English
Based on โThe AI Sandwich: Where Humans Excel in an AI Worldโ from Every Watch the original video
The AI Sandwich: How Humans Master the Future of Work with AI
In a world increasingly shaped by artificial intelligence, a profound question looms: what remains the unique domain of human endeavor? As AI tools become more sophisticated, automating tasks once thought exclusive to human minds, many fear a future where human work is rendered obsolete. Yet, a compelling new framework, dubbed the โAI Sandwich,โ offers a reassuring and empowering vision, positioning humans not as competitors to AI, but as its essential architects and refiners.
This insight, championed by Kieran, the General Manager of Kora and creator of the revolutionary Compound Engineering framework, suggests that while AI can handle the โfillingโ of our work, humans provide the crucial โbreadโ โ the beginning and the end that give work its true flavor, meaning, and distinction.
Compound Engineering: A Blueprint for Human-AI Collaboration
Kieranโs journey to this realization began with the development of Compound Engineering (CE), an innovative workflow designed to optimize how teams work with AI agents. Initially conceived for software development, the philosophy behind CE quickly proved applicable across a spectrum of knowledge work, from product design to general problem-solving.
The initial iteration of Compound Engineering comprised four core steps:
- Planning: Defining a clear, actionable plan for the AI to execute.
- Work: The AI agent performs the actual task, whether writing code, generating designs, or processing information.
- Review: Assessing the AIโs output for quality and identifying areas for improvement.
- Compound: Crucially, any lessons learned or mistakes identified during planning or review are โcompoundedโ โ stored as knowledge within the system. This allows AI agents to learn from past errors, preventing repetition and continuously improving their performance. Kieran highlights this compounding step as โby far the most powerful thingโ in the plugin, transforming AI from a mere tool into a learning collaborator.
However, as AI capabilities rapidly advanced, Kieran and his collaborator, Trevan, a key contributor to CE, noticed a striking trend: the โworkโ phase was becoming increasingly automated and reliable. โThe work phase is kind of done,โ Kieran remarks, explaining that with a well-defined plan, AI models are exceptionally good at โjust following steps, doing deep work, working for hours, days even now.โ This raised a provocative question: if AI could handle the core execution, and even improve its own planning and review, where did humans fit in? โDid I automate myself out of a job?โ Kieran mused.
The AI Sandwich: Where Humans Shine
The answer, they discovered, lies at the extremities of the work process, giving rise to the โAI Sandwichโ metaphor. โHumans are the bread in the sandwich and the AI is in the middle,โ explains Dan Shipper, the host of the Every podcast. โThe AI is whatever you put on your sandwich.โ
This means humans excel, and indeed are indispensable, in two critical phases:
The First Slice: Brainstorming, Ideation, and Setting the Frame
Before any plan can be executed, before any code can be written, or any design rendered, thereโs a crucial stage of exploration and definition. Trevan introduced โBrainstormโ and โIdeateโ steps to Compound Engineering, pushing the human role to the very beginning.
- Ideate: This phase involves โgoing wide,โ generating a multitude of ideas from diverse perspectives.
- Brainstorm: Here, humans engage deeply with a problem, even when its exact nature or solution isnโt fully clear. Itโs about asking probing questions and genuinely thinking hard, with AI serving as a support rather than the primary driver.
This initial human involvement is vital because, as Dan explains, humans are exceptionally good at โflipping and changing frames.โ Take, for instance, a simple problem: โMy knee hurts.โ An AI might suggest taking Advil (the immediate, obvious solution within a narrow frame). A human, however, can shift the frame: Is it an IT band issue? Am I running on hard surfaces too often? Each re-framing addresses the same problem at a deeper, more systemic level. โOur job is to set the frame or set the bounds within which we solve the problem,โ Dan emphasizes, a task that remains incredibly challenging for AI to do autonomously.
The Second Slice: Polishing, Elevating, and Infusing Soul
Once the AI has done its work, and even after it has been reviewed and tested for functionality, thereโs another critical human touchpoint: the final polish.
โThe beauty comes in when a human looks at it, clicks around and has a feel like, โOh, this doesnโt feel good. We can polish it even more. We can make it even better,โโ Kieran describes. This isnโt about fixing errors, but about elevating the output beyond mere correctness to something truly exceptional. Itโs about infusing โtasteโ and โfeelโ โ qualities that AI, despite its increasing sophistication, struggles to emulate.
Kieran draws an analogy to the Pomodoro technique, where staying with a task beyond its initial completion can lead to unexpected breakthroughs. In that extra time, โsomething beautiful happens because you will go deeper, you will go further than you would do.โ This final human touch ensures the output resonates, feels great, and stands out in a world where generic, AI-generated โslopโ is becoming increasingly common. โItโs very important to make it feel great because the bar is high, the bar will always get higher,โ Kieran warns.
Why Humans Remain Indispensable
The โAI Sandwichโ isnโt just a workflow; itโs a profound statement on the enduring value of human capabilities. Several deep reasons underscore why AI cannot fully replace the human โbreadโ:
- The Art of Ownership: โIf you ship something or do something or make a like a statement in the world, if you want it to be your ownโฆ you cannot fully automate everything. Like it is maybe a little bit like art,โ Kieran states. True art, and truly impactful work, requires a personal connection, a โfrom youโ quality that AI cannot replicate.
- Beyond Genericity: AI models, trained on vast datasets, tend to produce outputs that are, by nature, generic. While they can mimic styles, they lack the unique perspective, worldview, and specific context needed to solve a problem with genuine flair or to create something truly resonant. They are โsuper intelligences that have been kept in a box,โ as Dan puts it, needing human guidance to tune their output to exact, nuanced problems.
- Rare Expertise and Feedback Loops: Many critical decisions in complex work rely on rare expertise and data that is hard to come by. Humans accumulate this through years of experience and intuition, often through limited, high-stakes feedback loops that are difficult for AI to learn from.
- The Moving Target of Creativity: The boundary between โroteโ and โartโ is constantly shifting. As AI automates the rote, humans are freed to explore new frontiers of creativity, continuously pushing the envelope of what is considered โartโ or โbeautiful.โ
- The Music of the Soul: Kieran, with his background in classical composition, likens the human role to a musician or composer. While AI can generate melodies, it cannot capture the raw emotion of a live performance or the spark of creating something โout of nothing.โ The โmiddle partโ (practicing a piece 100 times) is rote, but the initial composition and the final performance are inherently human, infused with feeling and expression.
Riding the Wave: Adapting to the AI Era
For many, the rise of AI sparks anxiety about job security. However, the โAI Sandwichโ offers an optimistic outlook. Software engineers, often seen as the โcanary in the coal mine,โ are not losing their jobs but are evolving. They are becoming more like managers, orchestrating AI agents, and more like product people, focusing on the ultimate user experience and the โfeelโ of the product.
The key, according to Kieran, is to lean into what brings you joy and energy: โWhatever that means to you, that can mean beautiful code, beautiful abstractions, beautiful architecture, beautiful design, beautiful copy. I think itโs very important to lean into what is beautiful to you because then you will find a way to utilize an LLM to make something that gives you energy instead of drains you all the way.โ
This means embracing a future where we offload the mundane, repetitive tasks to AI, freeing ourselves to focus on the higher-order, more creative, and more uniquely human aspects of work โ the parts that involve deep thought, personal touch, and the pursuit of true excellence. Itโs about recognizing that the โfinal thing thatโs not automatable is art made by humans who feel something.โ
The AI Sandwich isnโt just a metaphor; itโs a practical framework for navigating the evolving landscape of work. It empowers us to leverage AIโs incredible capabilities while preserving and elevating the irreplaceable value of human creativity, intuition, and passion. By embracing our role as the โbread,โ we donโt just survive in an AI world โ we thrive, creating work that is not only efficient and effective but also deeply personal, beautiful, and profoundly human.
Based on โWhy Does Everyone Hate Rats? (Update) | Freakonomics Radioโ from Freakonomics Radio Network Watch the original video
The Unpopular Truth About Our Urban Co-Inhabitants: Why We Really Hate Rats
Itโs springtime in New York City, and as the city greens and sidewalk cafes bustle, another, less welcome, sign of the season emerges: the rats. Their sudden visibility often sparks a familiar wave of revulsion and a renewed call to arms. But this visceral hatred, as a recent Freakonomics Radio episode explores, is more complicated than it seems, rooted in centuries of misinformation, cultural biases, and our own urban failures.
The cityโs ongoing โwar on ratsโ took a particularly dramatic turn in 2022 when then-Mayor Eric Adams posted a job listing for a โcitywide director of rodent mitigation.โ The ideal candidate, it read, should be โhighly motivated and somewhat bloodthirsty,โ possessing a โswashbuckling attitude, crafty humor, and a general aura of badassery.โ This was no ordinary civil servant Adams sought; he was searching for a hero in his deeply personal crusade against an animal he openly detested. โRats do something to traumatize you,โ Adams declared, recalling the unforgettable horror of a rat running across his foot.
That hero turned out to be Katy Corradi, a biologist and urban sustainability expert with a masterโs degree in the field. As New York Cityโs first โrat czar,โ Corradi inherited a formidable challenge. While her title might evoke images of a ruthless exterminator, her approach is far more nuanced, focusing on the fundamental principles of pest management: sanitation.
New Yorkโs Unending Battle: The Rat Czarโs Perspective
Corradiโs journey to becoming the rat czar began with a childhood petition against rats behind her Brooklyn home, a testament to her early understanding of their habitat needs. โThey need a place to live and they need food to eat,โ she explains, highlighting how neglected spaces with overgrown brush provide ideal conditions for burrowing and nesting. Later, working in the cityโs Department of Education, she tackled rat mitigation by improving waste management in public schools, recognizing the inseparable link between garbage and rodents.
For rats, New York City is a paradise. Its dense human population generates an endless buffet of food waste, particularly visible in the overflowing trash bags on sidewalks and the proliferation of outdoor dining sheds that emerged during the COVID-19 pandemic. Estimates of the cityโs rat population vary wildly, from 3 million to 8 million, but Corradi wisely avoids specific numbers, deeming them โfutileโ and easily distorted.
Rats, or Rattus norvegicus (the brown rat we commonly encounter), are โcommensalโ animals, a term Corradi defines as literally meaning โsitting at the table with us.โ They are adept at exploiting and thriving in human-dense environments, demonstrating a remarkable savviness. โIn terms of adaptability to survive, thereโs few species greater,โ Corradi notes. Research even suggests rats exhibit empathy, altruism, and a form of โlaughing,โ challenging our simplistic view of them as mere vermin. They are survivors, and as Stephen Dubner puts it, โno one except humans exploits an urban space better.โ
The Ratโs Tarnished Image: Re-examining the Plague
Our deep-seated aversion to rats is not new. Their ancestors arrived in the Americas in the 18th century on European ships, but the historical timeline of their bad reputation is even more complex. Bethany Brookshire, a science journalist and author of โPests: How Humans Create Animal Villains,โ explains that while people disliked rats for spoiling food, they werenโt considered โdisgustingโ or major disease carriers until the 18th or 19th century.
The most famous association, of course, is with the Black Death, the devastating bubonic plague that ravaged Europe in the 14th century, killing half its population. Popular culture, from Netflix shows to classic horror films, continues to depict swarms of rats as harbingers of plague. But recent scientific inquiry challenges this long-held belief.
Nils Christian Stenseth, a professor of ecology and evolution at the University of Oslo and a leading plague researcher, argues that rats were likely not the primary culprits for the Black Death. His 2018 paper in the Proceedings of the National Academy of Sciences suggested that โhuman ectoparasites, like body lice and human fleas, might be more likely than rats to have caused the rapidly developing epidemics.โ The reasoning, as Harvard economist Ed Glaeser explains, is that rat-led plagues would spread more slowly because fleas only leave a rat when it dies. The rapid spread of the Black Death across Europe points to human-to-human transmission via parasites.
However, rats were implicated in the third major bubonic plague pandemic, which began in the 19th century and persists today in small numbers, particularly in India, Vietnam, and parts of the US. This โthird explosionโ did involve rats and their fleas transmitting the Yersinia pestis bacterium, which forms a biofilm in the fleaโs esophagus, causing it to regurgitate bacteria into every bite.
So, while the ratโs โguiltโ in the Black Death might be exaggerated, its reputation has been permanently scarred. โBlaming the rat is pretty much, you know, game over in terms of the ratโs global reputation,โ Glaeser observes. However, he quickly adds, โI think we should also just object to using the word guilt on rats. Itโs not like they know whatโs going on. Theyโre dying, tooโฆ Letโs go to Yersinia pestis itself. Thatโs where the evil lies.โ
Today, the public health risk from rats, while real, is often overstated. Leptospirosis, a bacterial illness transmitted through rat urine, saw 24 reported cases in New York City in 2023 โ a minor threat in a city of 8 million. However, Corradi points to โunrealized potential public health risks,โ such as novel viruses found on rats by Columbia University researchers, which pose a constant threat of mutation and host-jumping due to ratsโ close proximity to humans.
The Psychology of Disgust: Why We Hate Their Success
Beyond disease, our revulsion towards rats is deeply psychological and cultural. Bethany Brookshireโs work on โpestsโ highlights that our hatred for certain animals is โso subjective.โ Animals are simply โbeing animals,โ but we categorize them as โpestsโ when they are โnot where we want them to be.โ
Consider the pigeon, once a revered messenger, food source, and fertilizer provider. Colin Jerolmackโs research shows how the pigeon, over a century, devolved from โnoble, innocent, beautifulโ to โrats with wingsโ in the public eye. As we developed chemical fertilizers, email, and chicken, our utility for pigeons vanished, and their continued presence became an annoyance. โWe used to have such a use for them and now we donโt and we canโt fathom why they wonโt go away,โ Brookshire laments.
Rats, unlike pigeons, have never truly served a widespread human purpose in Western cultures. Yet, in other cultures, their perception is entirely different. The Karni Mata Temple in Deshnoke, India, houses around 25,000 black rats, considered sacred reincarnations of people. Devotees worship them, providing elaborate food and milk offerings, walking barefoot on marble floors shared with the holy rodents.
This stark contrast, Brookshire argues, stems from cultural narratives. Western cultures, often influenced by concepts like โdominion over animalsโ from texts like Genesis, expect to control their environment. When animals like rats thrive in our โpaved overโ spaces, it challenges our perceived control. โWe really hate them. We hate their success because their success feels like our failure,โ she concludes.
Our Urban Partners: Towards Sympathy for the Rat
Economist Ed Glaeser, an expert on cities, views rats as โagents of usually negative externalitiesโ โ spreading disease, damaging property, and reducing the โdensity level for peopleโ by making urban spaces less appealing. Rats are known to chew through wires, foundations, and food supplies, creating economic costs and, as Corradi emphasizes, significant mental well-being issues like stress, anxiety, and depression for those living in close quarters with them.
However, Glaeser also offers a more nuanced perspective. While acknowledging the need for โsome control,โ he cautions against โmaking a fetish out of complete eradication.โ He suggests a shift in perspective: โItโs hard not to think that rats have gotten something of a bad rapโฆ they sort of co-live with humans, theyโre in some sense our natural city partner.โ
This idea of โsympathy for the ratโ resonates with Dubner. He argues that rather than viewing them with horror, we should see them as โour urban partner.โ This perspective acknowledges their role in our shared ecosystem, even as we manage their populations.
New York Cityโs war on rats continues, with recent reports showing a 20% year-over-year decline in rat sightings. Current Mayor Zoran Momdani is pushing for citywide trash containerization by 2031, a policy Corradi would surely endorse as the fundamental solution to limiting ratsโ food access.
Ultimately, the story of why everyone hates rats is less about the animals themselves and more about us โ our history, our perceptions, and our choices in how we design and manage our urban environments. Perhaps itโs time to move beyond the โbloodthirstyโ rhetoric and cultivate a more informed, perhaps even empathetic, understanding of our most reviled urban co-inhabitants. After all, as Brookshire wisely notes, โwe have choices in the way that we treat other animals and we have choices in the way we treat each other. And we donโt need to live the way that we always have.โ
Based on โClaude Design is slow and I love it anyway (plus why I love ChatGPT Images 2.0)โ from How I AI Watch the original video
The AI Design Paradox: Embracing Claudeโs Deliberate Pace and ChatGPTโs Creative Leap
The world of AI-powered design is exploding, with new tools promising to revolutionize how we build and create. As a product leader and AI enthusiast, Claire Valle has been at the forefront, exploring these innovations to understand their true potential for businesses and creative workflows. This week, the spotlight falls on two major players: Anthropicโs Claude Design, a web-based design tool taking a swing at prototyping and presentations, and OpenAIโs new GPT Image 2 model, which claims a โnew era of image generation.โ
While both tools offer significant advancements, they also reveal a fascinating paradox in the current state of AI: the tension between groundbreaking capabilities and the sometimes-sluggish reality of their execution.
Claude Design: A New Approach to Prototyping
Anthropicโs Claude Design has entered the arena with a bold vision, aiming to simplify the creation of prototypes, wireframes, high-fidelity designs, slides, and even videos. The critical question on everyoneโs mind: Is this the Figma killer? Claireโs take is nuanced: it might not replace Figma entirely, but it certainly has the potential to disrupt the ecosystem of prototyping tools used before handing off designs to engineering.
Design Systems as a First-Class Citizen
What truly sets Claude Design apart is its revolutionary approach to design systems. Unlike many prototyping tools where integrating brand guidelines can be an afterthought or a cumbersome process, Claude Design makes the design system a โfirst-class citizen.โ This means the tool is designed from the ground up to understand and adhere to your brandโs visual language.
Claireโs initial test involved importing a design system for Lennyโs Newsletter. The process is surprisingly straightforward: provide HTML, logos, images, and fonts. Claude then embarks on a โreasoningโ journey, analyzing these materials to extract core colors, typography, components, and brand marks. This intelligent extraction aims to build a structured design system that AI can then use to render consistent, high-quality UI.
The tool even warns you that this process takes about five minutes, a detail that hints at the โdeliberate paceโ characteristic of current large language models. The result, however, is impressive, producing a system that closely mirrors the original brand. This focus on structured design systems is gaining traction across the industry; Google Labs, for instance, recently introduced the design.md standard, an effort to standardize how design systems are described for AI tools. This trend underscores a broader shift towards AI-native design workflows where consistency is paramount.
Building with Brand: Marketing Pages and Presentation Power
Claude Design truly shines when tasked with creating marketing-specific landing pages and websites that require strict adherence to a brandโs aesthetic. For marketers struggling to translate brand assets into beautiful, functional prototypes, Claude Design offers a compelling solution.
To demonstrate, Claire challenged Claude to create a landing page for a hypothetical โLennyDoc PRD builder,โ a direct competitor to her own Chat PRD, using Lennyโs Newsletter design system. The process involves a conversational Q&A, allowing users to guide the AI on audience, desired sections, interactivity, and even pricing. A standout feature is the ability to generate multiple variations (defaulting to three), offering diverse design directions without requiring iterative prompting. As Claire notes, โThis is really smart from a design tool perspective because often those cycles of no, make it better or make it different can be very slow, and most people come into Claude design probably donโt have the ability to articulate exactly the changes they make.โ
However, this is where the โdeliberate paceโ becomes a tangible challenge. The generation of a single landing page can take anywhere from 5 to 10 minutes. This slowness, coupled with Anthropicโs credit limits (Claire quickly hit hers, requiring a $200 top-up), highlights a significant hurdle for rapid iteration. โI think we underestimate how nice that is from a speed of iteration perspective when youโre building and designing things,โ Claire observes, contrasting it with Figmaโs instant drag-and-drop responsiveness. Despite the wait, the generated LennyDoc page impressively captures the brandโs essence, though Claude does have a โslop tellโ: an affinity for italicized serif fonts on landing pages.
Beyond landing pages, Claude Design excels at generating presentation slides. By simply dropping a PDF article and selecting a design system, the tool can craft a visually appealing and brand-aligned deck. For product marketers, this capability is a game-changer for training materials, enablement content, and customer presentations. The delightful inclusion of interactive elements, like a โcute fake terminalโ for code commands, hints at a future where slides are essentially code, offering dynamic and engaging experiences.
The Joy of โUglyโ Design: Unleashing Creativity
Perhaps the most surprisingly fun use case for Claude Design emerges when you remove the constraints of a design system. Claire experimented with asking Claude to create a 90s GeoCities-style version of Lennyโs Newsletter homepage. The result, dubbed โLennyโs Product Zone,โ is a hilariously accurate and intentionally โuglyโ rendition, complete with brick backgrounds and Comic Sans fonts.
This exercise reveals a โsecret powerโ of Claude Design: its exceptional copywriting. When given the freedom to generate content, it produces engaging and often humorous text, like โYour OKRs are cringe and seven ways to fix them before Q3.โ This showcases the AIโs ability to not just design visually but also to craft compelling narratives, an often-underappreciated aspect of great design.
ChatGPT Images 2.0: Thinking Beyond Pixels
Just as Claude Design pushes the boundaries of structured UI, OpenAIโs new GPT Image 2 model is redefining image generation. Billed as the โfirst model that can do thinking,โ its core advancements lie in accurately rendering text and objects within images, tackling long-standing challenges in AI art.
Text, Layout, and โThinkingโ: The New Frontier
The ability to accurately render text within an image has been a holy grail for AI image models, often resulting in garbled or nonsensical characters. GPT Image 2 appears to have largely overcome this, producing legible typography that doesnโt carry the tell-tale โAI image textโ look. Furthermore, its proficiency in layout and composition suggests a deeper understanding of visual hierarchy and design principles. This โthinkingโ capability allows it to arrange elements thoughtfully, moving beyond mere pixel generation to more intelligent design.
Building a Brand Kit from Scratch (and Reference Images)
One compelling use case Claire explored was generating a multi-page brand kit for her company, Chat PRD. Starting with a general prompt, GPT Image 2 produced a nine-grid layout of brand elements with clean typography. While good, it didnโt quite capture Chat PRDโs specific aesthetic.
This is where the iterative power of the new model comes into play. Claire uploaded reference images from Midjourney โ brighter, pinker, pixelated landscapes that defined her brand. The instruction was simple: โThatโs not really us. Here are some reference images. Update the brand kit.โ The model responded by generating a new brand kit that perfectly integrated the pixelated style and vibrant pinks, offering a fresh perspective on the brandโs visual identity. This workflow, combining initial generation with image-based iteration, addresses a common dissatisfaction among marketers who find existing image models struggle to align with specific brand assets and voice.
Personal Style Analysis: A Fun, Practical Use Case
Beyond corporate branding, GPT Image 2 also demonstrates its versatility with more personal applications. Claire uploaded a photo of herself and asked for a color analysis, a common practice to determine flattering shades. The AI provided an analysis identifying her as โwarm neutralโ with best colors in earthy tones, even rendering her image in these colors. While the initial facial rendering was โreal weirdโ and โteeth looking kind of funk,โ it was an impressive attempt.
More importantly, when Claire corrected the AI โ stating she was actually a โdark winterโ and clarifying her natural hair color โ the model course-corrected beautifully. It generated an accurate color palette for a dark winter, along with a sophisticated layout that combined text, color swatches, and images. The ability to perform such complex analysis, combine various visual and textual elements, and present them in a clean, professional layout suggests a significant leap in graphic generation. Claire even muses about switching her companyโs infographics to GPT Image 2 due to its โmore expensive and a little nicerโ aesthetic.
The Road Ahead: Patience, Progress, and Popping Tokens
The spring 2026 drop of AI design tools presents a thrilling glimpse into the future. Claude Design offers a robust framework for brand-consistent prototyping and presentations, while GPT Image 2 ushers in a new era of intelligent image generation, particularly in text and layout.
Yet, as Claire Valle aptly summarizes, โAt the end of the day, I donโt think any of these tools are dead yet. They all have their benefits. Theyโre all slow, and theyโre all making me top up my tokens.โ The journey to truly seamless AI-powered design is still unfolding, marked by the need for patience with processing times and the ongoing cost of advanced models. However, the progress is undeniable. From structured design systems to โthinkingโ image models, these tools are not just automating tasks; theyโre fundamentally changing how designers, marketers, and product builders approach creativity and execution. The future of design is here, and it promises to be both deliberate and delightfully disruptive.
Based on โInside Kash Patelโs F.B.I.โ from New York Times Podcasts Watch the original video
Inside the Storm: How Political Winds Rocked the FBI Under Kash Patel
For decades, the Federal Bureau of Investigation has strived to operate as an independent entity, dedicated to following facts โwithout fear or favorโ โ a commitment born from the scandals of Watergate. Yet, under the leadership of Kash Patel, appointed FBI Director during the Trump administration, this foundational principle appears to have been severely tested. A comprehensive investigation by New York Times journalists Emily Basilon and Rachel Poser, based on interviews with 45 current and former FBI employees, reveals an agency straining under unprecedented political pressure, its mission transformed, and its seasoned professionals deeply concerned for the safety and security of the nation.
The reporting paints a stark picture of an FBI leadership prioritizing political optics over fundamental law enforcement duties, leading to a palpable culture of fear, demotions, and resignations among agents and analysts across the country.
A Controversial Appointment and a Shift in Mission
Kash Patelโs selection as FBI Director was highly unusual. Lacking prior experience within the FBI or extensive federal law enforcement, Patelโs background included stints as a public defender and an intelligence official during the first Trump term. Those who knew him described him as ambitious, cocky, and full of bluster, with a history of spinning conspiracy theories about the bureau itself. He famously asserted that he would โshut down the FBI Hoover building on day one and reopening the next day as a museum of the deep state,โ a sentiment that immediately raised alarm bells both inside and outside the agency.
President Trump, who had openly called the FBI a โvery corrupt institutionโ and claimed to be โa victim of it,โ clearly signaled a desire for change. This directive was quickly echoed by the administration, which declared the โDepartment of Justice into the Department of Injusticeโ days were โover.โ For many within the bureau, this signaled a dramatic shift from their long-held tradition of political neutrality.
Tanya Ugoritz, then head of the Directorate of Intelligence at the FBI, earnestly tried to set Patel up for success. She prepared to deliver his morning intelligence briefings, even reading his book for insights. However, she quickly realized this was not business as usual.
Decisions Driven by Optics, Not Analysis
Patelโs tenure began with an immediate and impactful order: to move hundreds of field agents from Washington D.C. to field offices across the country. While some insiders acknowledged that the bureau had become too D.C.-heavy, the execution of this directive was described as โrandom and arbitrary,โ lacking the traditional โvery sober and considered analysisโ the FBI was known for. Ugoritz saw this as the first instance of a pattern: โdecisions first and then everybody scramble and figure out how to make it happen later.โ This signaled that Patel would make abrupt, sudden decisions affecting investigations without the usual meticulous planning.
The appointment of Dan Bongino as Patelโs deputy director further solidified these concerns. Bongino, a former Secret Service agent turned pro-Trump podcaster and Fox News commentator, had a history of conspiratorial views, once describing FBI agents as โthugs for the Democratic partyโ and advocating for the bureauโs disbandment. For agents like John Sullivan, an intelligence section chief, Bonginoโs selection was a critical turning point where he โrealized in that moment that it was not going to be okay.โ
Together, Patel and Bongino initiated a strong push for optics. Sources reported that President Trump, observing footage of raids without visible FBI โflack jackets,โ expressed anger, leading to a directive to prioritize such visual elements. Meanwhile, Patel himself participated in videos at Quantico, โcosplaying as Rambo,โ complete with explosions and helicopter rappelling. While intended to project an image of toughness, these actions were widely seen as โchildishโ and a drain on critical resources, diverting time and money from serious investigative work.
This prioritization of image over mission was glaringly evident in the investigation following the shooting of Charlie Kirk. Patel, unusually, took over an executive call during an active manhunt, berating the special agent in charge and dictating social media strategy. This rushed approach led to โmaking mistakesโ and โputting wrong information out into the world,โ deeply frustrating agents who understood the dangers of misdirection in an ongoing investigation.
The focus on optics even strained international alliances. At a secret intelligence conference in the UK, Patel reportedly attempted to post a photo taken with the king at Windsor Castle, despite clear instructions that some participants were โnon-disclosedโ intelligence operatives whose faces were not to appear in the press. This created a โminor international incident,โ underscoring how Patelโs priorities were at odds with the sensitive nature of intelligence work.
When confronted with these anecdotes, FBI spokesman Ben Williamson dismissed the reporting as โa regurgitation of fake narratives, conjecture, and speculation from anonymous sources who are disconnected from reality.โ However, the reporting highlighted that many sources were named, and the consistent narrative from numerous individuals painted a different picture.
The Rank and File: Reassignments, Red Lines, and Resignations
The transformation under Patelโs leadership profoundly impacted the day-to-day work of FBI employees. A significant trend was the Trump administrationโs emphasis on immigration, a domain where the FBI traditionally played no enforcement role. Agents and analysts across the country found themselves reassigned from their core dutiesโsuch as public corruption, cybercrime, white-collar crime, drug trafficking, and terrorismโto assist with immigration enforcement.
Jill Fields, a dedicated intelligence analyst in Los Angeles who specialized in violent crime, experienced this shift firsthand. When a major immigration push occurred, she and her team were tasked with unfamiliar duties. She recalled being told to pull analysts off other critical teams because โweโve got to do this for optics. Weโve got to make a show for the president.โ
The situation escalated when Fields was asked to cross what she considered a โred line.โ After a group of protesters filmed agents and used a megaphone to warn residents about ICE presence, Fieldsโ team was instructed to run a โpreassessment checkโโthe first step toward a criminal investigationโon these individuals. Her team reviewed cellphone video and concluded that the protesters were merely exercising their First Amendment rights and had committed no wrongdoing. Yet, they were explicitly ordered to open an investigation anyway.
Fields described the order as โludicrous,โ akin to a scenario in legal training where the correct answer would be to refuse. When she pushed back, she was told, โyou can get fired today or you can get fired in 40 years when another administration comes in and starts looking and seeing who had violated the law and who had followed and acquiesced.โ This chilling admission suggested that the leadership was aware of the questionable legality of their demands.
In response to her defiance, Fieldsโ squad was disbanded, she was reassigned, and informed she was โa problemโ and that the โseventh floorโโwhere the director sitsโwas aware of her. Feeling monitored and knowing her career was effectively over, Jill Fields decided to leave the bureau, taking with her a careerโs worth of dedication and expertise. โIt hurt to know that things had changed so much that I was not going to be able to continue doing the job that I loved,โ she lamented.
This incident also highlighted the erosion of the FBIโs โpost-Hoover mindset,โ a commitment to balancing its mission with American civil rights, ingrained in agents after historical abuses.
The Purge: Loyalty Tests and Political Dismissals
The political reorientation of the bureau also manifested in a systematic push to remove employees whose past work was deemed unfavorable. Patel had openly spoken about โpurging the agencyโ of anyone who had investigated the president.
Tanya Ugoritz found herself caught in this dragnet. In 2020, the FBI had issued an intelligence report about a second-hand tip alleging Chinese government involvement in creating fake IDs for the 2020 election. The report was later withdrawn due to questions about its credibility. Fast forward to 2025, Senator Chuck Grassley, chair of the Senate Judiciary Committee, began seeking internal FBI documents related to alleged โsuspect bad thingsโ the FBI did during the Biden administration. An email wrongly identified Ugoritz as the official who ordered the reportโs withdrawal.
Despite being the number two official, Dan Bongino told Ugoritz that the matter was โout of his hands.โ The next day, she was placed on administrative leave. An internal investigation and even a polygraph found no misconduct on her part. Yet, when she asked to return to her position, she was told that as a senior executive, her role was โat the discretion of the directorโโPatel. She was denied her job back, demoted, and ultimately decided to leave. Ugoritz expressed profound sadness at how โcasually discard[ing] such people as if they were like a used up tissueโ demonstrated a โcarelessโ disregard for the organizationโs integrity and the experience it was losing. She felt she was a โscapegoatโ to avoid awkward explanations to Senator Grassley.
Blair Tolman, another supervisory special agent, experienced a similar fate. She had led an elite public corruption unit, CR15, which investigated government employees and conducted the โArctic Frostโ investigation into President Trumpโs alleged interference in the 2020 election. This investigation became a major target for Trump and his critics. Tolman watched as agents she had personally selected for their skill and impartiality were fired. Eventually, her own termination letter arrived, citing โlack of judgment and lack of impartiality that led to the political weaponization of the government.โ For Tolman, it was clear: โwhat that meant was you were on this investigation and weโre firing you for it.โ
A Culture of Fear and a Compromised Mission
The firings and reassignments have instilled a โculture of fearโ and โparanoiaโ throughout the bureau. Dozens of employees have been dismissed, and the ongoing nature of these actions has made agents wary of taking on assignments that might be perceived as political, fearing they could be targeted by a future administration. This chilling effect means that if a tip about corruption or even a terrorist threat involves someone perceived as an ally of the administration, agents might hesitate to act, leading to missed threats and a less safe America.
The internal checks and balances, the โwatchdogs,โ and the channels for speaking to supervisors are widely perceived as either shut down or not functioning properly. This has driven many typically โbuttoned up, secretive agencyโ employees to speak to the media, viewing the courts and the press as the โlast open avenuesโ to convey their deep concern that the FBIโs core mission is being โseverely compromised.โ
Blair Tolman is now a leading plaintiff in a lawsuit challenging these dismissals, describing it as a โpolitical purge.โ The lawsuit seeks reinstatement and โdue process,โ but also aims to โmake sure that others are protected now and in the future.โ
The journalists concluded that while many within the FBI acknowledge the need for reform, the methods employed under Patelโs leadershipโcharacterized by abrupt changes, political prioritization, and a disregard for established protocolsโhave been detrimental. As one source stated, โthe ways in which you change an organization like the FBI are not by cutting off chunks of it and hoping that it grows back better.โ The fear is that โcertain threats just arenโt getting addressed because thereโs no one there to work them,โ leaving the nation vulnerable. โI donโt know how we rebuild after this,โ one agent confided, โI worry about what is being missed.โ
While Patelโs future remains uncertain, with mixed signals from the White House despite public statements of support, the profound impact of his tenure on the FBIโs independence, morale, and effectiveness is undeniable. The struggle for the soul of Americaโs premier law enforcement agency continues.
ํ๊ตญ์ด
โThe AI Sandwich: Where Humans Excel in an AI Worldโ โ Every ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI ์๋, ์ธ๊ฐ์ ์ง์ ํ ๊ฐ์น๋ฅผ ์ฐพ์๋ผ: โAI ์๋์์นโ ๋ชจ๋ธ์ ํต์ฐฐ
์ธ๊ณต์ง๋ฅ(AI) ๊ธฐ์ ์ด ์ ๋ก ์๋ ์๋๋ก ๋ฐ์ ํ๋ฉด์ ๋ง์ ์ฌ๋๋ค์ ๋ฏธ๋์ ์ ๋ฌด ํ๊ฒฝ๊ณผ ์ธ๊ฐ์ ์ญํ ์ ๋ํด ๋ถ์๊ฐ์ ๋๋ผ๊ณ ์์ต๋๋ค. ๊ณผ์ฐ AI๋ ์ฐ๋ฆฌ์ ์ผ์๋ฆฌ๋ฅผ ๋นผ์๊ณ ์ธ๊ฐ์ ์ญํ ์ ๋์ฒดํ ๊น์? ์๋ธ๋ฆฌ(Every) ์ฑ๋์ ์ ํ๋ธ ๋น๋์ค์์ ์ฝ๋ผ(Kora)์ ์ ๋๋ด ๋งค๋์ ์ด์ ์ปดํ์ด๋ ์์ง๋์ด๋ง(Compound Engineering)์ ์ฐฝ์์์ธ ํค์ด๋ฐ(Kieran)๊ณผ ํธ์คํธ ๋ ์ฌํผ(Dan Shipper)๋ ์ด๋ฌํ ์ง๋ฌธ์ ๋ํ ํ์ ์ ์ธ ํด๋ต์ ์ ์ํฉ๋๋ค. ๋ฐ๋ก โAI ์๋์์นโ ๋ชจ๋ธ์ ๋๋ค.
์ด ๋ชจ๋ธ์ AI ์๋์ ์ธ๊ฐ์ด ์ด๋์์ ํ์ํ ๋ฅ๋ ฅ์ ๋ฐํํ๊ณ , AI์ ์ด๋ป๊ฒ ํ๋ ฅํด์ผ ํ๋์ง์ ๋ํ ๋ช ํํ ์ฒญ์ฌ์ง์ ์ ๊ณตํฉ๋๋ค. ์ธ๊ฐ์ ์๋์์น์ โ๋นตโ์ฒ๋ผ ์์๊ณผ ๋์ ๋ด๋นํ๋ฉฐ, AI๋ ๊ทธ ์ฌ์ด์ โ์์ฌ๋ฃโ ์ญํ ์ ์ํํ๋ค๋ ๊ฒ์ด ํต์ฌ์ ๋๋ค.
โAI ์๋์์นโ์ ํ์ ๋ฐฐ๊ฒฝ: ์ปดํ์ด๋ ์์ง๋์ด๋ง(Compound Engineering) ์ฌ์
ํค์ด๋ฐ์ ์์ด์ ํธ(agent) ๊ธฐ๋ฐ์ ์ฝ๋ ์์ ์ ํ์ ํ๋ ์์ง๋์ด๋ง ํ๋ ์์ํฌ์ด์ ํ๋ฌ๊ทธ์ธ์ธ ์ปดํ์ด๋ ์์ง๋์ด๋ง์ ๊ฐ๋ฐํ๋ฉฐ โAI ์๋์์นโ ๋ชจ๋ธ์ ๋ํ ํต์ฐฐ์ ์ป๊ฒ ๋์์ต๋๋ค. ๊ทธ๋ ์ฝ๋ผ๋ฅผ ๊ตฌ์ถํ๋ฉด์ AI๋ฅผ ํ์ฉํด ๋ ๋น ๋ฅด๊ณ ํจ์จ์ ์ผ๋ก ์์ ํ๋ ๋ฐฉ๋ฒ์ ๊ณ ๋ฏผํ๊ณ , ์ด๊ธฐ ์ปดํ์ด๋ ์์ง๋์ด๋ง์ ๋ค ๊ฐ์ง ํต์ฌ ๋จ๊ณ๋ก ๊ตฌ์ฑ๋์์ต๋๋ค.
- ๊ณํ(Planning): ๋ฌด์์ ๋ง๋ค๊ณ ์ํํ ์ง ๋ช ํํ ๊ณํ์ ์ธ์๋๋ค.
- ์์ (Work): AI ์์ด์ ํธ๊ฐ ์ค์ ์์ ์ ์ํํ๊ณ ์ฝ๋๋ฅผ ์์ฑํ๊ฑฐ๋ ๋์์ธ ์์ ์ ์งํํฉ๋๋ค.
- ๊ฒํ (Review): ๊ฒฐ๊ณผ๋ฌผ์ ๊ฒํ ํ๊ณ ๊ฐ์ ์ ์ ์ฐพ์ต๋๋ค. ์ ํต์ ์ธ ์ฝ๋ ๋ฆฌ๋ทฐ์ ์ ์ฌํฉ๋๋ค.
- ๋ณตํฉํ(Compound): ๊ฒํ ๊ณผ์ ์ด๋ ๊ณํ ๋จ๊ณ์์ ์ป์ ์ค์ํ ํ์ต ๋ด์ฉ์ ์์คํ ์ ๋ค์ ํตํฉ(compound)ํฉ๋๋ค. ์ด๋ ์ง์์ผ๋ก ์ ์ฅ๋์ด ์์ด์ ํธ๊ฐ ๋ค์ ์์ ์ ์ด์ ์ ์ ์ง๋ฅธ ์ค์๋ฅผ ๋ฐ๋ณตํ์ง ์๋๋ก ๋๋ ๊ฐ์ฅ ๊ฐ๋ ฅํ ๊ธฐ๋ฅ์ ๋๋ค.
์ด ๊ณผ์ ์์ ํค์ด๋ฐ๊ณผ ๊ทธ์ ํ์ ํฅ๋ฏธ๋ก์ด ์ฌ์ค์ ๋ฐ๊ฒฌํ์ต๋๋ค. ์ผ๋จ ์ข์ ๊ณํ์ด ์ธ์์ง๋ฉด โ์์ โ ๋จ๊ณ๋ AI๊ฐ ๋งค์ฐ ์ ์ํํ๋ค๋ ๊ฒ์ ๋๋ค. ๋๊ท๋ชจ ์ธ์ด ๋ชจ๋ธ(LLM, Large Language Model)์ ์ง์๋ฅผ ๋ฐ๋ฅด๊ณ ๊น์ด ์๋ ์์ ์ ๋ช ์๊ฐ, ์ฌ์ง์ด ๋ฉฐ์น ๋์ ์ํํ๋ ๋ฐ ํ์ํ ๋ฅ๋ ฅ์ ๋ณด์์ต๋๋ค. โ์์ โ ๋จ๊ณ๊ฐ ๊ฑฐ์ โํด๊ฒฐโ๋ ์์ญ์ด ๋์, ํ์ ์ธ๊ฐ์ ์ญํ ์ด ์ด๋์ ๋จ์์๋์ง ์ง๋ฌธํ๊ธฐ ์์ํ์ต๋๋ค. โ๋ง์ฝ ๋ชจ๋ ๊ฒ์ด ์๋ํ๋ค๋ฉด, ๋ด ์ผ์๋ฆฌ๋ ์๋ํ๋๋ ๊ฒ์ธ๊ฐ?โ
์ธ๊ฐ๊ณผ AI์ ์ต์ ํ์ ๋ชจ๋ธ: ์๋์์น ๋น์ ์ ์ฌ์ธต ๋ถ์
์ด๋ฌํ ์ง๋ฌธ์ ๋ํ ๋ต์ ์ปดํ์ด๋ ์์ง๋์ด๋ง ํ๋ฌ๊ทธ์ธ์ ํฌ๊ฒ ๊ธฐ์ฌํ ์ ํ ์ ๋ฌธ๊ฐ ํธ๋ ๋ฐ ์ฐจ์ฐ(Trevan Chow)์ ์ถ๊ฐ ์์ ์์ ์์๋์์ต๋๋ค. ๊ทธ๋ โ๊ณํโ ๋จ๊ณ ์ด์ ์ โ๋ธ๋ ์ธ์คํ ๋ฐ(Brainstorm)โ๊ณผ โ์์ด๋์ด ๊ตฌ์(Ideate)โ ๋จ๊ณ๋ฅผ ์ถ๊ฐํ์ต๋๋ค.
AI์ ๊ฐ์ : ํจ์จ์ ์ธ ์ค๊ฐ ๋จ๊ณ ์๋ํ
AI๋ ์ ์ ์๋ ์์ ์ ์คํํ๋ ๋ฐ ํ์ํฉ๋๋ค. โAI ์๋์์นโ ๋ชจ๋ธ์์ AI๋ ์๋์์น์ โ์์ฌ๋ฃโ ์ญํ ์ ๋ด๋นํฉ๋๋ค. ์ฆ, ๊ณํ, ์์ , ๊ฒํ ์ ๊ฐ์ ์ค๊ฐ ๋จ๊ณ์ ๋๋ถ๋ถ์ ํจ์จ์ ์ผ๋ก ์๋ํํ ์ ์์ต๋๋ค. LLM์ ์ฃผ์ด์ง ์ง์๋ฅผ ์ ํํ ๋ฐ๋ฅด๊ณ , ๋ณต์กํ ์ฝ๋ ์์ฑ์ด๋ ๋ฐ์ดํฐ ๋ถ์ ๋ฑ ๋ฐ๋ณต์ ์ด๊ณ ์๊ฐ์ด ๋ง์ด ์์๋๋ ์์ ์ ์ ๋ขฐ์ฑ ์๊ฒ ์ํํฉ๋๋ค. ์ด๋ ์ธ๊ฐ์ด ์ง๋ฃจํ๊ณ ๋ฐ๋ณต์ ์ธ ์ ๋ฌด์์ ๋ฒ์ด๋ ๋ ๊ฐ์น ์๋ ์ผ์ ์ง์คํ ์ ์๋๋ก ๋์ต๋๋ค.
์ธ๊ฐ์ ๋ ๋ณด์ ์ญํ : ์๋์์น์ โ๋นตโ์ด ๋๋ ์๊ฐ
ํ์ง๋ง ์๋์์น๋ ๋นต ์์ด๋ ์์ฑ๋์ง ์์ต๋๋ค. ํค์ด๋ฐ๊ณผ ๋์ ์ธ๊ฐ์ ์ญํ ์ด ์๋์์น์ โ์์โ๊ณผ โ๋โ์ ์๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค.
1. ์์์ : ์์ด๋์ด ๋ฐ์๊ณผ ๋ฌธ์ ์ ์ (๋์ฐ ๋จ๊ณ)
๊ฐ์ฅ ์ฒซ ๋จ๊ณ์ธ โ์์ด๋์ด ๊ตฌ์โ๊ณผ โ๋ธ๋ ์ธ์คํ ๋ฐโ์ ์ธ๊ฐ์ ๊น์ ๊ฐ์ ์ด ํ์์ ์ ๋๋ค. ์ธ๊ฐ์ ๋ฌธ์ ์ ๋ณธ์ง์ ํ๊ณ ๋ค๊ณ , ๋ค์ํ ๊ฐ๋์์ ์ง๋ฌธ์ ๋์ง๋ฉฐ, ์ด๋ค ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ ์ง โํ๋ ์(frame)โ์ ์ค์ ํ๋ ๋ฐ ํ์ํฉ๋๋ค. AI๋ ์ด ๊ณผ์ ์์ ์ธ๊ฐ์ ์ง์ํ ์ ์์ง๋ง, ๊ถ๊ทน์ ์ผ๋ก ๋ฌธ์ ์ ๊ฒฝ๊ณ๋ฅผ ์ ํ๊ณ ํด๊ฒฐ ๋ฐฉํฅ์ ์ ์ํ๋ ๊ฒ์ ์ธ๊ฐ์ ๋ชซ์ ๋๋ค. ์ด๋ ๋จ์ํ ๋ฐ์ดํฐ ์ฒ๋ฆฌ ์ด์์ ๊ฒฝํ, ์ง๊ด, ๊ทธ๋ฆฌ๊ณ ๋งฅ๋ฝ์ ์ดํด๋ฅผ ์๊ตฌํ๋ ์์ญ์ ๋๋ค. ํค์ด๋ฐ์ โ์ธ๊ฐ์ AI๊ฐ ๋ฒ์ฑํ ์ ์๋ ํ๊ฒฝ์ ๊ตฌ์ถํด์ผ ํ๋ฉฐ, ์ฌ๋ฐ๋ฅธ ๊ฒ์ ์ ํํด์ผ ํ๋คโ๊ณ ๋งํฉ๋๋ค.
2. ์ข ์ฐฉ์ : ์ต์ข ๊ฒ์ฆ๊ณผ ์์ ์ ์์ฑ (ํด๋ฆฌ์ฑ ๋จ๊ณ)
๋ชจ๋ ์๋ํ๋ ์์ ์ด ๋๋ ํ์๋ ์ธ๊ฐ์ ์ญํ ์ ์ฌ์ ํ ์ค์ํฉ๋๋ค. AI๊ฐ ์์ฑํ ๊ฒฐ๊ณผ๋ฌผ์ด ๋ชจ๋ ์๊ตฌ ์ฌํญ์ ์ถฉ์กฑํ๊ณ ํ ์คํธ๋ฅผ ํต๊ณผํ๋๋ผ๋, ์ต์ข ์ ์ผ๋ก โ์ด๊ฒ์ด ์ข์๊ฐ?โ๋ฅผ ํ๋จํ๊ณ โ์๋ฆ๋ค์โ์ ๋ํ๋ ๊ฒ์ ์ธ๊ฐ์ ๋ชซ์ ๋๋ค. ์น ๋ธ๋ผ์ฐ์ ์๋ํ ํ ์คํธ๋ ๋ชจ๋ ๊ธฐ๋ฅ์ด ์๋ํ๋ค๊ณ ๋งํ ์ ์์ง๋ง, ์ธ๊ฐ์ด ์ง์ ํด๋ฆญํ๊ณ ์ฌ์ฉํด ๋ณด๋ฉฐ โ์ด๊ฒ์ ๋๋์ด ์ข์ง ์๋คโ, โ๋ ๋ค๋ฌ์ ์ ์๋คโ, โ๋ ์๋ฆ๋ต๊ฒ ๋ง๋ค ์ ์๋คโ๊ณ ํ๋จํ๋ ์๊ฐ์ด ์์ต๋๋ค.
ํค์ด๋ฐ์ ๋ฝ๋ชจ๋๋ก(Pomodoro) ํ ํฌ๋์ ์๋ก ๋ค๋ฉฐ, ์ ํด์ง ์๊ฐ ๋์ ํ๋์ ์์ ์ ๊น์ด ๋ชฐ์ ํ ๋ ์์์น ๋ชปํ โ์๋ฆ๋ค์ด ๊ฒฐ๊ณผโ๊ฐ ๋ํ๋๋ค๊ณ ์ค๋ช ํฉ๋๋ค. ์ด์ฒ๋ผ AI๊ฐ ๋ง๋ ๊ฒฐ๊ณผ๋ฌผ์ โ์ต์ข ์ ์ผ๋ก ๋์ด์ฌ๋ ค ๋์ฑ ํ๋ฅญํ๊ฒ ๋ง๋๋โ ํด๋ฆฌ์ฑ(polishing) ๋จ๊ณ๋ ์ธ๊ฐ์ ์ทจํฅ, ๊ฒฝํ, ๊ทธ๋ฆฌ๊ณ ์์ ์ ๊ฐ๊ฐ์ด ๋ฐํ๋๋ ์ง์ ์ ๋๋ค. โ๋ง์ฝ ์ฐ๋ฆฌ๊ฐ ์ด ์์ ์ ํ์ง ์๋๋ค๋ฉด, ๋ชจ๋ ๊ฒฐ๊ณผ๋ฌผ์ ๋๊ฐ๊ณ ํ๋ฒํด์ง ๊ฒ์ ๋๋ค.โ
AI ์๋, ์ธ๊ฐ์ ์ผ์๋ฆฌ๋ ์ฌ๋ผ์ง๊น? ์ํํธ์จ์ด ์์ง๋์ด์ ์ฌ๋ก
AI๊ฐ ์ธ๊ฐ์ ์ญํ ์ ๋์ฒดํ ๊ฒ์ด๋ผ๋ ์ฐ๋ ค์ ๋ํด ๋์ ์ํํธ์จ์ด ์์ง๋์ด๋ฅผ โํ๊ด ์ ์นด๋๋ฆฌ์(canary in the coal mine)โ์ ๋น์ ํ๋ฉฐ, ๊ทธ๋ค์ ๊ฒฝํ์ด ๋ค๋ฅธ ์ง์ ๋ ธ๋์๋ค์๊ฒ ์ค์ํ ํต์ฐฐ์ ์ ๊ณตํ๋ค๊ณ ๋งํฉ๋๋ค. ์๋ธ๋ฆฌ ๋ด๋ถ์์๋ AI ์์ด์ ํธ์ ๋์ ์๋ ๋ถ๊ตฌํ๊ณ ์ฌ์ ํ ์ํํธ์จ์ด ์์ง๋์ด๋ฅผ ๊ณ ์ฉํ๊ณ ์์ต๋๋ค. ๋ค๋ง, ๊ทธ๋ค์ ์ ๋ฌด ๋ฐฉ์์ ๋ณํํ์ต๋๋ค. ์ด์ ์์ง๋์ด๋ ์ฝ๋๋ฅผ ์ง์ ์์ฑํ๋ ๊ฒ๋ณด๋ค โ๊ด๋ฆฌโ์ ๊ฐ๊น์ด ์ญํ ์ ์ํํ๊ฒ ๋ฉ๋๋ค. ์ฆ, โAI ์๋์์นโ์ ์์๊ณผ ๋์ ์ฐธ์ฌํ๋ ๊ฒ์ด์ฃ .
์ด๋ ์นดํผ๋ผ์ดํ , ์ ๋ต ์๋ฆฝ, ๋์์ธ ๋ฑ ๋ค๋ฅธ ๋ชจ๋ ์ง์ ๋ ธ๋์๋ ๋์ผํ๊ฒ ์ ์ฉ๋ ๊ฒ์ ๋๋ค. ์ธ๊ฐ์ ๋ฌธ์ ์ โํ๋ ์โ์ ์ค์ ํ๊ณ ์ํฉ์ ๋ฐ๋ผ ํ๋ ์์ ์ ์ฐํ๊ฒ ์ ํํ๋ ๋ฐ ๋งค์ฐ ๋ฅ์ํฉ๋๋ค. ์๋ฅผ ๋ค์ด, ๋ฌด๋ฆ ํต์ฆ์ด๋ผ๋ ๋ฌธ์ ๊ฐ ์์ ๋, AI๋ ์ฆ์ ์งํต์ ๋ฅผ ์ถ์ฒํ ์ ์์ต๋๋ค. ํ์ง๋ง ์ธ๊ฐ์ ๋ ํฐ ํ๋ ์ ์์์ โIT ๋ฐด๋๋ฅผ ์คํธ๋ ์นญํด์ผ ํ ์๋ ์๋คโ, โ๋งค์ผ ๋ฑ๋ฑํ ํ๋ฉด์์ ๋ฌ๋ฆฌ๋ ๊ฒ์ ๋ฉ์ถฐ์ผ ํ๋คโ์ ๊ฐ์ด ๋ฌธ์ ์ ๊ทผ๋ณธ ์์ธ์ ํ์ ํ๊ณ ๋ ๋์ ์์ค์ ํด๊ฒฐ์ฑ ์ ์ ์ํ ์ ์์ต๋๋ค. AI ์์ด์ ํธ๊ฐ ์ด๋ฌํ ํ๋ ์ ์ ํ์ ์ค์ค๋ก ์ ์ํํ๊ธฐ๋ ๋งค์ฐ ์ด๋ ค์ธ ๊ฒ์ด๋ผ๋ ๊ฒ์ด ๊ทธ๋ค์ ์ฃผ์ฅ์ ๋๋ค.
AI์ ํ๊ณ์ ์ธ๊ฐ์ ๊ณ ์ ํ ๊ฐ์น: ์์ ๊ณผ ๊ฒฝํ์ ์์ญ
AI๋ 100๋ช ์ ๊ฐ์ ์ธ๋ฌผ์ ์๋ฎฌ๋ ์ด์ ํ์ฌ ์์ด๋์ด๋ฅผ ๊ตฌ์ํ๋ ๋ฑ ์๋ํ ์์ญ์ ๋ํ๊ฐ ๊ฒ์ ๋๋ค. ํ์ง๋ง ํค์ด๋ฐ์ โ๊ถ๊ทน์ ์ผ๋ก ์ธ์์ ๋ฌด์ธ๊ฐ๋ฅผ ๋ด๋๊ฑฐ๋, ์ด๋ค ์ฃผ์ฅ์ ํ๊ฑฐ๋, ๋น์ ๋ง์ ๊ฒ์ผ๋ก ๋ง๋ค๊ณ ์ถ๋ค๋ฉด, ๋ชจ๋ ๊ฒ์ ์์ ํ ์๋ํํ ์๋ ์๋คโ๊ณ ๋จ์ธํฉ๋๋ค. ์ด๋ ๋ง์น ์์ ์ ๋ง๋๋ ๊ฒ๊ณผ ๊ฐ์ต๋๋ค. ์ง์ ์ผ๋ก โ๋๋ง์ ๊ฒโ์ ๋ง๋ค๊ณ ์ถ๋ค๋ฉด, ๊ทธ๊ฒ์ ๋น์ ์๊ฒ์ ๋์ค๊ฑฐ๋ ์ด๋ค ์์ผ๋ก๋ ๋น์ ๊ณผ ์ฐ๊ฒฐ๋์ด์ผ ํฉ๋๋ค.
๋๊ท๋ชจ ์ธ์ด ๋ชจ๋ธ(LLM)์ ์ง๋ ์๋ ๊ฐ โ์์ ์์ ๊ฐํโ ํ์ตํ๊ธฐ ๋๋ฌธ์ ์ธ์์ ๋งฅ๋ฝ์ ์์ ํ ์ดํดํ์ง ๋ชปํ๋ฉฐ, ๊ทธ ๊ฒฐ๊ณผ๋ฌผ์ ๋ค์ โ์ ๋ค๋ฆญ(generic)โํ๊ณ ๊ฐ์ธ์ ์ธ ์ํฉ์ ๋ ๋ง์ถคํ๋ ์๋ฐ์ ์์ต๋๋ค. ์ง์ ์ผ๋ก ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๊ฑฐ๋ ์ฌ๋๋ค์๊ฒ ๊ณต๊ฐ์ ์ป๋ ์ ํ, ์์ ์ ๋ง๋ค๋ ค๋ฉด, ์ ํํ ๋ฌธ์ ์ ํํ์ ๋ง๊ฒ ์ธ๋ฐํ๊ฒ ์กฐ์ ๋์ด์ผ ํฉ๋๋ค. ๊ทธ๋ฆฌ๊ณ ์ด ๊ณผ์ ์์ LLM์ ์ธ๊ฐ์ ๋ง์ ๋์์ด ํ์ํฉ๋๋ค.
๋์ ์ธ๊ณต ์ผ๋ฐ ์ง๋ฅ(AGI, Artificial General Intelligence)์ ๊ธฐ์ค์ โ์์ด์ ํธ๊ฐ 24์๊ฐ ๋ด๋ด ๊ฒฝ์ ์ ์ผ๋ก ์์ต์ฑ์ด ์๊ฒ ์๋ํ๋ ์์ โ์ผ๋ก ์ ์ํ๋ฉฐ, ์ฐ๋ฆฌ๋ ์์ง ๊ทธ ๋จ๊ณ์ ํจ์ฌ ๋ชป ๋ฏธ์น๋ค๊ณ ๋งํฉ๋๋ค. ํ์ฌ์ AI๋ ์ ์ ์๋ ์์ ์ ์ํํ ์๋ ์์ง๋ง, ์ค์ค๋ก ํ๋ ์์ ๋ณ๊ฒฝํ๊ฑฐ๋, ํ ์์ ์ ๋ง์น๊ณ ๋ค์ ์์ ์ ์ ํํ๋ ๋ฑ ๋์์์ด ์ฐฝ์์ ์ธ ๊ฒฐ์ ์ ๋ด๋ฆฌ๋ ์์ค์๋ ๋๋ฌํ์ง ๋ชปํ์ต๋๋ค.
ํค์ด๋ฐ์ ์์ ๋ถ์ผ์ ๊ฒฝํ์ ๋ฐํ์ผ๋ก ์ด ์ ์ ๋์ฑ ๋ช ํํ ์ค๋ช ํฉ๋๋ค. ์์ ์ ์๊ณก(์์), ์ฐ์ต(์ค๊ฐ), ์ฐ์ฃผ(๋)์ ๊ณผ์ ์ ๊ฑฐ์นฉ๋๋ค. AI๊ฐ ์ฐ์ต์ฒ๋ผ ๋ฐ๋ณต์ ์ธ ์ค๊ฐ ๊ณผ์ ์ ์๋ํํ ์ ์์ง๋ง, ์๋ฌด๊ฒ๋ ์๋ ์ํ์์ ๋ฉ๋ก๋๋ฅผ ์ฐฝ์กฐํ๋ โ์๊ณกโ์ ์๊ฐ๊ณผ, ๊ด๊ฐ ์์์ ๊ฐ์ ์ ๋ด์ ๊ณก์ โ์ฐ์ฃผโํ๋ ๋ผ์ด๋ธ ํผํฌ๋จผ์ค์ ์๊ฐ์ ์ธ๊ฐ๋ง์ด ํ ์ ์๋ ํน๋ณํ ์์ญ์ ๋๋ค. ์ด์ฒ๋ผ AI ์๋์๋ ์ํํธ์จ์ด ๊ฐ๋ฐ์ด๋ ๋ค๋ฅธ ์ง์ ๋ ธ๋์์ โ์์โ๊ณผ โ๋โ์ ํน๋ณํ ์๊ฐ๋ค์ ์ธ๊ฐ์ ๋ชซ์ผ๋ก ๋จ์ ๊ฒ์ ๋๋ค.
๋ฏธ๋๋ฅผ ์ํ ์ ์ธ: ๋น์ ์ โ์๋ฆ๋ค์โ์ ์ฐพ์๋ผ
์ด๋ฌํ ๋ณํ๋ ๋๋ ค์ธ ์๋ ์์ง๋ง, ๋์์ ํฐ ๊ธฐํ๊ฐ ๋ ์ ์์ต๋๋ค. ๋ฐ๋ณต์ ์ด๊ณ ์ง๋ฃจํ ์ ๋ฌด๋ AI์ ๋งก๊ธฐ๊ณ , ์ธ๊ฐ์ ๋์ฑ ์ฐฝ์์ ์ด๊ณ ํฅ๋ฏธ๋ก์ด ์ผ์ ์ง์คํ ์ ์๊ฒ ๋ฉ๋๋ค. ํค์ด๋ฐ์ โ๋ฌด์์ด ๋น์ ์๊ฒ ๊ธฐ์จ์ ์ฃผ๊ณ , ๋ถ์ ์งํผ๊ณ , ํฅ๋ถํ๊ฒ ๋ง๋๋์งโ ์ฐพ์์ผ ํ๋ค๊ณ ์กฐ์ธํฉ๋๋ค. ๊ทธ๊ฒ์ด ์๋ฆ๋ค์ด ์ฝ๋๋ , ์ ๊ตํ ์ถ์ํ๋ , ๋งค๋ ฅ์ ์ธ ๋์์ธ์ด๋ , ์๋๋ฉด ํ๋ฅญํ ๋ฌธ๊ตฌ๋ ์๊ด์์ต๋๋ค.
AI ์๋์ ์์ง๋์ด๋ ๋จ์ํ ์ฝ๋๋ฅผ ์์ฑํ๋ ๊ฒ์ ๋์ด, ๊ด๋ฆฌ์๋ ์ ํ ์ฑ ์์์ ๊ฐ์ ๋ ๋์ ์ญํ ์ ์ํํ๊ฒ ๋ ๊ฒ์ ๋๋ค. ํต์ฌ์ โ์๋ฆ๋ค์ด ๊ฒโ์ ๋ง๋๋ ๋ฐ ์ง์คํ๊ณ , AI๋ฅผ ํ์ฉํ์ฌ ์๋์ง๋ฅผ ์๋ชจํ๋ ๋์ ์๋์ง๋ฅผ ์ป๋ ๋ฐฉ์์ผ๋ก ์ผํ๋ ๊ฒ์ ๋๋ค.
๊ฒฐ๋ก ์ ์ผ๋ก โAI ์๋์์นโ ๋ชจ๋ธ์ AI๊ฐ ์ธ๊ฐ์ ์ผ์๋ฆฌ๋ฅผ ์์ ํ ๋นผ์์ ๊ฒ์ด๋ผ๋ ๋ง์ฐํ ๋๋ ค์ ๋์ , ์ธ๊ฐ๊ณผ AI๊ฐ ์ํธ ๋ณด์์ ์ผ๋ก ํ๋ ฅํ์ฌ ๋ ํฐ ๊ฐ์น๋ฅผ ์ฐฝ์ถํ ์ ์๋ ๋ฏธ๋๋ฅผ ์ ์ํฉ๋๋ค. AI๋ ํจ์จ์ ์ธ ์กฐ๋ ฅ์์ด์ ๋๊ตฌ์ด๋ฉฐ, ์ธ๊ฐ์ ๊ทธ ๋๊ตฌ๋ฅผ ์ฌ์ฉํ์ฌ ์ธ์์ ์์ ๋ง์ ๋ ์ฐฝ์ ์ธ โ๋งโ๊ณผ โ์๋ฆ๋ค์โ์ ๋ํ๋ ์กด์ฌ๋ก ๊ฑฐ๋ญ๋ ๊ฒ์ ๋๋ค.
โWhy Does Everyone Hate Rats? (Update) | Freakonomics Radioโ โ Freakonomics Radio Network ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
๋ด์์ โ์ฅ ์ ์โ ๊ทธ ์ด๋ฉด: ํ์ค์ ์ญ์ฌ์ ๊ณต์กด์ ๊ฐ๋ฅ์ฑ
๋ด์์์ ๋ด์ ํธ๋ฅธ ๋๋ญ์๊ณผ ํ๊ธฐ ๋์น๋ ์ผ์ธ ์นดํ, ๊ทธ๋ฆฌ๊ณ ์ด๊น์์ด ๋ชจ์ต์ ๋๋ฌ๋ด๋ ์ฅ๋ค๋ก ํน์ง์ง์ด์ง๋๋ค. ๋์์ ๊ทธ๋ฆผ์ ์์์ ์ธ๊ฐ๊ณผ ๊ฐ์ฅ ๊ฐ๊น์ด ๊ณต์กดํ๋ ์ด ์์ ์๋ช ์ฒด๋ค์ ์ค๋ซ๋์ ํ์ค์ ๊ฒฝ๋ฉธ์ ๋์์ด์์ฃ . ํ์ง๋ง ๊ณผ์ฐ ์ฅ๋ค์ ์ฐ๋ฆฌ๊ฐ ์๊ฐํ๋ ๋งํผ โ์ ๋นโ์ผ๊น์? <ํ๋ฆฌ์ฝ๋ ธ๋ฏน์ค ๋ผ๋์ค(Freakonomics Radio)>๋ ๋ด์์์ โ์ฅ ์ ์โ์ ํตํด ์ฅ์ ๋ํ ์ฐ๋ฆฌ์ ๋ฟ๋ฆฌ ๊น์ ์ธ์์ ํํค์น๊ณ , ๊ทธ ์ด๋ฉด์ ์จ๊ฒจ์ง ์ญ์ฌ์ , ๋ฌธํ์ , ์๋ฌผํ์ ์ง์ค์ ํ๊ตฌํฉ๋๋ค.
๋ด์์์ โ์ฅ ์ ์โ: ๋์์ ๋์์๋ ๋์
2022๋ ๊ฐ์, ๋ด์์ ์ ๋ถ ์น์ฌ์ดํธ์๋ ์ด์์ ์ธ ์ฑ์ฉ ๊ณต๊ณ ๊ฐ ์ฌ๋ผ์์ต๋๋ค. โ์ด์์ ์ธ ํ๋ณด์๋ ๋์ ๋๊ธฐ ๋ถ์ฌ์ ๋ค์ ํผ์ ๊ตถ์ฃผ๋ฆฐ(bloodthirsty) ์ฑํฅ์ ๊ฐ์ง๊ณ ์์ผ๋ฉฐ, ๋ฐ์ดํฐ ์์ง, ๊ธฐ์ ํ์ , ๊ทธ๋ฆฌ๊ณ ๋๋ ํ์ด(wholesale slaughter)์ ํฌํจํ ๋ค์ํ ๊ฐ๋์์ ๋ชจ๋ ํด๊ฒฐ์ฑ ์ ๋ชจ์ํ ์์ง๊ฐ ์๋ ์ฌ๋์ด์ด์ผ ํ๋ค.โ ๋๋ ํ์ด์ ์๊ตฌํ๋ ์ ๋ถ ์ง์ฑ ์ด๋ผ๋, ๋๋ผ์์ ๊ธ์น ๋ชปํ ๋ด์ฉ์ด์์ฃ . ์ด ๊ณต๊ณ ์ ๋ฐฐํ์๋ ๋น์ ๋ด์ ์์ฅ์ด์๋ ์๋ฆญ ์๋ด์ค(Eric Adams)๊ฐ ์์์ต๋๋ค. ๊ทธ๋ ์ฅ์ ๋ํ ๋งน๋ ฌํ ํ์ค๊ฐ์ ๊ณต๊ณต์ฐํ ๋๋ฌ๋ด๋ฉฐ โ์ฅ๋ ๋น์ ์๊ฒ ํธ๋ผ์ฐ๋ง๋ฅผ ์๊ฒจ์ค๋ค. ๋๋ ์ฅ๋ฅผ ์ฆ์คํ๋คโ๊ณ ๋งํ์ต๋๋ค.
์๋ด์ค ์์ฅ์ โ์ฅ ์ ์โ์ ๋ด์์์ ๊ณ ์ง์ ์ธ ๋ฌธ์ ์ ๋ํ ๊ฐ๋ ฅํ ๋์์ฑ ์ผ๋ก ์์๋์์ต๋๋ค. ๊ทธ๋ฆฌ๊ณ ์ด โ์์ โ์ ์ฐพ๋ ๊ทธ์ ๋ ธ๋ ฅ์ ์ผ์ดํฐ ์ฝ๋ผ๋(Katy Corradi)๋ผ๋ ์ธ๋ฌผ์ ๋ง๋ ๊ฒฐ์ค์ ๋งบ์์ต๋๋ค. ๊ทธ๋ ๋ โ๋๋ดํ ํ๋, ๊ธฐ๋ฐํ ์ ๋จธ, ๊ทธ๋ฆฌ๊ณ ์ ๋ฐ์ ์ธ โ๋ฐฐ๋์ ์ค(badass)โ ๊ธฐ์ดโ์ ๊ฐ์ถ ์ธ๋ฌผ์ด๋ผ๋ ์ฑ์ฉ ๊ณต๊ณ ์ ๋ฌ์ฌ์ ์๋ฒฝํ๊ฒ ๋ถํฉํ๋ ๋ฏํ์ต๋๋ค. ์ฝ๋ผ๋๋ ๋ด์์์ โ์ ์ ์ญ ์ค์น๋ฅ ์ํ ์ฑ ์์(citywide director of rodent mitigation)โ, ์ผ๋ช โ์ฅ ์ฐจ๋ฅด(rat czar)โ๋ก ์๋ช ๋์์ต๋๋ค. ๊ทธ๋ ๋ ์ฅ ๋ฌธ์ ์ ๋ํด ๋ ๋ง์ ์ฌ๋์ด ์ด์ผ๊ธฐํ ์๋ก ํด๊ฒฐ์ ๋์์ด ๋ ๊ฒ์ด๋ผ๊ณ ๋ฏฟ์์ต๋๋ค.
๋ด์์ ๋น๋กฏํ ๋ง์ ๋๋์๋ค์ ์ฝ๋ก๋19 ํฌ๋ฐ๋ฏน ์ดํ ์ฅ ๊ฐ์ฒด์๊ฐ ๊ธ์ฆํ๋ ํ์์ ๊ฒช์์ต๋๋ค. ์ฝ๋ผ๋๋ ์ฅ ๋ฌธ์ ํด๊ฒฐ์ ํต์ฌ์ด ๋ฐ๋ก โ์์(sanitation)โ์ ์๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค. ๊ทธ๋ ๋ ์ด๋ฆฐ ์์ , ์ง ๋คํธ ์ฒ ๋ก๋ณ์ ๋ฌด์ฑํ ์ก์ด๊ฐ ์ฅ๋ค์ ์๋ฒฝํ ์์์ง๊ฐ ๋๋ ๊ฒ์ ๋ชฉ๊ฒฉํ๊ณ ์ด์๋ค๊ณผ ํจ๊ป ์ฒญ์ ์ด๋์ ๋ฒ์ฌ ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ ๊ฒฝํ์ด ์์ต๋๋ค. ์๋ฌผํ ํ์ฌ ํ์์ ๋์ ์ง์๊ฐ๋ฅ์ฑ ์์ฌ ํ์๋ฅผ ๊ฐ์ง ๊ทธ๋ ๋ ๋ด์์ ๊ต์ก๋ถ์ ์ง์๊ฐ๋ฅ์ฑ ๋ถ์์์ ์ฐ๋ ๊ธฐ ๊ฐ์ถ ํ๋ก๊ทธ๋จ์ ์ด์ํ๋ฉฐ ์ฅ ๋ฌธ์ ์ ๊น์ด ๊ด์ฌํ๊ฒ ๋์์ต๋๋ค. ๊ทธ๋ ๋ โ์ฐ๋ ๊ธฐ์ ์ฅ๋ ๋ผ๋ ค์ผ ๋ ์ ์๋ ๊ด๊ณโ๋ผ๋ฉฐ, ํ๊ต์ ์ฐ๋ ๊ธฐ ๊ด๋ฆฌ ๊ฐ์ ์ ํตํด ์ฅ๋ค์ ๋จน์ด ์ ๊ทผ์ฑ์ ์ค์ด๋ ๋ฐ ์ง์คํ์ต๋๋ค.
๋ด์์์ ์ฅ ๊ฐ์ฒด์๋ฅผ ์ ํํ ํ์ ํ๊ธฐ๋ ์ด๋ ต์ง๋ง, ์ผ๋ถ ์ถ์ ์น์ ๋ฐ๋ฅด๋ฉด 300๋ง ๋ง๋ฆฌ์ ๋ฌํ ์๋ ์๋ค๊ณ ํฉ๋๋ค. ์ฝ๋ผ๋๋ ์ซ์์ ์ง์ฐฉํ๊ธฐ๋ณด๋ค ์ฅ๊ฐ ์ธ๊ฐ๊ณผ โ์ํ์ ๊ณต์ ํ๋ ๋๋ฌผ(commensal)โ์ด๋ผ๋ ์ ์ ์ฃผ๋ชฉํฉ๋๋ค. ์ด๋ ์ฅ๋ค์ด ์ธ๊ฐ์ด ๋ง๋ค์ด ๋์ ๋์ ๊ณต๊ฐ์์ ์ธ๊ฐ์ ์์๋ฌผ ์ฐ๋ ๊ธฐ๋ฅผ ๋จน์ผ๋ฉฐ ์ด์๊ฐ๋ ์กด์ฌ์์ ์๋ฏธํฉ๋๋ค.
์ฅ์ ์ง๋ฅ๊ณผ ์์กด ์ ๋ต: ๋์์ ์จ๊ฒจ์ง ์ง๋ฐฐ์
์ฐ๋ฆฌ๋ ์ฅ๋ฅผ ํํ โํ์ค์ค๋ฌ์ดโ ์กด์ฌ๋ก ์ฌ๊ธฐ์ง๋ง, ๊ทธ๋ค์ ์ง๋ฅ๊ณผ ์์กด ๋ฅ๋ ฅ์ ๋๋ผ์ธ ์ ๋์ ๋๋ค. ๊ณผํ ์ ๋๋ฆฌ์คํธ์ด์ <ํด์ถฉ(Pests)>์ ์ ์์ธ ๋ฒ ์๋ ๋ธ๋ฃฉ์ (Bethany Brookshire)๋ ์ฅ๊ฐ โ์ธ๊ฐ์ ์ ์ธํ๊ณ ๋ ๋์ ๊ณต๊ฐ์ ๊ฐ์ฅ ์ ํ์ฉํ๋ ์๋ฌผโ์ด๋ผ๊ณ ๋งํฉ๋๋ค. ์ฝ๋ผ๋ ์ญ์ ์ฅ์ โ๊ตํํจ(savviness)โ๊ณผ โ์ ์๋ ฅ(adaptability)โ์ด ๋งค์ฐ ๋ฐ์ด๋๋ค๊ณ ํ๊ฐํ๋ฉฐ, ์ต๊ทผ ์ฐ๊ตฌ๋ค์ ์ฅ์ โ๊ณต๊ฐ(empathy)โ, โ์์(laughing)โ, โ์ดํ์ฃผ์(altruism)โ์ ๊ฐ์ ํน์ฑ๊น์ง ๋ฐํ๋ด๊ณ ์๋ค๊ณ ์ธ๊ธํฉ๋๋ค. ์ฌ์ง์ด ์ฅ๋ค์ ์๋ก์ด ๋จน์ด์์ ์ ๊ทผํ ๋ ๋ ์ง๋ฐฐ์ ์ธ ์ฅ๋ฅผ ๋จผ์ ๋ณด๋ด ์์ ์ฑ์ ํ์ธํ๋ ์๋ฆฌํ ์์กด ์ ๋ต์ ๊ตฌ์ฌํ๊ธฐ๋ ํฉ๋๋ค.
์ค๋๋ ์ฐ๋ฆฌ๊ฐ ํํ ๋ณด๋ ๊ฐ์ ์ฅ(Rattus norvegicus), ์ผ๋ช ๋ ธ๋ฅด์จ์ด ์ฅ๋ 18์ธ๊ธฐ ์ ๋ฝ ์ ๋ฐ์ ํตํด ๋ด์์ ๋ค์ด์จ ๊ฒ์ผ๋ก ์ถ์ ๋ฉ๋๋ค. ํ์ง๋ง ์ด ์ฅ์ ๊ธฐ์์ ๋ ธ๋ฅด์จ์ด๊ฐ ์๋ ๋ชฝ๊ณจ ์ง์ญ์ด๋ฉฐ, ์ฅ๋ฅผ ์ซ์ดํ๋ ์ฌ๋๋ค์ด ๊ทธ ์ด๋ฆ์ ์ซ์ดํ๋ ๋์์ ๋ถ์ธ ๊ฒฐ๊ณผ๋ผ๊ณ ๋ธ๋ฃฉ์ ๋ ์ค๋ช ํฉ๋๋ค.
์ฅ์ ๋ํ ํ์ค์ ์ญ์ฌ: ํ์ฌ๋ณ์ ์ค๋ช
์ฅ๊ฐ ์ธ๋ฅ์๊ฒ ํ์ค์ค๋ฌ์ด ์กด์ฌ๋ก ๊ฐ์ธ๋ ๊ฐ์ฅ ํฐ ์ด์ ๋ ๋ฐ๋ก โ์ง๋ณ ์ ํโ์ ๋ํ ์ธ์, ํนํ โํ์ฌ๋ณ(Black Death)โ๊ณผ์ ์ฐ๊ด์ฑ ๋๋ฌธ์ ๋๋ค. 14์ธ๊ธฐ ์ ๋ฝ์ ํฉ์ธ๋ฉฐ ์ธ๊ตฌ์ ์ ๋ฐ์ ์ฃฝ์์ผ๋ก ๋ชฐ์๋ฃ์๋ ํ์ฌ๋ณ์ ์ค๋ซ๋์ ์ฅ์ ์ฅ๋ฒผ๋ฃฉ์ด ์ฃผ๋ฒ์ผ๋ก ์ง๋ชฉ๋์ด ์์ต๋๋ค.
ํ์ง๋ง ์ต๊ทผ ๊ณผํ๊ณ์์๋ ์ฅ์ ํ์ฌ๋ณ์ ์ฐ๊ด์ฑ์ ๋ํ ๊ธฐ์กด์ ํต๋ ์ ๋์ ํ๋ ์ฐ๊ตฌ ๊ฒฐ๊ณผ๋ค์ด ๋์ค๊ณ ์์ต๋๋ค. ์ค์ฌ๋ก ๋ํ์ ์ํ ๋ฐ ์งํํ ๊ต์์ธ ๋์ค ํฌ๋ฆฌ์คํฐ์ ์คํ ์ธ์ค(Nils Christian Stenseth)๋ 2018๋ <๋ฏธ๊ตญ ๊ตญ๋ฆฝ๊ณผํ์ ํ๋ณด(Proceedings of the National Academy of Sciences)>์ ๋ฐํํ ๋ ผ๋ฌธ์์ ๋ค๋ฅธ ๋ชจ๋ธ์ ์ ์ํ์ต๋๋ค. ๊ทธ๋ 1300๋ ๋๋ถํฐ 1700๋ ๋๊น์ง์ ํ์ฌ๋ณ ์ฌ๋ง๋ฅ ๊ณผ ์ ํ ์๋๋ฅผ ๋ถ์ํ ๊ฒฐ๊ณผ, โ์ฅ๊ฐ ์ ๋ฝ์์ ํ์ฌ๋ณ ํ์ฐ์ ์ฃผ์ ์ญํ ์ ํ์ ๊ฐ๋ฅ์ฑ์ ๋งค์ฐ ๋ฎ๋คโ๊ณ ๊ฒฐ๋ก ๋ด๋ ธ์ต๋๋ค. ์ฅ๋ฒผ๋ฃฉ์ด ์ฅ์ ์ฃฝ์ ์ดํ์์ผ ์ธ๊ฐ์๊ฒ ์ฎ๊ฒจ๊ฐ๋ ์๋๋ฅผ ๊ฐ์ํ ๋, ํ์ฌ๋ณ์ ๋น ๋ฅธ ์ ํ ์๋๋ ์ฅ ๋ชจ๋ธ๋ก๋ ์ค๋ช ํ๊ธฐ ์ด๋ ต๋ค๋ ๊ฒ์ ๋๋ค. ๋์ ๊ทธ์ ์ฐ๊ตฌํ์ โ์ธ๊ฐ ์ฒด์ธ ๊ธฐ์์ถฉ(human ectoparasites)โ, ์ฆ โ๋ชธ๋(body lice)โ๋ โ์ธ๊ฐ ๋ฒผ๋ฃฉ(human fleas)โ์ด ๋ ์ ๋ ฅํ ์ ํ ๋งค๊ฐ์ฒด์์ ๊ฐ๋ฅ์ฑ์ด ๋๋ค๊ณ ์ฃผ์ฅํ์ต๋๋ค.
ํ๋ฒ๋ ๋ํ๊ต ๊ฒฝ์ ํ ๊ต์์ด์ ๋์ ์ ๋ฌธ๊ฐ์ธ ์๋ ๊ธ๋ ์ด์ (Ed Glaeser) ์ญ์ ์ด ๊ฒฌํด์ ๋์ํ๋ฉฐ, ์ฅ๊ฐ โ์ฃ์ฑ ๊ฐ(guilt)โ์ ๊ฐ์ง ๋ฆฌ ์์ผ๋ ๊ทธ ๋น๋์ ํ์ฌ๋ณ์ ์์ธ๊ท ์ธ โ์๋ฅด์๋์ ํ์คํฐ์ค(Yersinia pestis)โ์๊ฒ ๋๋ ค์ผ ํ๋ค๊ณ ๋งํฉ๋๋ค. ๋น๋ก ์ฅ๊ฐ 19์ธ๊ธฐ ์ดํ์ ์ธ ๋ฒ์งธ ํ์ฌ๋ณ ํฌ๋ฐ๋ฏน์๋ ์ค์ํ ์ญํ ์ ํ์์ง๋ผ๋, ๊ฐ์ฅ ์ ๋ช ํ ์ค์ธ ํ์ฌ๋ณ์ ์ฃผ๋ฒ์ ์๋๋ผ๋ ๊ฒ์ด์ฃ .
ํ๋ ๋์์์ ์ฅ๊ฐ ๋ฏธ์น๋ ์ํฅ: ๊ฒฝ์ ์ , ์ ์ ์ , ๊ทธ๋ฆฌ๊ณ ์ ์ฌ์ ์ํ
๊ทธ๋ ๋ค๋ฉด ํ๋ ๋์์์ ์ฅ๋ ์ด๋ค ์ํฅ์ ๋ฏธ์น ๊น์? ์ ์ง ์ฅ ์ฐจ๋ฅด ์ฝ๋ผ๋๋ ์ฅ๊ฐ ์ฌ๋ฌ ๊ฐ์ง ํํ๋ก โ๋ถ์ ์ ์ธ ์ธ๋ถ ํจ๊ณผ(negative externalities)โ๋ฅผ ์ ๋ฐํ๋ค๊ณ ์ง์ ํฉ๋๋ค.
- ๊ฒฝ์ ์ ํผํด: ์ฅ๋ ์ ์ , ๊ฑด๋ฌผ ๊ธฐ์ด ๋ฑ ๊ฑฐ์ ๋ชจ๋ ๊ฒ์ ๊ฐ์๋จน๋ ์ต์ฑ ๋๋ฌธ์ ์ฌ์ฐ์ ์๋นํ ํผํด๋ฅผ ์ ํ๋๋ค. ์ ์ฅ๋ ์๋๊ณผ ๊ณก๋ฌผ์๋ ์์์ ์ ํ ๊ฒฝ์ ์ ์์ค์ ์ด๋ํฉ๋๋ค.
- ์ ์ ์ ๊ฑด๊ฐ: ์ฅ๊ฐ ์ถ๋ชฐํ๋ ํ๊ฒฝ์ ์ฌ๋ ์ฌ๋๋ค์ ์คํธ๋ ์ค, ๋ถ์, ์ฐ์ธ์ฆ ๋ฑ ์ ์ ์ ๊ณ ํต์ ๊ฒช๋๋ค๋ ์ฐ๊ตฌ ๊ฒฐ๊ณผ๊ฐ ๋ง์ต๋๋ค. ์ฅ์ ๋ํ ํ์ค๊ฐ์ ๋ถ์พ๊ฐ๊ณผ ๋ถ์๊ฐ์ ์ ๋ฐํ๋ฉฐ ์ถ์ ์ง์ ๋จ์ด๋จ๋ฆฝ๋๋ค.
- ๊ณต์ค ๋ณด๊ฑด ์ํ: ์ฅ๋ ์๋ณ์ ํตํด โ๋ ํ ์คํผ๋ผ์ฆ(Leptospirosis)โ๊ณผ ๊ฐ์ ์ง๋ณ์ ์ ํํ ์ ์์ต๋๋ค. 2023๋ ๋ด์์์์ 24๋ช ์ ๋ ํ ์คํผ๋ผ์ฆ ํ์๊ฐ ๋ฐ์ํ์ง๋ง, 800๋ง ๋ช ์ด ๋๋ ๋์ ์ธ๊ตฌ๋ฅผ ๊ณ ๋ คํ ๋ ์น๋ช ์ ์ธ ์์ค์ ์๋๋๋ค. ํ์ง๋ง ์ฝ๋กฌ๋น์ ๋ํ์ ์ฐ๊ตฌ์ ๋ฐ๋ฅด๋ฉด ๋ด์์ ์ฅ๋ค์๊ฒ์๋ ๋ค์ํ ์ ์ข ๋ฐ์ด๋ฌ์ค์ ๋ณ์๊ท ์ด ๋ฐ๊ฒฌ๋์์ผ๋ฉฐ, ๋ฐ์ด๋ฌ์ค๊ฐ ๋์ฐ๋ณ์ด๋ฅผ ์ผ์ผ์ผ ์์ฃผ๋ฅผ ๋ฐ๊ฟ ์ ์ฌ์ ์ํ์ ํญ์ ์กด์ฌํฉ๋๋ค. ์ฅ๊ฐ ์ธ๊ฐ๊ณผ ๋งค์ฐ ๊ฐ๊น์ด ์ด๊ธฐ ๋๋ฌธ์ ์ด๋ฌํ ์ํ์ ๋์ฑ ์ปค์ง ์ ์์ต๋๋ค.
โํด์ถฉโ์ด๋ผ๋ ๊ฐ๋ ์ ๋ํ ๋นํ์ ๊ณ ์ฐฐ: ์ฅ๋ ์ ๋ง ์ ๋น์ธ๊ฐ?
๋ฒ ์๋ ๋ธ๋ฃฉ์ ๋ โํด์ถฉ(pest)โ์ด๋ผ๋ ๋จ์ด๊ฐ ์ฐ๋ฆฌ์ ์ฌํ์์ ๋ง์ ์ญํ ์ ํ๋ค๊ณ ๋งํฉ๋๋ค. โํด์ถฉ์ ์ฐ๋ฆฌ๊ฐ ์ํ์ง ์๋ ๊ณณ์ ์๋ ๋๋ฌผ์ ์ง์นญํ๋ ๋จ์ดโ์ด๋ฉฐ, ์ฐ๋ฆฌ๊ฐ ํน์ ๋๋ฌผ์ ์ซ์ดํ๋ ๊ฒ์ ๊ทธ ๋๋ฌผ์ด ์ด๋ค ํ๋์ ํ๋์ง๋ณด๋ค โ์ฐ๋ฆฌ๊ฐ ๋๋ฌผ์ด ์ด๋์ ์ํ๊ณ ๋ฌด์์ ํด์ผ ํ๋ค๊ณ ์๊ฐํ๋์งโ์ ๋ฌ๋ ค์๋ค๋ ๊ฒ์ ๋๋ค.
๊ทธ๋ ๋ ๋น๋๊ธฐ๋ฅผ ์๋ก ๋ค์ด ์ด๋ฌํ ์ฃผ๊ด์ฑ์ ์ค๋ช ํฉ๋๋ค. 8,000๋ ์ ๋ถํฐ ๊ธธ๋ค์ฌ์ง ๋น๋๊ธฐ๋ ํ๋ ์๋, ๋ฉ์ ์ (๋ก์ดํฐ ํต์ ์ ๋น๋๊ธฐ๋ฅผ ์ด์ฉํด ์ฃผ์ ์ ๋ณด๋ฅผ ์ ๋ฌํ๊ธฐ๋ ํ์ต๋๋ค), ์ฌ์ง์ด ๋น๋ฃ์ ์์ฒ์ผ๋ก์ ๋งค์ฐ ์ค์ํ์ต๋๋ค. ํ์ง๋ง ํํ ๋น๋ฃ, ์ด๋ฉ์ผ, ๋ญ๊ณ ๊ธฐ๊ฐ ๋ณดํธํ๋๋ฉด์ ๋น๋๊ธฐ๋ ๋ ์ด์ ์ธ๊ฐ์๊ฒ ์ ์ฉํ์ง ์๊ฒ ๋์๊ณ , โ๋ ๊ฐ ๋ฌ๋ฆฐ ์ฅ(rats with wings)โ๋ก ๋ถ๋ฆฌ๋ฉฐ โํด์ถฉโ์ผ๋ก ์ ๋ฝํ์ต๋๋ค. ์ด๋ ์ธ๊ฐ์ ํ์์ ์ธ์์ด ์ด๋ป๊ฒ ๋๋ฌผ์ ํํ์ ์์ ํ ๋ฐ๊ฟ ์ ์๋์ง๋ฅผ ๋ณด์ฌ์ฃผ๋ ๊ทน๋ช ํ ์ฌ๋ก์ ๋๋ค.
๋ธ๋ฃฉ์ ๋ ๋ํ ์ฅ์ ๋ํ ํ์ค๊ฐ ๋ฌธํ์ ๋ฐ๋ผ ์ผ๋ง๋ ๋ค๋ฅธ์ง๋ ๊ฐ์กฐํฉ๋๋ค. ์ธ๋ ๋ฐ์ฌ๋ ธํฌ(Deshnoke)์ ์๋ โ๊น๋ฅด๋ ๋งํ ์ฌ์(Karni Mata Temple)โ์๋ ์ฝ 2๋ง 5์ฒ ๋ง๋ฆฌ์ ๊ฒ์ ์ฅ๊ฐ ์ด๊ณ ์์ผ๋ฉฐ, ์ด ์ฅ๋ค์ โ์ ์ฑํ(sacred)โ ์กด์ฌ๋ก ์ฌ๊ฒจ์ง๋๋ค. ์ ์ค์ ๋ฐ๋ฅด๋ฉด, ์ด ์ฅ๋ค์ ๊น๋ฅด๋ ๋งํ ์ฌ์ ์ ๊ฐ์กฑ๋ค์ด ํ์ํ ๊ฒ์ด๋ฉฐ, ์ฃฝ์ผ๋ฉด ๋ค์ ์ธ๊ฐ์ผ๋ก ํ์ด๋๋ค๊ณ ๋ฏฟ์ต๋๋ค. ์ด ์ฌ์์ ์ฅ๋ค์ ์์ผ๋ก ์ฅ์๋ ๋ฌธ๊ณผ ๋๋ฆฌ์ ๋ฐ๋ฅ ์์์ ์ฐ์ ๋ฅผ ๋ง์๊ณ , ํต๋ฐ ๋นต, ๊ณผ์ผ, ์ฑ์ ๋ฑ ๊ฑด๊ฐํ ์๋จ์ ์ ๊ณต๋ฐ์ผ๋ฉฐ ์ญ๋ฐฐ๋ฐ์ต๋๋ค.
์ด๋ฌํ ๋ฌธํ์ ์ฐจ์ด๋ ์ฐ๋ฆฌ๊ฐ ๋๋ฌผ์ ๋ํ๋ ๋ฐฉ์์ ๋ํ ๊ทผ๋ณธ์ ์ธ ์ง๋ฌธ์ ๋์ง๋๋ค. ๋ธ๋ฃฉ์ ๋ ์ฐ๋ฆฌ๊ฐ ์ฅ๋ฅผ ์ซ์ดํ๋ ์ด์ ์ค ํ๋๊ฐ โ์ฐ๋ฆฌ๊ฐ ๋๋ฌผ๋ค์ด ์คํจํ ๊ฒ์ด๋ผ๊ณ ์์ํ๊ธฐ ๋๋ฌธโ์ด๋ผ๊ณ ๋งํฉ๋๋ค. ์ฐ๋ฆฌ๊ฐ ์ด๋ค ์ง์ญ์ ๊ฐ๋ฐํ๊ณ ๊ฑด๋ฌผ์ ์ธ์ฐ๋ฉด ๋๋ฌผ๋ค์ ๋ ๋ ๊ฒ์ด๋ผ๊ณ ์๊ฐํ์ง๋ง, ๋ง์ฝ ๊ทธ๋ค์ด ์ฌ์ ํ ๋จ์ ๋ฒ์ฑํ๋ค๋ฉด ์ฐ๋ฆฌ๋ ํ๋ฅผ ๋ธ๋ค๋ ๊ฒ์ ๋๋ค. ํนํ ๊ทธ๋ค์ด ์ฐ๋ฆฌ๊ฐ ๊ฐ์น ์๊ฒ ์ฌ๊ธฐ๋ ๊ฒ(์ฐ๋ฆฌ์ ์ ์, ๋์๋ฌผ, ์ฌ์ง์ด ๊ณ ์์ด)์ ํตํด ๋ฒ์ฑํ ๋ ๋์ฑ ๊ทธ๋ ์ต๋๋ค. โ์ฐ๋ฆฌ๋ ๊ทธ๋ค์ ์ฑ๊ณต์ ์ซ์ดํฉ๋๋ค. ์๋ํ๋ฉด ๊ทธ๋ค์ ์ฑ๊ณต์ด ์ฐ๋ฆฌ์ ์คํจ์ฒ๋ผ ๋๊ปด์ง๊ธฐ ๋๋ฌธ์ ๋๋ค.โ
์ฅ์์ ๊ณต์กด์ ๋ชจ์ํ๋ฉฐ: ๋์์ ๋๋ฐ์๋ฅผ ์ํ ์๋ก์ด ์๊ฐ
์ฅ์ ๋ํ ํ๊ณ ๋ ํ์ค๊ฐ์ ๋์ ํ๊ฒฝ ์์์ ์ฐ๋ ๊ธฐ์ ๋ถ๊ฒฐํจ์ ๋ํ ์ฐ๋ฆฌ์ ์ธ์๊ณผ ๊น์ด ์ฐ๊ด๋์ด ์์ต๋๋ค. ํ์ง๋ง ์๋ ๊ธ๋ ์ด์ ๋ ์ฅ๊ฐ ์ด์ ๋ ์ด์ ์ฃผ์ ์ง๋ณ ๋งค๊ฐ์ฒด๊ฐ ์๋ ํ, ๊ทธ๋ค์ ๋ํ ์ฐ๋ฆฌ์ ๋๋ ค์์ ์๋์ ๋ค๋จ์ด์ง ๊ฒ์ผ ์ ์๋ค๊ณ ์ง์ ํฉ๋๋ค. ๊ทธ๋ ์ฅ๊ฐ ๋์์์ ์ธ๊ฐ์ โ๋ถ์ ์ ์ธ ์ธ๋ถ ํจ๊ณผโ๋ฅผ ์ ๋ฐํ๋ ์กด์ฌ์ด๊ธด ํ์ง๋ง, โ๋งค์ฐ ์ฌ์ํ ๋ฌธ์ โ์ผ ์ ์๋ค๊ณ ๋ง๋ถ์ ๋๋ค.
๊ทธ๋ ๋ค๊ณ ์ฅ๋ฅผ ๋ฐฉ์นํ์๋ ์๋ฏธ๋ ์๋๋๋ค. ๊ธ๋ ์ด์ ๋ โ์ฅ ๊ฐ์ฒด์๋ฅผ ๊ด๋ฆฌ ๊ฐ๋ฅํ ์์ค์ผ๋ก ์ ์งํ๋ ์ ์ฑ ์๋ ์ฐฌ์ฑํ์ง๋ง, ์์ ํ ๋ฐ๋ฉธ์ ์ง์ฐฉํ ํ์๋ ์๋คโ๊ณ ๋งํฉ๋๋ค. ๊ทธ๋ ์ฅ๋ฅผ โ๋์์ ์์ฐ์ ์ธ ํํธ๋โ๋ก ๋ณด๊ณ , ๊ณผ๋ํ ๊ณตํฌ ๋์ โ์ฐ๋ฏผ(sympathy)โ์ ๊ฐ์ ธ์ผ ํ๋ค๊ณ ์ ์ํฉ๋๋ค.
๋ด์์์ โ์ฅ ์ ์โ์ ์ฌ์ ํ ์งํ ์ค์ ๋๋ค. ์ฅ ์ฐจ๋ฅด ์ผ์ดํฐ ์ฝ๋ผ๋์ ํ์์ธ ์กฐ๋ ๋ชธ๋ค๋(Zoran Momdani) ์์ฅ์ 2031๋ ๋ง๊น์ง ๋์ ์ ์ฒด์ ์ฐ๋ ๊ธฐ ์ปจํ ์ด๋ํ๋ฅผ ๋ฌ์ฑํ๊ฒ ๋ค๋ ๊ณํ์ ๋ฐํํ๋ฉฐ ์ฅ ๋ฌธ์ ํด๊ฒฐ์ ๋ํ ์์ง๋ฅผ ๋ณด์ด๊ณ ์์ต๋๋ค.
์ฅ๋ ์ธ๊ฐ์ด ๋ง๋ ๋์๋ผ๋ ์ํ๊ณ์์ ๋ฒ์ฑํ๋ ๊ฐ์ฅ ์ฑ๊ณต์ ์ธ ์๋ช ์ฒด ์ค ํ๋์ ๋๋ค. ๊ทธ๋ค์ ๋ฌด์กฐ๊ฑด์ ์ธ ํ์ค์ ๋์์ผ๋ก๋ง ๋ณผ ๊ฒ์ด ์๋๋ผ, ๋์ ํ๊ฒฝ์์ ์ธ๊ฐ๊ณผ ํจ๊ป ์ด์๊ฐ๋ โ๋๋ฐ์โ๋ก์ ์ดํดํ๊ณ ๊ด๋ฆฌํ๋ ์๋ก์ด ์๊ฐ์ด ํ์ํ ๋์ ๋๋ค. ์ฅ์ ๋ํ ์ฐ๋ฆฌ์ ์ธ์ ๋ณํ๋ ๊ฒฐ๊ตญ ์ธ๊ฐ์ด ๋ค๋ฅธ ์๋ช ์ฒด์ ์ด๋ป๊ฒ ๊ณต์กดํด์ผ ํ๋๊ฐ์ ๋ํ ๋ ๋์ ์ง๋ฌธ์ผ๋ก ์ด์ด์ง ๊ฒ์ ๋๋ค.
โClaude Design is slow and I love it anyway (plus why I love ChatGPT Images 2.0)โ โ How I AI ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI ๋์์ธ ๋๊ตฌ์ ์๋ก์ด ์งํ: Claude Design๊ณผ GPT Image 2 ์ฌ์ธต ๋ถ์
์ธ๊ณต์ง๋ฅ(AI)์ด ๋์์ธ ๋ถ์ผ์ ํ๋ช ์ ์ธ ๋ณํ๋ฅผ ๊ฐ์ ธ์ค๊ณ ์์ต๋๋ค. ์ฝ๋ ์์ฑ๋ถํฐ ๊ณ ๊ฐ ๋ฐ์ดํฐ ๋ถ์, ์ฌ์ง์ด ๊ณ ๊ฐ ์ง์์ ์ด๋ฅด๊ธฐ๊น์ง AI ๋๊ตฌ๋ค์ ์ด๋ฏธ ์ฐ๋ฆฌ ์ ๋ฌด ๋ฐฉ์์ ๋ณํ์ํค๊ณ ์์ฃ . ์ด๋ฒ ์ํผ์๋์์๋ Anthropic์ Claude Design๊ณผ OpenAI์ GPT Image 2 ๋ชจ๋ธ์ด ๋์์ธ ์์ ํ๋ฆ์ ์ด๋ป๊ฒ ๋ฐ๊พธ๊ณ ์๋์ง, ๊ทธ๋ฆฌ๊ณ ์ด ์๋ก์ด ๋๊ตฌ๋ค์ด ๊ฐ์ง ์ฅ๋จ์ ๊ณผ ์ค์ ํ์ฉ ์ฌ๋ก๋ค์ ์ฌ์ธต์ ์ผ๋ก ์ดํด๋ณด๊ฒ ์ต๋๋ค. ํธ์คํธ ํด๋ ์ด ๋ฐ๋ (Claire Valle)๋ AI ๊ธฐ์ ์ ํ์ฉํด ๋ ๋์ ๊ฒฐ๊ณผ๋ฌผ์ ๋ง๋ค๊ณ ์ ํ๋ ์ฌ๋๋ค์ ๋๊ธฐ ์ํด, ์ด ํ์ ์ ์ธ ๋๊ตฌ๋ค์ ๋ํ ์์งํ ๊ฒฌํด๋ฅผ ๊ณต์ ํฉ๋๋ค.
1. ๋์์ธ ์์คํ ์ ์ต์ฐ์ ์ผ๋ก: Anthropic์ Claude Design
Anthropic์ด ์ต๊ทผ ์ถ์ํ Claude Design์ ์น ๊ธฐ๋ฐ ๋์์ธ ๋๊ตฌ๋ก, ํ๋กํ ํ์ (prototype) ์ ์(์์ด์ดํ๋ ์(wireframe) ๋ฐ ๊ณ ํ์ง ํ๋กํ ํ์ ํฌํจ), ์ฌ๋ผ์ด๋, ๋น๋์ค ์ ์์ ์ค์ ์ ๋ก๋๋ค. ๋ง์ ์ฌ๋๋ค์ด Claude Design์ด Figma๋ฅผ ๋์ฒดํ ์ ์์์ง์ ๋ํด ๊ถ๊ธํดํ์ง๋ง, ํด๋ ์ด๋ ์ด ๋๊ตฌ๊ฐ ์์ง๋์ด๋ง ๋จ๊ณ๋ก ๋์ด๊ฐ๊ธฐ ์ ์ฌ์ฉํ๋ ์ผ๋ถ ํ๋กํ ํ์ดํ ๋๊ตฌ๋ค์ ๋์ฒดํ ์ ์๋ค๊ณ ์ ๋งํฉ๋๋ค. ํนํ ํฅ๋ฏธ๋ก์ด ์ ์ ์ ๋๋ฉ์ด์ ๋ฐ ๋น๋์ค ํ ํ๋ฆฟ๋ ์ ๊ณตํ๋ค๋ ๊ฒ์ ๋๋ค.
1.1. ํต์ฌ ๊ฐ์ : ๋์์ธ ์์คํ ํตํฉ
Claude Design์ด ๊ฐ์ฅ ์ฃผ๋ชฉ๋ฐ๋ ์ด์ ๋ ๋ฐ๋ก โ๋์์ธ ์์คํ (design system)์ ํต์ฌ ์์๋ก ๊ฒฉ์์์ผ ์ต์ฐ์ ์ผ๋ก ๋ค๋ฃฌ๋คโ๋ ์ ์ ๋๋ค. ํด๋ ์ด๋ ์ด ๋๊ตฌ๋ฅผ ์ฒ์ ์ฌ์ฉํ์ ๋ ๊ฐ์ฅ ๋จผ์ ์์ ์ ๋์์ธ ์์คํ ์ ๊ฐ์ ธ์ ๋ณด์๋ค๊ณ ๋งํฉ๋๋ค. ์ด๋ ๋งค์ฐ ์๋ฆฌํ ์ ๊ทผ ๋ฐฉ์์ธ๋ฐ, ์ ํ ๊ฐ๋ฐ ๊ณผ์ ์์ ํ๋กํ ํ์ ์ ์ฌ์ฉํ ๋ ์ฌ๋๋ค์ด ๊ฐ์ฅ ๋ง์ด ๋ถํํ๋ ๊ฒ์ด ๋ฐ๋ก โ์ฐ๋ฆฌ ๋ธ๋๋์ ๋ง์ง ์๋๋คโ, โ์ฐ๋ฆฌ ๋์์ธ ์์คํ ๊ณผ ์ผ์นํ์ง ์๋๋คโ๋ ์ ์ด๊ธฐ ๋๋ฌธ์ ๋๋ค. ๊ธฐ์กด์ ์ฌ๋ฌ ํ๋กํ ํ์ดํ ๋๊ตฌ๋ค๋ ๋์์ธ ์์คํ ๊ฐ์ ธ์ค๊ธฐ ๊ธฐ๋ฅ์ ์ ๊ณตํ์ง๋ง, ์ฝ๋ ๋ฒ ์ด์ค์ ์๋ ๊ธฐ์กด ๋์์ธ์ ์๋ฒฝํ๊ฒ ์ฌํํ๋ ๋ฐ๋ ํ๊ณ๊ฐ ์์์ต๋๋ค.
ํด๋ ์ด๋ Claude Design์ด ์ด ๋ฌธ์ ๋ฅผ ์ผ๋ง๋ ์ ํด๊ฒฐํ๋์ง ํ ์คํธํ๊ธฐ ์ํด ๋ ๋ ๋ด์ค๋ ํฐ(Lennyโs Newsletter)์ ๋์์ธ ์์คํ ์ ๊ฐ์ ธ์ค๋ ๊ณผ์ ์ ์์ฐํ์ต๋๋ค. ๋จ์ํ ๋ ๋์ ํํ์ด์ง HTML์ ์ ์ฅํ๊ณ , ๋ก๊ณ ์ ์ด๋ฏธ์ง๋ค์ ์ ๋ก๋ํ ํ ํฐํธ๋ฅผ ์ถ๊ฐํ๋ ๊ฒ๋ง์ผ๋ก Claude Design์ ํด๋น ์๋ฃ๋ค์ ํ์ํ๊ณ ๋์์ธ ์์คํ ์ ๊ตฌ์ถํ๊ธฐ ์์ํ์ต๋๋ค. ์ด ๊ณผ์ ์ ์ฝ 5๋ถ ์ ๋ ์์๋๋ฉฐ, Claude Design์ HTML๊ณผ ํต์ฌ ์ด๋ฏธ์ง๋ฅผ ๋ถ์ํ์ฌ ์๊ฐ์ ์ ๋ณด๋ฅผ ํ์ ํ๊ณ , ์ฃผ์ ์์ ๋ฐ ๊ตฌ์กฐ๋ฅผ ์ถ์ถํ์ฌ ์์ฒด์ ์ธ ๋์์ธ ์์คํ ์ผ๋ก ๊ตฌ์ฑํฉ๋๋ค.
์ด๋ฌํ ์ ๊ทผ ๋ฐฉ์์ ๋์์ธ ์์คํ
์ UI ํคํธ, ํ์ดํฌ๊ทธ๋ํผ(typography), ์์, ์ปดํฌ๋ํธ(component), ๋ธ๋๋ ๋งํฌ ๋ฑ์ผ๋ก ์ธ๋ถํํ์ฌ AI ๋๊ตฌ๊ฐ ์ผ๊ด์ฑ ์๊ณ ๊ณ ํ์ง์ UI๋ฅผ ๋ ๋๋งํ๋ ๋ฐ ํ์ฉํ ์ ์๋๋ก ํฉ๋๋ค. ์ค์ ๋ก ๊ตฌ๊ธ ๋ฉ์ค(Google Labs)๋ ์ต๊ทผ design.md๋ผ๋ ํ์ค์ ๋ฐํํ๋๋ฐ, ์ด๋ AI ์์ด์ ํธ(agent)๊ฐ ๋์์ธ ์์คํ
์ ์ดํดํ๊ณ ํ์ฉํ๋ ํ์ค ๋ฐฉ์์ ์ ์ํ๋ ค๋ ๋
ธ๋ ฅ์ ์ผํ์
๋๋ค. ์ด๋ AI๊ฐ ๋์์ธ ์์คํ
์ ๊ตฌ์กฐํ๋ ํํ๋ก ์ธ์ํ๊ณ ์ฌ์ฉํ๋ ๋ฐฉํฅ์ผ๋ก ์
๊ณ๊ฐ ๋์๊ฐ๊ณ ์์์ ๋ณด์ฌ์ค๋๋ค.
1.2. ์ค์ฉ์ ์ธ ํ์ฉ ์ฌ๋ก
Claude Design์ ํนํ ๋ค์๊ณผ ๊ฐ์ ์ํฉ์์ ๋งค์ฐ ์ ์ฉํ๊ฒ ํ์ฉ๋ ์ ์์ต๋๋ค.
-
๋ง์ผํ ๋๋ฉ ํ์ด์ง ๋ฐ ์น์ฌ์ดํธ: Claude Design์ ๋ง์ผํ ์ ์ฉ ๋๋ฉ ํ์ด์ง๋ ์น์ฌ์ดํธ์ ๋์์ธ ์์คํ ์ ์ ์ฉํ๋ ๋ฐ ํ์ํ ๋ฅ๋ ฅ์ ๋ณด์ฌ์ค๋๋ค. ๋ธ๋๋ ์์ฐ์ ์๋ฆ๋ค์ด ์ํฌํ๋ก์ฐ์ ํตํฉํ๊ธฐ ์ด๋ ค์ ํ๋กํ ํ์ดํ์ ๊บผ๋ ธ๋ ๋ง์ผํฐ๋ค์๊ฒ ํนํ ์ ์ฉํ ๊ฒ์ ๋๋ค.
- ์์: LennyDoc PRD ๋น๋ ํด๋ ์ด๋ ๋ ๋์ ๋์์ธ ์์คํ ์ ํ์ฉํ์ฌ โLennyDocโ์ด๋ผ๋ PRD(Product Requirements Document) ์์ฑ๊ธฐ ๋๋ฉ ํ์ด์ง๋ฅผ ๋ง๋ค์์ต๋๋ค. โ๋ ๋ ๋ด์ค๋ ํฐ์ ๋ชจ๋ ๋ฐ์ดํฐ๋ฅผ ๊ธฐ๋ฐ์ผ๋ก ํ PRD ์์ฑ๊ธฐ ๋ฐ AI PM ์ฝ์น๋ฅผ ์ํ ๋๋ฉ ํ์ด์ง๋ฅผ ๋ง๋ค์ด ๋ฌ๋ผ. ๋ ๋ ๋ด์ค๋ ํฐ ๋์์ธ ์์คํ ์ ๋ฐ๋ฅด๊ณ , ๋ฉ์ง๊ฒ ๋ง๋ค์ด ๋ฌ๋ผโ๋ ๊ฐ๋จํ ํ๋กฌํํธ(prompt)๋ง์ผ๋ก Claude Design์ ๋ค์ํ ์ง๋ฌธ๊ณผ ๋ต๋ณ์ ํตํด ๋์์ธ์ ๊ตฌ์ฒดํํ์ต๋๋ค. ํค๋๋ผ์ธ ์คํ์ผ, ๋ ์ด์์, ๋ฐฐ๊ฒฝ์, CTA(Call To Action) ๋ฌธ๊ตฌ ๋ฑ ์ฌ๋ฌ ์ต์ ์ ์ ๊ณตํ์ฌ ์ฌ์ฉ์๊ฐ ์ฝ๊ฒ ์ ํํ๊ณ ๋ณ๊ฒฝํ ์ ์๊ฒ ํฉ๋๋ค. ์ด๋ ๋์์ธ ๋๊ตฌ ๊ด์ ์์ ๋งค์ฐ ์๋ฆฌํ ๊ธฐ๋ฅ์ธ๋ฐ, ๋๋ถ๋ถ์ ์ฌ์ฉ์๋ ์ ํํ ์ด๋ค ๋ณ๊ฒฝ์ ์ํ๋์ง ๋ช ํํ ์ค๋ช ํ๊ธฐ ์ด๋ ต๊ธฐ ๋๋ฌธ์ ๋๋ค.
- ํผ๋๋ฐฑ ๋ฐ ์์ ๊ธฐ๋ฅ: Claude Design์ ํน์ ์ปดํฌ๋ํธ์ ๋๊ธ์ ๋ฌ์ AI๊ฐ ์ง์ ์์ ํ๋๋ก ์์ฒญํ ์ ์์ต๋๋ค. ์๋ฅผ ๋ค์ด, โCTA ํ ์คํธ๋ฅผ ๊ฐ์ ํด ๋ฌ๋ผโ๊ณ ์์ฒญํ๋ฉด AI๊ฐ ์ด๋ฅผ ๋ฐ์ํ์ฌ ๋ณ๊ฒฝํด ์ค๋๋ค. ์ด๋ ํ๋กํ ํ์ , ์ธ๋น์ (InVision)์ ๋๊ธ ๊ธฐ๋ฅ, ํผ๊ทธ์ผ(FigJam)์ ์์ด๋์ด ๊ณต์ , ํผ๊ทธ๋ง(Figma)์ ๋๊ธ ๊ธฐ๋ฅ ๋ฑ ๋ค์ํ ํ์ ๋๊ตฌ์ ์ฅ์ ์ AI ๊ธฐ๋ฐ์ผ๋ก ํตํฉํ ํํ์ ๋๋ค.
-
๊ณ ํ์ง ํ๋ ์ ํ ์ด์ ์ฌ๋ผ์ด๋: Claude Design์ ๊ธฐ์กด ์ฝํ ์ธ ์ ๋์์ธ ์์คํ ์ ๊ฒฐํฉํ์ฌ ์๋ฆ๋ค์ด ํ๋ ์ ํ ์ด์ ์ฌ๋ผ์ด๋๋ฅผ ๋ง๋๋ ๋ฐ๋ ๊ฐ์ ์ ๋ณด์ ๋๋ค. ํด๋ ์ด๋ ์์ ์ด ์์ฑํ Open Claw ๊ธฐ์ฌ PDF๋ฅผ ์ ๋ก๋ํ๊ณ , ์ด๋ฅผ ๋ฐํ์ผ๋ก Open Claw ์ค์ ๋ฐฉ๋ฒ์ ๊ฐ๋ฅด์น๋ ์ฌ๋ผ์ด๋ ๋ฑ์ ๋ง๋ค์ด ๋ฌ๋ผ๊ณ ์์ฒญํ์ต๋๋ค. ๊ฒฐ๊ณผ๋ฌผ์ ๋์์ธ ์์คํ ์ ์ถฉ์คํ๋ฉด์๋ ๋ด์ฉ ์ ๋ฌ์ ํจ๊ณผ์ ์ธ ๋งค์ฐ ๋ฉ์ง ํ๋ ์ ํ ์ด์ ์ด์์ต๋๋ค. ํฐ๋ฏธ๋(terminal) ๋ช ๋ น์ด๋ฅผ ์ํ ๊น๋นก์ด๋ ์ปค์๊ฐ ์๋ ๊ฐ์ง ํฐ๋ฏธ๋ ํ๋ฉด๊ณผ ๊ฐ์ ๊ท์ฌ์ด ๋์์ธ ์์๋ค์ ์ฌ๋ผ์ด๋๊ฐ ์ค์ ๋ก๋ ์ฝ๋ ๊ธฐ๋ฐ์ผ๋ก ๋ง๋ค์ด์ง๋ค๋ ๊ฒ์ ๋ณด์ฌ์ค๋๋ค. ์ด๋ ์ ํ ๋ง์ผํฐ๋ค์ด ๊ต์ก ์๋ฃ, ๊ณ ๊ฐ์ฉ ๋ฐํฌ ๋ฑ์ ๋ง๋ค ๋ ๋ธ๋๋ ์ผ๊ด์ฑ์ ์ ์งํ๋ฉด์๋ ํจ์จ์ ์ผ๋ก ์์ ํ ์ ์๊ฒ ํด ์ค ๊ฒ์ ๋๋ค.
-
์ฌ๋ฏธ์๋ ์น์ฌ์ดํธ ์ฌํด์: Claude Design์ ๋์์ธ ์์คํ ์ ์ ๊ณตํ์ง ์์ผ๋ฉด, ๋งค์ฐ ๋ ํนํ๊ณ ์ฐฝ์์ ์ธ ๊ฒฐ๊ณผ๋ฌผ์ ์ป์ ์ ์์ต๋๋ค. ํด๋ ์ด๋ โ90๋ ๋ ์ง์ค์ํฐ(GeoCities) ์คํ์ผ์ ๋ ๋ ๋ด์ค๋ ํฐ ํํ์ด์งโ๋ฅผ ๋ง๋ค์ด ๋ฌ๋ผ๊ณ ์์ฒญํ๋๋ฐ, โ๋ ๋์ ์ ํ ์กด(Lennyโs Product Zone)โ์ด๋ผ๋ ์ด๋ฆ์ผ๋ก ํ์ํ ๊ฒฐ๊ณผ๋ฌผ์ ์ ๋ง ๋๋ผ์ ์ต๋๋ค. ํนํ Claude Design์ ๋ฐ์ด๋ ์นดํผ๋ผ์ดํ ๋ฅ๋ ฅ๋ ๊ฐ์ง๊ณ ์์ด, โ๋น์ ์ OKR์ ๋์ฐํ๋ค(Your OKRs are cringe)โ์ ๊ฐ์ ์ฌ์น ์๋ ๋ฌธ๊ตฌ๋ค์ ์์ฑํด ๋ ๋๋ค. ์ด๋ ๋ ํผ๋ฐ์ค ์คํ์ผ์ ์ ์ดํดํ๊ณ ์๋ค๋ฉด, ๋์์ธ ์์คํ ์ ์ ์ฝ ์์ด AI๊ฐ ์์ ๋กญ๊ฒ ์ฐฝ์ํ๋๋ก ํ์ฌ ๋งค์ฐ ๋ ํนํ๊ณ ๋ฉ์ง ๋์์ธ์ ์ป์ ์ ์์์ ์์ฌํฉ๋๋ค.
1.3. Claude Design์ ํ๊ณ
Claude Design์ ๋ง์ ์ฅ์ ์ ๊ฐ์ง๊ณ ์์ง๋ง, ์์ง ํด๊ฒฐํด์ผ ํ ๊ณผ์ ๋ ์์ต๋๋ค. ๊ฐ์ฅ ํฐ ๋จ์ ์ ๋ฐ๋ก ๋๋ฆฐ ์๋์ ๋๋ค. ๋๋ฉ ํ์ด์ง ํ๋๋ฅผ ์์ฑํ๋ ๋ฐ 5๋ถ์์ 10๋ถ ์ ๋๊ฐ ์์๋ ์ ์์ต๋๋ค. ๋์์ธ ์์ ์์๋ ๋น ๋ฅธ ํผ๋๋ฐฑ ๋ฃจํ(feedback loop)๊ฐ ๋งค์ฐ ์ค์ํ๋ฐ, Figma์ ๊ฐ์ ๋๊ตฌ๋ ๋๋๊ทธ ์ค ๋๋กญ(drag and drop)์ผ๋ก ์ฆ์ ์์๋ฅผ ๋ณ๊ฒฝํ๊ณ ํฐํธ๋ฅผ ์์ ํ ์ ์์ต๋๋ค. ๋ฐ๋ฉด Claude Design์ LLM(Large Language Model) ํธ์ถ์ ๊ธฐ๋ค๋ฆฌ๊ฑฐ๋ ํฌ๋ ๋ง(credit)์ ์ถฉ์ ํด์ผ ํ๋ ๋ฑ ์ง์ฐ์ด ๋ฐ์ํฉ๋๋ค. ๋ชจ๋ธ์ด ๊ฐ์ ํ์ง ์๋ ๋น ๋ฅธ ๋ฐ๋ณต ์์ ์ ๊ฐ์น๋ฅผ ๊ณผ์ํ๊ฐํด์๋ ์ ๋ฉ๋๋ค. ๋ํ, ํฌ๋ ๋ง ์ ํ์ผ๋ก ์ธํด ์์ ๋์ค ์งํ์ด ๋งํ๋ ๊ฒฝ์ฐ๋ ์์ด ์ฌ์ฉ์ ๊ฒฝํ์ ์ํฅ์ ๋ฏธ ๋ฏธ์นฉ๋๋ค.
2. ์ด๋ฏธ์ง ์์ฑ์ ์๋ก์ด ์๋: OpenAI์ GPT Image 2
OpenAI๋ ์ต๊ทผ GPT Image 2 ๋ชจ๋ธ์ ์ถ์ํ๋ฉฐ ์ด๋ฏธ์ง ์์ฑ์ ์๋ก์ด ์๋๋ฅผ ์ด์์ต๋๋ค. ์ด ๋ชจ๋ธ์ โ์ฌ๊ณ ํ๋(thinking)โ ๋ฅ๋ ฅ์ ๊ฐ์ถ ์ต์ด์ ๋ชจ๋ธ๋ก, ํนํ ํ ์คํธ ๋ ๋๋ง(rendering)๊ณผ ๊ฐ์ฒด(object)์ ์ ํํ ํํ์ ์ค์ ์ ๋์์ต๋๋ค. ํด๋ ์ด๋ GPT Image 2๊ฐ ์ด ๋ ๊ฐ์ง ์ธก๋ฉด์์ ์ผ๋ง๋ ๋ฐ์ด๋ ์ฑ๋ฅ์ ๋ณด์ด๋์ง ๋ ๊ฐ์ง ํ์ฉ ์ฌ๋ก๋ฅผ ํตํด ๋ณด์ฌ์ฃผ์์ต๋๋ค.
2.1. ๋ธ๋๋ฉ ๋ฐ ๋ง์ผํ ํ์ฉ
- ๋ฉํฐ ํ์ด์ง ๋ธ๋๋ ํคํธ ์์ฑ: ํด๋ ์ด๋ ์์ ์ ํ์ฌ์ธ Chat PRD๋ฅผ ์ํ ๋ฉํฐ ํ์ด์ง ๋ธ๋๋ ํคํธ๋ฅผ ๋ง๋ค์ด ๋ฌ๋ผ๊ณ ์์ฒญํ์ต๋๋ค. ์ผ๋ฐ์ ์ธ ๋ธ๋๋ ๋์์ธ ์ง์นจ์ ์ ๊ณตํ์, GPT Image 2๋ ํ ์คํธ ํ์ง ๋ฉด์์ ๋งค์ฐ ํ๋ฅญํ ๊ฒฐ๊ณผ๋ฌผ์ ๋ด๋์์ต๋๋ค. ๊ธฐ์กด GPT ์ด๋ฏธ์ง ๋ชจ๋ธ์ ํน์ง์ ์ธ โAI ์ด๋ฏธ์ง ํ ์คํธโ ๋๋์ด ๊ฑฐ์ ์ฌ๋ผ์ก๊ณ , ํ์ดํฌ๊ทธ๋ํผ๋ ํฌ๊ฒ ๊ฐ์ ๋์์ต๋๋ค. 9๊ทธ๋ฆฌ๋ ๋ ์ด์์์ผ๋ก ๋ค์ํ ๋ธ๋๋ ์์๋ฅผ ๊ตฌ์ฑํ๋ ๋ฅ๋ ฅ๋ ์ธ์์ ์ด์์ต๋๋ค.
- ์ฐธ์กฐ ์ด๋ฏธ์ง๋ฅผ ํตํ ๊ฐ์ : ์ฒ์ ์์ฑ๋ ๋ธ๋๋ ํคํธ๋ Chat PRD์ ๊ธฐ์กด ๋ธ๋๋ ์ด๋ฏธ์ง์๋ ๋ค์ ๊ฑฐ๋ฆฌ๊ฐ ์์์ต๋๋ค. ์ด์ ํด๋ ์ด๋ ๋ฏธ๋์ ๋(Midjourney)์์ ์ฌ์ฉํ๋ ๋ฐ๊ณ ํํฌ์ ํค์ ํฝ์ ํ๋ ์ด๋ฏธ์ง๋ฅผ ์ฐธ์กฐ ์ด๋ฏธ์ง๋ก ์ ๊ณตํ๋ฉฐ ๋ธ๋๋ ํคํธ๋ฅผ ์ ๋ฐ์ดํธํด ๋ฌ๋ผ๊ณ ์์ฒญํ์ต๋๋ค. GPT Image 2๋ ์ด ์ฐธ์กฐ ์ด๋ฏธ์ง์ ์คํ์ผ์ ์ฑ๊ณต์ ์ผ๋ก ํ์ตํ์ฌ ํจ์ฌ ๋ ํํฌ์์ด๊ณ ํฝ์ ํ๋ ๋๋์ ๋ธ๋๋ ํคํธ๋ฅผ ์์ฑํ์ต๋๋ค. ์ด๋ ๋ง์ผํฐ๋ค์ด ๊ธฐ์กด ์ด๋ฏธ์ง์ ํ ์คํธ, ํ์ดํฌ๊ทธ๋ํผ, ๋ณด์ด์ค(voice)๋ฅผ ํตํฉํ๋ ๋ฐ ์ด๋ ค์์ ๊ฒช๋ ์ํฉ์์ GPT Image 2๊ฐ ํ๋ฅญํ ์์์ ์ด ๋ ์ ์์์ ๋ณด์ฌ์ค๋๋ค.
2.2. ๊ฐ์ธ์ ์ธ ํ์ฉ ์ฌ๋ก: ์ปฌ๋ฌ ๋ถ์
ํด๋ ์ด๋ GPT Image 2์ ๋ ๋ค๋ฅธ ์ฌ๋ฏธ์๋ ํ์ฉ ์ฌ๋ก๋ก โ์ปฌ๋ฌ ๋ถ์(color analysis)โ์ ์๊ฐํ์ต๋๋ค. ์์ ์ ์ฌ์ง์ ์ ๋ก๋ํ๊ณ ์ปฌ๋ฌ ๋ถ์์ ์์ฒญํ์, ๋ชจ๋ธ์ โ์ ๋ดํธ๋ด(warm neutral)โ ํค์ ์ ์ํ๋ฉฐ ์ด์ธ๋ฆฌ๋ ์์ ํ๋ ํธ๋ฅผ ์ ์ํ๊ณ , ์ฌ์ง์ด ๊ทธ ์์์ ์ท์ ์ ์ ํด๋ ์ด์ ๋ชจ์ต์ ํฉ์ฑํด ๋ณด์ฌ์ฃผ์์ต๋๋ค. ์ผ๊ตด ๋ถ๋ถ์ด ๋ค์ ๋ถ์์ฐ์ค๋ฝ๊ฒ ํํ๋๊ธฐ๋ ํ์ง๋ง, ์ ๋ฐ์ ์ธ ๋ ์ด์์, ์์, ํ ์คํธ ๋ถ์ ๋ฅ๋ ฅ์ ๋งค์ฐ ์ธ์์ ์ด์์ต๋๋ค.
ํด๋ ์ด๋ ์์ ์ด ์ค์ ๋ก๋ โ๋คํฌ ์ํฐ(dark winter)โ ํค์์ ์๋ ค์ฃผ๋ฉฐ ๋ค์ ๋ถ์์ ์์ฒญํ๊ณ , GPT Image 2๋ ์ ํํ ์์ ํ๋ ํธ์ ๋ ์ด์์์ ์์ฑํด ์ฃผ์์ต๋๋ค. ํ ์คํธ, ์์, ์ด๋ฏธ์ง ๋ถ์, ๋ ์ด์์์ ์ด์ฒ๋ผ ํจ๊ณผ์ ์ผ๋ก ๊ฒฐํฉํ๋ ๋ชจ๋ธ์ ๋๋ฌผ๋ฉฐ, ๊ธฐ์กด์ ์ธํฌ๊ทธ๋ํฝ(infographic)์ด๋ ๊ทธ๋ํฝ ์์ฑ ๋๊ตฌ๋ณด๋ค ํจ์ฌ ๋ ๊ณ ๊ธ์ค๋ฝ๊ณ ๊น๋ํ ๊ฒฐ๊ณผ๋ฌผ์ ์ ๊ณตํฉ๋๋ค.
3. AI ๋์์ธ ๋๊ตฌ์ ๋ฏธ๋์ ์์ฌ์
์ด๋ฒ ์ํผ์๋์์ ์ดํด๋ณธ Claude Design๊ณผ GPT Image 2๋ AI๊ฐ ๋์์ธ ๋ถ์ผ์์ ์ผ๋ง๋ ๋น ๋ฅด๊ฒ ๋ฐ์ ํ๊ณ ์๋์ง๋ฅผ ์ฌ์คํ ๋ณด์ฌ์ค๋๋ค.
- Claude Design์ ๋์์ธ ์์คํ ์ ๊ธฐ๋ฐ์ผ๋ก ๋ง์ผํ ๋๋ฉ ํ์ด์ง, ์ฌ๋ผ์ด๋ ๋ฐํฌ๋ฅผ ํจ์จ์ ์ผ๋ก ๋ง๋ค๊ณ , ๊ธฐ์กด ์น์ฌ์ดํธ๋ฅผ ์ฌํด์ํ๋ ๋ฐ ๊ฐ์ ์ ๋ณด์ ๋๋ค.
- GPT Image 2๋ ๋ ์ด์์๊ณผ ํ์ดํฌ๊ทธ๋ํผ์์ ํ์ ์ ์ธ ๋ฐ์ ์ ์ด๋ฃจ์์ผ๋ฉฐ, โ์ฌ๊ณ ํ๋โ ๋ฅ๋ ฅ์ ํตํด ๋์์ธ ํ์ง์ ํ ๋จ๊ณ ๋์ด์ฌ๋ ธ์ต๋๋ค.
๋ฌผ๋ก , ์ด ๋ชจ๋ ๋๊ตฌ๋ค์ ์์ง ์ด๊ธฐ ๋จ๊ณ์ด๋ฉฐ, ๋๋ฆฐ ์๋์ ํฌ๋ ๋ง ๋น์ฉ๊ณผ ๊ฐ์ ๊ณตํต์ ์ธ ํ๊ณ๋ฅผ ๊ฐ์ง๊ณ ์์ต๋๋ค. ํ์ง๋ง ๊ตฌ๊ธ์ design.md ํ์คํ ๋
ธ๋ ฅ์์ ๋ณผ ์ ์๋ฏ์ด, AI๊ฐ ๋์์ธ ์์คํ
๊ณผ ๋ธ๋๋๋ฅผ ์ดํดํ๋ ๋ฐฉ์์ ๊ณ์ํด์ ๋ฐ์ ํ๊ณ ์์ต๋๋ค.
๊ฒฐ๋ก ์ ์ผ๋ก, ์ด ์๋ก์ด AI ๋์์ธ ๋๊ตฌ๋ค์ ๊ฐ์์ ์ฅ์ ์ ๊ฐ์ง๊ณ ์์ผ๋ฉฐ, ์ด๋ค ๋๊ตฌ๋ โ์ฃฝ์๋คโ๊ณ ๋งํ๊ธฐ๋ ์ด๋ฆ ๋๋ค. ์ด๋ค์ ๋ชจ๋ ๋๋ฆฌ์ง๋ง ๊พธ์คํ ๋ฐ์ ํ๊ณ ์์ผ๋ฉฐ, ์ฐ๋ฆฌ์ ๋์์ธ ์์ ๋ฐฉ์์ ๋์ฑ ํ์ ์ ์ผ๋ก ๋ณํ์์ผ ๋๊ฐ ๊ฒ์ ๋๋ค. ์์ผ๋ก AI ๋๊ตฌ๋ค์ด ์ด๋ป๊ฒ ๋ฐ์ ํด ๋๊ฐ์ง ๊ธฐ๋๋ฉ๋๋ค.
โInside Kash Patelโs F.B.I.โ โ New York Times Podcasts ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
์นด์ฌ ํํ ์ FBI: ์ ์นํ๋ ์กฐ์ง, ํ๋ค๋ฆฌ๋ ์ฌ๋ช
๋ด์ํ์์ฆ์ ํ์บ์คํธ โ๋ ๋ฐ์ผ๋ฆฌโ๋ ์นด์ฌ ํํ ์ด ์ฐ๋ฐฉ์์ฌ๊ตญ(FBI) ๊ตญ์ฅ์ผ๋ก ์๋ช ๋ ์ดํ ๊ทธ์ ๋ฆฌ๋์ญ์ด ๋ถ๋ฌ์จ ๋ ผ๋๊ณผ ์ฐ๋ ค๋ฅผ ์ฌ์ธต ๋ณด๋ํ์ต๋๋ค. ์๋ฐ๋ฆฌ ๋ฐ์ค๋ก (Emily Basilon)๊ณผ ๋ ์ด์ฒผ ํฌ์ (Rachel Poser) ๋ ๊ธฐ์๋ ์์ญ ๋ช ์ ํ์ง ๋ฐ ์ ์ง FBI ์ง์๋ค๊ณผ์ ์ธํฐ๋ทฐ๋ฅผ ํตํด ์กฐ์ง์ด ์ด๋ป๊ฒ ๋ณ๋ชจํ๋์ง ๋ฉด๋ฐํ ์ถ์ ํ์ต๋๋ค. ์ด๋ค์ ๋ณด๊ณ ์๋ ํธ๋ผํ ํ์ ๋ถ ํ์์ FBI๊ฐ ๊ฒช์๋ ์ ์น์ ์๋ ฅ๊ณผ ๊ทธ๋ก ์ธํ ๋ด๋ถ ํผ๋, ๊ทธ๋ฆฌ๊ณ ๊ตญ๊ฐ ์๋ณด์ ๋ฏธ์น ์ ์๋ ์ ์ฌ์ ์ํ์ ์์ํ๊ฒ ๋ณด์ฌ์ค๋๋ค.
์นด์ฌ ํํ , ๋ ผ๋์ ์ค์ฌ์ ์๋ค
์นด์ฌ ํํ ์ FBI ๊ตญ์ฅ์ด๋ผ๋ ์ค์ฑ ์ ๋งค์ฐ ์ด๋ก์ ์ธ ์ธ๋ฌผ์ด์์ต๋๋ค. ๊ทธ๋ FBI์์ ์ผํ ๊ฒฝํ์ด ์ ๋ฌดํ๊ณ , ์ฐ๋ฐฉ ๋ฒ ์งํ ๊ธฐ๊ด์์์ ๊ฒฝํ๋ ๋ฏธ๋ฏธํ์ต๋๋ค. ์ ์ง ๊ตญ์ ๋ณํธ์ธ์ด์ ํธ๋ผํ ํ์ ๋ถ 1๊ธฐ์์ ์ ๋ณด ๊ณ ์ ๊ด๋ฃ๋ก ์ผํ๋ ๊ทธ๋, ์ฃผ๋ณ ์ฌ๋๋ค๋ก๋ถํฐ โ๋ฑํ ์ด์ฌํ ์ผํ์ง๋ ์์์ง๋ง ๋งค์ฐ ๊ฑด๋ฐฉ์ง๊ณ ์ผ์ฌ๋ง๋งํ๋ฉฐ ํํ์ด ์ฌํโ ์ธ๋ฌผ๋ก ๋ฌ์ฌ๋ฉ๋๋ค.
๋ฌด์๋ณด๋ค ํํ ์ ๊ณผ๊ฑฐ FBI์ ๋ํ ์๋ชจ๋ก ์ ์ ๊ธฐํ๋ฉฐ ์กฐ์ง์ ๋ํ ๊น์ ๋ถ์ ์ ๋๋ฌ๋์ต๋๋ค. ๊ทธ๋ 2020๋ ๋์ ์ด ์กฐ ๋ฐ์ด๋ (Joe Biden)์ ์ํด ์กฐ์๋์๋ค๊ณ ์ฃผ์ฅํ์ผ๋ฉฐ, 1์ 6์ผ ๋ฏธ๊ตญ ๊ตญํ์์ฌ๋น ๋์ ์ฌํ์๋ FBI ์์๋ค์ด ๊ฐ์ ํ๋ค๊ณ ๋ฏฟ์์ต๋๋ค. ํธ๋ผํ ์ ๋ํต๋ น์ด ๊ทธ๋ฅผ ๋ฐํํ ์ด์ ์ค ํ๋๋ ๋ฐ๋ก ์ด๋ฌํ ์๊ฐ ๋๋ฌธ์ด์์ต๋๋ค. ํํ ์ ์ฌ์ง์ด FBI ๋ณธ๋ถ๋ฅผ ํ์ํ๊ณ โ๋ฅ ์คํ ์ดํธ(deep state) ๋ฐ๋ฌผ๊ดโ์ผ๋ก ์ฌ๊ฐ์ฅํด์ผ ํ๋ค๊ณ ์ฃผ์ฅํ๊ธฐ๋ ํ์ต๋๋ค.
์ด๋ฌํ ๋ฐฐ๊ฒฝ ๋๋ฌธ์ ํํ ์ ์๋ช ์ FBI ๋ด์ธ๋ถ์์ ํฐ ํ์๊ฐ๊ณผ ์ฐ๋ ค๋ฅผ ๋ถ๋ฌ์ผ์ผ์ผฐ์ต๋๋ค. ์ ๋ณด๊ตญ์ฅ ํ๋ ์ฐ๊ณ ๋ฆฌ์ธ (Tanya Ugoritz)์ ๊ฐ์ ๊ณ ์ ๊ฐ๋ถ๋ค์ ํํ ์ ๋ฐฐ๊ฒฝ๊ณผ ์ ๋ฌด ๋ฐฉ์์ ๋ํ ์ดํด๋ฅผ ๋์ด๊ธฐ ์ํด ๊ทธ์ ์ฑ ์ ์ฝ๊ณ ํ์บ์คํธ๋ฅผ ์ฒญ์ทจํ๋ ๋ฑ ๊ฐ์ธ์ ์ธ ๋ ธ๋ ฅ์ ๊ธฐ์ธ์์ง๋ง, ๊ณง ์ด๊ฒ์ด ํ์์ ๋ค๋ฅธ ์ํฉ์์ ๊นจ๋ซ๊ฒ ๋ฉ๋๋ค.
๊ธ์ง์ ๋ณํ์ ์กฐ์ง ๋ด๋ถ์ ๋์
ํํ ์ ๋ฆฌ๋์ญ ํ์์ FBI๋ ๊ณผ๊ฑฐ์ ๋ ๋ฆฝ์ฑ์ ์์คํ๊ณ ์ ์น์ ์๋ ฅ์ ์๋ฌ๋ฆฌ๋ ์กฐ์ง์ผ๋ก ๋ณ๋ชจํ์ต๋๋ค.
1. ๋ฌด๋ถ๋ณํ ์กฐ์ง ๊ฐํธ๊ณผ ์ธ์ฌ (Staffing Changes and Appointments)
ํํ ์ ์ด๊ธฐ ์ง์ ์ค ํ๋๋ ์๋ฐฑ ๋ช ์ ํ์ฅ ์์์ ์์ฑํด D.C. ๋ณธ๋ถ์์ ๊ฐ ์ง๋ฐฉ ์ฌ๋ฌด์๋ก ์ฌ๋ฐฐ์นํ๋ ๊ฒ์ด์์ต๋๋ค. ์ด๋ D.C.์ ๋๋ฌด ๋ง์ ์ธ๋ ฅ์ด ์ง์ค๋์ด ์๋ค๋ FBI ๋ด๋ถ์ ์๊ฒฌ๋ ์์๊ธฐ์ ์ผ๋ถ์์๋ ๊ธ์ ์ ์ผ๋ก ํ๊ฐ๋๊ธฐ๋ ํ์ต๋๋ค. ๊ทธ๋ฌ๋ ๋ฌธ์ ๋ ๊ทธ ๋ฐฉ์์ด์์ต๋๋ค. ์ฌ๋ฐฐ์น๊ฐ ๋ฌด์์์ ์ด๊ณ ์์์ ์ผ๋ก ์ด๋ฃจ์ด์ก์ผ๋ฉฐ, ์์๋ค์ด ์๋ก์ด ํ๊ฒฝ์์ ์ ๋๋ก ์ ๋ฌด๋ฅผ ์ํํ ์ ์์์ง์ ๋ํ ๊ณ ๋ ค๊ฐ ๋ถ์กฑํ์ต๋๋ค. ํ๋ ์ฐ๊ณ ๋ฆฌ์ธ ๋ ์ด๋ฅผ โ๋ณธ๋ถ์ ์ ๋ฌด์ ๋ํ ์ดํด ๋ถ์กฑโ์์ ๋น๋กฏ๋ ๊ฒฐ์ ์ด๋ผ๊ณ ์ง์ ํ๋ฉฐ, โ์ผ๋จ ๊ฒฐ์ ํ๊ณ ๋์ค์ ์์ตํ๋โ ํจํด์ ์์์ด์๋ค๊ณ ํ๊ณ ํ์ต๋๋ค.
์ด๋ฌํ ๊ฒฝํฅ์ ํํ ์ด ๋ ๋ด๊ธฐ๋ ธ(Dan Bongino)๋ฅผ ๋ถ๊ตญ์ฅ์ผ๋ก ์๋ช ํ๋ฉด์ ๋์ฑ ์ฌํ๋์์ต๋๋ค. ์ ์ง ๋น๋ฐ๊ฒฝํธ๊ตญ ์์์ด์ ์น(่ฆช)ํธ๋ผํ ํ์บ์คํฐ, ํญ์ค ๋ด์ค(Fox News) ํ๋ก ๊ฐ์๋ ๋ด๊ธฐ๋ ธ๋ FBI ์์๋ค์ โ๋ฏผ์ฃผ๋น์ ๊นกํจโ๋ผ๊ณ ๋ฌ์ฌํ๊ณ FBI ํด์ฒด๋ฅผ ์ฃผ์ฅํ๋ ์ธ๋ฌผ์ด์์ต๋๋ค. ์กด ์ค๋ฆฌ๋ฒ(John Sullivan)๊ณผ ๊ฐ์ ์ ๋ณด ๋ถ์์ ์์๋ค์ ๋ด๊ธฐ๋ ธ์ ์๋ช ์ ๋ณด๋ฉฐ โ์ด์ ๋ชจ๋ ๊ฒ์ด ๊ด์ฐฎ์ง ์์ ๊ฒโ์ด๋ผ๊ณ ์ง๊ฐํ์ต๋๋ค.
2. โ๋ณด์ฌ์ฃผ๊ธฐ์โ ํ์ ๊ณผ ์๋ฌด์ ์ ์นํ (Prioritizing Optics and Politicized Mission)
ํํ ๊ณผ ๋ด๊ธฐ๋ ธ์ ๋ฆฌ๋์ญ์์ ๊ฐ์ฅ ๋๋๋ฌ์ง ํน์ง์ โ๋ณด์ฌ์ฃผ๊ธฐโ ์ฆ, ํ๋ณด ํจ๊ณผ๋ฅผ ์ง๋์น๊ฒ ์ค์ํ๋ค๋ ์ ์ ๋๋ค. ํธ๋ผํ ์ ๋ํต๋ น์ด ์์ฌ ํ์ฅ์์ FBI ๋ฐฉํ์กฐ๋ผ(flak jackets)๋ฅผ ๋ณด์ง ๋ชปํด ๋ถ๋ง์ ํ์ถํ ์ผํ๋ ์ด๋ฌํ ๊ฒฝํฅ์ ๋จ์ ์ผ๋ก ๋ณด์ฌ์ค๋๋ค. ํํ ์ ์ฝดํฐ์ฝ(Quantico) ํ๋ จ์ฅ์์ โ๋๋ณดโ์ฒ๋ผ ํญ๋ฐ ์ฅ๋ฉด๊ณผ ํฌ๋ฆฌ์ฝฅํฐ ๋ ํ ํ๋ จ์ ๋ฐฐ๊ฒฝ์ผ๋ก ์ดฌ์ํ ๋น๋์ค๋ฅผ ๊ณต๊ฐํ๊ธฐ๋ ํ์ต๋๋ค. ์ด๋ FBI๊ฐ ๊ฐ์ธํ ์กฐ์ง์ด๋ผ๋ ์ด๋ฏธ์ง๋ฅผ ์ฌ์ด์ฃผ๊ธฐ ์ํจ์ด์์ง๋ง, ์ค์ ๋ก๋ ๊ท์คํ ์์๊ณผ ์๊ฐ์ ๋ญ๋นํ๊ณ ์กฐ์ง์ ์ ๋ฌธ์ฑ์ ํผ์ํ๋ ํ์๋ก ๋นํ๋ฐ์์ต๋๋ค.
์ด๋ฌํ โ๋ณด์ฌ์ฃผ๊ธฐ์โ ํ์ ์ ์ฐฐ๋ฆฌ ์ปคํฌ(Charlie Kirk) ์ด๊ฒฉ ์ฌ๊ฑด ์์ฌ์์ ์ฌ๊ฐํ ๋ฌธ์ ๋ฅผ ์ผ๊ธฐํ์ต๋๋ค. FBI ๊ตญ์ฅ์ผ๋ก์๋ ์ด๋ก์ ์ผ๋ก ํํ ์ ์์ฌ ์ด๊ธฐ ๋จ๊ณ๋ถํฐ ๊ฐ์ ํ์ฌ ํ์ฅ ์์๋ค์ ์ง์ฑ ํ๊ณ ์์ ๋ฏธ๋์ด ์ ๋ต์ ์ง์ ์ง์ํ์ต๋๋ค. ์ด ๊ณผ์ ์์ ์๋ชป๋ ์ ๋ณด๊ฐ ์ ํฌ๋์ด ์์ฌ์ ํผ์ ์ ์ด๋ํ๊ณ , ํ์ฅ ์์๋ค์ ์ฌ์ค ํ์ธ๋ณด๋ค ์ธ๋ก ๋์์ ๊ธ๊ธํ ์ํฉ์ ๊น์ ์ข์ ๊ฐ์ ๋๊ผ์ต๋๋ค. ์์ฌ ์ค ์ฌ์ง ์ดฌ์์ ๊ฐ์ํ๋ ๋ฑ์ ์ง์๋ ์์ ๋ณด์๊ณผ ์์๋ค์ ์์ ๊น์ง ์ํํ์ต๋๋ค.
์ฌ์ง์ด ํด์ธ ๋๋งน๊ตญ๊ณผ์ ๊ด๊ณ์๋ ์ ์ํฅ์ ๋ฏธ์ณค์ต๋๋ค. ํํ ์ ์๊ตญ์์ ์ด๋ฆฐ ๋น๋ฐ ์ ๋ณด ํ์์ ์ฐธ์ํ ๋ค, ์ฐธ๊ฐ์๋ค์ ์ ์์ด ๊ณต๊ฐ๋์ง ์๋๋ก ์์ฒญ๋ ๋จ์ฒด ์ฌ์ง์ ์์ ์ ์์ ๋ฏธ๋์ด์ ๊ฒ์ํ๋ ค ํด ๊ตญ์ ์ ์ธ๊ต ๋ฌธ์ ๋ฅผ ์ผ์ผํฌ ๋ปํ์ต๋๋ค. ์๊ตญ ์ ๋ถ๊ฐ ๊ฐ๋ ฅํ๊ฒ ๋ง๋ฅํ ํ์์ผ ๊ฒ์๋ฅผ ์ฒ ํํ์ง๋ง, ์ด๋ ํํ ์ โ๋ณด์ฌ์ฃผ๊ธฐโ ์ง์ฐฉ์ด ์ผ๋ง๋ ์ฌ๊ฐํ ์์ค์ด์๋์ง๋ฅผ ๋ณด์ฌ์ฃผ๋ ์ฌ๋ก์ ๋๋ค.
FBI ๋๋ณ์ธ ๋ฒค ์๋ฆฌ์์จ(Ben Williamson)์ ์ด ๋ชจ๋ ๋ณด๋๋ฅผ โํ์ค๊ณผ ๋๋จ์ด์ง ์ต๋ช ์ ์ถ์ฒ์์ ๋์จ ๊ฐ์ง ์ด์ผ๊ธฐ์ ์ถ์ธกโ์ด๋ผ๊ณ ์ผ์ถํ์ง๋ง, ๊ธฐ์๋ค์ ์ฌ๋ฌ ๋ช ์ ์ค๋ช ์ถ์ฒ๋ฅผ ํตํด ํํ ์ ๋ฆฌ๋์ญ์ด ์ค์ง โ๋ํต๋ น์ด๋ผ๋ ๋จ ํ ๋ช ์ ์ฒญ์คโ์ ๋ง์กฑ์ํค๊ธฐ ์ํด ์ ์น์ ์ด๋ฏธ์ง๋ฅผ ์ต์ฐ์ ์ผ๋ก ์ผ์์์ ๋ฐํ๋์ต๋๋ค.
์๋ฌด ์ฌ์ ์์ โ์ ์น์ ์์ฒญโ
ํํ ์ ๋ฆฌ๋์ญ์ FBI์ ํต์ฌ ์๋ฌด๋ฅผ ์ฌ์ ์ํ๊ณ , ์กฐ์ง ๋ด๋ถ์ โ์ ์น์ ์์ฒญโ์ด๋ผ๋ ์์ธํ ๊ทธ๋ฆผ์๋ฅผ ๋๋ฆฌ์ ์ต๋๋ค.
1. ์ด๋ฏผ ๋จ์์ผ๋ก์ ์๋ฌด ์ ํ (Shift to Immigration Enforcement)
ํธ๋ผํ ํ์ ๋ถ์ ์ด๋ฏผ ๋จ์ ๊ฐํ ๊ธฐ์กฐ์ ๋ฐ๋ผ FBI์ ์ผ์์ ์ธ ์ ๋ฌด๋ ํฌ๊ฒ ๋ณ์ง๋์์ต๋๋ค. ์ ํต์ ์ผ๋ก ์ด๋ฏผ ๋จ์์ ๊ด์ฌํ์ง ์๋ FBI๋ ์ ๊ตญ์ ์ผ๋ก ์๋ง์ ์์๊ณผ ๋ถ์๊ฐ๋ค์ ์ด๋ฏผ ๊ด๋ จ ์ ๋ฌด์ ์ฌ๋ฐฐ์นํ์ต๋๋ค. ์ด๋ก ์ธํด ๊ณต๊ณต ๋ถํจ, ์ฌ์ด๋ฒ ๋ฒ์ฃ, ํ์ดํธ์นผ๋ผ ๋ฒ์ฃ, ๋ง์ฝ ๋ฐ๋งค, ํ ๋ฌ๋ฆฌ์ฆ ๋ฑ FBI์ ํต์ฌ ์ฐ์ ์์์๋ ๋ถ์ผ์์ ์ธ๋ ฅ ๊ณต๋ฐฑ์ด ๋ฐ์ํ์ต๋๋ค.
๋ก์ค์ค์ ค๋ ์ค์์ ๊ฐ๋ ฅ ๋ฒ์ฃ๋ฅผ ๋ด๋นํ๋ ์ ๋ณด ๋ถ์๊ฐ ์ง ํ์ฆ(Jill Fields)์ ์ฌ๋ก๋ ์ด๋ฌํ ๋ณํ์ ๋จ๋ฉด์ ๋ณด์ฌ์ค๋๋ค. ๊ทธ๋ ๋ ๋จํธ๊ณผ ํจ๊ป FBI์ ํ์ ํ๋ฉฐ ์๋ ์๊ฒ๋ โ์ฃผ๋์ด ์์โ ์ฌํท์ ์ ํ ์ ๋๋ก ์กฐ์ง์ ๋ํ ์๋ถ์ฌ์ด ์ปธ์ต๋๋ค. ๊ทธ๋ฌ๋ ํธ๋ผํ ํ์ ๋ถ์ ์ด๋ฏผ ๋จ์์ด ๊ฐํ๋๋ฉด์ ๊ทธ๋ ๋ ์ด๋ฏผ ๊ด๋ จ ์ ๋ฌด์ ํฌ์ ๋์๊ณ , ๊ฒฐ๊ตญ โ๋ ๋ ๋ผ์ธโ์ ๋๋ ์ง์๋ฅผ ๋ฐ๊ฒ ๋ฉ๋๋ค.
๊ทธ๋ ๋ ์์๋๊ฐ ์ด๋ฏผ์ธ๊ด์งํ๊ตญ(ICE) ์์๋ค์ ์ฒดํฌ๋ฅผ ๋ฐฉํดํ๋ค๋ ์ฃผ์ฅ์ ๋ํด ์๋น ํ๊ฐ(preassessment)๋ฅผ ์ค์ํ๋ผ๋ ์ง์๋ฅผ ๋ฐ์์ต๋๋ค. ์์๋๋ ํ์ฑ๊ธฐ๋ฅผ ์ฌ์ฉํ์ง๋ง, ์์๋ค์ด ์ค์ ํ ์ ์ ๋์ง ์์์ผ๋ฉฐ ๋ฒ์ ์ ์ด๋๋ ์ด๋ ํ ํ์๋ ํ์ง ์์๋ค๋ ๊ฒ์ด ํ์ฅ ์์๋ค์ ํ๋จ์ด์์ต๋๋ค. ๊ทธ๋ผ์๋ ๋ถ๊ตฌํ๊ณ ํ์ฆ๋ ์์ฌ๋ฅผ ๊ฐํํ๋ผ๋ ์๋ ฅ์ ๋ฐ์์ต๋๋ค. ๊ทธ๋ ๋ ์ด๋ฅผ โํ๋ฒ์ผ๋ก ๋ณดํธ๋ฐ๋ ํ๋โ์ด๋ผ๋ฉฐ ๊ฑฐ๋ถํ๊ณ , โ์ง๊ธ ํด๊ณ ๋๊ฑฐ๋, ์๋๋ฉด 40๋ ํ์ ๋ค๋ฅธ ํ์ ๋ถ๊ฐ ๋ค์ด์ ๋ฒ์ ์๋ฐํ ์ฌ๋๋ค์ ์ฐพ์๋ผ ๋ ํด๊ณ ๋ ๊ฒโ์ด๋ผ๋ ๊ฒฝ๊ณ ๋ฅผ ๋ฐ์์ต๋๋ค. ๊ฒฐ๊ตญ ๊ทธ๋ ๋ ๋ถ์์์ ๋ฐฐ์ ๋๊ณ ๊ฐ์๋ฅผ ๋ฐ๋ค๊ฐ FBI๋ฅผ ๋ ๋๊ฒ ๋ฉ๋๋ค.
์ด ์ฌ๊ฑด์ FBI๊ฐ ๊ณผ๊ฑฐ J. ์๋๊ฑฐ ํ๋ฒ(J. Edgar Hoover) ๊ตญ์ฅ ์์ ์๋ฏผ๋ค์ ํ๋ฒ์ ๊ถ๋ฆฌ๋ฅผ ์นจํดํ๋ ๊ณผ์ค๋ฅผ ๋ํ์ดํ ์ ์๋ค๋ ์ฐ๋ ค๋ฅผ ๋ถ๋ฌ์ผ์ผ์ผฐ์ต๋๋ค. ์ฝดํฐ์ฝ ์์นด๋ฐ๋ฏธ์์๋ถํฐ ์๋ฌด์ ์๋ฏผ๊ถ์ ๊ท ํ ์๊ฒ ๋ค๋ฃจ๋๋ก ๊ต์ก๋ฐ์๋ FBI ์์๋ค์๊ฒ ์ด๋ฌํ ์ง์๋ ์กฐ์ง์ ๊ทผ๋ณธ์ ์ธ ๊ฐ์น๋ฅผ ํผ์ํ๋ ๊ฒ์ด์์ต๋๋ค.
2. ๊ณผ๊ฑฐ ์ ๋ฌด๊ฐ โ์กฑ์โ๊ฐ ๋๋ค (Past Work as a Liability)
ํํ ์ ๊ณผ๊ฑฐ ๋ํต๋ น์ ์์ฌํ๋ ์ง์๋ค์ ์กฐ์ง์์ ์์ฒญํ๊ฒ ๋ค๊ณ ๊ณต์ธํ์ต๋๋ค. ์ด๋ฌํ โ์ ์น์ ์์ฒญโ์ ํ๋ ์ฐ๊ณ ๋ฆฌ์ธ ์ ์ฌ๋ก์์๋ ๋ช ํํ ๋๋ฌ๋ฉ๋๋ค.
2020๋ 9์, FBI๋ ์ค๊ตญ ์ ๋ถ๊ฐ 2020๋ ๋์ ์์ ์กฐ ๋ฐ์ด๋ ์ ์ํด ๊ฐ์ง ์ ๋ถ์ฆ์ ๋ง๋ค์๋ค๋ ์ ๋ณด ๋ณด๊ณ ์๋ฅผ ๋ฐํํ์ต๋๋ค. ์ด ์ ๋ณด์ ์ ๋ขฐ์ฑ์ ๋ํ ๋ ผ๋ ๋์ FBI๋ ๋ณด๊ณ ์๋ฅผ ์ฒ ํํ์ต๋๋ค. ์ดํ ์ฒ ๊ทธ๋์ฌ๋ฆฌ(Chuck Grassley) ์์์์์ด FBI์ ๋ด๋ถ ๋ฌธ์๋ฅผ ์๊ตฌํ๋ ๊ณผ์ ์์, ํ FBI ์ง์์ด ํ๋ ์ฐ๊ณ ๋ฆฌ์ธ ๋ฅผ ์ด ๋ณด๊ณ ์ ์ฒ ํ๋ฅผ ์ง์ํ ์ธ๋ฌผ๋ก ์๋ชป ์ง๋ชฉํ ์ด๋ฉ์ผ์ด ๋ฐ๊ฒฌ๋์์ต๋๋ค.
ํํ ๊ณผ ๋ด๊ธฐ๋ ธ๋ ์ด ์ฌ์์ด โ์์ ๋ค์ ์์ ๋ ๋ฌ๋คโ๊ณ ๋งํ๋ฉฐ ๊ทธ๋ ๋ฅผ ํ์ ํด์ง(administrative leave)์ ์ฒํ์ต๋๋ค. ์ดํ ๋ ์ฐจ๋ก์ ์ฌ๋ฌธ๊ณผ ๊ฑฐ์ง๋ง ํ์ง๊ธฐ ์กฐ์ฌ๋ฅผ ๊ฑฐ์ณค์ง๋ง, ๋ด๋ถ ์กฐ์ฌ ๊ฒฐ๊ณผ ๊ทธ๋ ์ ์ด๋ ํ ์๋ฒ ํ์๋ ๋ฐ๊ฒฌ๋์ง ์์์ต๋๋ค. ๊ทธ๋ผ์๋ ๋ถ๊ตฌํ๊ณ ๊ทธ๋ ๋ ์ด์ ์ง์ฑ ์ผ๋ก ๋ณต๊ทํ๋ ๊ฒ์ ๊ฑฐ๋ถ๋นํ๊ณ , ๋ค๋ฅธ ์ง๋ฐฉ ์ฌ๋ฌด์๋ก ์ ์ถ ๋ฐ ๊ฐ๋ฑ๋ ๊ฒ์ด๋ผ๋ ํต๋ณด๋ฅผ ๋ฐ์์ต๋๋ค. ๊ฒฐ๊ตญ ๊ทธ๋ ๋ 24๋ ๊ฐ ํ์ ํ๋ FBI๋ฅผ ๋ ๋ ์๋ฐ์ ์์์ต๋๋ค. ๊ธฐ์๋ค์ ์ด๋ฅผ โ๋ค์ฝํ ์ด์ผ๊ธฐ์ ํฌ์์โ์ด์ โ๋๊ตฌ์ ๋จธ๋ฆฌ๋ ๊ตด๋ฌ๊ฐ์ผ ํ๋คโ๋ ํํ ๊ณผ ๋ด๊ธฐ๋ ธ์ ๋ฉ์์ง์๋ค๊ณ ๋ถ์ํ์ต๋๋ค.
๋ธ๋ ์ด ํจ๋จผ(Blair Tolman)์ ์ฌ๋ก๋ ์ ์ฌํฉ๋๋ค. ๊ทธ๋ ๋ ํธ๋ผํ ํ์ ๋ถ 2๊ธฐ ์ด์ ์ FBI์ ์๋ฆฌํธ ๊ณต๊ณต ๋ถํจ ์์ฌํ(CR15)์ ํ์ฅ์ด์์ต๋๋ค. ์ด ํ์ ํธ๋ผํ ์ ๋ํต๋ น์ 2020๋ ๋์ ๊ฐ์ ์ฌ๋ถ๋ฅผ ์์ฌํ๋๋ฐ, ์ด๋ โ์ํฌํฑ ํ๋ก์คํธ(Arctic Frost)โ๋ผ๋ ์ฝ๋๋ค์์ผ๋ก ๋ถ๋ ธ์ต๋๋ค. ์ด ์์ฌ๋ ํธ๋ผํ์ ๊ทธ๋์ฌ๋ฆฌ ์์์์์ ์ฃผ์ ํ์ ์ด ๋์๊ณ , ํจ๋จผ์ ์์ ์ด ์ ๋ฐํ๋ ์ ๋ฅํ ์์๋ค์ด ํด๊ณ ๋๋ ๊ฒ์ ์ง์ผ๋ด์ผ ํ์ต๋๋ค. ๊ฒฐ๊ตญ ๊ทธ๋ ์์ ๋ โํ๋จ๋ ฅ ๋ถ์กฑ๊ณผ ๊ณต์ ์ฑ ๊ฒฐ์ฌโ๋ฅผ ์ด์ ๋ก ํด๊ณ ๋์๋๋ฐ, ์ด๋ ์ฌ์ค์ ์์ ์ด ๋ด๋นํ๋ ์์ฌ ๋๋ฌธ์ด์๋ค๊ณ ๊ทธ๋ ๋ ํ๊ณ ํ์ต๋๋ค.
ํ๋ค๋ฆฌ๋ FBI์ ๋ฏธ๋
์ด๋ฌํ ์ผ๋ จ์ ์ฌ๊ฑด๋ค์ FBI ๋ด๋ถ์ ๊น์ ๊ณตํฌ์ ํธ์ง์ฆ์ ๋ฌธํ๋ฅผ ์กฐ์ฑํ์ต๋๋ค. ํํ ์ด ๊ตญ์ฅ์ผ๋ก ์ทจ์ํ ์ดํ ์์ญ ๋ช ์ ์ง์์ด ํด๊ณ ๋์๊ณ , ์ด๋ฌํ ํด๊ณ ๋ ๊ณ์๋๊ณ ์์ต๋๋ค. ์์๋ค์ ๋ค์ ํ์ ๋ถ๊ฐ ์์ ์ ์ ๋ฌด๋ฅผ ์ ์น์ ์ผ๋ก ํด์ํ์ฌ ํด๊ณ ์ฌ์ ๋ก ์ผ์๊น ๋ด ์ ์น์ ์ผ๋ก ๋ฏผ๊ฐํ ์๋ฌด๋ฅผ ๋งก๋ ๊ฒ์ ์ฃผ์ ํ๊ฒ ๋์์ต๋๋ค. ์ด๋ ๊ฒฐ๊ตญ ๋ถํจ๋ ํ ๋ฌ ์ํ๊ณผ ๊ฐ์ ์ค์ํ ์ ๋ณด๊ฐ ์ํ๋ ์ ์์ผ๋ฉฐ, ๋ฏธ๊ตญ์ธ๋ค์ ์์ ์ด ์ํ๋ฐ์ ์ ์์์ ์๋ฏธํฉ๋๋ค.
FBI ์ง์๋ค์ ์กฐ์ง์ ๋ด๋ถ ๊ฒฌ์ ์์คํ ์ด๋ ์๊ธ์์์ ์ํต ์ฑ๋์ด ์ ๋๋ก ์๋ํ์ง ์๋๋ค๊ณ ๋๋ผ๋ฉด์, ํ์์๋ ๋น๋ฐ์ค๋ฌ์ด ์ด๋ค์ด ์ธ๋ก ๊ณผ ๋ฒ์์ ์์งํ๊ฒ ๋์์ต๋๋ค. ๋ธ๋ ์ด ํจ๋จผ์ ํ์ฌ FBI๋ฅผ ์๋๋ก ํ ์ง๋จ ์์ก์ ์ฃผ์ ์๊ณ ์ค ํ ๋ช ์ ๋๋ค. ์ด ์์ก์ โ์ ์น์ ์์ฒญโ์ผ๋ก ์ธํ ํด๊ณ ์ ์ด์๋ฅผ ์ ๊ธฐํ๊ณ , ์ ๋นํ ์ ์ฐจ๋ฅผ ์๊ตฌํ๋ฉฐ, ๋ค๋ฅธ ์์๋ค์ ๋ณดํธํ๊ธฐ ์ํ ๊ฒ์ ๋๋ค.
๊ธฐ์๋ค์ FBI๊ฐ ์๋ฒฝํ์ง ์์ผ๋ฉฐ ๊ฐํ์ด ํ์ํ๋ค๋ ์ ์๋ ๋ชจ๋๊ฐ ๋์ํ์ง๋ง, ํ์ฌ ์งํ๋๋ ๋ณํ ๋ฐฉ์์ ์กฐ์ง์ ๊ทผ๋ณธ์ ์ธ ๋ ๋ฆฝ์ฑ์ ํผ์ํ๊ณ ๊ตญ๊ฐ๋ฅผ ๋ ์์ ํ๊ฒ ๋ง๋ค๊ณ ์๋ค๊ณ ์ง์ ํฉ๋๋ค. ์ด๋ค์ โ์ฐ๋ฆฌ๊ฐ ๋ฌด์์ ๋์น๊ณ ์๋์ง, ์ด๋ค ์ํ์ ๋ณด์ง ๋ชปํ๋์ง ์๋ฌธ์ด ๋ ๋ค. ์ด ์ดํ์ ์ด๋ป๊ฒ ์กฐ์ง์ ์ฌ๊ฑดํ ์ ์์์ง ๋ชจ๋ฅด๊ฒ ๋คโ๋ฉฐ FBI์ ๋ฏธ๋์ ๋ํ ๊น์ ์ฐ๋ ค๋ฅผ ํํ์ต๋๋ค. ์นด์ฌ ํํ ์ ๋ฆฌ๋์ญ ํ์ FBI๋ ๊ทธ ์ด๋ ๋๋ณด๋ค ์ ์นํ๋๊ณ ๋ถ์ด๋์์ผ๋ฉฐ, ๊ทธ ์ฌํ๋ ์์ผ๋ก๋ ์ค๋ซ๋์ ์ง์๋ ๊ฒ์ผ๋ก ๋ณด์ ๋๋ค.