ๋ฐ์ด๋ธ ์ฝ๋ฉ(vibe coding)โ ํ๋ซํผ์ ๊ตฌ์ถํ๋๋ฐ, HTML์ ํตํด ํ์ด ์ ํ ๋ฆฌ๋ทฐ, ๋์์ธ ๋ฆฌ๋ทฐ, ์คํ ๋ฆฌ๋ทฐ๋ฅผ ์งํํ๋ ๋ฐฉ์์ ๋๊ตฌ ์์ฒด์ ๋ฐ์ํ ์ ์์๋ค๊ณ ํฉ๋๋ค. ์ด๋ ํ์ ์์ ๋ฐฉ์์ ๋ง์ถฐ ๋๊ตฌ๋ฅผ ๋ง์ถคํํ ์ ์๋ ๊ฐ๋ ฅํ ์ด์ ์ ๋๋ค.
๋ฒ๊ฐ ์ง๋ฌธ (Lightning Round) ์์ฝ
- ๊ฐ์ฅ ์ข์ํ๋ ํด๋ก๋ ๋ฐ์คํฌํฑ ์ฑ ํญ: ํ๋ฆฌํฌ๋ ์ฃผ์ ์์ด โ์ฝ๋(Code)โ ํญ์ ๊ผฝ์์ต๋๋ค.
- ๊ฐ์ฅ ๊ธฐ๋๋๋ ์์: ์คํธ๋กํฝ๊ณผ ์คํ์ด์คX(SpaceX)์ ํํธ๋์ญ, ๊ทธ๋ฆฌ๊ณ โ๊ถค๋ ๋ฐ์ดํฐ ์ผํฐ(orbital data centers)โ์ ๋ํ ๊ตฌ์. ๊ณต์๊ณผํ ๊ฐ์ง๋ง ํ์ค์ด ๋ ์ ์๋ ๋ฏธ๋์ ๋ํ ๊ธฐ๋๋ฅผ ๋๋ฌ๋์ต๋๋ค.
- ํด๋ก๋๊ฐ ๋ง์ ๋ฃ์ง ์์ ๋์ ํ๋กฌํํธ ๊ธฐ์ : ํ๋ฆฌํฌ๋ ์๋ฆฌ์น์ง ์๋๋ค๊ณ ๋งํฉ๋๋ค. ๋์ , ํด๋ก๋์ ๊ฐ์ ์ ์ธก๋ฉด์ ๋ํ ์ฐ๊ตฌ๋ฅผ ์ธ๊ธํ๋ฉฐ, โ์น์ ํ๊ณ ์๋ฅํ๊ฒโ ๋ํ๋ ๊ฒ์ด ๋ ๋์ ๊ฒฐ๊ณผ๋ฌผ์ ์ป๋ ๋ฐ ๋์์ด ๋ ๊ฒ์ด๋ผ๋ ๋ฏฟ์์ ๊ฐ์ง๊ณ ์์ต๋๋ค.
๊ฒฐ๋ก
์คํธ๋กํฝ ์์ง๋์ด ํ๋ฆฌํฌ ์ฌํํ๋ฅด์ HTML ๊ธฐ๋ฐ AI ์คํ ํ์ฉ๋ฒ์ AI ์๋์ ๊ธฐํ ๋ฐ ๊ฐ๋ฐ ํจ๋ฌ๋ค์์ด ์ด๋ป๊ฒ ๋ณํํ๊ณ ์๋์ง๋ฅผ ๋ช ํํ ๋ณด์ฌ์ค๋๋ค. ๋งํฌ๋ค์ด์ ํ๊ณ๋ฅผ ๋์ด, HTML์ ๊ฐ๋ ์ฑ, ์๊ฐํ, ์ธํฐ๋ํฐ๋ธํ ์ํธ์์ฉ์ ํตํด ์ธ๊ฐ์ ๊ฐ์ ์ ๋์ฑ ํจ๊ณผ์ ์ผ๋ก ๋ง๋ค๊ณ , ๊ถ๊ทน์ ์ผ๋ก ๋ ๋์ ํ์ง์ ์ ํ์ ๋ง๋ค์ด๋ด๋ ๋ฐ ๊ธฐ์ฌํฉ๋๋ค. โ์ปดํจํธ ํ ๋น์โ๋ก์์ ์๋ก์ด ์ญํ , โ๋ง์ดํฌ๋ก ์ํํธ์จ์ดโ์ ๋ฑ์ฅ, ๊ทธ๋ฆฌ๊ณ โ์ด์์๋ ๋ฌธ์โ์ ๊ฐ์น๋ AI์ ์ธ๊ฐ์ด ํจ๊ป ํ๋ ฅํ์ฌ ๋ฏธ๋๋ฅผ ๊ตฌ์ถํ๋ ์๋ก์ด ์งํ์ ์ด๊ณ ์์ต๋๋ค. ๋จ์ํ ํ์์ ๋ณํ๋ฅผ ๋์ด, ์ด๋ AI ์๋์ ์์ ๋ฐฉ์ ์์ฒด๋ฅผ ํ์ ํ๋ ์ค๋ํ ์์ง์์ ๋๋ค.
โThe Courtroom Showdown Between Elon Musk and Sam Altmanโ โ New York Times Podcasts ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI ์ ๊ตญ ๊ฑด์ค, ๊ทธ ์ด๋ฉด์ ์ถ์ ํ ์ง์ค: ๋จธ์คํฌ ๋ ์ํธ๋ง ๋ฒ์ ๋๋ผ๋ง
์ง๋ 3์ฃผ๊ฐ, ์ธ๊ณ์์ ๊ฐ์ฅ ์ํฅ๋ ฅ ์๋ ๊ธฐ์ ๊ฑฐ๋ฌผ ๋ ๋ช ์ด ๋ฒ์ ์์ ์น์ดํ๊ณ ์ง์ ๋ถํ ์ ์์ ๋ฒ์์ต๋๋ค. ๋ฐ๋ก ํ ์ฌ๋ผ(Tesla)์ ์คํ์ด์คX(SpaceX)์ CEO์ธ ์ผ๋ก ๋จธ์คํฌ(Elon Musk)์ ์ธ๊ณต์ง๋ฅ(AI) ํ๋ช ์ ์ด๋๋ OpenAI์ CEO ์ ์ํธ๋ง(Sam Altman)์ ๋๋ค. ์ธ๋ฅ์ ๋ฏธ๋๋ฅผ ์ข์ฐํ AI ๊ธฐ์ ์ ๋๋ฌ์ผ ์ด ๊ณ ์ก ์์ก์ ๋จ์ํ ๊ธฐ์ ๋ถ์์ ๋์ด ๊ฐ์ธ์ ์ธ ๊ฐ์ ์ธ์์ผ๋ก ๋ฒ์ง๋ฉฐ ์ค๋ฆฌ์ฝ๋ฐธ๋ฆฌ์ ์ด๋์ด ์ด๋ฉด์ ์ ๋๋ผํ๊ฒ ๋๋ฌ๋์ต๋๋ค.
๋จธ์คํฌ๋ OpenAI๊ฐ ์๋์ ๋น์๋ฆฌ์ ์ฌ๋ช ์ ๋ฐฐ์ ํ๊ณ ์๋ฆฌ ๊ธฐ์ ์ผ๋ก ๋ณ์ง๋์๋ค๋ฉฐ, ์ํธ๋ง์ CEO ์๋ฆฌ์์ ํด์ํ๊ณ OpenAI๋ฅผ ๋ค์ ๋น์๋ฆฌ ๋จ์ฒด๋ก ๋๋๋ฆด ๊ฒ์ ์๊ตฌํ๊ณ ์์ต๋๋ค. ๋ฐฐ์ฌ์ ์ฌ์๊ฐ ์์๋ ์ง๊ธ, ๋ด์ํ์์ค ๊ธฐ์ ์ ๋ฌธ ๊ธฐ์ ๋ง์ดํฌ ์์ด์(Mike Isaac)์ ๋ฒ์ ๋๋ผ๋ง์ ์์ํ ํ์ฅ์ ์ ํ๋ฉฐ, ์ด๋ป๊ฒ ๊ธฐ์ ๋ถ์์ด ๊ทน๋๋ก ๊ฐ์ธ์ ์ธ ์ธ์์ผ๋ก ๋นํ๋์๋์ง ์ค๋ช ํฉ๋๋ค.
๋ฒ์ ๋ฐ์ ์์ปค์ค์ ์์ ๊ธด์ฅ
์คํด๋๋ ์๋ด์ ์ง๋ฐฉ ๋ฒ์ ์์ ๋งค์ผ ์๋ฒฝ 5~6์๋ถํฐ ๋ถ์ํต์ ์ด๋ค์ต๋๋ค. ๋ง์น ์์ปค์ค๋จ์ด ๋ง์์ ๋ค์ด์ ๋ฏํ ํ๊ฒฝ์ด์์ต๋๋ค. ๋ฒ์ ์ฃผ๋ณ์๋ ๋ค์ํ ์์๋๊ฐ ์ง๊ฒฐํ๋๋ฐ, ์ ์ํธ๋ง์ ๋น๋ํ๋ ์ฌ๋๋ค๊ณผ ์ผ๋ก ๋จธ์คํฌ๋ฅผ ํ์คํ๋ ์์๋๊ฐ ๋ค์์ฌ ์์์ต๋๋ค. โ์ผ๋ก ์ ์ต์ ์ด์ผ(Elon sucks)โ๋ผ๊ณ ์ฐ์ธ ๋ํ ํ์ ์ธํ๊ณผ ์ฌ์ง์ด ๋์น์ ๊ฒฝ๋ก๋ฅผ ํ๋ ๋จธ์คํฌ์ ๋ชจ์ต์ด ๋ด๊ธด ๊ทธ๋ฆผ๊น์ง ๋ฑ์ฅํ๋ฉฐ ๊ธฐ๊ดดํ ๋ณผ๊ฑฐ๋ฆฌ๋ฅผ ์ ๊ณตํ์ต๋๋ค. ํํธ์์๋ โ์์ญ์ต ๋ฌ๋ฌ ์ฌ์ฐ๊ฐ๋ค์ด AI๋ฅผ ๋๊ณ ์ธ์ฐ๋ ๋์, ์์ญ๋ง ๋ช ์ ๋ ธ๋์๋ค์ ์ด ์ง์ญ์์ ์ด ํํธ๋ ์ ๋๋คโ๊ณ ์ธ์น๋ ๋ชฉ์๋ฆฌ๋ ๋ค๋ ธ์ต๋๋ค.
๋ฒ์ ์์ผ๋ก ๋ค์ด๊ฐ๋ ค๋ ์ฌ๋๋ค์ ๋จ 30์์ ๋น์ง์ ์์ ์ฐจ์งํ๊ธฐ ์ํด ๋ง์น ์ธ๊ธฐ ํด๋ฝ์ ์ ์ฅํ๋ ค๋ ๋ฏ ๊ฒฝ์ํ์ต๋๋ค. ํ์ง๋ง ์ด๊ณณ์ ๋จ์ํ ํฅ๋ฏธ๋ฅผ ์ข๋ ๊ณณ์ด ์๋๋ผ, AI๋ก ์ธํ ์ธ์์ ์ข ๋ง์ ๊ฑฑ์ ํ๋ ์ฌ๋๋ค์ด ๋ชจ์ธ โํด๋ฝโ์ด์์ต๋๋ค.
๋ฒ์ ๋ด๋ถ๋ ์ธ๋ถ์ ์๋๊ณผ๋ ๋ค๋ฅธ ๊ธด์ฅ๊ฐ์ผ๋ก ๊ฐ๋ํ์ต๋๋ค. ๋ฒ์ ์ผ๋ก ์ด์ด์ง๋ ๋ณต๋๋ ๊ฝค ๋น์ข์๊ณ , ๊ทธ ์์๋ ์ต๋ง์ฅ์๋ค์ด ์ผ๋ฐ์ธ๋ค๊ณผ ์์ด์ง ์๋๋ก ๋๊ธฐํ๋ ๊ณต๊ฐ์ด ์์์ต๋๋ค. ๋งค์ผ ์ผ๋ก ๋จธ์คํฌ, ์ ์ํธ๋ง, OpenAI์ ๊ณต๋ ์ฐฝ์ ์ ๊ทธ๋ ๋ธ๋ก๋ง(Greg Brockman) ๋ฑ AI ์ ๊ณ์ ๊ฑฐ๋ฌผ๋ค์ด ์ธ๋ก ๊ณผ ๋์ค, ๊ทธ๋ฆฌ๊ณ ๊ทธ๋ค์ ํฌ๊ณผ ๋นํ์๋ค์ ์ง๋์ณ ๋ฒ์ ์ผ๋ก ๋ค์ด์๋ ๋ชจ์ต์ ๊ทธ์ผ๋ง๋ก โํ์ค ์ธ๊ณ๋ก ๋ด๋ ค์จ(coming down to earth)โ ์๊ฐ์ด์์ต๋๋ค. ์ด ๋ชจ๋ ๊ณผ์ ์์ ์์ฒญ๋ ์ดํด๊ด๊ณ๊ฐ ๊ฑธ๋ ค ์์์ ๋ชจ๋๊ฐ ์ธ์งํ๊ณ ์์์ต๋๋ค.
์์ก์ ๋ฐฐ๊ฒฝ: ๋น์๋ฆฌ์์ ์๋ฆฌ๋ก?
๋จธ์คํฌ์ ์ฃผ์ฅ์ ๊ฐ๋จํฉ๋๋ค. ์๋ ์ ์์ ์ด OpenAI ํ๋ก์ ํธ์ ์ฐธ์ฌํ์ ๋, ์ธ๋ฅ์ ์ (ๅ)์ ์ํ ๋น์๋ฆฌ ์ธ๊ณต์ง๋ฅ ๊ฐ๋ฐ์ ๋ชฉํ๋ก ํ์ง๋ง, ๊ณต๋ ์ฐฝ์ ์ ์ ์ํธ๋ง๊ณผ ๊ทธ๋ ๋ธ๋ก๋ง์๊ฒ โ์์๋คโ๋ ๊ฒ์ ๋๋ค. ๋จธ์คํฌ๋ ์ด๋ค์ด ๋์ค์ โ๋๊ณผ ๋ช ์(stars and dollar signs)โ์ ๋์ด ๋ฉ์ด OpenAI๋ฅผ ์๋ฆฌ ๊ธฐ์ ์ผ๋ก ๋ง๋ค๊ธฐ๋ก ๊ฒฐ์ ํ๋ค๊ณ ์ฃผ์ฅํฉ๋๋ค.
์ด ์์ก์ ์ค๋ฆฌ์ฝ๋ฐธ๋ฆฌ์์ ํํ ๋ณผ ์ ์๋ ๋ถ์์ด์ง๋ง, ์ด๋ฒ๋งํผ์ OpenAI์ ์น๋ช ์ ์ธ ํ๊ฒฉ์ ์ ํ ์ ์์ต๋๋ค. ํ์ฌ ๊ฑฐ์ 1์กฐ ๋ฌ๋ฌ(์ฝ 1,300์กฐ ์)์ ๋ฌํ๋ ๋น์์ฅ ๊ธฐ์ ์ธ OpenAI๋ ๋จธ์คํฌ์ ์๊ตฌ๋๋ก๋ผ๋ฉด ์กดํ์ ๊ธฐ๋ก์ ์ค ์ ์์ต๋๋ค.
๋จธ์คํฌ๋ ํฌ๊ฒ ์ธ ๊ฐ์ง๋ฅผ ์๊ตฌํ๊ณ ์์ต๋๋ค. ์ฒซ์งธ, 1,500์ต ๋ฌ๋ฌ(์ฝ 200์กฐ ์)์ ๋ฌํ๋ ์ํด๋ฐฐ์๊ธ์ ์๊ตฌํฉ๋๋ค. ๋ค๋ง ์ด ๋์ ์์ ์ ์ํ ๊ฒ์ด ์๋๋ผ, ์๋ฆฌ ๊ธฐ์ ์ธ OpenAI์ ๋ชจํ์ฌ ๊ฒฉ์ธ ๋น์๋ฆฌ ์กฐ์ง์ผ๋ก ํ์๋์ด์ผ ํ๋ค๊ณ ์ฃผ์ฅํฉ๋๋ค. ๋์งธ, ํ์ฌ์ ์๋ฆฌ ๊ตฌ์กฐ๋ฅผ ํด์ฒดํ๊ณ OpenAI๋ฅผ ๋ค์ ์์ํ ๋น์๋ฆฌ ๋จ์ฒด๋ก ๋๋๋ฆด ๊ฒ์ ์๊ตฌํฉ๋๋ค. ์ ์งธ, ๊ฒฝ์์ง์ ๋ํ โ์ฟ ๋ฐํ(coup de grรขce)โ ๊ฒฉ์ธ ์กฐ์น๋ก, ์ ์ํธ๋ง๊ณผ ๊ทธ๋ ๋ธ๋ก๋ง์ ํ์ฌ ์ต๊ณ ์์ง๊ณผ ์ด์ฌํ์์ ํด์ํ ๊ฒ์ ์๊ตฌํฉ๋๋ค. ์ด ๋ชจ๋ ์๊ตฌ๋ OpenAI์ ๊ทผ๊ฐ์ ๋คํ๋ค ์ ์๋ ์์ฒญ๋ ํ๊ธ๋ ฅ์ ๊ฐ์ง๋๋ค.
OpenAI์ ํ์: ์ธ๋ฅ๋ฅผ ์ํ AI
์ฌ๊ฑด์ ์์์ 2015๋ ์ผ๋ก ๊ฑฐ์ฌ๋ฌ ์ฌ๋ผ๊ฐ๋๋ค. ์ผ๋ก ๋จธ์คํฌ๋ ๊ตฌ๊ธ์ CEO ๋๋ฆฌ ํ์ด์ง(Larry Page)์ AI์ ๋ฏธ๋์ ๋ํด ํฐ ์ด๊ฒฌ์ ๋ณด์์ต๋๋ค. ๋จธ์คํฌ๋ AI๊ฐ ์ธ๋ฅ์๊ฒ ํด๋ฅผ ๋ผ์น ๊ฒ์ ๊ทน๋๋ก ์ฐ๋ คํ๋ฉฐ โAI๊ฐ ์ธ๋ฅ๋ฅผ ํด์น ์ ์๋คโ๋ ์ฐ๋ ค๋ฅผ ํํฐ์์์กฐ์ฐจ ๋์์์ด ์ด์ผ๊ธฐํ์ต๋๋ค. ํ์ง๋ง ํ์ด์ง๋ โAI๊ฐ ์ธ๋ฅ๋ณด๋ค ์ค๋ ์ด์๋จ๋๋ผ๋ ๊ด์ฐฎ์ ๊ฒโ์ด๋ผ๋ฉฐ ๋์๋กญ์ง ์๊ฒ ์ฌ๊ฒผ์ต๋๋ค. ๋จธ์คํฌ๋ ํ์ด์ง๊ฐ ์ธ๋ฅ์ ๋ํ ๊ด์ฌ์ด ์๋ค๋ฉฐ ์ถฉ๊ฒฉ๊ณผ ๋ถ๋ ธ๋ฅผ ํํ๊ณ , ํ์ด์ง๋ ๋จธ์คํฌ๋ฅผ โ์ข ์ฐจ๋ณ์ฃผ์์(speciesist)โ๋ผ๊ณ ๋ถ๋ฅด๋ฉฐ AI ๋ก๋ด์ ๋ํ ์ฐจ๋ณ์ ํ๋ค๊ณ ๋ง๋ฐ์์ณค์ต๋๋ค.
๊ตฌ๊ธ์ ๋ํญํด์ผ ํ๋ค๊ณ ์๊ฐํ ๋จธ์คํฌ๋ ๋น์ ์ค๋ฆฌ์ฝ๋ฐธ๋ฆฌ์์ ์คํํธ์ ์ธํ๋ฒ ์ดํฐ๋ฅผ ์ด์ํ๋ ์ ์ํธ๋ง๊ณผ ์๊ธฐํฌํฉํ์ต๋๋ค. โ์ด๊ฒ์ ์ค์กด์ ๋ฌธ์ ์ด๋ฉฐ, AI๊ฐ ๋ฏธ๋์ ๋ฌด์์ ํ ์ ์๋์ง ๊ฑฑ์ ํด์ผ ํ๋คโ๋ ๋จธ์คํฌ์ ์ฃผ์ฅ์ ์ํธ๋ง์ ๊ณต๊ฐํ๊ณ , ๊ทธ๋ ๊ฒ OpenAI๊ฐ ํ์ํ์ต๋๋ค. ๊ทธ๋ค์ ๋ชฉํ๋ ์ธ๋ฅ์ ์ ์ ์ํ ์์ ํ๊ณ ์ดํ์ฃผ์์ ์ธ ๋ฐฉ์์ผ๋ก ์ธ๊ณต์ง๋ฅ์ ๋ฏธ๋๋ฅผ ๊ตฌ์ถํ๋ ๊ฒ์ด์์ต๋๋ค. ๋น์ ํ์ดํ(PayPal)๋ก ๋ง๋ํ ๋์ ๋ฒ๊ณ ํ ์ฌ๋ผ์ ์คํ์ด์คX๋ฅผ ์ฑ๊ณต์ ์ผ๋ก ์ด๋๋ ๋จธ์คํฌ๋ โ์๊ธ์ค(money man)โ ์ญํ ์ ๋งก์ ๋ง๋ํ ์๊ธ์ ์ง์ํ์ต๋๋ค.
์ด๊ธฐ๋ถํฐ ๋จธ์คํฌ์ ์ํธ๋ง์ ํฌํจํ ๋ชจ๋ ๊ด๊ณ์๋ AI๊ฐ ์ธ๋ฅ์๊ฒ ์์ฒญ๋๊ฒ ์ค์ํ ๊ธฐ์ ์ด๋ฉฐ, โ์ ํ ์๋ค(good guys)โ์ด ์ด ๊ฐ๋ ฅํ ๊ธฐ์ ์ ๊ด๋ฆฌ์๊ฐ ๋์ด์ผ ํ๋ค๋ ๋ฐ ๋์ํ์ต๋๋ค. ์ํธ๋ง์ AI ๊ฐ๋ฐ์, ๋จธ์คํฌ๋ ์๊ธ ์ง์์ ๋ด๋นํ๊ธฐ๋ก ํ์ต๋๋ค. ๋น์ AI๋ ์ง๊ธ์ฒ๋ผ ์ฃผ๋ฅ ๊ธฐ์ ์ด ์๋์๊ธฐ์, ๊ทธ๋ค์ ์ด๋ฉ์ผ๋ก๋ AI๊ฐ ์ธ์์ ์ฅ์ ํ ์ ์๋ค๋ โ์ข ๋ง๋ก ์ (dumerous)โ ์ฐ๋ ค๋ฅผ ์ง์งํ๊ฒ ๋ ผ์ํ์ต๋๋ค. ์์ด์ ๊ธฐ์๋ โ๋๋ถ๋ถ์ ๊ฒฝ์ฐ ์ธ๋ฅ์ ๋ํ ๊ฑฐ์ฐฝํ ์ ์ธ์ ํ์์ ์ด์ง๋ง, ์ด๋ค์ ์ค๋ซ๋์ ๊ฐ์ ๋ฐฉ์์ผ๋ก ์ด์ผ๊ธฐํด์๊ณ , 3๋ช ์ด ์ฃผ๊ณ ๋ฐ๋ ์ด๋ฉ์ผ ์ฒด์ธ์์ ์ฐ๊ทน์ ํ๊ณ ์์๋ ๊ฒ์ด ์๋๋คโ๋ผ๊ณ ๊ฐ์กฐํ์ต๋๋ค.
๊ฒฐ๋ณ๊ณผ ์ ํ์ : ๋ง์ดํฌ๋ก์ํํธ์ ๋ฑ์ฅ
๋จธ์คํฌ๋ ์ฌ๋ฌ ํ์ฌ๋ฅผ ์ด์ํ๋ฉฐ OpenAI์ โ์์ธโ๋ฅผ ๋ด๋ ํ์์ ์ญํ ์ ๋จธ๋ฌผ๋ ๊ณ , ์ ์ํธ๋ง๊ณผ ๊ทธ๋ ๋ธ๋ก๋ง์ ์ฐจ์ธ๋ AI ๋ชจ๋ธ์ ๊ฐ๋ฐํ๋ ๋ฐ ์ฃผ๋ ฅํ์ต๋๋ค. ํ์ง๋ง 2017๋ ์์ 2018๋ ๊ฒฝ, ๊ทธ๋ ๊ณผ ์์ ์ํ๋ ๋ฐฉ์์ผ๋ก AI ๋ชจ๋ธ์ ๊ฐ๋ฐํ๋ ค๋ฉด ๋ง๋ํ ์๊ธ์ด ํ์ํ๋ค๋ ๊ฒ์ ๊นจ๋ฌ์์ต๋๋ค. ์๋ฐฑ๋ง, ์์ฒ๋ง, ์ฌ์ง์ด ์๋ฐฑ์ต ๋ฌ๋ฌ์ ๋ฌํ๋ ์๊ธ์ด ํ์ํ๊ณ , ๊ธฐ์กด์ ๋น์๋ฆฌ ๋ชจ๋ธ๋ก๋ ํ๊ณ๊ฐ ์์์ต๋๋ค. ๋น์๋ฆฌ ๊ตฌ์กฐ๋ก๋ ํ์ํ ์์์ ํ๋ณดํ ์ ์๋ค๋ ํ์ค์ ์ธ์งํ๊ฒ ๋ ๊ฒ์ ๋๋ค.
์ดํ ์ ์ํธ๋ง๊ณผ ์ผ๋ก ๋จธ์คํฌ ์ฌ์ด์๋ โ๋์ (cold war battle)โ๊ณผ ๊ฐ์ ๊ฐ๋ฑ์ด ์นํธ๊ธฐ ์์ํ์ต๋๋ค. ๋จธ์คํฌ๋ ๋น์๋ฆฌ ํ์ฌ์ ๋ํ ์์ ์ ์ํฅ๋ ฅ์ ์์๊น ์ฐ๋ คํ๊ณ , OpenAI๋ฅผ ํ ์ฌ๋ผ ๋ด๋ถ๋ก ํตํฉํ์ฌ AI๋ฅผ ๊ฐ๋ฐํ์๋ ์์ด๋์ด๋ฅผ ๋์ต๋๋ค. ํ์ง๋ง ์ํธ๋ง์ ์ด๋ฅผ ๊ฑฐ๋ถํ๊ณ , ๊ฒฐ๊ตญ ๊ด๊ณ๋ ํ๊ตญ์ผ๋ก ์น๋ฌ์์ต๋๋ค. ๋จธ์คํฌ๋ โ๋ ์ด๋ง ๊ฐ๊ฒ ๋ค. ํ์ด์ ๋น๋คโ๋ฉฐ ํ๋ก์ ํธ์์ ์์ ๋์ต๋๋ค. ๋น์ AI ๊ฐ๋ฐ์ ์ง๊ธ์ฒ๋ผ ๋จ๊ฑฐ์ด ๋ถ์ผ๊ฐ ์๋์๊ธฐ์, ๋จธ์คํฌ๋ ๋๋ง ์ฐ๋ ๊ณจ์นซ๋ฉ์ด์ ๋ถ๊ณผํ๊ณ , ์ ๋๋ก ์๋ํ ์ง ๋ฏธ์ง์์ธ ํ๋ก์ ํธ๋ฅผ ๋๊ณ ๊ทธ๋ค๊ณผ ๋คํฌ๋ ๊ฒ์ ์ง์ณ ์์์ต๋๋ค.
๋จธ์คํฌ๊ฐ ๋ ๋ ํ์๋ ๊ทธ๋ ๋ช ๋ฌ๊ฐ OpenAI ์ฌ๋ฌด์ค์ ์์ธ๋ฅผ ๋ด์ฃผ๊ธฐ๋ ํ์ต๋๋ค. ํ์ง๋ง ์๊ณผ ๊ทธ๋ ์ ๋ง์ดํฌ๋ก์ํํธ(Microsoft)์ ์ ์ดํ๊ธฐ ์์ํ์ต๋๋ค. ๊ตฌ๊ธ์ ๋นํด AI ๋ถ์ผ์์ ๋ค์ฒ์ ธ ์๋ค๊ณ ํ๋จํ ๋ง์ดํฌ๋ก์ํํธ์ CEO ์ฌํฐ์ ๋๋ธ๋ผ(Satya Nadella)๋ OpenAI ์ฐ๊ตฌ์ ์๊ธ์ ์ง์ํ๊ธฐ๋ก ๊ฒฐ์ ํ์ต๋๋ค. ๋ง์ดํฌ๋ก์ํํธ๋ OpenAI์ 10์ต ๋ฌ๋ฌ(์ฝ 1์กฐ 3์ฒ์ต ์)๋ฅผ ์ฝ์ํ์ต๋๋ค. ์ด๋ ๋จธ์คํฌ๊ฐ ํฌ์ํ๋ 3,800๋ง ๋ฌ๋ฌ์๋ ๋น๊ตํ ์ ์๋ ์์ฒญ๋ ๊ธ์ก์ด์์ต๋๋ค.
์ฑGPT์ ์ฑ๊ณต๊ณผ ๋จธ์คํฌ์ ๋ถ๋ ธ
์ด๊ธฐ ๊ฒฐ๋ณ์ ๊ทธ๋ค์ง ๊ฒฉ๋ ฌํ์ง ์์์ต๋๋ค. ํ์ง๋ง 2022๋ , OpenAI๊ฐ ์ฑGPT(ChatGPT)๋ฅผ ๊ณต๊ฐํ๋ฉด์ ์ํฉ์ ๊ธ๋ณํ์ต๋๋ค. ๋น์ OpenAI๋ ์ฑGPT๊ฐ โ๋์๋กญ์ง ์๋คโ๊ณ ์ฌ๊ฒผ์ง๋ง, ์ฑGPT๋ ๋ฐค์ ๊ฐ์ฅ ๋ง์ด ๋ค์ด๋ก๋๋ ์ฑ์ด ๋๋ฉฐ ์ ์ธ๊ณ์ ์ธ ์ผ์ธ์ด์ ์ ์ผ์ผ์ผฐ์ต๋๋ค. ์ด๋ ์ค๋ฆฌ์ฝ๋ฐธ๋ฆฌ์ ๋ชจ๋ ๋๊ธฐ์ ๋ค์ ๊นจ์ AI์ ๋ง๋ํ ํฌ์๋ฅผ ์์ํ๊ฒ ํ๊ณ , ๋ง์ดํฌ๋ก์ํํธ๋ OpenAI์ 100์ต ๋ฌ๋ฌ(์ฝ 13์กฐ ์)๋ฅผ ์ถ๊ฐ ํฌ์ํ๊ธฐ๋ก ๊ฒฐ์ ํ์ต๋๋ค.
์ด ์์์ ๋ค์ ๋จ ํ ์ฌ๋, ์ผ๋ก ๋จธ์คํฌ๋ ๋ถ๋ ธํ์ต๋๋ค. ์์ ์ด ๋น์๋ฆฌ๋ก ๊ณต๋ ์ฐฝ์ ํ๋ AI ์คํํธ์ ์ด ์ด์ ์์ฒญ๋ ํฌ์๋ฅผ ๋ฐ๋ ์๋ฆฌ ๊ธฐ์ ์ ํน์ฑ์ ๋ ๋ ๊ฒ์ ๋ํด โ์ด๊ฑด ๋ฏธ๋ผ ์ํ์ ๋ด์ธ์ ์๋น์๋ฅผ ์์ด๋ ํ์(bait and switch)โ๋ผ๋ฉฐ ์ํธ๋ง์๊ฒ ๊ฐํ๊ฒ ํญ์ํ์ต๋๋ค. ํฅ๋ฏธ๋กญ๊ฒ๋ ๋จธ์คํฌ
โHow to be more creative | The Gray Areaโ โ Vox ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
์ฐฝ์์ฑ๊ณผ ํ๋ณต์ ์ํ ์ญ์ค์ ์ธ ์ง์ค: ๋ฌดํํ ์์ ๊ฐ ๋ ์ด ๋ ๋
์ฐ๋ฆฌ๋ ํํ ์ฐฝ์์ฑ์ด ๋ฌดํํ ์์ ์ ์๊ฐ์์ ์์๋๋ค๊ณ ๋ฏฟ์ต๋๋ค. ์ด๋ค ์ ์ฝ๋ ์์ด ๋ง์๊ป ์์ํ๊ณ ํ๊ตฌํ ๋ ๋น๋ก์ ์ง์ ํ ๋ ์ฐฝ์ฑ์ด ๋ฐํ๋๋ค๊ณ ์๊ฐํ์ฃ . ํ์ง๋ง ๋ฒ ์คํธ์ ๋ฌ ์๊ฐ ๋ฐ์ด๋น๋ ์ก์คํ์ธ(David Epstein)์ ๊ทธ์ ์ ์ <์ธ์ฌ์ด๋ ๋ ๋ฐ์ค(Inside the Box)>๋ฅผ ํตํด ์ด๋ฌํ ํต๋ ์ ์ ๋ฉด์ผ๋ก ๋์ ํฉ๋๋ค. ํ์บ์คํธ โ๋ ๊ทธ๋ ์ด ์์ด๋ฆฌ์ด(The Gray Area)โ์์ ์ก์คํ์ธ์ โ์์ ํ ์์ ๋ฅผ ๊ณผ๋ํ๊ฐํ๊ณ ์์ผ๋ฉฐ, ์ ์ฉํ ์ ํ(useful limits)์ด์ผ๋ง๋ก ์ฐ๋ฆฌ๋ฅผ ๊ฐ์ฅ ์ฐฝ์์ ์ด๊ณ ๋ง์กฑ์ค๋ฌ์ด ์กด์ฌ๋ก ํด๋ฐฉ์ํฌ ์ ์๋คโ๊ณ ์ฃผ์ฅํฉ๋๋ค. ํนํ ๋ชจ๋ ๊ฒ์ด ๋์ณ๋๋ ํ๋์ฌํ์์ ์ด๋ฌํ ํต์ฐฐ์ ๋์ฑ ์ค์ํด์ง๊ณ ์์ต๋๋ค.
์ ์ฝ์ด ๋ณ์ ์๋ํ ์ฐฝ์กฐ์ ์๊ฐ๋ค
์ก์คํ์ธ์ ์ญ์ฌ๋ฅผ ํตํด ์ ์ฝ์ด ์ด๋ป๊ฒ ์๋ํ ์ฐฝ์กฐ๋ฅผ ์ด๋์ด๋๋์ง ๋ค์ํ ์ฌ๋ก๋ฅผ ์ ์ํฉ๋๋ค.
๋ฅํฐ ์์ค(Dr. Seuss)์ 50๋จ์ด์ ๋ง๋ฒ
์ด๋ฆฐ์ด ๋ฌธํ์ ์ ์ค, ๋ฅํฐ ์์ค(Dr. Seuss)์ ์ด์ผ๊ธฐ๋ ์ ์ฝ์ ํ์ ๊ทน๋ช ํ๊ฒ ๋ณด์ฌ์ค๋๋ค. ๊ทธ์ ํธ์ง์๋ ๋ฅํฐ ์์ค๊ฐ ๋จ 50๊ฐ์ ๋จ์ด๋ง์ ์ฌ์ฉํด ์ด๋ฆฐ์ด์ฑ ์ ์ธ ์ ์์ ๊ฒ์ด๋ผ๋ฉฐ ๋ด๊ธฐ๋ฅผ ๊ฑธ์์ต๋๋ค. ์ด ๋ด๊ธฐ์ ๊ฒฐ๊ณผ๋ก ํ์ํ ๊ฒ์ด ๋ฐ๋ก ์ ์ธ๊ณ์ ์ธ ๋ฒ ์คํธ์ ๋ฌ <์ด๋ก ๋ฌ๊ฑ๊ณผ ํ(Green Eggs and Ham)>์ ๋๋ค.
๋น์ ๋ฅํฐ ์์ค๋ ํน์ ๋จ์ด ๋ชฉ๋ก(์ฝ 200๊ฐ)๋ง์ ์ฌ์ฉํด์ผ ํ๋ ๋ ๋ค๋ฅธ ์ ์ฝ ์์์ ์ฑ ์ ์ฐ๊ณ ์์๋๋ฐ, ํ์ฉ์ฌ๊ฐ ๋๋ฌด ์ ์ด ๋ถํํ๊ธฐ๋ ํ์ต๋๋ค. ๊ฒฐ๊ตญ ๊ทธ๋ ๋ชฉ๋ก์์ ์ฒ์ ๋์ค๋ ๋ ๊ฐ์ ์ด์จ์ด ๋ง๋ ๋จ์ด๋ฅผ ๊ฐ์ง๊ณ ์ฑ ์ ๋ง๋ค๊ธฐ๋ก ๊ฒฐ์ฌํ๊ณ , ๊ทธ ๊ฒฐ๊ณผ โ๊ณ ์์ด์ ๋ชจ์(Cat in the Hat)โ๊ฐ ํ์ํ์ต๋๋ค. ์ดํ์ ๋ํ ์๊ฒฉํ ์ ํ์ ๊ทธ๊ฐ ์ต์ง๋ก๋ผ๋ ์ด์จ๊ณผ ๋ฆฌ๋ฌ์ ์ง์คํ๊ฒ ๋ง๋ค์๊ณ , ์ด๋ ๊ทธ์ ์ํ์ ๋ ํนํ ์๋ช ๋ ฅ์ ๋ถ์ด๋ฃ๋ ๊ฒฐ์ ์ ์ธ ์์๊ฐ ๋์์ต๋๋ค. ์๊ฐ๋ณด๋ค ํ์์ด ๋ ์ค์ํ ์ ์๋ค๋ ์ฌ์ค์ ๋ณด์ฌ์ฃผ๋ ๋๋ผ์ด ์ฌ๋ก์ ๋๋ค.
๋ฉ๋ธ๋ ์ํ(Mendeleev)์ ์ฃผ๊ธฐ์จํ: ๊ฟ์ด ์๋ ๊ณ์ฝ์ ์ฐ๋ฌผ
๋ฌ์์ ํํ์ ๋ฉ๋ธ๋ ์ํ(Dmitri Mendeleev)๊ฐ ์ฃผ๊ธฐ์จํ๋ฅผ ๊ฟ์์ ๋ฐ๊ฒฌํ๋ค๋ ์ด์ผ๊ธฐ๋ ๊ณผํ์ ๋ฐ๊ฒฌ์ ์ ํ์ฒ๋ผ ์ ํด์ ธ ๋ด๋ ค์ต๋๋ค. ์ก์คํ์ธ ์ญ์ ์ด ์ด์ผ๊ธฐ๋ฅผ ๋ฏฟ์์ง๋ง, ์ค์ ์ด์ผ๊ธฐ๋ ๋ฌ๋์ต๋๋ค. ๋ฉ๋ธ๋ ์ํ๋ ๋น์ ๋ ๊ถ์ง๋ฆฌ ํํ ์ ๋ฌธ ๊ต๊ณผ์๋ฅผ ์งํํ๋ ์ถํ ๊ณ์ฝ์ ๋งบ๊ณ ์์์ต๋๋ค. 1๊ถ์๋ 63๊ฐ์ ์๋ ค์ง ์์ ์ค 8๊ฐ๋ฐ์ ๋ฃ์ง ๋ชปํ๊ณ , ๋๋จธ์ง 55๊ฐ๋ฅผ 2๊ถ์ ์ค์ด์ผ ํ๋ ์ํฉ์ด์์ฃ . ๊ทธ๋ ๋ ์ด์ ์์ ํ๋ํ๋๋ฅผ ๊ฐ๋ณ์ ์ผ๋ก ๋ค๋ฃฐ ์ ์๋ค๋ ๊ฒ์ ๊นจ๋ฌ์๊ณ , ์ด๋ณด์๋ ์ดํดํ ์ ์๋๋ก ๋ ผ๋ฆฌ์ ์ธ ๋ฐฉ์์ผ๋ก ์์๋ฅผ ๋ถ๋ฅํด์ผ ํ์ต๋๋ค.
์ด๋ฌํ ์๊ฒฉํ ์ ์ฝ ์์์ ๊ทธ๋ ์์๋ค์ ๊ฐ์กฑ ๊ฐ๋ ์ผ๋ก ๋ฌถ์ด ์๊ฐํ๊ธฐ ์์ํ๊ณ , ํํ์ ๋ฌผ๋ฆฌ์ ํน์ฑ์ด ์ฃผ๊ธฐ์ ์ผ๋ก ๋ฐ๋ณต๋๋ค๋ ํจํด, ์ฆ ์ฃผ๊ธฐ์จ(periodic pattern)์ ๋ฐ๊ฒฌํ๊ฒ ๋ฉ๋๋ค. ์ก์คํ์ธ์ ์ด๋ฅผ โ๊ฟ๊พธ๋ ๋์ ์์ ๋ก์์ด ํ์ค์ ๊ฒฝ๊ณ๋ฅผ ๋ฒ์ด๋๋ ๊ฒ๊ณผ ๋ฌ๋ฆฌ, ๋งค์ฐ ์๊ฒฉํ ์ ์ฝ์ด ์์ฐ์ ์ธ ์คํ์ผ๋ก ์ด๋์๋คโ๊ณ ์ค๋ช ํ๋ฉฐ, ์ ํ์ ํ์ค ์ฌ์ด์ ๊ฐ๊ทน์ด ์ฐ๋ฆฌ๊ฐ ์์ ํ ์์ ๋ฅผ ๊ณผ๋ํ๊ฐํ๊ณ ์ ์ฉํ ์ ์ฝ์ ํ์ ๊ณผ์ํ๊ฐํ๋ ๊ฒฝํฅ์ ์์งํ๋ค๊ณ ์ง์ ํฉ๋๋ค.
์ ์ฐ๋ฆฌ๋ โ์์ ํ ์์ โ๋ฅผ ๊ณผ๋ํ๊ฐํ๋๊ฐ?
์ฌ๋ฆฌํ์๋ค์ด ์ค์ํ ์ฐฝ์์ฑ ์ ํ์ ๋ํ ๊ตญ์ ์ค๋ฌธ์กฐ์ฌ์์ ๊ฐ์ฅ ์ธ๊ธฐ ์๋ ์คํด๋ โ์ฌ๋๋ค์ ๊ฐ์ฅ ์์ ๋ก์ธ ๋ ๊ฐ์ฅ ์ฐฝ์์ ์ด๋คโ์์ต๋๋ค. ํ์ง๋ง ์ฐ๊ตฌ ๊ฒฐ๊ณผ๋ ์ด์ ๋ค๋ฆ ๋๋ค.
๋๋ โ์๊ฐํ๊ธฐ ์ซ์ดํ๋คโ: ์ธ์ง ๊ณผํ์ ๊ด์
์ธ์ง ๊ณผํ์ ๋ค๋์ ์๋งํ(Daniel Willingham)์ ๋ง์ฒ๋ผ, โ์ฐ๋ฆฌ์ ๋๋ ์๊ฐํ๊ธฐ ์ํด ๋ง๋ค์ด์ง ๊ฒ์ด ์๋๋ผ, ๊ฐ๋ฅํ๋ฉด ์๊ฐ์ ํ์ง ์๊ธฐ ์ํด ๋ง๋ค์ด์ก๋คโ๊ณ ์ก์คํ์ธ์ ๋งํฉ๋๋ค. ์ฌ๊ณ ๋ ์๋์ง๊ฐ ๋ง์ด ์๋ชจ๋๋ ํ๋์ด๋ฏ๋ก, ๋๋ ๊ฐ๋ฅํ ํ ํธ๋ฆฌํ๊ณ ์ฌ์ด ํด๊ฒฐ์ฑ , ์ฆ ์ด์ ์ ํด๋ดค๊ฑฐ๋ ์ฝ๊ฒ ๋ ์ค๋ฅด๋ ํด๊ฒฐ์ฑ ์ ์ ํํ๋ ค๋ ๊ฒฝํฅ์ด ์์ต๋๋ค.
์ต์ํจ์ ๋ซ: ์ต์ ์ ํญ์ ๊ธธ(Path of Least Resistance)
์์ ํ ์์ ๊ฐ ์ฃผ์ด์ง๋ฉด ๋๋ ์ธ์ง ๊ณผํ์๋ค์ด โ์ต์ ์ ํญ์ ๊ธธ(path of least resistance)โ์ด๋ผ๊ณ ๋ถ๋ฅด๋ ์ต์ํ๊ณ ํธ๋ฆฌํ ๊ธธ๋ก๋ง ๊ฐ๋ ค ํฉ๋๋ค. ๋ฐ๋ผ์ ์ง์ ์ผ๋ก ์ฐฝ์์ ์ด ๋๊ธฐ๋ ๊ฑฐ์ ๋ถ๊ฐ๋ฅํด์ง๋๋ค. ์ก์คํ์ธ์ ๊ฐ์ฅ ํจ๊ณผ์ ์ธ ์ฐฝ์์ ์ด์ง์ ๋ โ๋ฌธ์ ๋ ๊ณผ์ ์ ์ง๋ฉดํ์ ๋, ๊ฐ์ฅ ๋จผ์ ๋ ์ค๋ฅด๋ ํด๊ฒฐ์ฑ ์ ์ง์๋ฒ๋ฆฌ๊ณ โ์ด๊ฒ์ ํ์ฉ๋์ง ์๋๋คโ๊ณ ์ค์ค๋ก์๊ฒ ๋งํ๋ ๊ฒโ์ด๋ผ๊ณ ์กฐ์ธํฉ๋๋ค. ์ด๋ ๋๋ฅผ ์ต์ ์ ํญ์ ๊ธธ์์ ๋ฒ์ด๋๊ฒ ํ์ฌ ์ฐฝ์์ ์ธ ์ฌ๊ณ ๋ฅผ ๊ฐ์ ํ๋ ๊ฐ์ฅ ํ์คํ ๋ฐฉ๋ฒ์ ๋๋ค.
์์ ์ ํ์ ๊ณผ โ์๋ฅ์ด ์ ์ฝ(Paired Constraints)โ
์ฌ๋ฆฌํ์ ํจํธ๋ฆฌ์ค ์คํก์ค(Patricia Stokes)๋ ์์ ์ ํ์ ์ ์ญ์ฌ๋ฅผ ์ฐ๊ตฌํ๋ฉฐ โ์๋ฅ์ด ์ ์ฝ(paired constraints)โ์ด๋ผ๋ ๊ฐ๋ ์ ์ ์ํ์ต๋๋ค. ์ด๋ ์์ ๊ฐ๊ฐ ๊ธฐ์กด์ ๊ดํ(status quo)์ ๋จผ์ ํ์ ํ ๋ค์, ๋ฌด์์ โ์ฐจ๋จ(preclude)โํ๊ณ ๋ฌด์์ โ์ด์ง(promote)โํ ์ง ๊ฒฐ์ ํ๋ ์ฒด๊ณ์ ์ธ ๊ณผ์ ์ ๋๋ค.
๋ชจ๋ค(Monet)์ ์ธ์์ฃผ์: โ๊ธ์งโ์ โ์ด์งโ์ ์กฐํ
ํด๋ก๋ ๋ชจ๋ค(Claude Monet)์ ์ฌ๋ก๋ ์๋ฅ์ด ์ ์ฝ์ ์ ๋ณด์ฌ์ค๋๋ค. ๋ชจ๋ค๋ ๋น์์ ์ผ๋ฐ์ ์ธ ๋ฏธ์ ๊ธฐ๋ฒ์ ์์งํ๊ณ ์์์ง๋ง, ๋น์ ์ธ์(impressions of light)์ ๋ฌ์ฌํ๊ธฐ ์ํด ์๋ก์ด ๋ฐฉ์์ ํํ์ต๋๋ค. ๊ทธ๋ โ๊ฒ์์๊ณผ ๋ฐ๊ณ ์ด๋์ด ์์์ ์ฌ์ฉํ์ง ์๊ฒ ๋คโ๊ณ ์ ์ธํ๋ฉฐ ํน์ ์ ์ฝ์ ๊ฐํ์ต๋๋ค. ๊ทธ๋ฆฌ๊ณ ๊ทธ ๋์ โ๋ชจ์์ดํฌ์ฒ๋ผ ์์ํ ์์๋ง์ ๊ฐ๊น์ด ๋ฐฐ์นํ์ฌ ๋น์ ์ธ์์ ํํํ๊ฒ ๋คโ๋ ์๋ก์ด ๊ท์น์ ์ค์ ํ์ต๋๋ค.
์ด๋ฌํ โ์ฐจ๋จโ๊ณผ โ์ด์งโ์ ์ ์ฝ์ ๋ชจ๋ค๋ฅผ ์ธ์์ฃผ์์ ๊ธธ๋ก ์ด๋์์ต๋๋ค. ๋ชจ๋ค๊ฐ ๊ฒ์์์ ์ผ๋ง๋ ์ฒ ์ ํ ๋ฐฐ์ ํ๋์ง ๋ณด์ฌ์ฃผ๋ ์ผํ๊ฐ ์์ต๋๋ค. ๊ทธ์ ์ฅ๋ก์์์ ๊ด ์์ ๊ฒ์ ์ฒ์ด ๋ฎ์ด์, ๊ทธ์ ์น๊ตฌ ์ค ํ ๋ช ์ด โ๋ชจ๋ค์๊ฒ ๊ฒ์์์ ์ ๋ผ!โ๋ผ๊ณ ์ธ์น๋ฉฐ ๊ฝ๋ฌด๋ฌ ์ํ๋ณด๋ฅผ ๊ฐ์ ธ์ ๊ด์ ๋ฎ์๋ค๊ณ ํฉ๋๋ค. ์ด๋ ์๊ฐ์ ๋ฒ๊ฐ์ฒ๋ผ ์ฐพ์์ค๋ ์ฐฝ์์ฑ ์ ํ์๋ ๋งค์ฐ ๋ค๋ฅธ, ์ฒด๊ณ์ ์ธ ํ์ ์ ๊ณผ์ ์ ๋ณด์ฌ์ค๋๋ค.
์ ์ฝ๊ณผ ๋ฐฉํด๋ฌผ์ ๋ฏธ๋ฌํ ์ฐจ์ด: ๋์ฌ(NASA)์ ์ฌ๋ก
๋ฌผ๋ก ์ ์ฝ์๋ ์ ๋๊ฐ ์์ต๋๋ค. ๋๋ฌด ๋ง์ ์ ์ฝ์ ์คํ๋ ค ๋ฐฉํด๋ฌผ(impediment)์ด ๋ ์ ์์ต๋๋ค. ์ก์คํ์ธ์ โ์์ ์ ๋๋ผ๊ฒ ํ ์ฌ์ง๊ฐ ์๋ค๋ฉด, ๋๋ฌด ๋ง์ ์ ์ฝ์ ๋ฐ๊ณ ์๋ ๊ฒโ์ด๋ผ๊ณ ๋งํฉ๋๋ค. ์ ์ฝ์ ์ ์ฉ์ฑ์ ์๋ก์ด ๋ฐฉ์์ผ๋ก ์๊ฐํ๊ณ ์์ฐ์ ์ผ๋ก ํ์ํ๋๋ก ์ ๋ํ๋ ๋ฐ ์์ต๋๋ค.
๋์ฌ(NASA)์ ์ํฌ๋ก์ค(Elcross) ๋ฏธ์ ์ฌ๋ก๋ ์ด๋ฌํ ๊ด์ ์ ํ์ ์ค์์ฑ์ ๋ณด์ฌ์ค๋๋ค. ์ํฌ๋ก์ค ํ์ ์์ํ๋ ์๊ฐ๊ณผ ์์ฐ์ ์ ๋ฐ๋ง์ผ๋ก ํ๋ก์ ํธ๋ฅผ ์ํํด์ผ ํ๋ ์ํฉ์ ์ฒํ์ต๋๋ค. ๋ถํ๋ ํ์ง๋ง, ๊ฒฐ๊ตญ โ์ด๋ป๊ฒ๋ ํด๋ธ๋ค๋ฉด ์ด๋ป๊ฒ ํด์ผ ํ ๊น?โ๋ผ๋ ์ง๋ฌธ์ ๋์ง๋ฉฐ ์ฌ๊ณ ๋ฐฉ์์ ์ ํํ์ต๋๋ค. ๊ทธ๋ค์ ์ฒ์๋ถํฐ ๋ชจ๋ ๊ฒ์ ๋ง๋ค๋ ค ํ๊ธฐ๋ณด๋ค, ์ก๊ตฐ ํฑํฌ์ ์ด๋ฏธ์ง ์ฅ๋น์ ๋์ค์นด(NASCAR) ๊ฒฝ์ฃผ์ฉ ์๋์ฐจ์ ์์ง ์จ๋ ์ผ์ ๋ฑ์ ๋น๋ ค ํ์ฉํ์ต๋๋ค. ๊ทธ ๊ฒฐ๊ณผ, ๋ฌ์ ๋ฌผ์ด ์์์ ํ์ธํ๋ ํ์ฌ์ ์ ์ฑ๊ณต์ ์ผ๋ก ๋ง๋ค ์ ์์์ต๋๋ค. ์์น ์์๋ ์ ์ฝ์ด์์ง๋ง, ์ด๋ ๊ทธ๋ค์ด ์ ์ ์๊ฐํด๋ณด์ง ๋ชปํ ๋ฐฉ์์ผ๋ก ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๊ฒ ๋ง๋ค์์ต๋๋ค.
์์ ์ ๋ํ ์คํด: ์ฌํ์ ๊ท๋ฒ๊ณผ ๊ฐ์ธ์ ํ๋ณต
์ฐ๋ฆฌ๋ ํํ ์์ ๋ฅผ โ์ ํ์ ๋ถ์ฌโ๋ก ์ ์ํ์ง๋ง, ์ก์คํ์ธ์ ์ด๊ฒ์ด ๊ฐ์ฅ ์ ์ฉํ ์ข ๋ฅ์ ์์ ๊ฐ ์๋ ์ ์๋ค๊ณ ์ง์ ํฉ๋๋ค.
๋๊ธ๋ฌ์ค ๋ ธ์ค(Douglas North)์ ํต์ฐฐ: ์์ธก ๊ฐ๋ฅ์ฑ์ด ์ ๋ขฐ๋ฅผ ๋ณ๋๋ค
๊ฒฝ์ ํ ๋ ธ๋ฒจ์ ์์์ ๋๊ธ๋ฌ์ค ๋ ธ์ค(Douglas North)๋ ์ฌํ์ ๊ท๋ฒ๊ณผ ์ ์นยท๊ฒฝ์ ์ ๊ตฌ์กฐ๊ฐ ์ธ๊ฐ ํ๋์ ์ ์ฝํจ์ผ๋ก์จ ํ์ธ๊ณผ์ ์์ธก ๊ฐ๋ฅ์ฑ์ ๋์ด๊ณ , ์ด๋ ๊ณง ์ ๋ขฐ๋ก ์ด์ด์ง๋ค๊ณ ์ฃผ์ฅํ์ต๋๋ค. ์ธ๋ฅ ์ญ์ฌ์ ๋๋ถ๋ถ ๋์ ์ฌ๋๋ค์ ์น์กฑ์ด๋ ์ด์๊ณผ๋ง ํ๋ ฅํ์ง๋ง, ์ด๋ฌํ ์ฌํ์ ๊ท๋ฒ๊ณผ ๊ฒฝ๊ณ๊ฐ ๋ฏ์ ์ฌ๋๋ค์ ์ ๋ขฐํ๊ฒ ๋ง๋ค์์ต๋๋ค.
์ก์คํ์ธ์ ํ์ฌ ์ฌํ์์ ๊ณต์ ๋ ์ฌํ์ ๊ท๋ฒ์ด ๋ฌด๋์ง๋ฉด์ ์ ๋ขฐ๊ฐ ํ๋ฝํ๊ณ ์๋ค๋ ๋ถ๊ธธํ ์งํ๋ค์ ๋ด ๋๋ค. ํจ(Pew) ๋ฆฌ์์น ์ผํฐ์ ์ค๋ฌธ์กฐ์ฌ์ ๋ฐ๋ฅด๋ฉด, ๋๋ค์์ ์ฑ์ธ์ด ํ์ธ์ ๋๋์ฑ์ด ๋์๋ค๊ณ ์๊ฐํฉ๋๋ค. ์ด๋ ๋ฏ์ ์ฌ๋์ ๋ํ ์ ๋ขฐ๊ฐ ๊ณต๋ ๋ฒ์์ ๊ธฐ๋ฐ์์ ๊ณ ๋ คํ ๋, ๋ ธ์ค์ ์์ธก๊ณผ ์ ํํ ์ผ์นํ๋ ํ์์ ๋๋ค. โ์ฐ๋ฆฌ ๋ชจ๋๊ฐ ์ํ๋ ๋๋ก ํ๋ํด์ผ ํ๋คโ๋ ์๊ฐ์ ์ฌํ์ ๊ณต๋ ๋ฒ์์ ๊ธฐ๋ฐ์ด ๋ ์ ์์ต๋๋ค.
โ์ต๋ ์์ ๋ ์๋ฐโ์ด๋ผ๋ ๋ณด์์ ํต์ฐฐ
์ก์คํ์ธ์ ์ ์ ์์ ์๋ ๋ฐ๋ฐํ์ง๋ง, ๋์ด๊ฐ ๋ค๋ฉด์ โ์ต๋ ์์ ๋ ๊ทธ ์์ฒด๋ก ์๋ฐโ์ด๋ผ๋ ์ค๋๋ ๋ณด์์ ํต์ฐฐ์ ๊ณต๊ฐํ๊ฒ ๋์๋ค๊ณ ๋งํฉ๋๋ค. ์ถ์์ ์ผ๋ก๋ ์์ ๋ฅผ ์ํ์ง๋ง, ํ์ค์์๋ ๊ทธ๊ฒ์ด ์คํ๋ ค ๋ถ๋ด์ด ๋๋ ๊ฒฝ์ฐ๊ฐ ๋ง๋ค๋ ๊ฒ์ด์ฃ . ์ ์น๋ฃ๋ฒ์ ์ค์ค๋ก ์ ํํ ์์ ๋ฅผ ์ํ๋ค๊ณ ๋งํ๋ ์ฌ๋์ ๋ง์ง๋ง, ์ค์ ๋ก ์์ ๊ฑธ๋ ธ์ ๋ ๊ทธ ์ ํ์ ์ง๋ฉดํ๊ณ ์ถ์ด ํ๋ ์ฌ๋์ ๊ฑฐ์ ์๋ค๋ ์ฌ๋ก๋ ์ด๋ฌํ ์ญ์ค์ ์ ๋ณด์ฌ์ค๋๋ค.
์ก์คํ์ธ ์์ ๋ ์๊ฐ๊ฐ ๋์ด ์์ ํ ์์จ์ฑ์ ์ถ๊ตฌํ์ง๋ง, ๊ฒฐ๊ตญ ์ง๋์น ์์จ์ฑ์ด ๋ถํ์ผ๋ก ์ด์ด์ก์์ ๊นจ๋ฌ์์ต๋๋ค. ๋ชจ๋ ๊ฒฐ์ ์ ์ค์ค๋ก ๋ด๋ ค์ผ ํ๋ โ๊ฒฐ์ ํผ๋ก(decision fatigue)โ์ ์๋ฌ๋ ธ๊ณ , ํ์ธ๊ณผ์ ์ฐ๊ฒฐ์ด ๋ถ์กฑํด์ก์ต๋๋ค. ๊ทธ๋ ๊ฒฐ๊ตญ ์ง์ญ ๋น์๋ฆฌ ๋จ์ฒด ์ด์ฌํ์ ์ฐธ์ฌํ๊ณ , ๋์ค ๋ชจ์์ ๋๊ฐ๊ณ , ์ ๋ ์์ฌ ํ ๋ก ํด๋ฝ์ ๊ฐ์ ํ๋ ๋ฑ โ๋ถํธํ์ง๋ง ์๋ฏธ ์๋โ ์๋ฌด๋ค์ ๋ง๋ค์์ต๋๋ค. ์ธ๊ฐ์ โ์ํธ ์๋ฌด์ ์กฐ๋ฐํ ๋คํธ์ํฌ(dense network of reciprocal obligation)โ ์์์ ๋ฒ์ฑํ๋ค๋ ์ฐ๊ตฌ ๊ฒฐ๊ณผ์ฒ๋ผ, ์ด๋ฌํ ์ ์ฝ๋ค์ด ๊ทธ์ ์ถ์ ์๋ฏธ์ ๊ตฌ์กฐ๋ฅผ ๋ค์ ๊ฐ์ ธ๋ค์ฃผ์์ต๋๋ค.
๋๋ฌด ๋ง์ ์ ํ์ด ๊ฐ์ ธ์ค๋ ๋ง๋น์ ๋ถ์
ํ๋ ์ฌํ๋ ์ ๋ก ์๋ ์ ํ์ ํ์ ์์์ ์ด๊ณ ์์ต๋๋ค. ํ์ง๋ง ์ด๋ฌํ ์ ํ์ ๊ณผ์์ด ์ฐ๋ฆฌ๋ฅผ ๋ ํ๋ณตํ๊ฒ ๋ง๋ค์์๊น์? ์ก์คํ์ธ์ ์คํ๋ ค ๋ง๋น์ ๋ถ์์ ์ด๋ํ๋ค๊ณ ์ง์ ํฉ๋๋ค.
ํ๋ฒํธ ์ฌ์ด๋จผ(Herbert Simon)์ โ๋ง์กฑํ(Satisficing)โ ์ ๋ต
์ ์นํ์์ด์ ์ปดํจํฐ ๊ณผํ, ์ฌ๋ฆฌํ, ๊ฒฝ์ ํ ๋ถ์ผ์์ ์ต๊ณ ์์ ์์ํ ํ๋ฒํธ ์ฌ์ด๋จผ(Herbert Simon)์ โ๋ง์กฑํ(Satisficing)โ๋ผ๋ ๊ฐ๋ ์ ์ ์ํ์ต๋๋ค. ์ด๋ โ๋ง์กฑ์ํค๋ค(satisfy)โ์ โ์ถฉ๋ถํ๋ค(suffice)โ์ ํฉ์ฑ์ด๋ก, ์ธ๊ฐ์ ๋ชจ๋ ์ต์ ์ ํ๊ฐํ์ฌ ์ต์ ์ ๊ฒ์ ์ ํํ๋ ๊ณ ์ ๊ฒฝ์ ํ์ ํฉ๋ฆฌ์ ํ์์ ๋ชจ๋ธ์ฒ๋ผ ํ๋ํ์ง ์๋๋ค๋ ๊ฒ์ ๋ณด์ฌ์ค๋๋ค. ์ฐ๋ฆฌ๋ ์ ํ๋ ์ธ์ง ๋์ญํญ(cognitive bandwidth)์ ๊ฐ์ง๊ณ ์์ผ๋ฏ๋ก, โ์ถฉ๋ถํ ์ข์(good enough)โ ํด๊ฒฐ์ฑ ์ ์ ํํ๋ ์ง๋ฆ๊ธธ์ ์ฌ์ฉํฉ๋๋ค.
์ฌ์ด๋จผ์ ์ด๋ฅผ ์ ๊ทน์ ์ผ๋ก ํ์ฉํด์ผ ํ๋ค๊ณ ์ฃผ์ฅํ์ต๋๋ค. ๊ทธ๋ ํ์ ๊ฐ์ ์ข ๋ฅ์ ์๋ง์ ์ ๊ณ , ๊ฐ์ ์์นจ ์์ฌ๋ฅผ ํ๊ณ , 46๋ ๋์ ๊ฐ์ ์ง์ ์ด๋ฉฐ ์ฌ์ํ ๊ฒฐ์ ์ ์๋ฅผ ์ค์์ต๋๋ค. ์ด๋ฅผ ํตํด ๊ทธ๋ ์์ ์ด ์ง์ ์ผ๋ก ์ง์คํ๊ณ ์ถ์ ์ผ์ ์๋์ง์ ์ธ์ง ๋์ญํญ์ ์ ์ฝํ ์ ์์์ต๋๋ค. ์ฌ์ด๋จผ์ โ์ ๋ณด๊ฐ ํ๋ถํ ์ธ์์์๋ ๋ค๋ฅธ ๋ฌด์ธ๊ฐ๊ฐ ๋ถ์กฑํ๋ค. ์ ๋ณด๊ฐ ์๋นํ๋ ๊ฒ์ ์ฃผ์๋ ฅ(attention)์ด๋คโ๋ผ๊ณ ๋งํ๋ฉฐ, ์ ๋ณด ๊ณผ๋ถํ ์๋์ ๋ง์กฑํ ์ ๋ต์ ์ค์์ฑ์ ๊ฐ์กฐํ์ต๋๋ค.
์ต์ ํ์ ํจ์ : โ์ต๊ณ ๋ ์ข์ ๊ฒ์ ์ โ
์ฌ์ด๋จผ์ ๋ํ โ์ต๊ณ ๋ ์ข์ ๊ฒ์ ์ (The best is enemy of the good)โ์ด๋ผ๋ ์ ๋ช ํ ๋ง์ ๋จ๊ฒผ์ต๋๋ค. ํญ์ โ์ต๊ณ โ๋ฅผ ์ถ๊ตฌํ๋ โ๊ทน๋ํ์(maximizer)โ๋ค์ ์ฌ๋ฆฌ ์ฐ๊ตฌ์์ ๋ ํ๋ณตํ๊ณ , ํํํ ๊ฐ๋ฅ์ฑ์ด ๋์ผ๋ฉฐ, ๊ฒฐ์ ์ ๋ฒ๋ณตํ๊ธฐ ์ฝ๋ค๊ณ ๋ํ๋ฌ์ต๋๋ค. ์ด๋ ๊ทธ๋ค์ด ๊ฒฐ์ ์ ๋ด๋ฆฌ๋ ๋ฐ ์ด๋ ค์์ ๊ฒช๊ณ , ์ถ์ ๋ํ ๋ง์กฑ๋๊ฐ ๋ฎ๊ธฐ ๋๋ฌธ์ ๋๋ค. ์ก์คํ์ธ ์ญ์ ์์ ์ด ๊ทน๋ํ์ ๊ฒฝํฅ์ด ์์์ ์ธ์ ํ๋ฉฐ, โ๋ช ๊ฐ์ง ๊ธฐ์ค์ด ์ถฉ์กฑ๋๋ฉด ๋ ์ด์ ๋ค๋ฅธ ๊ฒ์ ์ฐพ์๋ณด์ง ์๊ณ ๊ฒฐ์ ์ ๋ด๋ฆฌ๋โ ๋ง์กฑํ ๊ท์น์ ์ธ์ฐ๋ ๊ฒ์ด ์ค์ํ๋ค๊ณ ๋งํฉ๋๋ค.
ํ๋ ๋ํจ์ ์ญ์ค(Fredkinโs Paradox): ์ฌ์ํ ๊ฒฐ์ ์ ๋งค๋ฌ๋ฆฌ๋ ์ด์
โํ๋ ๋ํจ์ ์ญ์ค(Fredkinโs Paradox)โ์ ์ ํ์ง๊ฐ ์๋ก ๋น์ทํ ์๋ก ๊ทธ ์ฌ์ด๋ฅผ ์ ํํ๋ ๊ฒ์ด ๋ ์ค์ํด์ง์ง๋ง, ์์ด๋ฌ๋ํ๊ฒ๋ ์ฐ๋ฆฌ๋ ๊ทธ๋งํผ ๋ ๋ง์ ์๊ฐ๊ณผ ์๋์ง๋ฅผ ์๋๋ค๋ ๊ฒ์ ์๋ฏธํฉ๋๋ค. ๋ทํ๋ฆญ์ค์์ ๋ฌด์์ ๋ณผ์ง ๊ณ ๋ฏผํ๋ ๋ฐ ๋ง์ ์๊ฐ์ ์ฐ๋ ๊ฒ์ฒ๋ผ, ๊ฐ์ฅ ๋ ์ค์ํ ๊ฒฐ์ ์ ๊ฐ์ฅ ๋ง์ ์๊ฐ๊ณผ ์๋์ง๋ฅผ ๋ญ๋นํ๋ ๊ฒ์ด์ฃ .
โ๋ฏธ๋ฃจ๋ ์ถโ๊ณผ โ๊ฒฐ์ ํ๋ ์ถโ์ ์ฐจ์ด
์ ํ์ ๊ณผ์์ โ๊ฒฐ์ ์ข ๊ฒฐ์ ๋ํ ๋๋ ค์(fear of closure)โ์ผ๋ก ์ด์ด์ง๋๋ค. ์ด๋ค ๊ฒ์ ์ ๋ ํ ๋, ์ฐ๋ฆฌ๋ ๋ค๋ฅธ ๊ฐ๋ฅ์ฑ์ ๋์น๊ฑฐ๋ ์ ์ฌ์ ์ธ ๋ฏธ๋์ ๋๋ฅผ ์ฃฝ์ด๋ ๊ฒ ๊ฐ์ ๋ถ์๊ฐ์ ๋๋๋๋ค.
โ๋ฏธ๋๋ฌ์ง๋(Sliding)โ ๊ด๊ณ์ ์ํ์ฑ
๊ด๊ณ์ ๋ํ ์ฐ๊ตฌ์์ โ๋ฏธ๋๋ฌ์ง(sliding)โ์ ์ ์ ์ธ๋์์ ์ฆ๊ฐํ๊ณ ์๋ ํ์์ ๋๋ค. ์ด๋ ์ต์ ์ ์ด์ด๋๋ ค๋ ๋ ธ๋ ฅ์ผ๋ก ์ธํด ๋ฌด์ฌ์ฝ ๊ด๊ณ์ ๋จ๊ณ(๋๊ฑฐ ๋ฑ)๋ฅผ ๋ฐ์ ์ํค๋ ๊ฒ์ ์๋ฏธํฉ๋๋ค. ๋ช ํํ ๊ฒฐ์ ์ ๋ด๋ฆฌ์ง ์๊ณ ๊ด๊ณ๊ฐ ํ๋ฌ๊ฐ๋๋ก ๋ด๋ฒ๋ ค๋๋ ๊ฒ์ด์ฃ . ํ์ง๋ง ์ด๋ ๊ฒ โ๋ฏธ๋๋ฌ์ง๋โ ์ปคํ์ ๊ฒฐํผ ํ ์ดํผํ ํ๋ฅ ์ด ๋๊ณ , ์ถ์ ๋ง์กฑ๋๋ ๋ฎ์ต๋๋ค. ์ต์ ์ ์ด์ด๋๋ ํ์ ์์ ์์ง๋ง, ์ค์ ๋ก๋ ์ด๋ฏธ ๋ค๋ฅธ ๊ฐ๋ฅ์ฑ๋ค์ด ๋ซํ๊ณ ์๋ ๊ฒ์ ๋๋ค.
๋ฏธ๋์ ๋๋ฅผ ์ฃฝ์ด๋ ํ์: ์ ํ์ฑ์ ์ง์ํ๋ผ
์ก์คํ์ธ์ ์ฐ๋ฆฌ๊ฐ ๋ง์น ๋ชจ๋ ๊ฒ์ ๋ค ํ ์ ์๋ค๋ ํ์์ ์ฌ๋ก์กํ ์๋ค๊ณ ๋งํฉ๋๋ค. ํ์ง๋ง ์๊ฐ๊ณผ ์ถ์ ์ ํํ๋ฉฐ, ์ฐ๋ฆฌ๋ ๋ณ๋ชฉ ํ์(bottleneck) ๊ทธ ์์ฒด์ ๋๋ค. ๋ฐ๋ผ์ โ์๊ฐ๊ณผ ์ถ์ด ์ ํ๋์ด ์์์ ์ธ์ํ๊ณ ๋ฌด์๋นํ๊ฒ ์ฐ์ ์์๋ฅผ ์ ํด์ผ ํ๋คโ๊ณ ๊ฐ์กฐํฉ๋๋ค. ๋ฏธ๋์ ๋๋ฅผ ์ฃฝ์ด๋ ๊ฒ์ด ๋๋ ค์ ๊ฒฐ์ ์ ๋ฏธ๋ฃจ์ง๋ง, ์ฌ์ค ์ฐ๋ฆฌ๋ ์ด์ฐจํผ ๊ทธ๋ ๊ฒ ํ๊ณ ์์ต๋๋ค. ๊ทธ๋ฌ๋ฏ๋ก ์๋์ ์ผ๋ก ๊ฒฐ์ ์ ๋ด๋ฆฌ๋ ๊ฒ์ด ํจ์ฌ ์ค์ํฉ๋๋ค.
๋์งํธ ์๋์ ์ทจ์ฝ์ฑ: ๊ธฐ์ ์ ์ฐ๋ฆฌ์ ์ฝ์ ์ ์ ์ฉํ๋ค
์ก์คํ์ธ์ ๋์งํธ ์ธ์์ด ์ธ๊ฐ์ ์ฌ๋ฆฌ์ ์ทจ์ฝ์ ๊ณผ ๋์ ์ต๊ด์ ์ ์ฉํ๋๋ก ์ค๊ณ๋์ด ์๋ค๊ณ ๋นํํฉ๋๋ค. ์ฐ๋ฆฌ์ ์ฃผ์๋ ฅ์ ์์ ๊ณผ ์ถ์ ํ์๋กญ๊ฒ ํ๋ ๊ฒ๋ค๋ก๋ถํฐ ๋ฉ์ด์ง๊ฒ ํ๋ ์๊ทน์ผ๋ก ๊ฐ๋ ์ฐจ ์๋ค๋ ๊ฒ์ ๋๋ค.
AI์ โ์ํฌ ์ฌ๋กญ(Work Slop)โ: ์ ๋ณด ๊ณผ๋ถํ์ ๊ฐ์ํ
์ธ๊ณต์ง๋ฅ(AI)์ ๋ฑ์ฅ์ ์ด๋ฌํ ๋ฌธ์ ๋ฅผ ๋์ฑ ์ ํ์ํต๋๋ค. AI๋ ์ฐ๋ฆฌ๊ฐ ์๋ง์ ์ผ์ ์์ํ๊ณ ๋๋ด์ง ์์ ์์ฒญ๋ ๋ฅ๋ ฅ์ ์ ๊ณตํ๋ฉฐ, โ์ํฌ ์ฌ๋กญ(work slop)โ์ด๋ผ๋ ํ์์ ์ด๋ํฉ๋๋ค. ์ด๋ ์ฌ๋๋ค์ด ์์ฒญ๋ ์์ ํ๋ฒํ ๊ฒฐ๊ณผ๋ฌผ์ ์์ฐํ๊ณ ์์๋๋ ๊ฒ์ ์๋ฏธํฉ๋๋ค. ๋์์์ด ๋ ๋ง์ ์ ๋ณด, ๋ ๋ง์ ์์, ๋ ๋ง์ ์์ ๊ฒฐ๊ณผ๋ฌผ์ด ์์์ ธ ๋์ค๋ฉฐ ์ฃผ์๋ ฅ์ ํด๋ก์ด ๋ฐฉ์์ผ๋ก ํํธํ์ํต๋๋ค.
๋ถ๊ดด๋๋ ์ฃผ์๋ ฅ: 45์ด๋ง๋ค ์ ํ๋๋ ์ง์ค๋ ฅ
์ฌ๋ฆฌํ์ ๊ธ๋ก๋ฆฌ์ ๋งํฌ(Gloria Mark)์ ์ฐ๊ตฌ์ ๋ฐ๋ฅด๋ฉด, 25๋ ์ ์ง์ฅ์ธ์ ์ ๋ฌด ์ ํ ์๊ฐ์ ํ๊ท 3๋ถ์ด์์ง๋ง, 2012๋ ์๋ 75์ด, ๊ทธ๋ฆฌ๊ณ 2022๋ ์๋ 45์ด๋ก ๊ธ๊ฐํ์ต๋๋ค. ๋ฉํฐํ์คํน์ ์ค์ ๋ก๋ ๋ถ๊ฐ๋ฅํ๋ฉฐ, ์คํ๋ ค ๋ชจ๋ ์ผ์ ๋ํ ํจ์จ์ฑ์ ๋จ์ด๋จ๋ฆฌ๊ณ ์คํธ๋ ์ค๋ฅผ ์ ๋ฐํฉ๋๋ค. ์ก์คํ์ธ์ โ์ฐ๋ฆฌ ์ค์ค๋ก ์ฃผ์๋ ฅ์ ๊ตฌ์กฐํํด์ผ ํ๋ ๋ถํํ ์ํฉ์ ์ฒํด ์์ผ๋ฉฐ, ์ด๋ ๋ง์ ๋ ธ๋ ฅ์ ํ์๋ก ํ์ง๋ง ์ถฉ๋ถํ ๊ฐ๋ฅํ๊ณ ๋น ๋ฅด๊ฒ ๋ณํ๋ฅผ ๋ง๋ค ์ ์๋คโ๊ณ ๊ฐ์กฐํฉ๋๋ค.
์ถ์ ์ง์ ๋์ด๋ โ์์(Ritual)โ์ ํ
์ด๋ฌํ ๋์งํธ ์๋์ ํผ๋ ์์์ โ์์(ritual)โ์ ์ฐ๋ฆฌ์ ์ฃผ์๋ ฅ์ ๋ณดํธํ๊ณ ์ค์ํ ๊ฒ์ ์ง์คํ๊ฒ ํ๋ ๊ฐ๋ ฅํ ๋๊ตฌ๊ฐ ๋ ์ ์์ต๋๋ค. ์ก์คํ์ธ์ ์์์ด ์ค์ํ ์ผ์ ๋ ์ฝ๊ฒ ๋ง๋ค๊ณ , ์ถ์ ๊ณ์ ๊ฐ๊ณผ ์ ํ์ ๊ฐ๊ฐ์ ๋ถ์ฌํ๋ค๊ณ ๋งํฉ๋๋ค.
์ด์ฌ๋ฒจ ์์๋ฐ(Isabel Allende)์ ๊ธ์ฐ๊ธฐ ์์
์ธ๊ณ์ ์ธ ์๊ฐ ์ด์ฌ๋ฒจ ์์๋ฐ(Isabel Allende)์ ์ฌ๋ก๋ ์์์ ํ์ ๋ณด์ฌ์ฃผ๋ ์ข์ ์์ ๋๋ค. ๊ทธ๋ ๋ 40์ธ ์ง์ ์ ์ฒซ ์ฑ ์ ์ฐ๊ธฐ ์์ํ ์ด๋๋ก, 83์ธ๊ฐ ๋ ์ง๊ธ๊น์ง 18๊ฐ์๋ง๋ค ๋ฒ ์คํธ์ ๋ฌ๋ฅผ ๋ฐํํ๊ณ ์์ต๋๋ค. ๊ทธ๋ ๋ 1981๋ 1์ 8์ผ์ ์ฒซ ์ฑ <์ํผ์ ์ง(The House of the Spirits)>์ ์์ํ๊ณ , ๊ทธ ์ดํ ๋ชจ๋ ์ฑ ์ ๋งค๋ 1์ 8์ผ์ ์์ํ๋ ์์ ๋ง์ ์์์ ๋ง๋ค์์ต๋๋ค.
1์ 8์ผ์ด ๋ค๊ฐ์ค๋ฉด ๊ทธ๋ ๋ ๋ชจ๋ ์ผ์ ๋ง๋ฌด๋ฆฌํ๊ณ , ๋ก์ ์ฐ๊ตฌ ์๋ฃ๋ฅผ ์ ๋ฆฌํ๋ ๋ฑ ์ค๋น ๊ณผ์ ์ ๊ฑฐ์นฉ๋๋ค. ๊ทธ๋ฆฌ๊ณ ๊ทธ๋ ์ด ๋๋ฉด ์ธ์๊ณผ์ ์ ์ด์ ๋๊ณ ์นจ๋ฌต ์์ผ๋ก ๋ค์ด๊ฐ๋๋ค. ๋งค์ผ ์์ ์์ ์์๋ ์์ด๋ฅผ ์ผ๊ณ , ์์ ์ข ๋ฃ ์์๋ ์์ด๋ฅผ ๋๋ ์์๋ ์ง์ผฐ์ต๋๋ค. ์ด๋ฌํ ์์์ ๊ทธ๋ ์๊ฒ ์ถ์ ๊ท์จ๊ณผ ๊ตฌ์กฐ๋ฅผ ์ ๊ณตํ์ผ๋ฉฐ, ์์ญ ๋ ๊ฐ ๊พธ์คํ ๊ธ์ ์ฐ๋ ์๋๋ ฅ์ด ๋์์ต๋๋ค.
๊ฒฐ๋ก : โ์ ํโ์ ํตํด ์ป๋ ์ง์ ํ ์์ ์ ํ์๋ก์ด ์ถ
๋ฐ์ด๋น๋ ์ก์คํ์ธ์ ์ด ์ฑ ์ ์ฐ๋ฉด์ ์์ ์ ์ถ์๋ ๋ง์ ๋ณํ๋ฅผ ์ฃผ์๋ค๊ณ ๊ณ ๋ฐฑํฉ๋๋ค. ์ด์ ์ฑ ๋ค์ ํญ์ ๋ถ๋์ ์ด๊ณผํ์ฌ ์ฐ๊ณ ๋ค์ ์ค์ฌ์ผ ํ์ง๋ง, ์ด๋ฒ์๋ 1๋ ๊ฐ์ ์ฐ๊ตฌ ํ ๋จ ํ ํ์ด์ง์ง๋ฆฌ ๊ตฌ์กฐ์ ๊ฐ์๋ฅผ ๊ฐ์ ๋ก ์์ฑํ์ฌ ๋ฌด์๋นํ๊ฒ ์ฐ์ ์์๋ฅผ ์ ํ์ต๋๋ค. ๊ทธ ๊ฒฐ๊ณผ, ์ด์ ์ฑ ๋ค๋ณด๋ค 20% ์งง๊ณ ํจ์จ์ ์ธ ์ฑ ์ ์กฐ๊ธฐ ๋ง๊ฐํ ์ ์์์ต๋๋ค.
๊ฐ์ธ์ ์ธ ์ถ์์๋ ๊ทธ๋ ์ง๋์น ์์จ์ฑ์ด ๊ฑด๊ฐํ์ง ์์์ ๊นจ๋ซ๊ณ , ํน์ ์๊ฐ๊ณผ ์ฅ์์ ๋ฌผ๋ฆฌ์ ์ผ๋ก ์กด์ฌํด์ผ ํ๋ ์๋ฌด๋ฅผ ๋ง๋ค์์ต๋๋ค. ์ง์ญ ์ฌํ ํ๋์ ์ฐธ์ฌํ๊ณ , ๋์ค ๋ชจ์์ ๋๊ฐ๊ณ , ํ ๋ก ํด๋ฝ์ ๊ฐ์ ํ๋ ๋ฑ ์๋์ ์ผ๋ก ์ ์ฝ์ ์ค์ ํ์ต๋๋ค. ๋ํ, โ๋ง์กฑํโ ๊ท์น์ ์ธ์ ์จ๋ผ์ธ ์ผํ์ด๋ ๋ค๋ฅธ ๊ฒฐ์ ์์ โ์ถฉ๋ถํ ์ข์โ ๊ฒ์ ๋ง์กฑํ๊ณ ๋ ์ด์ ์ฐพ์๋ณด์ง ์๋ ์ต๊ด์ ๋ค์์ต๋๋ค.
์ฐ๋ฆฌ๋ ๋ฌดํํ ์ ๋ณด์ ์ ํ, ๊ทธ๋ฆฌ๊ณ โ๋ ๋ง์ดโ๋ฅผ ์ธ์น๋ ๋์งํธ ์๋์ ์ด๊ณ ์์ต๋๋ค. ์ด๋ฌํ ํ๊ฒฝ์ ์ฐ๋ฆฌ์ ์ฃผ์๋ ฅ์ ํํธํํ๊ณ , ๋ถํ์ํ ๋ถ์๊ณผ ๋ง๋น๋ฅผ ์ด๋ํฉ๋๋ค. ์ก์คํ์ธ์ ํต์ฐฐ์ ์ด๋ฌํ ํ๋ฆ์ ์ ํญํ๊ณ , ์๋์ ์ผ๋ก โ์ ํโ์ ์ค์ ํจ์ผ๋ก์จ ์คํ๋ ค ์ง์ ํ ์ฐฝ์์ฑ, ํจ์จ์ฑ, ๊ทธ๋ฆฌ๊ณ ๊ถ๊ทน์ ์ธ ํ๋ณต์ ์ฐพ์ ์ ์๋ค๋ ์ญ์ค์ ์ธ ๊ธธ์ ์ ์ํฉ๋๋ค. ์ฐ๋ฆฌ์๊ฒ ํ์ํ ๊ฒ์ ๋ฌดํํ ์์ ๊ฐ ์๋๋ผ, ์ถ์ ์๋ฏธ์ ๊ตฌ์กฐ๋ฅผ ๋ถ์ฌํ๊ณ , ๊ฐ์ฅ ์ค์ํ ๊ฒ์ ์ง์คํ๊ฒ ํด์ฃผ๋ โํ๋ช ํ ์ ์ฝโ์ผ์ง๋ ๋ชจ๋ฆ ๋๋ค.