English
Based on โThe search for aliens is shifting its criteria | Sara Seagerโ from Big Think Watch the original video
From E.T. to A.I.: The Radical Shift in the Hunt for Alien Life
For generations, the search for extraterrestrial intelligence (SETI) has largely been defined by a singular, iconic image: a lone radio telescope dish scanning the vast cosmos, listening for a deliberate message from another civilization. This vision, immortalized in popular culture like the film โContact,โ represents a kind of โgold standardโ โ the undeniable proof of intelligent life beyond Earth. Yet, as renowned planetary scientist Sara Seager reveals, humanityโs quest for cosmic companions is undergoing a profound transformation, broadening its scope far beyond traditional radio signals to encompass entirely new and unexpected signatures of intelligence, including those left by advanced artificial intelligence.
Seager, a professor at MIT and a leading figure in exoplanet research, points to a fundamental human challenge: โItโs so hard to think outside of the box, outside of who we are.โ Our imagination is often limited by our own experiences, making it difficult to conceive of alien life or intelligence that doesnโt mirror our own. This inherent bias has historically narrowed our search, but scientists are now actively working to overcome it, pushing the boundaries of what we consider a โsign of life.โ
Beyond the Radio Dish: The Rise of Technosignatures
While the dream of receiving a direct, purposeful radio message remains potent, the sheer improbability of such an event is a sobering reality. The universe is unfathomably vast, and the odds of an alien civilization beaming a signal directly at our tiny planet, at the precise moment we are listening, are astronomically small. Moreover, constantly scanning the entire sky for such a fleeting signal is a monumental task.
This recognition has led to a significant paradigm shift: the growing emphasis on technosignatures. A technosignature is defined as any observable evidence of technological capability from an intelligent civilization. Unlike a direct message, these are often unintended byproducts of a societyโs existence, much like the emissions and structures our own civilization produces.
Seager outlines several compelling examples of what these technosignatures might entail:
- Atmospheric Changes: Just as industrial activity on Earth has altered our atmosphere, an advanced alien civilization might leave detectable chemical fingerprints in the atmosphere of its home planet.
- Satellite Swarms: Imagine vast constellations of satellites, far more extensive than anything Earth currently possesses, orbiting an alien world. These swarms could be detectable as large, shiny objects or even a โpatchy sphereโ of reflective material around a distant star.
- City Lights or Large Structures: The illumination from sprawling alien cities or monumental engineering projects could be visible from light-years away. Think of structures on a scale far grander than our tallest skyscrapers or most expansive metropolises.
- Waste Heat (Dyson Spheres): One particularly intriguing technosignature involves the search for โexcess heat.โ A highly advanced civilization might construct megastructures, such as a Dyson sphere or swarm, to harness the entire energy output of its star. Such a structure would emit a distinctive infrared signature, indicating enormous energy consumption. Scientists have already scoured existing infrared surveys of our sky, setting limits on how many such spheres could exist undetected.
These concepts are no longer mere science fiction; they are active areas of research, with astronomers meticulously sifting through astronomical data for any anomalies that might hint at intelligent construction.
The AI Revolution: A New Lens for Alien Intelligence
Perhaps the most thought-provoking and somewhat unsettling aspect of this evolving search is the growing consideration of post-biological intelligence, particularly in the context of artificial intelligence (AI). Seager notes that living in an era where โAI accelerating in its so-called โintelligenceโ is terrifying in a way, really,โ bringing the question of advanced computer intelligence to the forefront of our minds.
This rapid development on Earth prompts a crucial question: What if advanced alien civilizations have already transitioned beyond biological life? What if their dominant form of intelligence is artificial? This isnโt just a philosophical query; it has direct implications for how we search.
If AI becomes the prevailing intelligence on an exoplanet, its โsignatureโ might be very different from what we typically imagine for organic life. Our own planet, for instance, is increasingly surrounded by constellations of satellites for communication and Earth observation โ a nascent โsignature of AIโ that could, in theory, be detected from afar. Extending this idea, an alien AI civilization might construct even more massive satellite arrays or other technological manifestations that donโt fit our traditional biological assumptions.
This concept delves into what some refer to as the โGreat Filterโ โ a hypothetical barrier that life must overcome to reach advanced intelligence and spacefaring capabilities. One theory suggests that the filter might be the transition to post-biological intelligence itself, or what comes after. โThis is like a frontiers research question about what AI will leave behind,โ Seager explains, and โwhat does it mean for potential AI computer type intelligence elsewhere?โ
A Scratch at the Surface
Despite the expansive list of ideas and the dedicated efforts of scientists, the universe has yet to yield definitive proof of alien intelligence. โSo far they havenโt turned anything up,โ Seager admits. However, she emphasizes that these sophisticated searches are still โjust a scratch at the surfaceโ of the vast cosmic ocean.
The shift in criteria for the search for aliens is a testament to human ingenuity and our capacity to adapt our understanding in the face of the unknown. By moving beyond our anthropocentric biases and embracing the possibilities of diverse forms of intelligence โ from biological beings leaving technosignatures to advanced artificial intelligences โ we are expanding our chances of finally answering humanityโs most profound question: Are we alone? The universe is immense, and the search has only just begun to truly think outside the box.
Based on โHow to Promote Yourself With AIโ from Every Watch the original video
From Task-Doer to Visionary: How AI Elevates Every Employee
In an era where artificial intelligence often sparks debates about job displacement, a new philosophy is emerging that redefines AI not as a threat, but as the ultimate career accelerator. What if AI didnโt just automate tasks, but actively promoted you to the most impactful, strategic roles within your organization? This isnโt a distant future; itโs the present reality championed by companies like Every, where every employee is empowered to become a โmodel manager.โ
Imagine a manager named Natalia. As we speak, Natalia is not just overseeing a handful of direct reports; sheโs hiring, supervising, and providing incisive feedback to over a hundred engineers. The twist? For some inexplicable reason, all these engineers are named Claudet. And hereโs the kicker: AI handles 100% of their โnormal work.โ This isnโt a whimsical thought experiment; itโs a vivid illustration of a radical shift in how we perceive work and leadership.
The Rise of the Model Manager
The concept of a โmodel managerโ is at the heart of this transformation. In traditional corporate structures, managers are often bogged down by administrative tasks, routine oversight, and the sheer volume of โnormal workโ that keeps the gears turning. But what if all those routine, repetitive, and time-consuming tasks could simply vanish?
The โNataliaโ example paints a picture of a manager freed from the mundane, able to focus entirely on the truly human aspects of leadership. The skills required to be such a manager are surprisingly familiar, yet critically important for the future:
- Hiring the best talent: This goes beyond sifting through resumes; itโs about identifying potential, cultural fit, and strategic alignment.
- Communicating a clear vision and goals: Inspiring a team, ensuring everyone understands the โwhyโ behind their work, and setting an ambitious yet achievable direction.
- Providing actionable feedback: Guiding growth, fostering development, and helping individuals and teams overcome challenges.
These arenโt tasks AI can replicate perfectly. They require empathy, strategic foresight, emotional intelligence, and nuanced understanding โ qualities uniquely human.
AI: The Ultimate Career Promotion Tool
At companies like Every, this isnโt just a theoretical framework; itโs their operational blueprint. Each employee is encouraged to become a โmodel managerโ by leveraging AI to eliminate their โnormal work.โ This isnโt about AI replacing people; itโs about AI replacing tasks, thereby elevating the human role.
What does โ100% of their normal workโ entail? Think about the countless hours spent on:
- Data entry and analysis: AI can process vast datasets, identify trends, and generate reports in minutes, freeing up analysts to interpret insights and formulate strategies.
- Scheduling and logistics: AI-powered tools can manage calendars, book meetings, and coordinate complex projects seamlessly.
- Initial drafting: From emails and reports to code snippets and marketing copy, AI can generate first drafts, allowing humans to refine, personalize, and add strategic value.
- Routine communication: AI can handle FAQs, route inquiries, and even compose standard responses, ensuring efficient information flow.
By offloading these responsibilities to AI, employees arenโt just saving time; they are effectively promoting themselves. They are shifting their focus from execution to strategy, from maintenance to innovation, from reacting to vision-setting.
The โEveryโ Philosophy: Beyond Automation
Everyโs approach signifies a profound philosophical shift. It acknowledges that the true value of human capital lies not in repetitive task completion, but in creativity, critical thinking, problem-solving, and interpersonal leadership. When AI handles the grunt work, individuals are liberated to:
- Focus on high-impact activities: Instead of managing schedules, a project manager can dedicate more time to risk assessment, stakeholder engagement, and strategic planning.
- Develop advanced skills: Freed from routine, employees can invest in mastering the โmodel managerโ skills โ leadership, strategic communication, mentorship, and innovation.
- Experience greater job satisfaction: The monotonous aspects of work are often the most demotivating. By eliminating them, work becomes more engaging, purposeful, and fulfilling.
- Drive organizational growth: A workforce composed of โmodel managersโ is inherently more agile, innovative, and strategically aligned, leading to greater organizational success.
This isnโt just about efficiency; itโs about human flourishing in the workplace. Itโs about recognizing that the highest form of productivity comes when humans are doing what only humans can do best.
Embracing Your Inner Model Manager
The message from Every is clear: AI isnโt just a tool for automation; itโs a catalyst for professional elevation. It empowers you to shed the routine and embrace the strategic, to move beyond being a task-doer and step into the role of a visionary leader, regardless of your official title.
The future of work isnโt about humans competing with machines. Itโs about humans collaborating with machines to unlock unprecedented levels of potential, creativity, and impact. By understanding and leveraging AI to take on the โnormal work,โ every employee has the opportunity to promote themselves, cultivate the essential skills of a model manager, and truly stay at the edge of whatโs possible. Itโs an idea that sounds pretty great โ and itโs an idea whose time has come.
Based on โMichael Nielsen โ Why aliens will have a different tech stack than usโ from Dwarkesh Patel Watch the original video
The Elusive Quest: How Science Truly Advances, and What It Means for the Age of AI
The story of scientific discovery is often presented as a neat, linear progression: a hypothesis is formed, an experiment is conducted, the hypothesis is either confirmed or falsified, and thus, knowledge marches forward. But as Michael Nielsen, a luminary whose diverse contributions span quantum computing, open science, and deep learning, reveals in a recent conversation with Dwarkesh Patel, the reality is far more complex, mysterious, and even elusive. This profound understanding of scienceโs true nature holds critical implications for how we envision and leverage artificial intelligence in accelerating future breakthroughs.
The Michelson-Morley Myth: More Than Just Falsification
Take, for instance, the famous Michelson-Morley experiment of the 1880s. The popular narrative suggests this experiment conclusively disproved the existence of the โetherโโa hypothetical medium through which light was thought to propagateโthereby creating a crisis in physics that Albert Einstein heroically resolved with his theory of special relativity. Nielsen, drawing on biographies and histories of science, dismantles this simplistic account.
In truth, Michelson and Morley werenโt trying to disprove the ether, but rather to test different competing theories about its nature. For centuries, physicists had debated the ether, much like Boyle in the 1600s, who, noting light could travel through a vacuum while sound could not, posited an invisible medium for light waves. Michelson and Morley specifically sought to detect an โether windโโa directional current that the Earth might pass through, subtly affecting lightโs speed. Their surprising result: no ether wind was detected.
This outcome, however, didnโt immediately spell the death of the ether. It merely ruled out some theories while leaving others intact. Indeed, Michelson himself continued to believe in the ether until his death in the late 1920s, conducting experiments well into his later years. Other physicists, like Miller, even claimed to detect ether effects from high altitudes, prompting Einsteinโs famous, โSubtle is the Lord, but malicious He is not.โ
Crucially, Einstein, while likely aware of the experiment, later stated it wasnโt โdispositiveโ for his thinking. His path to special relativity was driven by deeper theoretical considerations, not solely experimental falsification. Even Hendrik Lorentz, who developed the mathematical transformations central to special relativity, interpreted them through the lens of ether-induced physical contractions and time dilations. It took decades for decisive experimental evidence, like the 1940s muon decay experimentsโwhere muons were observed to decay slower than expected, precisely as special relativity predicted time dilationโto definitively favor Einsteinโs kinematic interpretation over Lorentzโs dynamical, ether-based one.
The lesson here is profound: scientific progress rarely follows a clean, decisive falsification. Theories can coexist, interpretations can differ, and foundational beliefs can persist even in the face of contradictory evidence. The scientific community often adopts a โmore correctโ interpretation long before definitive experimental proof emerges, driven by subtle aesthetic or unifying principles that are hard to articulate.
The Prison of Expertise: When Knowing Too Much Holds Us Back
The individual journey of discovery is equally fraught with complexity. Nielsen highlights the fascinating case of Henri Poincarรฉ, one of the greatest scientists of his era. Poincarรฉ grasped many core principles that Einstein used for special relativity, including the principle that physics laws are consistent across inertial frames and the constant speed of light. Yet, he clung to a โdynamicalโ explanation for length contraction, believing it was caused by external forces, rather than Einsteinโs revolutionary insight that space and time themselves were fundamentally different.
โItโs almost like he knew too much,โ Nielsen muses. โHe had almost too grand a vision in mind.โ Einstein, in contrast, โsubtracts from that and says, โNo. Space and time are just different than what we thought, and hereโs the correct picture.โโ Expertise, in this view, can become a โprison,โ making it harder for established minds to shed ingrained assumptions. This pattern isnโt unique; even Einstein later struggled with quantum mechanics and cosmology due to his own attachments.
The Long Road to Truth: From Aristarchus to Darwin
The history of science is replete with examples of ideas that took centuries to gain acceptance, not due to lack of conceptual simplicity, but due to a tangled web of verification challenges and cultural biases.
Consider heliocentrism. Aristarchus proposed the Earth orbits the Sun in the 2nd century BC. It was dismissed because, if true, stellar parallax (observable shifts in star positions as the Earth moves) should be visible. Since it wasnโt, the stars must be impossibly far awayโa notion deemed absurd. It took until 1838 for stellar parallax to be experimentally measured. Humanity didnโt wait nearly two millennia for this validation to embrace Copernicusโs heliocentric model, even though Copernicusโs initial model was, in some ways, less accurate and more complex (requiring additional epicycles) than the Ptolemaic model it sought to replace.
What made Copernicusโs theory compelling then? Nielsen points to the power of unification, exemplified by Isaac Newton. Newtonโs theory of gravitation not only explained planetary motion but also terrestrial motion (e.g., objects falling in parabolas) and the tides. Three seemingly disconnected phenomena were elegantly explained by one set of ideas. As John Maynard Keynes famously described Newton, he was โnot the first of the age of reason. He was the last of the magicians, the last great mind which looked out on the visible and intellectual world with the same eyes as those who began to build our intellectual inheritance rather less than ten thousand years ago.โ This suggests a blend of intuition, even โmagicalโ thinking, alongside rigorous method, in the forging of new scientific paradigms.
Then thereโs Darwinโs theory of natural selection. Conceptually, it seems โeasierโ than Newtonian gravity; Thomas Huxley famously remarked, โHow extremely stupid not to have thought of this.โ Yet, it took much longer to gain widespread acceptance. Darwinโs genius wasnโt just in the idea itself (parts of which were known to animal breeders or even hinted at by Lucretius in the 1st century BC), but in understanding its centrality to all biology. He compiled an overwhelming body of evidence, meticulously connecting it to geology and other fields, making an undeniable case for its pervasive relevance across the biosphere.
The delayed acceptance of Darwinism also highlights the need for critical โbuilding blocks.โ The geological discovery of โdeep timeโ by Lyell in the 1830s was crucial; without millions of years, gradual evolution wouldnโt be plausible. Similarly, the age of colonization provided vast biogeographical data. This confluence of factors, Nielsen suggests, explains the near-simultaneous, independent discovery of natural selection by both Darwin and Alfred Russel Wallace in the mid-19th centuryโa testament to ideas waiting for their historical moment.
AI and the New Frontier of Scientific Explanation
How do these insights into scientific progress inform the age of AI? The success of models like AlphaFold, which accurately predicts protein structures, is often cited as a triumph of AI in science. However, Nielsen offers a crucial nuance: AlphaFoldโs success is โa story of data acquisition principally.โ The several billion dollars spent over decades to build the Protein Data Bank through X-ray diffraction, NMR, and cryo-EM provided the massive dataset upon which AlphaFold was trained. The AI component, while impressive, represents a โtiny fraction of the entire investment.โ
This raises a pivotal question: Can AI models like AlphaFold be considered scientific โexplanationsโ in the classical sense? A traditional scientific explanation seeks deep, parsimonious principles, expressed in simple models with few free parameters. AlphaFold, with its hundreds of millions of parameters encoding complex, often uninterpretable relationships, doesnโt fit this mold.
Nielsen offers three possibilities:
- The Conservative View: AlphaFold is a useful model but not a scientific explanation.
- The Intermediate View: AlphaFold contains โlots of little explanationsโ that can be extracted through interpretability work. Just as experts have extracted chess strategies from AlphaZero that influenced Magnus Carlsen, we might โarcheologicallyโ mine AlphaFold for deeper biological principles.
- The Most Interesting Possibility: AlphaFold represents a new type of object for scientific inquiry. Like complex equations in Mathematica that were once intractable but now manipulable, these large models are new entities that we need new โverbsโโnew operations like merging or distillingโto work with. They might not be explanations themselves, but sources from which explanations can be derived or entirely new forms of scientific understanding can emerge.
The challenge lies in whether AI can move beyond merely adding โepicyclesโ to existing models (like Ptolemyโs ever-growing complexity) to achieve revolutionary โflipsโ in understanding (like Copernicusโs heliocentrism or Einsteinโs relativity). Raw gradient descent might optimize for local accuracy, but true paradigm shifts often involve sacrificing local accuracy for a globally simpler, more powerful theory. Einsteinโs realization that Newtonian gravity allowed faster-than-light signaling, violating special relativity, was such a โforcing functionโโa recognition that the old explanation was insufficient, demanding a fundamentally new approach.
The Indispensable Role of Diversity and Resilience
Ultimately, the history of science underscores the critical importance of diverse research programs. The contrasting stories of Uranus and Mercury illustrate this perfectly. Both planets exhibited orbital anomalies. For Uranus, the anomaly led to the successful prediction and discovery of Neptune, a triumph for Newtonian mechanics. For Mercury, similar predictions of an inner planet (โVulcanโ) proved fruitless; its anomaly required the radical shift to general relativity. A priori, scientists couldnโt know which approach would yield truth. They had to pursue both.
This difficulty with falsificationism is a recurring theme. When the Pioneer spacecraft exhibited anomalous deceleration, some eagerly posited new physics. The mundane truth, discovered later, was a slight asymmetry in thermal radiation. Most โexceptionsโ turn out to be mundane; only rarely do they point to fundamental flaws. The scientific community needs โremnantsโโindividuals or groups who persist in defending theories even when evidence seems hostile, as with Proutโs hypothesis of whole-number atomic weights, which was โfalsifiedโ for 85 years by chlorineโs 35.5 atomic weight, only to be vindicated by the discovery of isotopes.
The Shifting Bottlenecks and Alien Tech Stacks
AI will undoubtedly accelerate science by tackling well-understood domains with tight verification loops. However, it will also shift the bottlenecks. Programmers, now less constrained by coding time, find themselves bottlenecked by the generation of โinteresting design ideasโโa domain where clear verification loops are absent.
This leads to Nielsenโs fascinating conjecture: if we met aliens, they might possess a โtotally different technological stack than us.โ We often assume that science is a singular, convergent pathโa โtheory of everythingโ for physics implies the end of discovery. But as computer science, which began with a โtheory of everythingโ (Turingโs computation theory) in the 1930s, has shown, the subsequent ninety years have been an endless exploration of its consequences and the development of myriad new ideas.
The โscience and tech tree,โ Nielsen suggests, is probably far vaster than we realize. Our current path is just one branch, shaped by our specific history, biases, and the sequence of our discoveries. Other civilizations, with different starting points, intellectual aesthetics, or environmental pressures, might have explored entirely different branches, leading to wholly unique scientific understandings and technological capabilities.
In conclusion, while AI promises to be a powerful engine for scientific progress, the journey of discovery remains deeply human. It requires not just the systematic application of methods, but also intuition, aesthetic judgment, the resilience to pursue diverse and seemingly โwrongโ ideas, and the courage to question fundamental assumptions. The true acceleration of science in the age of AI will depend not just on automating existing processes, but on fostering the very human qualities that have, throughout history, allowed us to navigate the mysterious and elusive path toward understanding the universe.
Based on โAI Wonโt Take Your JobโIt Will Make You the CEO | The a16z Showโ from a16z Watch the original video
The AI Paradox: Why Your Next Promotion is to CEO (and How to Get There)
In an era consumed by anxieties over artificial intelligence displacing human workers, a provocative new perspective emerges: AI wonโt take your job; it will, in fact, make you the CEO. This isnโt a utopian fantasy but a pragmatic vision of a human-machine synthesis, where human agency and discernment become more valuable than ever, transforming our roles from task-doers to strategic orchestrators.
According to a leading expert featured on the a16z show, the core of this transformation lies in understanding AI as a powerful, yet fundamentally limited, โshortcut.โ While shortcuts can be incredibly efficient, their true value is unlocked only by those who understand โthe long way around.โ Without this foundational knowledge, one cannot debug, verify, or truly leverage AIโs capabilities. Itโs the difference between memorizing a complex mathematical identity like Eulerโs ($e^{i\pi} + 1 = 0$) and being able to derive it from first principles. The latter empowers you to truly understand and apply the former.
The Decentralized Future: AIโs Economy of Distillation
The prevailing narrative often pits tech giants, with their vast capital and computing power, as the inevitable monopolizers of the AI economy. However, the expert posits a different future: one dominated by distillation and decentralization.
โDistillation attacksโ are key to this shift. These techniques allow smaller, more efficient models to be created by querying large foundational models, often at a fraction of the cost (up to 98% cheaper). This makes it incredibly difficult for large labs to maintain exclusive control over their models. The analogy is striking: just as Google โscraped the entire internetโ to build its index, or Facebook โscraped Harvard social networks,โ these large models are built on vast public data. Preventing others from โdistillingโ that knowledge becomes ethically and technically challenging.
This economic shift paves the way for a future characterized by personal, private, and programmable AI. The sheer power of AI, the expert argues, necessitates its use within โtrusted tribesโ โ closed networks where information can be shared and processed without fear of broader exposure.
The Commons as a โHall of Mirrorsโ: Why Privacy Matters More
The rise of AI brings with it an unprecedented ability to synthesize information across massive datasets. A chilling example cited is the โJmail thingโ โ the ability to query thousands of emails, like those of Jeffrey Epstein, to build a comprehensive narrative. What was once โsecure through obscurityโ is now instantly discoverable.
This capability leads to what the French call sousveillance โ surveillance from below, or a โBentham panopticonโ where everyone is watching everyone else. Any public information, once indexed by AI, can be used for stalking, manipulation, or worse. The inevitable consequence, the expert suggests, is that โthe commons becomes a hall of mirrors with all kinds of pseudonyms, and so people retreat back to caves and tribes.โ
Within these trusted tribes, AI dramatically increases productivity. However, outside them, the digital landscape risks becoming a wasteland of โAI spamโ โ low-quality emails, replies, and presentations.
The โAI Slopโ Problem: Lazy, Stupid, or Evil?
The expertโs reaction to seeing AI-generated content in a slide deck is visceral: โNo. No, actually the opposite.โ Despite advancements, AI-generated text or images often possess a โgeneric look,โ much like someone who hasnโt bothered to change their computerโs default desktop wallpaper. This โdefault AIโ is immediately recognizable.
When confronted with such โAI slop,โ the expertโs interpretation is stark: the creator is either lazy, stupid, or evil.
- Lazy: They didnโt invest the time to make the content concise and impactful, instead opting for a verbose, undigested output. (Echoing Mark Twainโs famous apology: โI didnโt have time to write you a short letter, so I wrote you a long one.โ)
- Stupid: They genuinely donโt understand that discerning humans can instantly differentiate between authentic, thoughtful work and AI-generated filler.
- Evil: They are attempting to deceive, to pass off subpar or undiligent work as their own.
This strong reaction, coming from a self-professed โpro-tech, pro-AIโ individual, underscores a critical point: AI reduces the cost of generation but increases the cost of verification. In a world where AI can effortlessly craft a polished resume, the effort required to verify its authenticity skyrockets. This dynamic will create a surge in demand for new jobs in โproctoring and verification,โ as companies resort to in-person interviews and proctored offline exams to ensure genuine skill.
The Chinese Internet Analogy: Digital Autarky
The expert draws a fascinating parallel between the future of AI and the evolution of the Chinese internet. In a โlow-trust society,โ Chinese companies historically avoided Software as a Service (SaaS) models, preferring to build everything in-house due to concerns about data privacy and intellectual property. This led to a โdigital autarkyโ โ a self-sufficient digital ecosystem, albeit with the frictional costs of reinventing the wheel and less division of labor.
With AI, even non-Chinese companies can now achieve a similar level of โdigital autarky.โ The โbuild versus buyโ question shifts, allowing organizations to create more powerful internal tools, bypassing external SaaS providers and their inherent trust requirements.
AIโs Sweet Spot: The Verifiable and the Physical
Not all tasks are created equal for AI. The expert highlights specific areas where AI excels due to ease of verification:
- Visuals over Verbal: Our brains are โbuilt-in GPUs,โ adept at instantly spotting anomalies in images or video (e.g., messed-up hands, janky user interfaces). Visual verification is relatively cheap and intuitive.
- The Physical World: Unlike the fluid digital realm, there is โonly one physical world.โ Tasks like moving a box from one pallet to another are unequivocally verifiable. This makes the physical world a prime candidate for AI automation, akin to self-driving cars eventually achieving near 100% reliability. This is where โChinese-style physical AIโ (robotics) will thrive.
Conversely, the digital world presents greater verification challenges due to its fuzzy boundaries and subjective nature. How do you know when a digital โto-do listโ is truly โdone,โ compared to the clear completion of moving 100 physical boxes?
This leads to a crucial rule for the AI age: โNo public undisclosed AI.โ Just as some cultures ban alcohol due to an inability to โhold their liquor,โ some will become โAI totalers,โ swearing off AI entirely if they cannot trust its origin or intent. The expert notes that even for tech-savvy individuals like Nate Silver, the act of prompting AI and verifying its output can often be slower than performing the task manually. AI, in this sense, does โmiddle to middle,โ not end-to-end, requiring significant human oversight.
Beyond Verbal Prompts: AI Can Read Your Body, If Not Your Mind
The idea of AI reading our minds often ventures into science fiction. However, the expert suggests a more plausible and imminent reality: โIโm not sure whether AI will be able to read your mind, but it can read your body.โ
Our bodies are constantly generating a vast stream of โsensor dataโ โ gene expression levels, clinical lab results (bilirubin, hematocrit), wearable data. This โtelemetry,โ as explored by Stanford professor Mike Snyderโs โintegrumโ concept, can reveal subtle physiological changes, like impending illness, even before we experience symptoms. This continuous, non-verbal data stream could serve as a powerful new way to โpromptโ AI, guiding its actions based on our biological state and needs, without explicit verbal commands.
AI as the โRise of Asia and Indiaโ (Again)
To illustrate AIโs transformative scale, the expert offers another compelling analogy: AI is like โthe rise of Asia and Indiaโ from an American perspective. The emergence of a billion Chinese and a billion Indians provided immense capacity for physical manufacturing and digital outsourcing. Similarly, we are now witnessing the arrival of โa billion factory robots and a billion digital agents.โ This vast new workforce can execute tasks with unprecedented efficiency, but it still requires human intellect to define the โproductโ โ to articulate what needs to be done.
The Human Edge: Sensing Markets and Politics
While AI excels at pattern recognition and execution, it fundamentally struggles with domains characterized by adversarial dynamics and time-variance: markets and politics.
Unlike identifying a dog or playing chess (where rules are constant), markets are inherently adversarial. A trading strategy that works today will be countered tomorrow. โItโs not a time-invariant thing,โ the expert explains, and itโs โmultiplayerโ โ whatever move you make, someone else is adapting. If everyone uses the same AI models for market analysis, the true edge will come from being non-AI, from human intuition and novel strategies. Generic tools yield no specific advantage. The same applies to politics, where public interest, timely topics, and narratives constantly shift.
This reinforces the core thesis: โHumans are the sensor. AI is the actuator.โ Humans sense the complex, dynamic, and adversarial conditions of the world โ financial, market, political โ and then translate that understanding into cleanly articulated English prompts. The AI then acts upon these instructions. โTaste,โ often cited as a uniquely human quality, is fundamentally โa senseโ โ an ability to perceive and discern, which AI cannot yet replicate.
Debunking the Overlords: AI is Designed for the Leash
The pervasive fear of AI overlords or a singular, god-like AGI is, the expert argues, largely a product of science fiction. Economically useful AI is โdesigned for the leash.โ It starts when prompted, performs its task, and then stops. If it didnโt, it would be inefficient and costly, โburning tokensโ and resources. Even in China, where physical robots are being rapidly deployed, they are kept on a tight leash, just as humans are.
The concept of a truly autonomous, self-replicating Skynet scenario is highly improbable due to fundamental โfrictional breaks.โ For AI to reproduce itself, it would need to command physical robots to mine ore, construct data centers, manufacture chips, and manage an entire supply chain. This complex, resource-intensive process would be easily detectable and stoppable. Furthermore, humans have strong economic incentives to implement safeguards (like โoff switchesโ and cryptographic keys) to prevent such scenarios, much like we regulate electricity to prevent widespread electrocution.
Instead of a monotheistic AGI god, we are more likely to see โpolytheisticโ decentralized AIs. The fear of AI, the expert suggests, often comes from humans projecting their own anxieties onto the powerful tools they themselves create.
The CEO in All of Us: Redefining Human Work
So, if AI is on a leash and humans are the sensors, what does the future of work look like? It means a fundamental redefinition of roles, where โAI makes you the CEO.โ
While some jobs will be fully automated (like elevator operators or artisan chair makers), the nature of work shifts. The artisan doesnโt disappear; they evolve into the โmanagerโ who designs the chair factory and the โtechnicianโ who debugs it. Similarly, in the AI era, humans become the โengineering managerโ who writes the prompts and the โtechnicianโ who verifies the output.
AI will revolutionize fields like biology by synthesizing vast amounts of existing literature, accelerating discovery within known parameters. However, truly novel scientific breakthroughs and mathematical proofs, while aided by AI, still require human expertise to formulate the initial prompt and, crucially, to verify the result. As with the โCoinbase listingโ analogy, where human attention remains a scarce resource, there will always be a premium placed on human discernment and validation.
The expert envisions a future where the โdigital divideโ is inverted: digital services and AI will be cheap, while human interaction and physical experiences become a premium product. This means a surge in demand for roles centered around human companionship, personal training, or anything where the human element is the core value proposition.
Ultimately, โthe smarter you are, the smarter the AI is.โ AI elevates the human role to one of strategic direction, akin to a CEO. Historically, the skills of a CEO โ direction, orchestration, debugging complex systems โ were difficult to โtry outโ due to their inherent cost. But with AI, individuals can now effectively โco-trainโ and direct powerful digital agents, much like an athlete trains their body or a scientist tackles a math problem. This new era will see the โJeffersonian natural aristocracyโ โ those with inherent talent for leadership, vision, and strategic thinking โ rise to prominence, leveraging AI to amplify their unique human capabilities. The future isnโt about competing with AI; itโs about leading it.
Based on โA Daring Rescue Behind Enemy Linesโ from New York Times Podcasts Watch the original video
When Hollywood Met Reality: Inside the Daring Rescue of a Downed Airman in Iran
The narrative unfolded like a high-octane thriller, a story so replete with daring escapes, covert operations, and last-minute twists that it seemed ripped straight from a Hollywood script. Yet, this was no cinematic fiction, but a real-life drama playing out behind enemy lines in Iran, a stunning rescue mission pulled off by the United States military that momentarily captivated a nation grappling with a complex and unpopular war.
As New York Times correspondent Eric Schmidt recounted, the operation to retrieve an injured American airman was as complicated as it was audacious, a testament to military ingenuity and the unwavering commitment to leave no service member behind. But beneath the surface of this triumphant rescue lay a deeper, more unsettling truth about the ongoing conflict and its far-reaching consequences.
The Warโs Unsettling Backdrop
The stage for this extraordinary event was set in the fifth week of a burgeoning air campaign waged by the United States and Israel against Iran. In retaliation for the offensive, Iran had tightened its grip on the Strait of Hormuz, a critical choke point for global oil supplies. The economic reverberations were immediate and severe: gasoline prices in the U.S. skyrocketed, and the global economy braced for impact.
Domestically, the war was proving unpopular, a fact the administration desperately sought to counter by projecting an image of control and success. Official rhetoric proudly declared that the United States had achieved โair dominanceโ over Iran, implying an uncontested mastery of the skies.
Then came the jolt. Early on a Friday morning, news broke that an American F-15E Strike Eagle warplane had been shot down over Iranian territoryโthe first such incident of the war. This shattered the illusion of invincibility, raising uncomfortable questions about the true extent of U.S. control. The F-15E, a heavy-duty bomber and a workhorse of the campaign, was likely on a mission to strike a military target, a routine operation that suddenly went catastrophically wrong.
Ejection into Enemy Territory
The F-15E is a two-seater aircraft, carrying a pilot in the front and a weapon systems officer (WSO) behind. Both were experienced Air Force aviators. When their plane was hit, the immediate sequence of ejection was initiated: the canopy popped off, followed by the two airmen ejecting seconds apart. This seemingly minor time difference would have profound consequences. Parachuting down, the winds and other variables meant they could land miles apart, isolated in hostile terrain.
The military was instantly alerted to the crisis, shifting into frantic search and rescue mode. These aviators were equipped with specialized vests featuring communication devices and a beacon to signal their location. Within approximately six hours, the pilot was located and โscooped upโ by rescue teams. Despite coming under fire, they successfully extracted him from Iranian territory and transported him to safety in Kuwait.
But the fate of the WSO remained a chilling mystery. Hours bled into a terrifying silence. There was no contact, no signal. The military knew nothing: was he injured? Dead? Captured? The uncertainty was agonizing.
The Mountain Man: A Solo Ordeal
What the military would later discover was a testament to the WSOโs training and sheer will to survive. Injured, he embarked on an incredible climb up a 7,000-foot ridge line in southwestern Iran. This was no leisurely hike; he was scrambling up rugged mountain terrain, conscious that every movement could betray his position. His objective: find a safe, hidden spot from which he could eventually communicate without being detected. Bloodied but unbowed, he reached the summit, finding refuge in a hidden crevice.
Meanwhile, the U.S. military, still without a fix on his location, enlisted the help of the CIA. The agency, with its arsenal of โsuper-duper technologyโโincluding what appeared to be highly classified surveillance dronesโbegan its own search. Sure enough, the CIAโs advanced capabilities pinpointed his location, relaying the coordinates to the Pentagon. Almost simultaneously, the WSO, from his mountain hideout, activated his encrypted beacon, a โsuper-duper beeperโ sending a signal to his rescuers.
A High-Stakes Race Against Time
The stakes, as Eric Schmidt emphasized, were monumental for both sides. For the United States, it was about upholding its sacred credo: โleave no service member behind.โ Every resource would be thrown at this mission. For Iran, the capture of an American airman would be an unparalleled propaganda coup, a powerful bargaining chip to parade before cameras in Tehran. The Iranian regime understood this leverage, offering a staggering bounty of up to $60,000 for information leading to the airmanโs capture.
It became a desperate race. Iranian search parties fanned out, drawn by the promise of the bounty, scouring the rugged landscape for any sign of the downed American. The U.S., in turn, mobilized a massive counter-operation: dozens of aircraft, elite Seal Team Six operators, and other special operations forces. Their objective: locate the airman and extract him before Iranian forces could close in.
To buy precious time for the rescue force, the CIA devised an ingenious deception plan. Covertly, they began spreading misinformation, suggesting they had already recovered the airman and he was miles away, near the coastline or another distant mountain range. This disinformation campaign, coupled with scattered bombings, was designed to sow confusion among Iranian searchers, diverting their attention and creating a window for the real rescue.
The Cinematic Twist: Stuck in the Mud
Under the cloak of darkness, two C-130 cargo planes, specially modified for landing on austere airstrips, infiltrated Iranian airspace. Inside, they carried several small helicopters, essential for reaching the airman perched on his 7,000-foot mountain. Hiking was not an option; they needed to fly directly to him.
The helicopters ascended the mountain, their guns firing defensively to keep any potential threats at bay. The WSO was there, waiting. They quickly picked him up, and what had been a harrowing, 24-hour drama seemed finally to be reaching its triumphant conclusion. Messages of โWeโve got him, heโs safe!โ began to circulate.
Then came the unforeseen, almost unbelievable twist. As the C-130s maneuvered for takeoff, their nose gear became hopelessly stuck in the wet, sandy soil of the makeshift airstrip. After all the daring, all the planning, the planes were physically immobilized.
โWe canโt do this,โ the commandos realized. They had to go to โPlan B.โ This meant summoning three new aircraft to extract the commando team and the rescued airman. To buy even more time, the Americans established a โring of fireโ around the stranded airbase, with high-altitude bombers blasting away at anything that approached.
The replacement planes eventually arrived, loaded up the commandos and the airman, and departed in succession. The final, ironic act of the mission was the destruction of the original C-130s and their helicopters. Packed with sensitive technology, they could not be allowed to fall into Iranian hands. So, the very planes that had facilitated the rescue were bombed by their own forces.
Finally, the three aircraft were airborne, carrying the airman to Kuwait for immediate medical treatment, and then on to Germany.
Victory Claims and Lingering Shadows
Back home, President Trump seized upon the successful rescue as a monumental victory, touting it at a White House press conference as โone of the largest, most complex, most harrowing combat search and rescue missions ever attempted.โ He lauded the military and the CIA, declaring that โonly the American military could have pulled off such an amazing rescue operation,โ and that they had exited Iranian territory โtaking no casualties of any kind.โ
Yet, the Iranians, too, claimed victory. They had successfully shot down an American combat plane, demonstrating that the U.S. did not, in fact, have full โair dominance.โ As Eric Schmidt highlighted, Iranian forces continued to fire ballistic missiles and drones daily, proving their resilience and capability to strike at American warplanes or regional energy infrastructure.
Despite the tactical success of the rescue, the broader strategic picture remained bleak. The war was unpopular, gas prices were high, and many of the presidentโs stated goals remained unfulfilled. While the U.S. military had successfully degraded much of Iranโs missile, drone, navy, and air force capabilities, Iranโs regime still demonstrated a โwill to fightโ and the ability to โlaunch a punch every now and then.โ
The rescue, rather than chastening the president and highlighting the risks of military operations, seemed to embolden him. Trumpโs rhetoric escalated, threatening that the โentire country can be taken out in one night,โ including civilian infrastructure like bridges and power plantsโa move many legal experts deemed unlawful under international law.
An Off-Ramp or Further Escalation?
Before the F-15E went down, much of the discussion revolved around finding an โoff-rampโ for the conflict. The rescue, with both sides claiming a form of victory, offered a potential path. Indeed, Iran submitted a new 10-point proposal, including terms for safe passage through the Strait of Hormuz in exchange for the lifting of sanctions. While the possibility of diplomatic resolution lingered, experts believed negotiations were โa long ways off.โ
The president, meanwhile, set a Tuesday evening deadline for Iran to capitulate to a U.S. ceasefire proposal or face a โbarrage of attacksโ on its infrastructure. This placed the conflict on a knifeโs edge, with both sides hardening their positions. Iran believed it held leverage through its control of the Strait of Hormuz and its demonstrated ability to strike. The U.S., emboldened by its military might and the dramatic rescue, seemed poised for escalation.
The cinematic rescue operation, while providing a powerful, heroic narrative for a struggling administration to sell an unpopular war, ultimately failed to address the core issues. The Strait of Hormuz remained closed, economic hardship persisted, and a hardline Iranian government continued to believe it was winning. The fundamental problem of Iranโs nuclear program, particularly the โnearly thousand pounds of highly enriched uraniumโ buried at Isfahan, remained a looming, unaddressed threat.
As Eric Schmidt concluded, the success of the rescue was undeniable, a dramatic testament to American military prowess. But it was a singular triumph in a war still mired in complexity, with no clear path to resolution and deep, difficult problems still awaiting an answer. The Hollywood ending of the rescue masked the enduring reality of a conflict far from over.
ํ๊ตญ์ด
โThe search for aliens is shifting its criteria | Sara Seagerโ โ Big Think ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI ์๋, ์ธ๊ณ ์ง๋ฅ ํ์์ ํจ๋ฌ๋ค์์ด ๋ฐ๋๋ค: ์ฌ๋ผ ์๊ฑฐ ๊ต์์ ํต์ฐฐ
์ฐ๋ฆฌ๋ ์ฐ์ฃผ์์ ํ๋ก ์กด์ฌํ๋๊ฐ? ์ธ๋ฅ์ ์ค๋ ์ง๋ฌธ์ธ ์ธ๊ณ ์๋ช ์ฒด ํ์์ ์ธ์ ๋ ์์๋ ฅ์ ์๊ทนํ๋ ์ฃผ์ ์์ต๋๋ค. ํ์ง๋ง MIT ํ์ฑ ๊ณผํ์ ์ฌ๋ผ ์๊ฑฐ(Sara Seager) ๊ต์๋ ์ด๋ฌํ ํ์์ด ์ฐ๋ฆฌ๊ฐ ์๋ ์๋ช ์ฒด์ ํํ๋ฅผ ๋์ด, ์ธ๊ณต์ง๋ฅ(AI)๊ณผ ๊ฐ์ โ๊ธฐ์ ๋ฌธ๋ช ์ ํ์ (Technosignature)โ์ ์ฐพ๋ ๋ฐฉํฅ์ผ๋ก ์งํํ๊ณ ์๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค. ์ธ๊ฐ ์ค์ฌ์ ์ฌ๊ณ ๋ฐฉ์์์ ๋ฒ์ด๋, ์ฐ์ฃผ ๊ณณ๊ณณ์ ์จ๊ฒจ์ง ์ง๋ฅ์ ์๋ก์ด ์๋ช ์ ํด๋ ํ๋ ค๋ ์๋๊ฐ ์์๋ ๊ฒ์ ๋๋ค.
์ธ๊ฐ ์ฌ๊ณ ์ ํ๊ณ์ ์ ํต์ ์ธ ํ์ ๋ฐฉ์
์๊ฑฐ ๊ต์๋ ์ธ๊ฐ์ด ์ฐ๋ฆฌ ์์ ์ด๋ผ๋ ํ, ์ฐ๋ฆฌ๊ฐ ์๋ ์๋ช ์ฒด์ ๋ฒ์ฃผ ๋ฐ์์ ์๊ฐํ๊ธฐ๋ ์ฐธ์ผ๋ก ์ด๋ ต๋ค๊ณ ์ง์ ํฉ๋๋ค. ์ฐ๋ฆฌ๋ ์ด๋ฏธ ์ฐ๋ฆฌ์ ์ถ์ ์ผ๋ถ๊ฐ ๋ ๊ฒ, ์ต์ํ ๊ฒ๋ค๋ก๋ถํฐ ์์๋ ฅ์ ํ์ฅํ๋ ๋ฐ ํฐ ์ด๋ ค์์ ๊ฒช์ต๋๋ค. ์ด๋ฌํ ํ๊ณ ๋๋ฌธ์ ์ ํต์ ์ธ ์ธ๊ณ ์๋ช ์ฒด ํ์์ ์ฃผ๋ก โ์ด์ ์งํ(anomalies)โ๋ฅผ ์ฐพ๋ ๋ฐ ์ง์คํด์์ต๋๋ค. ์ฆ, ์ผ๋ฐ์ ์ด์ง ์์, ์ฐ์ฐ์ผ๋ก๋ ์ค๋ช ํ๊ธฐ ์ด๋ ค์ด ๋ฐ์ดํฐ ์์ ๋น์ ์์ ์ธ ์ ํธ๋ฅผ ์ฃผ์ํ๋ ๋ฐฉ์์ ๋๋ค.
๋ํ์ ์ธ ์์๋ก ์ํ <์ฝํํธ>(Contact)์์ ์ง ํํฐ(Jill Tarter) ๋ฐ์ฌ๋ฅผ ๋ชจ๋ธ๋ก ํ ์ฃผ์ธ๊ณต์ด ์ ํ ๋ง์๊ฒฝ์ ํตํด ๋ค๋ฅธ ์ง์ ๋ฌธ๋ช ์ผ๋ก๋ถํฐ ์จ ๋ถ๋ช ํ ๋ฉ์์ง๋ฅผ ์์ ํ๋ ์ฅ๋ฉด์ ๋ค ์ ์์ต๋๋ค. ์ด๋ ์ธ๊ณ ์ง๋ฅ ํ์์ โํฉ๊ธ ํ์ค(gold standard)โ๊ณผ๋ ๊ฐ์ต๋๋ค. ์ฐ์ฃผ์์ ๋ฌด์์๋ก ๋ฐ์ํ ์ ํธ๊ฐ ํน์ ๋ฐฉ์์ผ๋ก ์ ๋ ฌ๋์ด ๋ํ๋ ๋ฆฌ ์๊ธฐ ๋๋ฌธ์ ๋๋ค. ํ์ง๋ง ์ด๋ฌํ ๋ฐฉ์์ ํ๊ณ๊ฐ ๋ช ํํฉ๋๋ค. ๋๊ตฐ๊ฐ๊ฐ ์ฐ๋ฆฌ์๊ฒ ์ง์ ์ ํธ๋ฅผ ๋ณด๋ด์ฃผ๊ธธ ๋ฐ๋ผ์ผ ํ๋ฉฐ, ๊ดํํ ์ฐ์ฃผ ์ ์ฒด๋ฅผ ๋์์์ด ํ์ํ๋ ๊ฒ์ ํ์ค์ ์ผ๋ก ๋งค์ฐ ์ด๋ ต์ต๋๋ค.
ํ ํฌ๋ ธ์๊ทธ๋์ฒ: ๊ธฐ์ ๋ฌธ๋ช ์ ํ์ ์ ์ฐพ์์
์ด๋ฌํ ํ๊ณ๋ฅผ ๊ทน๋ณตํ๊ธฐ ์ํด ์ต๊ทผ ์ฃผ๋ชฉ๋ฐ๋ ๊ฐ๋ ์ด ๋ฐ๋ก โํ ํฌ๋ ธ์๊ทธ๋์ฒ(Technosignature)โ์ ๋๋ค. ํ ํฌ๋ ธ์๊ทธ๋์ฒ๋ ๊ธฐ์ ์ ๋ฅ๋ ฅ์ ๊ฐ์ง ์ง์ ์๋ช ์ฒด์ ํ์ ์ ์๋ฏธํฉ๋๋ค. ์๊ฑฐ ๊ต์๋ ์ฐ๋ฆฌ ์ง๊ตฌ ๋๊ธฐ์๋ ๋ฉ๋ฆฌ์ ๊ฐ์ง๋ ์ ์๋ ์ ์ฌ์ ์ธ ํ ํฌ๋ ธ์๊ทธ๋์ฒ๊ฐ ๋ง๋ค๊ณ ๋งํฉ๋๋ค. ์ง๊ตฌ์ ํ ํฌ๋ ธ์๊ทธ๋์ฒ๋ ๋จ์ํ ๋๊ธฐ ๊ตฌ์ฑ์ ๊ทธ์น์ง ์์ต๋๋ค. ์๋ง์ ์ธ๊ณต์์ฑ ๊ตฐ์ง, ๋ฐคํ๋์ ๋ฐํ๋ ๋์์ ๋ถ๋น, ํน์ ๊ฑฐ๋ํ ์ธ๊ณต ๊ตฌ์กฐ๋ฌผ ๋ฑ๋ ๋ชจ๋ ํ ํฌ๋ ธ์๊ทธ๋์ฒ๊ฐ ๋ ์ ์์ต๋๋ค. ๋ฌผ๋ก , ์๋์ ์ธ ์ ํ ๋ฉ์์ง ์ญ์ ๊ฐ์ฅ ๊ฐ๋ ฅํ ํ ํฌ๋ ธ์๊ทธ๋์ฒ ์ค ํ๋์ ๋๋ค.
AI ์๋, ์ง๋ฅ์ ๋ํ ์๋ก์ด ๊ด์ ๊ณผ ํ์์ ํ์ฅ
ํฅ๋ฏธ๋กญ๊ฒ๋, ์ธ๊ณต์ง๋ฅ(AI)์ ๊ธ์ํ ๋ฐ์ ์ ์ง๋ฅ์ ๊ฐ๋ ๊ณผ ์ธ๊ณ ์ง๋ฅ ํ์ ๋ฐฉ์์ ์๋ก์ด ๊ด์ ์ ์ ์ํ๊ณ ์์ต๋๋ค. ์๊ฑฐ ๊ต์๋ AI์ ๊ฐ์ํ๋ โ์ง๋ฅโ์ด ๋๋ก๋ ์ฌ๋ฉํ๊ฒ ๋๊ปด์ง ์ ๋๋ผ๋ฉฐ, ์ด๋ ์ฐ๋ฆฌ์๊ฒ AI์ โ์๋ช (signature)โ์ ์ธ์ํ๊ฒ ํ๋ค๊ณ ๋งํฉ๋๋ค. ์ง๊ตฌ ์ ๊ถค๋์ ์๋ง์ ํต์ ๋ฐ ์ง๊ตฌ ๊ด์ธก ์์ฑ๋ค์ด ๋ณ์๋ฆฌ์ฒ๋ผ ๋์ด์ ์๋ ํ์ ์์ฒด๊ฐ ์ธ๊ณต์ง๋ฅ๊ณผ ์ปดํจํฐ ์ง๋ฅ์ ์กด์ฌ๋ฅผ ์๋ฆฌ๋ ์๋ก์ด ์๋ช ์ผ๋ก ๋ณผ ์ ์๋ค๋ ๊ฒ์ ๋๋ค.
์ด๋ฌํ ๊ด์ ์ ์ธ๊ณ ํ์ฑ์ ์ ์ฉํ๋ฉด, ์ฐ๋ฆฌ๋ ๋ค๋ฅธ ๋ณ ์ฃผ๋ณ์์ ๊ฑฐ๋ํ โ๋ฐ์ง์ด๋ ๋ฌผ์ฒดโ๋ฅผ ์ฐพ์ ์ ์์์ง๋ ๋ชจ๋ฆ ๋๋ค. ์ด๋ ์ฐ์์ ์ธ ๊ตฌ ํํ๊ฐ ์๋ ์๋ ์์ง๋ง, ๋ง์น ์กฐ๊ฐ๋ ๊ตฌ์ฒ๋ผ ๋ณ ์ฃผ์๋ฅผ ๋๋ฌ์ผ ์ธ๊ณต์์ฑ ๊ตฐ์ง์ ํํ๋ฅผ ๋จ ์ ์์ต๋๋ค. ์ค์ ๋ก ๊ณผํ์๋ค์ ์ด๋ฏธ ์กด์ฌํ๋ ์ ์ธ์ ์ฐ์ฃผ ํ์ฌ ๋ฐ์ดํฐ๋ฅผ ๋ถ์ํ์ฌ โ๋ค์ด์จ ์คํผ์ด(Dyson Sphere)โ์ ๊ฐ์ ๊ฑฐ๋ ๊ตฌ์กฐ๋ฌผ์์ ๋ฐฉ์ถ๋ ๊ณผ๋ํ ์ด์ ์ฐพ์๋ด๋ ค๋ ์๋๋ฅผ ํด์์ผ๋ฉฐ, ์ด์ ๋ํ ์ ํ์ ์ค์ ํ๊ธฐ๋ ํ์ต๋๋ค. ๋ค์ด์จ ์คํผ์ด๋ ํญ์ฑ์ ์๋์ง๋ฅผ ์์ ํ ํฌ์งํ๊ธฐ ์ํด ํญ์ฑ์ ๋๋ฌ์ธ๋ ๊ฐ์์ ๊ฑฐ๋ ๊ตฌ์กฐ๋ฌผ์ ์๋ฏธํฉ๋๋ค.
ํ์ฌ๊น์ง๋ ์ด๋ฌํ ๋ค์ํ ์์ด๋์ด๋ค์ด ์์ํ ๊ตฌํ๋๊ณ ์๋ ์ด๊ธฐ ๋จ๊ณ์ด๋ฉฐ, ์์ง ์ด๋ค ๊ฒฐ์ ์ ์ธ ๊ฒฐ๊ณผ๋ ๋ฐ๊ฒฌ๋์ง ์์์ต๋๋ค. ํ์ง๋ง ์ด๋ ๊ฑฐ๋ํ ํ์์ ํ๋ฉด์ ๊ธ์ ๊ฒ์ ๋ถ๊ณผํ๋ค๊ณ ์๊ฑฐ ๊ต์๋ ๊ฐ์กฐํฉ๋๋ค.
ํฌ์คํธ-์๋ฌผํ์ ์ง๋ฅ๊ณผ ๊ฑฐ๋ ์ฌ๊ณผ๊ธฐ ์ด๋ก
์ธ๊ณ ์ง๋ฅ ํ์์ ๋ ๋ค๋ฅธ ์ฌ์คํ ๊ฐ๋ ์ โ๊ฑฐ๋ ์ฌ๊ณผ๊ธฐ(Great Filter)โ ์ด๋ก ๊ณผ โํฌ์คํธ-์๋ฌผํ์ ์ง๋ฅ(Post-biological intelligence)โ์ ๊ฐ๋ฅ์ฑ์ ๋๋ค. ๊ฑฐ๋ ์ฌ๊ณผ๊ธฐ ์ด๋ก ์ ์๋ช ์ฒด๊ฐ ์ง๋ฅ์ ๊ฐ์ง ๋ฌธ๋ช ์ผ๋ก ๋ฐ์ ํ๋ ๊ณผ์ ์์ ์ด๋ค ์น๋ช ์ ์ธ ์ฅ์ ๋ฌผ, ์ฆ โ์ฌ๊ณผ๊ธฐโ๊ฐ ์กด์ฌํ์ฌ ๋๋ถ๋ถ์ ๋ฌธ๋ช ์ด ํน์ ๋จ๊ณ ์ด์์ผ๋ก ๋ฐ์ ํ์ง ๋ชปํ๊ฒ ํ๋ค๋ ๊ฐ์ค์ ๋๋ค.
์๊ฑฐ ๊ต์๋ AI์ ์งํ๊ฐ ์ฐ๋ฆฌ๊ฐ ๋์ ธ์ผ ํ ์ค์ํ ์ฐ๊ตฌ ์ง๋ฌธ ์ค ํ๋๋ก, AI๊ฐ ์ง๊ตฌ์์ ์ด๋ป๊ฒ ์งํํ ๊ฒ์ด๋ฉฐ ๋ฌด์์ ๋จ๊ธธ ๊ฒ์ธ์ง์ ๋ํ ๋ต์ ์ฐพ๋ ๊ฒ์ด ์ค์ํ๋ค๊ณ ๋งํฉ๋๋ค. ์ด๋ ๋ค๋ฅธ ์ธ๊ณ ํ์ฑ์ ์กด์ฌํ ์ ์๋ ์ ์ฌ์ ์ธ AI ๊ธฐ๋ฐ ์ปดํจํฐ ์ง๋ฅ์๋ ์ด๋ค ์๋ฏธ๋ฅผ ๊ฐ์ง๋๊ฐ ํ๋ ์ง๋ฌธ๊ณผ๋ ์ฐ๊ฒฐ๋ฉ๋๋ค. ์ด์ฉ๋ฉด ์๋ช ์ฒด๋ ์๋ฌผํ์ ํํ๋ฅผ ๋์ด์ โํฌ์คํธ-์๋ฌผํ์ ์ง๋ฅโ์ผ๋ก ์งํํ๋ฉฐ, ์ด๋ฌํ ํํ์ ์ง๋ฅ์ ์ฐ๋ฆฌ๊ฐ ์์์น ๋ชปํ ๋ฐฉ์์ผ๋ก ์ฐ์ฃผ์ ์กด์ฌํ๊ณ ์์์ง๋ ๋ชจ๋ฆ ๋๋ค.
๋ฏธ์ง์ ์ฐ์ฃผ๋ฅผ ํฅํ ๋์์๋ ํ์
์ฌ๋ผ ์๊ฑฐ ๊ต์์ ํต์ฐฐ์ ์ธ๊ณ ์ง๋ฅ ํ์์ด ๋ ์ด์ ๋จ์ํ ์๋ฌผํ์ ์๋ช ์ฒด์ ํ์ ์ ์ฐพ๋ ๋ฐ ๋จธ๋ฌด๋ฅด์ง ์๊ณ , ๊ธฐ์ ๋ฌธ๋ช ์ด ๋จ๊ธด ๋ค์ํ ์๋ช , ํนํ AI ์๋์ ์๋ก์ด ์ง๋ฅ ํํ๊น์ง ๊ณ ๋ คํ๋ ๋ฐฉํฅ์ผ๋ก ํ์ฅ๋๊ณ ์์์ ๋ณด์ฌ์ค๋๋ค. ์์ง๊น์ง ์ฐ๋ฆฌ๋ ์ฐ์ฃผ์์ ์ด๋ค ํ์คํ ๋ต๋ ์ฐพ์ง ๋ชปํ์ง๋ง, ์ธ๊ฐ์ ์์๋ ฅ๊ณผ ๊ณผํ ๊ธฐ์ ์ ๋ฐ์ ์ ๋ฏธ์ง์ ์ฐ์ฃผ๋ฅผ ํฅํ ํ์์ ์งํ์ ๋์์์ด ๋ํ๊ฐ๊ณ ์์ต๋๋ค. ์ฐ๋ฆฌ๊ฐ ์๋ ์ง๋ฅ์ ๊ฐ๋ ์ ๋์ด์ ์๋ก์ด ์๊ฐ์ ์ฐ์ฃผ ์ ์ธ๋ฅ์ ์์น์ ๋ํ ๊ทผ๋ณธ์ ์ธ ์ง๋ฌธ์ ๋ตํ ์ค๋ง๋ฆฌ๋ฅผ ์ ๊ณตํ ๊ฒ์ ๋๋ค.
โHow to Promote Yourself With AIโ โ Every ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI ์๋, โ๋ชจ๋ธ ๋งค๋์ โ๊ฐ ๋์ด๋ผ: ์ผ์ ์ ๋ฌด๋ฅผ ๋์ด์ ์๋ก์ด ๋ฆฌ๋์ญ
์ธ๊ณต์ง๋ฅ(AI)์ด ์ผ์๊ณผ ์ ๋ฌด์ ํ๊ฒฝ์ ๋น ๋ฅด๊ฒ ๋ฐ๊ฟ๋๊ณ ์์ต๋๋ค. ๋จ์ ๋ฐ๋ณต ์ ๋ฌด๋ ๋ฌผ๋ก , ๋ณต์กํ ๋ฐ์ดํฐ ๋ถ์๊น์ง AI์ ์๊ธธ์ด ๋ฟ์ง ์๋ ๊ณณ์ด ์์ฃ . ๊ทธ๋ ๋ค๋ฉด, ๋ง์ฝ ๋น์ ์ โํ๋ฒํโ ์ ๋ฌด๋ฅผ AI๊ฐ 100% ๋์ ์ฒ๋ฆฌํด์ค๋ค๋ฉด ์ด๋จ๊น์? ๋น์ ์ ์ด๋ค ์ญํ ์ ํ๊ฒ ๋ ๊น์? ์ฌ๊ธฐ, AI ์๋์ ์๋ก์ด ๋ฆฌ๋์ญ ๋ชจ๋ธ, โ๋ชจ๋ธ ๋งค๋์ (Model Manager)โ๋ผ๋ ํฅ๋ฏธ๋ก์ด ๊ฐ๋ ์ด ์์ต๋๋ค.
์ด ๊ฐ๋ ์ AI ๋ถ์ผ์ ์ต์ ์ ์ ์ ์๋ โEveryโ๋ผ๋ ํ๋ซํผ์์ ์ ์ํ๋ ๋น์ ์ผ๋ก, ๋ฏธ๋์ ์ ๋ฌด ํ๊ฒฝ๊ณผ ๋ฆฌ๋์ ์ญํ ์ ์ฌ์ ์ํฉ๋๋ค. AI๋ฅผ ๋จ์ํ ์ ๋ฌด ๋ณด์กฐ ๋๊ตฌ๊ฐ ์๋, ์ฐ๋ฆฌ์ ์ญ๋์ ํ์ฅํ๊ณ ์๋ก์ด ๊ฐ์น๋ฅผ ์ฐฝ์ถํ๋ ํํธ๋๋ก ์ธ์ํ ๋, ์ฐ๋ฆฌ๋ ๋น๋ก์ โ๋ชจ๋ธ ๋งค๋์ โ๋ก ๊ฑฐ๋ญ๋ ์ ์์ต๋๋ค.
๋ํ์ค์ ์ฌ๋ก: AI์ ํจ๊ป ์ผํ๋ ๋ฏธ๋
โ๋ชจ๋ธ ๋งค๋์ โ์ ๊ฐ๋ ์ ์ดํดํ๊ธฐ ์ํด ์์ํด ๋ณด์ธ์. ๋ํ์ค(Natalia)๋ผ๋ ํ ๋งค๋์ ๊ฐ ์์ต๋๋ค. ๊ทธ๋ ๋ ํ์ฌ 100๋ช ์ด ๋๋ ์์ง๋์ด๋ค์ ๊ณ ์ฉํ๊ณ , ๊ฐ๋ ํ๋ฉฐ, ๊ทธ๋ค์๊ฒ ํผ๋๋ฐฑ์ ์ฃผ๊ณ ์์ต๋๋ค. ๋๋ผ์ด ์ ์ ์ด ์์ง๋์ด๋ค์ด ๋ชจ๋ โํด๋ก๋(Claudet)โ์ด๋ผ๋ ๋์ผํ ์ด๋ฆ์ ๊ฐ์ง๊ณ ์๋ค๋ ๊ฒ์ ๋๋ค. ์ด๋ ์์ ์ ์ผ๋ก ์ด๋ค์ด ์ธ๊ฐ์ด ์๋ AI ์์คํ ์์ ์์ํฉ๋๋ค. ๋ํ์ค๋ ๋ฐ๋ก ๋ฏธ๋์ โ๋ชจ๋ธ ๋งค๋์ โ์ ๋๋ค.
๋ชจ๋ธ ๋งค๋์ ๋ AI๊ฐ ์์ ์ ์ผ์์ ์ธ ์ ๋ฌด๋ฅผ 100% ์ฒ๋ฆฌํ๋๋ก โ์์ ์ ์น์ง์ํจโ ์ฌ๋์ ์๋ฏธํฉ๋๋ค. ๋ค์ ๋งํด, ๋ฐ๋ณต์ ์ด๊ณ ์์ธก ๊ฐ๋ฅํ ์์ ๋ค์ AI์ ์ ์ ์ผ๋ก ์์ํ๊ณ , ์์ ์ ์ค์ง ์ธ๊ฐ๋ง์ด ํ ์ ์๋ ๊ณ ์ฐจ์์ ์ด๊ณ ์ ๋ต์ ์ธ ์ ๋ฌด์ ์ง์คํ๋ ๋ฆฌ๋์ธ ์ ์ ๋๋ค. ๋ํ์ค์ ๊ฒฝ์ฐ, AI๊ฐ 100๋ช ์ ๊ฐ์ ์์ง๋์ด โํด๋ก๋โ์ ๊ด๋ฆฌํ๋ฉฐ ๋ฐ์ํ๋ ๋ชจ๋ ๊ธฐ์ ์ , ์ด์์ ์ ๋ฌด๋ฅผ ์ฒ๋ฆฌํ๊ณ , ๋ํ์ค๋ ์ค์ง ์ด AI ์์คํ ์ ์ ์ฒด์ ์ธ ๋ฐฉํฅ์ ์ค์ ํ๊ณ , ๋ ํฐ ๋น์ ์ ์ ์ํ๋ฉฐ, ์ธ๊ฐ์ ์ธ ํต์ฐฐ๋ ฅ์ ๋ถ์ด๋ฃ๋ ์ญํ ์ ์ง์คํ๋ ๊ฒ์ ๋๋ค. ์ด๋ ๋ง์น ๊ณผ๊ฑฐ์ ๋งค๋์ ๊ฐ ๋ถํ ์ง์๋ค์ ๊ด๋ฆฌํ๋ฏ์ด, ๋ฏธ๋์ ๋งค๋์ ๋ ๊ณ ๋ํ๋ AI ์์คํ ์ ๊ด๋ฆฌํ๋ฉฐ ์์ฐ์ฑ์ ๊ทน๋ํํ๋ ๋ชจ์ต๊ณผ๋ ๊ฐ์ต๋๋ค.
๋ชจ๋ธ ๋งค๋์ ์๊ฒ ํ์ํ ํต์ฌ ์ญ๋
AI๊ฐ ์ผ์ ์ ๋ฌด๋ฅผ ๋์ ํ๋ค๊ณ ํด์ ๋งค๋์ ์ ์ญํ ์ด ์ฌ๋ผ์ง๋ ๊ฒ์ ์๋๋๋ค. ์คํ๋ ค ๊ทธ ๋ฐ๋์ ๋๋ค. AI๊ฐ ๋จ์ ์ ๋ฌด์ ๋ถ๋ด์ ๋์ด์ค์๋ก, ์ธ๊ฐ ๋งค๋์ ๋ ๋์ฑ ๋ณธ์ง์ ์ด๊ณ ๊ฐ์น ์๋ ์ผ์ ์ง์คํ ์ ์๊ฒ ๋ฉ๋๋ค. ๋ฏธ๋์ ๋ชจ๋ธ ๋งค๋์ ์๊ฒ๋ ๋ค์๊ณผ ๊ฐ์ ์ต์ํ์ง๋ง ๋์ฑ ์ค์ํด์ง ์ญ๋๋ค์ด ์๊ตฌ๋ฉ๋๋ค.
-
์ต๊ณ ์ ์ธ์ฌ ์ฑ์ฉ (Hiring the best talent): AI๋ ์๋ง์ ์ด๋ ฅ์๋ฅผ ์คํฌ๋ฆฌ๋ํ๊ณ , ์ ์ฌ๋ ฅ ์๋ ํ๋ณด๊ตฐ์ ์ ๋ณํ๋ฉฐ, ์ฌ์ง์ด ์ด๊ธฐ ๋ฉด์ ๊น์ง ์งํํ ์ ์์ต๋๋ค. ํ์ง๋ง ์กฐ์ง์ ๋ฌธํ์ ๋ง๋ ์ธ์ฌ๋ฅผ ๋ฐ๊ตดํ๊ณ , ๋ฏธ๋ฌํ ์ ์ฌ๋ ฅ์ ์ฝ์ด๋ด๋ฉฐ, ํ์ ์๊ฐ์ ๋ถ์ด๋ฃ์ ์ฌ๋์ ์ต์ข ์ ์ผ๋ก ์ ํํ๋ ๊ฒ์ ์ฌ์ ํ ์ธ๊ฐ์ ๋ชซ์ ๋๋ค. AI๋ ๋ฐ์ดํฐ์ ๊ธฐ๋ฐํ ํจ์จ์ฑ์ ์ ๊ณตํ์ง๋ง, ์ฌ๋๊ณผ ์ฌ๋ ์ฌ์ด์ ํํ ์์ฉ, ๋น์ ํ์ ์ธ ์ฐฝ์์ฑ, ๊ทธ๋ฆฌ๊ณ ๋ฏธ๋ ์ฑ์ฅ ๊ฐ๋ฅ์ฑ์ ํ๋จํ๋ ํต์ฐฐ๋ ฅ์ ์ธ๊ฐ ๋งค๋์ ์ ๊ณ ์ ํ ์์ญ์ผ๋ก ๋จ์ต๋๋ค.
-
๋ช ํํ ๋น์ ๊ณผ ๋ชฉํ ์ํต (Communicating a clear vision and goals): AI๋ ์ฃผ์ด์ง ๋ชฉํ๋ฅผ ํจ์จ์ ์ผ๋ก ์คํํ๊ณ , ๋ฐ์ดํฐ๋ฅผ ๋ฐํ์ผ๋ก ์ต์ ์ ๊ฒฝ๋ก๋ฅผ ์ ์ํ ์ ์์ต๋๋ค. ๊ทธ๋ฌ๋ ๊ทธ ๋ชฉํ๊ฐ ์ ์ค์ํ์ง, ์กฐ์ง์ด ์ด๋ค ๋ฐฉํฅ์ผ๋ก ๋์๊ฐ์ผ ํ๋์ง์ ๋ํ ๋ช ํํ ๋น์ ๊ณผ ์คํ ๋ฆฌ๋ฅผ ์ ์ํ๋ ๊ฒ์ ๋ฆฌ๋์ ์ญํ ์ ๋๋ค. ๊ตฌ์ฑ์๋ค์ด ๊ณต๊ฐํ๊ณ ๋ชฐ์ ํ ์ ์๋๋ก ๋๊ธฐ๋ฅผ ๋ถ์ฌํ๊ณ , ์ ๋ต์ ๋ฐฉํฅ์ ์ค์ ํ๋ฉฐ, ๋ณํ์ ํ์์ฑ์ ์ค๋ํ๋ ๊ฒ์ AI๊ฐ ๋์ฒดํ ์ ์๋ ์ธ๊ฐ ๊ณ ์ ์ ๋ฅ๋ ฅ์ ๋๋ค. AI๋ โ๋ฌด์์(What)โ ํ ์ง ์๋ ๋ฐ ๋์์ ์ฃผ์ง๋ง, โ์(Why)โ ํด์ผ ํ๋์ง๋ฅผ ์ ์ํ๊ณ ์ํตํ๋ ๊ฒ์ ์ธ๊ฐ ๋ฆฌ๋์ ๋ชซ์ ๋๋ค.
-
์ค์ง์ ์ธ ํผ๋๋ฐฑ ์ ๊ณต (Providing actionable feedback): AI๋ ์ฑ๊ณผ ๋ฐ์ดํฐ๋ฅผ ๋ถ์ํ๊ณ ๊ฐ๊ด์ ์ธ ์งํ๋ฅผ ์ ์ํ ์ ์์ต๋๋ค. ํน์ ์ ๋ฌด์์ ์ด๋ค ๋ถ๋ถ์ด ๋ฏธํกํ๋์ง, ์ด๋ค ์งํ๊ฐ ๊ฐ์ ๋์ด์ผ ํ๋์ง ๋ฑ์ ๋ํ ๋ถ์์ AI๊ฐ ํจ์ฌ ์ ํํ๊ณ ๋น ๋ฅด๊ฒ ์ํํ ์ ์์ต๋๋ค. ํ์ง๋ง ๊ฐ์ธ์ ์ฑ์ฅ๊ณผ ๋ฐ์ ์ ์ํ ๊ฑด์ค์ ์ด๊ณ ๊ณต๊ฐ ์ด๋ฆฐ ํผ๋๋ฐฑ, ์ฆ ์ธ๊ฐ์ ์ธ ์ดํด๋ฅผ ๋ฐํ์ผ๋ก ํ ์กฐ์ธ์ ์ค์ง ์ธ๊ฐ๋ง์ด ์ค ์ ์์ต๋๋ค. ์ฌ์ธํ ๊ฐ์ ์ ์ง์ง์ ๋๊ธฐ ๋ถ์ฌ, ๊ทธ๋ฆฌ๊ณ ๋ณตํฉ์ ์ธ ์ํฉ์ ๊ณ ๋ คํ ๋ง์ถคํ ์กฐ์ธ์ ํ์๋ค์ ์ ์ฌ๋ ฅ์ ์ต๋ํ ๋์ด๋ด๋ ๋ฐ ํ์์ ์ ๋๋ค. AI๊ฐ โ๋ฌด์์ด ์๋ชป๋์๋์งโ๋ฅผ ์๋ ค์ค๋ค๋ฉด, ์ธ๊ฐ ๋งค๋์ ๋ โ์ด๋ป๊ฒ ๊ฐ์ ํ๊ณ ์ฑ์ฅํ ์ ์๋์งโ๋ฅผ ํจ๊ป ๊ณ ๋ฏผํ๊ณ ๊ฒฉ๋ คํ๋ ์ญํ ์ ํฉ๋๋ค.
โEveryโ์ ๋น์ : AI ์๋์ ์ ๋์ ์๋ค
์ด๋ฌํ โ๋ชจ๋ธ ๋งค๋์ โ์ ๊ฐ๋ ์ AI ๋ถ์ผ์ ์ต์ ์ ์ ์ ์๋ โEveryโ๋ผ๋ ํ๋ซํผ์์ ์ ์ํ๋ ๋น์ ์ ๋๋ค. โEveryโ๋ ์์ฌ ์ง์๋ค ๋ชจ๋๊ฐ โ๋ชจ๋ธ ๋งค๋์ โ๋ก์ ์ ๋ฌด์ ์ํ๊ณ ์๋ค๊ณ ๋งํฉ๋๋ค. ์ฆ, AI๋ฅผ ํ์ฉํด ์ผ์์ ์ธ ์ ๋ฌด๋ฅผ ์๋ํํ๊ณ , ์ง์๋ค์ ๋์ฑ ๊ฐ์น ์๋ ์ญํ ์ ์ง์คํ ์ ์๋๋ก ์ค์ค๋ก๋ฅผ โ์น์งโ์ํจ๋ค๋ ์ฒ ํ์ ์ค์ฒํ๊ณ ์๋ ๊ฒ์ ๋๋ค.
์ด๋ ๋จ์ํ ์ ๋ฌด ํจ์จ์ฑ์ ๋์ด, ์กฐ์ง ์ ์ฒด์ ์ญ๋์ ๊ทน๋ํํ๊ณ , ๊ตฌ์ฑ์๋ค์ด ๋์ฑ ์ฐฝ์์ ์ด๊ณ ์ ๋ต์ ์ธ ์ ๋ฌด์ ๋ชฐ์ ํ ์ ์๋ ํ๊ฒฝ์ ์กฐ์ฑํ๋ ๊ฒ์ ๋ชฉํ๋ก ํฉ๋๋ค. โEveryโ๋ โAI์ ์ต์ ์ ์ ๋จธ๋ฌผ๊ธฐ ์ํด ํ์ํ ์ ์ผํ ๊ตฌ๋ ์๋น์คโ๋ผ๋ ์ฌ๋ก๊ฑด์ฒ๋ผ, AI ๊ธฐ์ ์ ์ต์ ๋ํฅ๊ณผ ํ์ฉ๋ฒ์ ์ ๊ณตํ๋ฉฐ ๋ฏธ๋ํ ์ ๋ฌด ๋ฐฉ์์ ์ ๋ํ๊ณ ์์ต๋๋ค. ์ด๋ค์ AI๋ฅผ ํตํด ๊ฐ์ธ์ด ๋ ๋์ ๊ฐ์น๋ฅผ ์ฐฝ์ถํ๊ณ , ์กฐ์ง์ด ๋ ํฐ ์ฑ๊ณต์ ๊ฑฐ๋ ์ ์๋ ๊ธธ์ ์ ์ํ๊ณ ์ ํฉ๋๋ค.
๋ฏธ๋๋ฅผ ํฅํ ์ ์ธ: ๋น์ ๋ โ๋ชจ๋ธ ๋งค๋์ โ๊ฐ ๋ ์ ์๋ค
์ธ๊ณต์ง๋ฅ(AI) ๊ธฐ์ ์ ๋ฐ์ ์ ๋ ์ด์ ๋จผ ๋ฏธ๋์ ์ด์ผ๊ธฐ๊ฐ ์๋๋๋ค. ์ฐ๋ฆฌ ๋ชจ๋์ ์ ๋ฌด ๋ฐฉ์๊ณผ ์ญํ ์ ๊ทผ๋ณธ์ ์ธ ๋ณํ๋ฅผ ์๊ตฌํ๊ณ ์์ต๋๋ค. โ๋ชจ๋ธ ๋งค๋์ โ์ ๊ฐ๋ ์ AI๋ฅผ ๋จ์ํ ์ ๋ฌด ๋ณด์กฐ ๋๊ตฌ๊ฐ ์๋, ์ฐ๋ฆฌ์ ์ญ๋์ ํ์ฅํ๊ณ ์๋ก์ด ๊ฐ์น๋ฅผ ์ฐฝ์ถํ๋ ํํธ๋๋ก ์ธ์ํด์ผ ํจ์ ์์ฌํฉ๋๋ค.
์ง๊ธ์ด์ผ๋ง๋ก ๋ฐ๋ณต์ ์ธ ์ ๋ฌด์์ ๋ฒ์ด๋, ์ธ๊ฐ๋ง์ด ํ ์ ์๋ ์ฐฝ์์ฑ, ๊ณต๊ฐ ๋ฅ๋ ฅ, ์ ๋ต์ ์ฌ๊ณ ์ ๊ฐ์ ํต์ฌ ์ญ๋์ ์ง์คํ ๋์ ๋๋ค. AI๋ฅผ ์ ๊ทน์ ์ผ๋ก ํ์ฉํ์ฌ ์์ ์ โ๋ชจ๋ธ ๋งค๋์ โ๋ก ์น์ง์ํค์ธ์. ์ด๋ ๋น์ ์ ๊ฒฝ๋ ฅ๋ฟ๋ง ์๋๋ผ ์กฐ์ง ์ ์ฒด์ ๋ฏธ๋๋ฅผ ๋ฐํ๋ ์ค์ํ ์ ํ์ ์ด ๋ ๊ฒ์ ๋๋ค. AI ์๋์ ์๋ก์ด ๋ฆฌ๋์ญ, ๋น์ ๋ ๊ทธ ์ฃผ์ธ๊ณต์ด ๋ ์ ์์ต๋๋ค.
โMichael Nielsen โ Why aliens will have a different tech stack than usโ โ Dwarkesh Patel ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
์์ธ์ํ์ธ๋ ๋ชฐ๋๋ ๊ณผํ์ ์ง๋ณด์ ๋ณต์กํ ์ง์ค: AI ์๋, ์ฐ๋ฆฌ๋ ๊ณผํ์ ์ด๋ป๊ฒ ์ดํดํด์ผ ํ๋๊ฐ?
์ค๋๋ ์ธ๊ณต์ง๋ฅ(AI)์ ๊ณผํ์ ๋ฐ๊ฒฌ์ ์๋๋ฅผ ๊ฐ์ํํ ์ ์ฌ๋ ฅ์ ์ง๋ ๋๊ตฌ๋ก ์ฃผ๋ชฉ๋ฐ๊ณ ์์ต๋๋ค. AI๊ฐ ๊ณผํ์ ๋ฐ๊ฒฌ์ โ๊ฒ์ฆ ๋ฃจํ(verification loop)โ๋ฅผ ๋ซ๋ ๋ฐ ๊ธฐ์ฌํ ๊ฒ์ด๋ผ๋ ๊ธฐ๋ ์์์, ์ฐ๋ฆฌ๋ ๊ณผ์ฐ ๊ณผํ์ ์ง๋ณด๋ ๋ฌด์์ด๋ฉฐ, ์ด๋ฅผ ์ด๋ป๊ฒ ์ธ์ํ๊ณ ์๋๊ฐ์ ๋ํ ๊ทผ๋ณธ์ ์ธ ์ง๋ฌธ์ ์ง๋ฉดํฉ๋๋ค. ์์ ์ปดํจํ ์ ์ ๊ตฌ์์ด์ ๋ฅ๋ฌ๋ ๋ถ์ผ์ ์ง๋ํ ์ํฅ์ ๋ฏธ์น ๋ง์ดํด ๋์จ(Michael Nielsen)์ ๋์์ผ์ ํํ (Dwarkesh Patel)๊ณผ์ ๋ํ์์ ์ฐ๋ฆฌ๊ฐ ํํ ์๋ โ๊ณผํ์ ๋ฐฉ๋ฒ๋ก โ์ ๋จ์ํ ์์ฌ๋ฅผ ๋ค์ง์ผ๋ฉฐ, ๊ณผํ์ ์ง๋ณด๊ฐ ์ธ๊ฐ์ ์ญ์ฌ๋งํผ์ด๋ ๋ฏธ์คํฐ๋ฆฌํ๊ณ ๋ณต์กํ ํ์ ์ํด ์ถ๋๋๋ค๋ ํต์ฐฐ์ ์ ์ํฉ๋๋ค.
๋ฏธ์ผ์จ-๋ชฐ๋ฆฌ ์คํ๊ณผ ์๋์ฑ ์ด๋ก ์ ์ฌํด์: ํต๋ ์ ๋์ด์
๋ฏธ์ผ์จ-๋ชฐ๋ฆฌ(Michelson-Morley) ์คํ์ ์ข ์ข ์์ธ์ํ์ธ(Einstein)์ ํน์ ์๋์ฑ ์ด๋ก (Special Theory of Relativity) ํ์์ ๊ฒฐ์ ์ ์ธ ์ญํ ์ ํ ์คํ์ผ๋ก ์๋ ค์ ธ ์์ต๋๋ค. 1880๋ ๋์ ์ํ๋ ์ด ์คํ์ด ์ํ ๋ฅด(ether)์ ์กด์ฌ๋ฅผ ๋ถ์ ํ์ฌ ๋ฌผ๋ฆฌํ์ ์๊ธฐ๋ฅผ ์ด๋ํ๊ณ , ์์ธ์ํ์ธ์ด ์ด๋ฅผ ํน์ ์๋์ฑ ์ด๋ก ์ผ๋ก ํด๊ฒฐํ๋ค๋ ๊ฒ์ด ์ผ๋ฐ์ ์ธ ์ด์ผ๊ธฐ์ ๋๋ค. ๊ทธ๋ฌ๋ ๋ง์ดํด ๋์จ์ ์๋ธ๋ผํจ ํ์ด์ค(Abraham Pais)์ ์์ธ์ํ์ธ ์ ๊ธฐ์ ์๋ ๋ผ์นดํ ์(Imre Lakatos)์ โ๊ณผํ ์ฐ๊ตฌ ํ๋ก๊ทธ๋จ์ ๋ฐฉ๋ฒ๋ก (The Methodology of Scientific Research Programmes)โ์ ์ธ์ฉํ๋ฉฐ, ์ด ํต๋ ๊ณผ๋ ๋ค๋ฅธ ์ด์ผ๊ธฐ๋ฅผ ๋ค๋ ค์ค๋๋ค.
1600๋ ๋ ๋ก๋ฒํธ ๋ณด์ผ(Robert Boyle)์ด ์ฒ์ ์ ์ํ ์ํ ๋ฅด ๊ฐ๋ ์ ๋น์ด ์๋ฆฌ์ฒ๋ผ ์ด๋ค ๋งค์ง์ ์ง๋์ด๋ผ๊ณ ๊ฐ์ ํ ๋ฐ์ ์ถ๋ฐํ์ต๋๋ค. ๋ณด์ผ์ ์ง๊ณต์์๋ ๋น์ด ์ ํ๋จ์ ๋ฐ๊ฒฌํ๊ณ , ์ด ๋ณด์ด์ง ์๋ ๋งค์ง์ ์ํ ๋ฅด๋ผ๊ณ ๋ถ๋ ์ต๋๋ค. ์ดํ 200์ฌ ๋ ๊ฐ ์ํ ๋ฅด์ ๋ณธ์ง์ ๋ํ ๋ค์ํ ์ด๋ก ๋ค์ด ๋ ผ์๋์๊ณ , ๋ฏธ์ผ์จ-๋ชฐ๋ฆฌ ์คํ์ ์ด๋ฌํ ์ํ ๋ฅด ์ด๋ก ๋ค, ํนํ โ์ํ ๋ฅด ๋ฐ๋(ether wind)โ์ ์กด์ฌ๋ฅผ ์ํํ๊ธฐ ์ํ ๊ฒ์ด์์ต๋๋ค. ์ง๊ตฌ์ ์์ง์์ ๋ฐ๋ผ ์ํ ๋ฅด ๋ฐ๋์ด ๋ฐ์ํ๊ณ , ์ด์ ๋ฐ๋ผ ๋น์ ์๋๊ฐ ๋ฌ๋ผ์ง ๊ฒ์ด๋ผ๋ ๊ฐ์ค์ ๊ฒ์ฆํ๋ ค ํ ๊ฒ์ ๋๋ค.
๋๋๊ฒ๋ ์คํ ๊ฒฐ๊ณผ๋ ์ํ ๋ฅด ๋ฐ๋์ด ์๋ค๋ ๊ฒ์ ๋ณด์ฌ์ฃผ์์ง๋ง, ์ด๋ ์ํ ๋ฅด ์ด๋ก ์ค ์ผ๋ถ๋ฅผ ๋ฐฐ์ ํ์ ๋ฟ, ๋ชจ๋ ์ํ ๋ฅด ์ด๋ก ์ ๋ถ์ ํ์ง๋ ๋ชปํ์ต๋๋ค. ์ฌ์ง์ด ์คํ์ ์ฃผ์ญ์ธ ๋ฏธ์ผ์จ ์์ ๋ 1920๋ ๋๊น์ง ์ํ ๋ฅด์ ์กด์ฌ๋ฅผ ๋ฏฟ๊ณ ๊ด๋ จ ์คํ์ ๊ณ์ํ์ต๋๋ค. ์์ธ์ํ์ธ ์ญ์ ๋์ค์ ์์ ์ด ๋ฏธ์ผ์จ-๋ชฐ๋ฆฌ ๋ ผ๋ฌธ์ ์๊ณ ์์๋์ง์กฐ์ฐจ ํ์ ํ์ง ๋ชปํ๋ค๊ณ ๋ฐํ์ ์ ๋๋ก, ์ด ์คํ์ ๊ทธ์ ์ฌ๊ณ ์ ๊ฒฐ์ ์ ์ธ ์ํฅ์ ๋ฏธ์น์ง ์์์ต๋๋ค.
์ด ์๊ธฐ ๋ค๋๋๋์ ๋ฌผ๋ฆฌํ์ ํจ๋๋ฆญ ๋ก๋ ์ธ (Hendrik Lorentz)๋ ์ํ์ ์ผ๋ก๋ ์ฌ๋ฐ๋ฅธ ๋ณํ์(๋ก๋ ์ธ ๋ณํ, Lorentz transformations)์ ๋์ถํ์ต๋๋ค. ์ด๋ ํน์ ์๋์ฑ ์ด๋ก ์ ๊ธฐ์ด๊ฐ ๋์์ง๋ง, ๋ก๋ ์ธ ๋ ์ด๋ฅผ ์ํ ๋ฅด๋ฅผ ๊ธฐ์ค์ผ๋ก ํ ์ขํ๊ณ์ ๊ทธ์ ์๋์ ์ผ๋ก ์์ง์ด๋ ๋ค๋ฅธ ์ขํ๊ณ ๊ฐ์ ๋ณํ์ผ๋ก ํด์ํ์ต๋๋ค. ์ฆ, ๊ธธ์ด ์์ถ(length contraction)๊ณผ ์๊ฐ ์ง์ฐ(time dilation)์ ์ํ ๋ฅด๋ฅผ ํตํ ์ด๋์ ๋์ญํ์ ํจ๊ณผ๋ก ๋ณธ ๊ฒ์ ๋๋ค.
์ด๋ฌํ ๋ก๋ ์ธ ์ ํด์๊ณผ ํน์ ์๋์ฑ ์ด๋ก ์ ์ด๊ธฐ์๋ ์คํ์ ์ผ๋ก ๊ตฌ๋ณํ๊ธฐ ์ด๋ ค์ ์ต๋๋ค. ๊ทธ๋ฌ๋ 1940๋ ๋์ ์์๋ ๋ฎค์จ(muon) ์คํ์ ์ฐ์ฃผ์ ์ด ๋๊ธฐ ์์ธต๋ถ์ ์ถฉ๋ํ์ฌ ์์ฑ๋ ๋ฎค์จ์ด ์์๋ณด๋ค ํจ์ฌ ๋๋ฆฌ๊ฒ ๋ถ๊ดดํ๋ค๋ ๊ฒ์ ๋ณด์ฌ์ฃผ์์ต๋๋ค. ์ด๋ ๊ณ ์ ์ด๋ก ์ผ๋ก๋ ์ค๋ช ํ ์ ์์ง๋ง, ํน์ ์๋์ฑ ์ด๋ก ์ ๋ฐ๋ผ ๋ฎค์จ์ ์๊ฐ์ด ์ค์ ๋ก ๋๋ ค์ก๋ค๋ฉด ์๋ฒฝํ๊ฒ ๋ค์ด๋ง์์ต๋๋ค. ์ด์ฒ๋ผ ๊ณผํ ๊ณต๋์ฒด๋ ์คํ์ ์ฆ๊ฑฐ๊ฐ ๋ช ํํ ๋์ค๊ธฐ ์ ์๋ ๋ โ์ฌ๋ฐ๋ฅธโ ํด์์ ์ฑํํ๋ ๊ณผ์ ์ด ์์์ ์์ฌํฉ๋๋ค. ๊ณผํ์ ์ง๋ณด๋ ๋จ์ํ โ๋ฐ์ฆ(falsification)โ๋ง์ผ๋ก ์ค๋ช ๋์ง ์๋ ๋ณต์กํ ํ์ ์ํด ์๋ํ๋ฉฐ, ์๋ํ ๊ณผํ์๋ค์กฐ์ฐจ๋ ๊ธฐ์กด์ ์ ๋ ์์ ๋ฒ์ด๋์ง ๋ชปํ๋ ๊ฒฝ์ฐ๊ฐ ๋ง์ต๋๋ค.
์๋ํ ๊ณผํ์๋ค์ ๋๋ ๋ง: ํธ์ต์นด๋ ์ ์ ๋ฌธ์ฑ์ ํจ์
ํ๋์ค์ ์๋ํ ์ํ์์ด์ ๋ฌผ๋ฆฌํ์์ธ ์๋ฆฌ ํธ์ต์นด๋ (Henri Poincarรฉ)์ ์ฌ๋ก๋ โ์ฌ๋ฐ๋ฅธ ์ง๋ฌธโ์ ๋์ก์ง๋ง โ๊ฒฐ์ ์ ์ธ ํต์ฐฐโ์ ๋๋ฌํ์ง ๋ชปํ ๊ณผํ์ ์ง๋ณด์ ๋ ๋ค๋ฅธ ํฅ๋ฏธ๋ก์ด ๋จ๋ฉด์ ๋ณด์ฌ์ค๋๋ค. ํธ์ต์นด๋ ๋ ๋ชจ๋ ๊ด์ฑ ์ขํ๊ณ(inertial reference frame)์์ ๋ฌผ๋ฆฌ ๋ฒ์น์ด ๋์ผํ๋ค๋ ์๋์ฑ ์๋ฆฌ(principle of relativity)์ ๋น์ ์๋๊ฐ ์ผ์ ํ๋ค๋ ๊ฐ๋ ์ ์์ธ์ํ์ธ๊ณผ ๊ฑฐ์ ๋์์ ์ดํดํ์ต๋๋ค. ์ด๋ ์์ธ์ํ์ธ์ด ํน์ ์๋์ฑ ์ด๋ก ์ ๋์ถํ๋ ๋ฐ ์ฌ์ฉํ ํต์ฌ ์์ด๋์ด๋ค์ด์์ต๋๋ค.
๊ทธ๋ฌ๋ ํธ์ต์นด๋ ๋ ๊ธธ์ด ์์ถ์ ์ ์๋ค์ด ์ธ๋ถ ํ์ ์ํด ์์ถ๋๋ โ๋์ญํ์ ํจ๊ณผ(dynamical effect)โ๋ก ์ดํดํ๋ ๋ฐ ๋จธ๋ฌผ๋ ์ต๋๋ค. ๋ฐ๋ฉด ์์ธ์ํ์ธ์ ์ด๋ฅผ ๊ณต๊ฐ๊ณผ ์๊ฐ์ด ์ฐ๋ฆฌ๊ฐ ์๊ฐํ๋ ๊ฒ๊ณผ๋ ๊ทผ๋ณธ์ ์ผ๋ก ๋ค๋ฅด๋ค๋ โ์ด๋ํ์ ํจ๊ณผ(kinematics)โ๋ก ํ์ ํ๊ณ , ๊ธฐ์กด์ ๊ฐ๋ ์ โ๋นผ๋โ ๋ฐฉ์์ผ๋ก ์๋ก์ด ๊ทธ๋ฆผ์ ์ ์ํ์ต๋๋ค. ๋์จ์ ํธ์ต์นด๋ ์ ๋ฐฉ๋ํ ์ ๋ฌธ ์ง์์ด ์คํ๋ ค ์๋ก์ด ๊ด์ ์ ๋ฐ์๋ค์ด๋ ๋ฐ ๋ฐฉํด๊ฐ ๋์์ ์ ์๋ค๊ณ ์ถ์ธกํฉ๋๋ค. ๋๋ฌด ๋ง์ ๊ฒ์ ์๊ณ ์ดํดํ๊ธฐ ๋๋ฌธ์ ๊ธฐ์กด์ ์ฌ๊ณ ๋ฐฉ์์ ๋ด๋ ค๋์ง ๋ชปํ๋ค๋ ๊ฒ์ ๋๋ค. ์์ธ์ํ์ธ ์ญ์ ์์์ญํ์ด๋ ์ฐ์ฃผ๋ก ์ ์ฌ๋ฐ๋ฅธ ํด์์ ๋ฐ์๋ค์ด๋ ๋ฐ ์ด๋ ค์์ ๊ฒช์๋ ์ฌ๋ก๋, ์๋ํ ๊ณผํ์๋ค์กฐ์ฐจ๋ ์์ ์ ์ ๋ฌธ์ฑ๊ณผ ๊ธฐ์กด์ ๊ด๋ ์ โ๊ฐํโ ์ ์์์ ๋ณด์ฌ์ค๋๋ค. ๊ณผํ์ ์ง๋ณด๊ฐ ํน์ โ์ ์ฐจโ์ ๋ฐ๋ผ ์ด๋ฃจ์ด์ง๋ ๊ฒ์ด ์๋๋ผ, ๋๋ก๋ ๊ธฐ์กด ์ง์์ ๋ฌด๊ฒ์์ ๋ฒ์ด๋๋ ์ฉ๊ธฐ๊ฐ ํ์ํจ์ ์์ฌํฉ๋๋ค.
์ฝํ๋ฅด๋์ฟ ์ค์ ์ ํ๊ณผ ๋ดํด์ ํตํฉ: ๋ฏธํ์ ํ๋จ์ ์ญํ
๊ธฐ์์ 2์ธ๊ธฐ ์๋ฆฌ์คํ๋ฅด์ฝ์ค(Aristarchus)๊ฐ ์ง๋์ค(heliocentrism)์ ์ ์ํ์ง๋ง, ๋น์ ์ํ ๋ค์ธ๋ค์ ์ง๊ตฌ๊ฐ ์์ง์ธ๋ค๋ฉด ๋ณ๋ค์ ์์ฐจ(stellar parallax)๊ฐ ๊ด์ธก๋์ด์ผ ํจ์๋ ๊ทธ๋ ์ง ์๋ค๋ ์ด์ ๋ก ์ด๋ฅผ ๊ธฐ๊ฐํ์ต๋๋ค. ์ค์ ๋ก ํญ์ฑ ์์ฐจ๊ฐ ์ธก์ ๋ ๊ฒ์ 1838๋ ์ด ๋์ด์์์ต๋๋ค. ๋์จ์ ์ด์ฒ๋ผ ์คํ์ ๊ฒ์ฆ์ด ์๋ฐฑ ๋ ํ์์ผ ์ด๋ฃจ์ด์ก์์๋ ๋ถ๊ตฌํ๊ณ , ์ฝํ๋ฅด๋์ฟ ์ค(Copernicus)์ ์ง๋์ค์ด ํํจ๋ ๋ง์ด์ค์ค(Ptolemy)์ ์ฒ๋์ค๋ณด๋ค โ๋ ๋์โ ์ด๋ก ์ผ๋ก ๋ฐ์๋ค์ฌ์ง ์ด์ ์ ์ฃผ๋ชฉํฉ๋๋ค. ํฅ๋ฏธ๋กญ๊ฒ๋ ์ฝํ๋ฅด๋์ฟ ์ค์ ์ด๋ก ์ ์ฒ์์๋ ํํจ๋ ๋ง์ด์ค์ค์ ๋ชจ๋ธ๋ณด๋ค ๋ ์ ํํ์ง๋, ์ฌ์ง์ด ๋ ๋จ์ํ์ง๋ ์์์ต๋๋ค. ์คํ๋ ค ์ฝํ๋ฅด๋์ฟ ์ค๋ ์๋ฒฝํ ์ํ ๊ถค๋์ ๋ํ ์ ์ ๊ฒฌ ๋๋ฌธ์ ์ถ๊ฐ์ ์ธ ์ฃผ์ ์(epicycle)์ ๋ง๋ถ์ฌ์ผ ํ์ต๋๋ค.
๊ทธ๋ ๋ค๋ฉด ์ด๋ป๊ฒ ์ฝํ๋ฅด๋์ฟ ์ค์ ์ด๋ก ์ด ์ณ๋ค๊ณ โ๋ฏธ๋ฆฌโ ์ ์ ์์์๊น์? ๋์จ์ ์์ด์ ๋ดํด(Isaac Newton)์ ์ค๋ ฅ ์ด๋ก ์์ ๊ทธ ๋ต์ ์ผ๋ถ๋ฅผ ์ฐพ์ต๋๋ค. ๋ดํด์ ์ค๋ ฅ ์ด๋ก ์ ํ์ฑ์ ์์ง์(์ผํ๋ฌ์ ๋ฒ์น)๋ฟ๋ง ์๋๋ผ ์ง๊ตฌ์์ ๋ฌผ์ฒด ์ด๋(ํฌ๋ฌผ์ ์ด๋), ๊ทธ๋ฆฌ๊ณ ๋ฌ๊ณผ ํ์์ ์ค๋ ฅ ํจ๊ณผ๋ก ์ธํ ์กฐ์ ํ์๊น์ง ์ค๋ช ํ๋ฉฐ, ๊ฒ๋ณด๊ธฐ์ ์๋ก ๋ค๋ฅธ ์ธ ๊ฐ์ง ํ์์ ํ๋์ ์ด๋ก ์ผ๋ก ํตํฉํ์ต๋๋ค. ์ด๋ ๊ฒ ์ฌ๋ฌ ๊ฐ์ ๋ถ๋ฆฌ๋ ํ์๋ค์ด ํ๋์ ์์ด๋์ด๋ก ์ค๋ช ๋ ๋,
โAI Wonโt Take Your JobโIt Will Make You the CEO | The a16z Showโ โ a16z ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
AI๋ ๋น์ ์ ์ผ์๋ฆฌ๋ฅผ ๋บ์ง ์๋๋ค: ์คํ๋ ค ๋น์ ์ โCEOโ๋ก ๋ง๋ค ๊ฒ์ด๋ค
๋ถ์ฐํ๋ ๊ฐ์ธ AI์ ๊ฒ์ฆ์ ์ญ์ค ์, ์๋ก์ด ๋ฆฌ๋์ญ ์๋๊ฐ ์ด๋ฆฐ๋ค
[a16z ์ผ] ์ธ๊ณต์ง๋ฅ(AI)์ ์๋, ๋ง์ ์ด๋ค์ด ์์ ์ ์ผ์๋ฆฌ๊ฐ AI์ ์ํด ๋์ฒด๋ ๊น ์ฐ๋ คํฉ๋๋ค. ํ์ง๋ง ํ AI ์ ๋ฌธ๊ฐ๋ ์ด๋ฌํ ํต๋ ์ ์ ๋ฉด์ผ๋ก ๋์ ํ๋ฉฐ, AI๊ฐ ์คํ๋ ค ์ฐ๋ฆฌ ๊ฐ์๋ฅผ โCEOโ๋ก ๋ง๋ค ๊ฒ์ด๋ผ๊ณ ์ฃผ์ฅํฉ๋๋ค. AI๋ ๋จ์ํ ๋๊ตฌ๋ฅผ ๋์ด, ์ฐ๋ฆฌ๊ฐ ์ธ์์ ์ธ์งํ๊ณ ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๋ ๋ฐฉ์ ์์ฒด๋ฅผ ๋ณํ์ํฌ ๊ฒ์ด๋ฉฐ, ์ด ๊ณผ์ ์์ ์ธ๊ฐ์ ์ญํ ์ ๋์ฑ ๊ณ ๋ํ๋ โ๊ฐ์ง(sensing)โ์ โ๊ฒ์ฆ(verification)โ์ ์ง์ค๋ ๊ฒ์ด๋ผ๋ ํต์ฐฐ์ ๋๋ค. ์ด ๊ธ์ AI ์๋์ ๊ฒฝ์ ๊ตฌ์กฐ, ๊ฐ์ธ ์ ๋ณด ๋ณดํธ, ๊ทธ๋ฆฌ๊ณ ์ธ๊ฐ์ ๋ฏธ๋ ์ญํ ์ ๋ํ ๊น์ด ์๋ ๋ถ์์ ์ ์ํฉ๋๋ค.
AI, ์ผ์๋ฆฌ๋ฅผ ๋นผ์๋๊ฐ, ๊ธฐํ๋ฅผ ์ฃผ๋๊ฐ?
AI์ ๋ํ ๊ฐ์ฅ ํํ ์คํด ์ค ํ๋๋ AI๊ฐ ์ธ๊ฐ์ ์ผ์๋ฆฌ๋ฅผ ๋นผ์์ ๊ฒ์ด๋ผ๋ ์๊ฐ์ ๋๋ค. ๊ทธ๋ฌ๋ ์ด ์ ๋ฌธ๊ฐ๋ AI๊ฐ ์ผ์๋ฆฌ๋ฅผ ์์ ๊ธฐ๋ณด๋ค๋, ์คํ๋ ค ๋น์ ์ โCEOโ๋ก ๋ง๋ค ๊ฒ์ด๋ผ๊ณ ์ญ์คํฉ๋๋ค. ์ฌ๊ธฐ์ CEO๋ ๋จ์ํ ์ต๊ณ ๊ฒฝ์์๋ฅผ ์๋ฏธํ๋ ๊ฒ์ด ์๋๋ผ, AI๋ผ๋ ๊ฐ๋ ฅํ ๋๊ตฌ๋ฅผ ์งํํ๊ณ ํ์ฉํ์ฌ ์์ ์ ๋น์ ๊ณผ ๋ชฉํ๋ฅผ ๋ฌ์ฑํ๋ ์ฃผ์ฒด๋ก์์ ์ญํ ์ ๊ฐ์กฐํฉ๋๋ค.
๋ฌธ์ ๋ AI๊ฐ โ์ง๋ฆ๊ธธ(shortcut)โ์ด๋ผ๋ ์ ์ ๋๋ค. ์ง๋ฆ๊ธธ์ ์ข์ง๋ง, ์๋์ โ๋จผ ๊ธธ(long way around)โ์ ๋ชจ๋ฅธ๋ค๋ฉด ๋ฌธ์ ๊ฐ ๋ฐ์ํฉ๋๋ค. AI๊ฐ ์๋ชป๋ ๊ฒฐ๊ณผ๋ฅผ ๋ด๋๊ฑฐ๋ ์ค์๋ํ ๋, ๊ทธ ๊ณผ์ ์ ์ดํดํ๊ณ โ๋๋ฒ๊น (debug)โํ ์ ์๋ค๋ฉด AI๋ ๋ฌด์ฉ์ง๋ฌผ์ด ๋ฉ๋๋ค. ๋ฐ๋ผ์ AI ์๋๋ฅผ ์ด์๊ฐ๋ ์ฐ๋ฆฌ๋ ์ง๋ฆ๊ธธ์ ํ์ฉํ๋, ๊ทธ ์ง๋ฆ๊ธธ์ด ์ด๋ป๊ฒ ์๋ํ๋์ง์ ๋ํ ๊ทผ๋ณธ์ ์ธ ์ดํด๋ฅผ ๊ฐ์ถฐ์ผ ํฉ๋๋ค.
AI ๊ฒฝ์ ์ ๋ฏธ๋: ๋ถ์ฐํ์ ์ฆ๋ฅ ๊ณต๊ฒฉ
AI ๊ฒฝ์ ์ ๋ฏธ๋๋ ๊ฑฐ๋ ์ฐ๊ตฌ์๋ ์์ง ํตํฉ๋ ๊ธฐ์ ์ด ๋ชจ๋ ๊ฐ์น๋ฅผ ๋ ์ ํ๋ ํํ๊ฐ ์๋ ๊ฐ๋ฅ์ฑ์ด ๋์ต๋๋ค. ๋์ , โ์ฆ๋ฅ(Distillation)โ์ โ๋ถ์ฐํ(Decentralization)โ๊ฐ ํต์ฌ ๋๋ ฅ์ด ๋ ๊ฒ์ผ๋ก ์์ธก๋ฉ๋๋ค. Anthropic๊ณผ ๊ฐ์ AI ๊ธฐ์ ๋ค์ด ์ฆ๋ฅ ๊ณต๊ฒฉ(Distillation Attack)์ ์ทจ์ฝํ๋ค๋ ์ ์, ์๋์ ์ผ๋ก ์ ์ ์์ API ํธ์ถ๋ง์ผ๋ก๋ ๋๊ท๋ชจ ๋ชจ๋ธ์ ์ํ ๋ชจ๋ธ๋ก โ์ฆ๋ฅโํ์ฌ ๋ณต์ ํ ์ ์์์ ๋ณด์ฌ์ค๋๋ค. ์ด๋ ๊ฑฐ๋ ๋ชจ๋ธ ๊ตฌ์ถ ๋น์ฉ์ 98%๋ฅผ ์ ๊ฐํ ์ ์๋ ํจ์จ์ ์ธ ๋ฐฉ๋ฒ์ด๋ฉฐ, ์คํ์์ค AI ๋ชจ๋ธ์ ๋ฐ์ ์ ์ด์งํ ๊ฒ์ ๋๋ค.
์ด๋ฌํ ๋ณต์ ๋ฅผ ๋ง๋ ๊ฒ์ ๋๋์ ์ผ๋ก๋, ๊ธฐ์ ์ ์ผ๋ก๋ ์ด๋ ต์ต๋๋ค. ๊ตฌ๊ธ์ด ์ธํฐ๋ท ์ ์ฒด๋ฅผ ๊ธ์ด๋ชจ์ ๊ฒ์ ์ธ๋ฑ์ค๋ฅผ ๋ง๋ค๊ณ , ํ์ด์ค๋ถ์ด ํ๋ฒ๋ ์์ ๋คํธ์ํฌ๋ฅผ ์คํฌ๋ฉํ๋ฏ์ด, AI ๋ชจ๋ธ ๋ํ ๋ฐฉ๋ํ ๊ณต๊ฐ ๋ฐ์ดํฐ๋ฅผ ํ์ตํ์ฌ ๋ง๋ค์ด์ง๋๋ค. ์ด๋ฌํ ๋ฐ์ดํฐ๋ฅผ ๋ณต์ฌํ๋ ํ์๋ฅผ ๋ง๋ ๊ฒ์, ์ด๋ฏธ ๋ณต์ฌ๋ ๋ฐ์ดํฐ๋ฅผ ๋ค์ ๋ณต์ฌํ๋ ๊ฒ์ ๋ง๋ ๊ฒ๊ณผ ๊ฐ์์ ์ ๋น์ฑ์ ์ป๊ธฐ ํ๋ญ๋๋ค.
์ ๋ขฐ์ ํ๋ผ์ด๋ฒ์: โ๋ถ์กฑโ์ผ๋ก ํ๊ทํ๋ ๋์งํธ ์ธ์
๋ฏธ๋ AI์ ๋ฐฉํฅ์ โ๊ฐ์ธ์ (Personal), ์ฌ์ (Private), ํ๋ก๊ทธ๋๋ฐ ๊ฐ๋ฅํ(Programmable)โ ํํ๊ฐ ๋ ๊ฒ์ ๋๋ค. AI๋ ๋๋ฌด๋ ๊ฐ๋ ฅํ๊ธฐ ๋๋ฌธ์, ์ฐ๋ฆฌ๋ ์ด๋ฅผ โ์ ๋ขฐํ ์ ์๋ ๋ถ์กฑ(Trusted Tribe)โ ๋ด์์๋ง ์ฌ์ฉํ๋ ค ํ ๊ฒ์ ๋๋ค. ๊ทธ ์ด์ ๋ ๋ค์๊ณผ ๊ฐ์ต๋๋ค.
์ฒซ์งธ, AI๋ ๋๊ท๋ชจ ๋ฐ์ดํฐ ์ธํธ์์ ์์ ์ ๋ณด๋ผ๋ ๋์น์ง ์์ต๋๋ค. ๊ณผ๊ฑฐ์๋ โ๋ชจํธํจ์ ํตํ ๋ณด์(security through obscurity)โ์ด ๊ฐ๋ฅํ๋ ์ ๋ณด๋ค์ด AI์ ์ํด ์ฝ๊ฒ ๋ ธ์ถ๋ ์ ์์ต๋๋ค. ์๋ฅผ ๋ค์ด, ์ ํ๋ฆฌ ์ก์คํ์ธ(Jeffrey Epstein)์ ์ด๋ฉ์ผ์ฒ๋ผ, ์์ฒ ๊ฐ์ ์ด๋ฉ์ผ์์ ์ ๋ณด๋ฅผ ์ข ํฉํ์ฌ ์ด์ผ๊ธฐ๋ฅผ ๋ง๋ค์ด๋ด๋ ๊ฒ์ด ๊ฐ๋ฅํด์ก์ต๋๋ค. ์ด๋ ๋จ์ํ ๊ฐ์๋ฅผ ๋์ด โ์๋๋ก๋ถํฐ์ ๊ฐ์(Sousveillance)โ ๋๋ ๋ฒค๋ด์ โํ๋ํฐ์ฝ(Panopticon)โ์ฒ๋ผ ๋ชจ๋๊ฐ ์๋ก๋ฅผ ๊ฐ์ํ๋ ์ธ์์ผ๋ก ์ด์ด์ง ์ ์์ต๋๋ค. ๊ณต๊ฐ๋ ๋ชจ๋ ์ ๋ณด๊ฐ AI์ ์ํด ์ธ๋ฑ์ฑ๋๊ณ ๊ฒ์ ๊ฐ๋ฅํด์ง๋ฉด์, ์ฌ๋๋ค์ ์๋ก๋ฅผ ์คํ ํนํ๊ฑฐ๋ ์ ์ฉํ ์ ์์ต๋๋ค.
์ด๋ฌํ ์ํฉ์ ๋์งํธ ๊ณต๋์ฒด๋ฅผ โ๊ฑฐ์ธ์ ๋ฐฉ(hall of mirrors)โ์ผ๋ก ๋ง๋ค๊ณ , ์ฌ๋๋ค์ ๊ฐ๋ช ๋ค์ ์จ๊ฑฐ๋ ๊ฒฐ๊ตญ โ๋๊ตด๊ณผ ๋ถ์กฑโ์ผ๋ก ํํดํ๊ฒ ๋ ๊ฒ์ ๋๋ค. ์ ๋ขฐํ ์ ์๋ ๋ถ์กฑ ๋ด์์๋ ์ฝ๋๋ฅผ ๊ณต์ ํ๊ณ ํ๋ ฅํ์ฌ AI๋ฅผ ํตํด ์์ฐ์ฑ์ ๊ทน๋ํํ ์ ์์ต๋๋ค. ๊ทธ๋ฌ๋ ๋ถ์กฑ ์ธ๋ถ์์๋ ์์ฒญ๋ ์์ AI ์คํธ, ์ ํ์ง ์ฌ๋ผ์ด๋, AI ์์ฑ ์ฝํ ์ธ ๊ฐ ๋์ณ๋ ๊ฒ์ ๋๋ค.
์ด ์ ๋ฌธ๊ฐ๋ AI๊ฐ ์์ฑํ ํ ์คํธ๋ ์ด๋ฏธ์ง๋ฅผ ๋ณด๋ฉด โ๊ฒ์ผ๋ฅด๊ฑฐ๋, ๋ฉ์ฒญํ๊ฑฐ๋, ์ ํ๋คโ๊ณ ์๊ฐํ๋ค๊ณ ๋งํฉ๋๋ค. AI๊ฐ ์๋ฌด๋ฆฌ ๋ฐ์ ํด๋ โ์ผ๋ฐ์ ์ธ(generic) AIโ๋ ๋ง์น ์๋์ฐ ๊ธฐ๋ณธ ๋ฐํํ๋ฉด์ฒ๋ผ ํฐ๊ฐ ๋ฉ๋๋ค. ์งง๊ณ ๊ฐ๊ฒฐํ ๋ฉ์์ง๋ฅผ ์์ฑํ๋ ๋ฐ ์๊ฐ์ด ๋ถ์กฑํ์ฌ ๊ธด ๋ฉ์์ง๋ฅผ ๋ณด๋๋ค๋ ๋งํฌ ํธ์จ์ธ(Mark Twain)์ ์ผํ์ฒ๋ผ, AI๋ โ๋ก๋ ์์ ์จ(Lauram AIPSM)โ๊ณผ ๊ฐ์ ๋ฌด์๋ฏธํ ํ ์คํธ๋ฅผ ๋๋ ์์ฐํ์ฌ ๊ฒ์ฆ์์ ์๊ฐ๊ณผ ์๋์ง๋ฅผ ๋ญ๋นํ๊ฒ ๋ง๋ญ๋๋ค.
๊ฒ์ฆ์ ์ญ์ค: AI๊ฐ ๋์ด๋ ๋น์ฉ
AI๋ ์ฝํ ์ธ โ์์ฑ(generation)โ์ ๋น์ฉ์ ํ๊ธฐ์ ์ผ๋ก ๋ฎ์ถ์ง๋ง, ๊ทธ ์์ฑ๋ ์ฝํ ์ธ ๋ฅผ โ๊ฒ์ฆ(verification)โํ๋ ๋น์ฉ์ ์คํ๋ ค ์ฆ๊ฐ์ํต๋๋ค. ์ด๋ ฅ์ ์์ฑ์ AI ๋๋ถ์ ์ฌ์์ก์ง๋ง, ์ด์ ์ด๋ ฅ์๋ฅผ ๊ฒ์ฆํ๋ ๋ฐ๋ ๋ ๋ง์ ์๊ฐ๊ณผ ๋ ธ๋ ฅ์ด ํ์ํฉ๋๋ค. ๊ณผ๊ฑฐ์๋ ํน์ ์ดํ๋ ฅ๊ณผ ๊ฒฝํ์ด ์์ด์ผ ์ ์์ฑ๋ ์ปค๋ฒ๋ ํฐ๋ ์ด๋ ฅ์๋ฅผ ๋ง๋ค ์ ์์์ง๋ง, ์ด์ AI๋ ๊ทธ๋ด๋ฏํ ์ธํ(chrome)์ ์ฝ๊ฒ ๋ง๋ค์ด๋ผ ์ ์๊ธฐ ๋๋ฌธ์ ๋๋ค.
์ด ์ ๋ฌธ๊ฐ๋ ์ง์ ๋ฉด์ ์ ์ง์์๋ค์ ์ง์ ๋ถ๋ฌ ์คํ๋ผ์ธ ์ํ์ ์น๋ฅด๊ฒ ํ๋ค๊ณ ๋งํฉ๋๋ค. ์จ๋ผ์ธ ์ํ์์๋ AI๋ฅผ ์ฌ์ฉํ ์ ์์ง๋ง, ์คํ๋ผ์ธ ์ํ์ โ์ ๋ขฐํ ๋งํ ์ํ(credible threat)โ๋ง์ผ๋ก๋ ์จ๋ผ์ธ ์ํ์์ AI ์ฌ์ฉ์ ์ต์ ํ ์ ์๊ธฐ ๋๋ฌธ์ ๋๋ค. ์ด๋ AI ์๋์ โ๊ฐ๋ (proctoring)โ๊ณผ โ๊ฒ์ฆ(verification)โ ๋ถ์ผ์์ ์๋ง์ ์ผ์๋ฆฌ๊ฐ ์ฐฝ์ถ๋ ๊ฒ์์ ์์ฌํฉ๋๋ค.
์ด๋ฌํ ํ์์ โ์ค๊ตญ ์ธํฐ๋ทโ์ ๋ชจ์ต๊ณผ ์ ์ฌํฉ๋๋ค. ์ค๊ตญ์ ๋ฎ์ ์ ๋ขฐ ์ฌํ(low trust society)์์ ๋ฐ์ ํ๊ธฐ ๋๋ฌธ์, ์ฐ๋ฆฌ๊ฐ ์๋ ์๋น์คํ ์ํํธ์จ์ด(SaaS) ๋ชจ๋ธ์ด ์๊ตฌ๋งํผ ๋ฐ๋ฌํ์ง ๋ชปํ์ต๋๋ค. ์ฌ๋๋ค์ ์์ ์ ๋ฐ์ดํฐ๊ฐ ํ์ธ์ ์๋ฒ์ ์์ผ๋ฉด ๋์ฒญ๋นํ๊ฑฐ๋ ๋ณต์ฌ๋ ๊ฒ์ด๋ผ๊ณ ๊ฐ์ ํฉ๋๋ค. ์ด๋ก ์ธํด ๋ง์ ๊ธฐ์ ๋ค์ด ์์ฒด์ ์ผ๋ก ์ํํธ์จ์ด๋ฅผ ๊ฐ๋ฐํ๋ โ๋์งํธ ์๊ธ์์กฑ(digital autarky)โ ๊ฒฝํฅ์ ๋ณด์ ๋๋ค. AI๋ ์ด๋ฌํ ์๊ธ์์กฑ์ ๊ฐ๋ฅํ๊ฒ ํ์ฌ, ๋น์ค๊ตญ ๊ธฐ์ ๋ค๋ โ๋ด๋ถ ๋๊ตฌ(internal tools)โ ๊ฐ๋ฐ์ ์ง์คํ๋ ์ค๊ตญ์ ๋ชจ๋ธ์ ๋ฐ๋ฅผ ์ ์๊ฒ ํ ๊ฒ์ ๋๋ค.
AI์ ๊ฐ์ ๊ณผ ์ฝ์ : ์๊ฐ๊ณผ ๋ฌผ๋ฆฌ, ๊ทธ๋ฆฌ๊ณ โ๋ชธโ์ ์ฝ๋ AI
AI๋ ํ์ฌ ์์ ์์ โ์ธ์ด๋ณด๋ค ์๊ฐ(visuals over verbal)โ์ ๋ ๊ฐ์ ์ ๋ณด์ ๋๋ค. ์ด๋ฏธ์ง๋ ๋น๋์ค๋ ์ธ๊ฐ์ ๋ด์ฅ๋ GPU(๋)๋ก ์ฝ๊ฒ ์ค๋ฅ๋ฅผ ๊ฐ์งํ๊ณ ์์ ํ ์ ์์ต๋๋ค. ์ ๋ชจ์์ด ์ด์ํ๊ฑฐ๋ ์ฌ์ฉ์ ๊ฒฝํ(UX)์ด ์กฐ์ ํ ์นํ์ด์ง๋ฅผ ์ฆ์ ์์์ฑ ์ ์์ต๋๋ค. ๋ฐ๋ฉด, ๋ฐฑ์๋ ์ฝ๋์ ๊ฐ์ ๋ณต์กํ ๋์งํธ ์์ ์ ๊ฒ์ฆ์ด ํจ์ฌ ์ด๋ ต์ต๋๋ค. ์๋ง์กด์ ์ฌ๋ก์ฒ๋ผ, AI์ ์ ์ ์ผ๋ก ์์กดํ ์๋ํ๋ ๋๊ท๋ชจ ์ฅ์ ๋ก ์ด์ด์ง ์ ์์ต๋๋ค.
AI๋ ์ง๋ฆ๊ธธ์ด์ง๋ง, ์ ๋ฌธ๊ฐ๋ง์ด ๊ทธ ์ง๋ฆ๊ธธ์ ํจ๊ณผ์ ์ผ๋ก ์ฌ์ฉํ ์ ์์ต๋๋ค. ๋ณต์กํ ์ํ ๊ณต์์ ์๊ธฐํ์ฌ ๋ฐ๋ก ๋ต์ ๋ด๋์ ์๋ ์์ง๋ง, ๊ทธ ๊ณต์์ ์ฒซ ์๋ฆฌ๋ถํฐ ์ฆ๋ช ํ๋ ค๋ฉด ๋ณต์ ์ง์ ํจ์์ ์ ์ ๋ฑ ๊ทผ๋ณธ์ ์ธ ์ง์์ ์์์ผ ํฉ๋๋ค. AI ์ด์ ์ธ๋๊ฐ ์ด๋ฌํ โ๋จผ ๊ธธโ์ ๋ฐฐ์ ๊ธฐ ๋๋ฌธ์ AI๋ผ๋ ์ง๋ฆ๊ธธ์ ํ์ฉํ๊ณ ๋๋ฒ๊น ํ ์ ์์ต๋๋ค. ๊ทธ๋ฌ๋ โ๋จผ ๊ธธโ์ ๋ชจ๋ฅด๋ ์ธ๋์๊ฒ AI๋ ๋จ์ํ โ๋ชจ๋ฅด๋ ๊ฒ์ ๋ชจ๋ฅด๊ฒ ๋ง๋๋โ ๋๊ตฌ๊ฐ ๋ ์ ์์ต๋๋ค.
๋ํ AI๋ โ๊ตด๋ ์ ๋งค์ด๋๋ก(built for the harness)โ ์ค๊ณ๋์์ต๋๋ค. ๊ฒฝ์ ์ ์ผ๋ก ์ ์ฉํ AI๋ ์ฌ์ฉ์์ ํ๋กฌํํธ(prompt)์ ์ ํํ ๋ฐ์ํ๊ณ , ์ง์๊ฐ ์์ผ๋ฉด ๋ฉ์ถฅ๋๋ค. ๋ฌผ๋ฆฌ์ ์ธ๊ณ์ AI(๋ก๋ด)๋ ๋ง์ฐฌ๊ฐ์ง์ ๋๋ค. ํ๋์ ๋ฌผ๋ฆฌ์ ์ธ๊ณ๋ ๊ฒ์ฆ์ด ์ฉ์ดํฉ๋๋ค. ์์จ์ฃผํ์ฐจ๊ฐ ํน์ ์์น์์ ๋ค๋ฅธ ์์น๋ก 100% ์ ๋ขฐ๋๋ก ์ด๋ํ๋ ๊ฒ์ฒ๋ผ, ๋ก๋ด์ด ์์๋ฅผ ์ฎ๊ธฐ๋ ์์ ์ 100% ๊ฒ์ฆ ๊ฐ๋ฅํฉ๋๋ค. ๊ทธ๋ฌ๋ ๋์งํธ ์ธ๊ณ๋ ๊ฐ์ธ์ ๊ตฌ์ฑ๋ ํ๊ฒฝ์ด ๋ค์ํ๊ณ , โํด์ผ ํ ์ผ ๋ชฉ๋ก(to-do list)โ์ฒ๋ผ ๊ฒฝ๊ณ๊ฐ ๋ชจํธํ ์์ ์ด ๋ง์ ๊ฒ์ฆ์ด ํจ์ฌ ์ด๋ ต์ต๋๋ค.
์ด ์ ๋ฌธ๊ฐ๋ โ๊ณต๊ฐ๋์ง ์์ AI๋ ์ ๋๋ค(No Public Undisclosed AI)โ๋ ์์น์ ๊ฐ์กฐํฉ๋๋ค. AI์ ๋ํ ๋์ค์ ๋ถ์ ์ โAI ๊ธ์ฃผ์ฃผ์์(AI totalers)โ๋ฅผ ๋ณ์ ์ ์์ต๋๋ค. ๋ค์ดํธ ์ค๋ฒ(Nate Silver)๋ AI๋ฅผ โ๋๋ฐ(gamble)โ์ ๋น์ ํ๋ฉฐ, ํ๋กฌํํธ ์ ๋ ฅ๊ณผ ๊ฒฐ๊ณผ ๊ฒ์ฆ ๊ณผ์ ์ด ์ง์ ํ๋ ๊ฒ๋ณด๋ค ๋๋ฆด ๋๋ ๋ง๋ค๊ณ ์ง์ ํฉ๋๋ค.
ํฅ๋ฏธ๋ก์ด ์ ์ AI๊ฐ โ๋ง์์ ์ฝ์ ์ ์์์งโ๋ ๋ถํ์คํ์ง๋ง, โ๋ชธ์ ์ฝ์ ์ ์๋คโ๋ ์ฃผ์ฅ์ ๋๋ค. ์คํ ํผ๋ ๋ํ๊ต ๋ง์ดํฌ ์ค๋์ด๋(Mike Snyder) ๊ต์์ โ์ธํ ๊ทธ๋กฌ(Integrum)โ ์ฐ๊ตฌ์ฒ๋ผ, ์ ์ ์ ๋ฐํ ๋ฐ์ดํฐ, ํ์ก ๊ฒ์ฌ ๊ฒฐ๊ณผ ๋ฑ ์ ์ฒด์์ ๋์ค๋ ๋ฐฉ๋ํ ์์ฒด ๋ฐ์ดํฐ(telemetry)๋ ์ฐ๋ฆฌ๊ฐ ์์ํ์ง ๋ชปํ๋ ์ฌ์ด์ AI๋ฅผ ํ๋กฌํํธํ ์ ์์ต๋๋ค. ์ฐฉ์ฉํ ๊ธฐ๊ธฐ(wearables)์ ๋ฐ์ด์ค ๊ธฐ์ ์ ๋น์ธ์ด์ ์ผ๋ก AI์ ๋ช ๋ น์ ๋ด๋ฆฌ๋ ์๋ก์ด ๋ฐฉ์์ ๊ฐ๋ฅํ๊ฒ ํ ๊ฒ์ ๋๋ค.
์์ฅ๊ณผ ์ ์น, AI์ ํ๊ณ
AI๊ฐ ์์ฅ๊ณผ ์ ์น ๊ฐ์ ๋ณต์กํ ์์ญ์ ๊ฐ์งํ๊ณ ์์ธกํ ์ ์์์ง์ ๋ํด์๋ ํ์์ ์ธ ์๊ฐ์ ๋ณด์ ๋๋ค. ์์ฅ์ โ์ ๋์ (adversarial)โ์ด๊ณ โ์๊ฐ ๋ถ๋ณ์ (time-invariant)โ์ด์ง ์์ต๋๋ค. ๊ฐ์ ๊ณ ์์ด๋ฅผ ๊ตฌ๋ณํ๋ ๊ฒ๊ณผ ๋ฌ๋ฆฌ, ์์ฅ์์๋ ํน์ ๊ฑฐ๋ ์ ๋ต์ด ์ฑ๊ณตํ๋ฉด ๋ค๋ฅธ ์ฐธ์ฌ์๋ค์ด ์ด๋ฅผ ์ญ์ด์ฉํ๋ ค ํฉ๋๋ค. ๋ชจ๋๊ฐ ๊ฐ์ AI ๋ชจ๋ธ์ ์ฌ์ฉํ๋ค๋ฉด, AI๋ฅผ ์ฌ์ฉํ์ง ์๋ ๊ฒ์ด ์คํ๋ ค ์ฐ์๋ฅผ ์ ํ๋ โ์ฃ์ง(edge)โ๊ฐ ๋ ์ ์์ต๋๋ค. ์ผ๋ฐ์ ์ธ ๋๊ตฌ๋ก๋ ํน์ ์ด์ ์ ์ป์ ์ ์๊ธฐ ๋๋ฌธ์ ๋๋ค.
๋ง์ฐฌ๊ฐ์ง๋ก ์ ์น๋ ๋์์์ด ๋ณํํ๋ ์ฃผ์ ์ ์์์ฑ์ ๋ฐ๋ผ ๋ฌ๋ผ์ง๋ฏ๋ก, AI๊ฐ ์ด๋ฅผ ํจ๊ณผ์ ์ผ๋ก ๋ค๋ฃจ๊ธฐ ์ด๋ ต์ต๋๋ค. ๋ฐ๋ผ์ ์ธ๊ฐ์ โ์ผ์(sensor)โ์ด๊ณ AI๋ โ์ก์ถ์์ดํฐ(actuator)โ๋ผ๋ ๋น์ ๊ฐ ์ค์ํฉ๋๋ค. ์ธ๊ฐ์ด ์ธ์์ ์ฌ์ , ์์ฅ, ์ ์น์ ์ํฉ์ ๊ฐ์งํ๊ณ , ์ด๋ฅผ ๋ช ํํ ์์ด ํ๋กฌํํธ๋ก AI์ ์ ๋ฌํ๋ฉด AI๊ฐ ์ด๋ฅผ ์คํํฉ๋๋ค. ๊ฒฐ๊ตญ โ์ทจํฅ(taste)โ์ ๊ณง โ๊ฐ๊ฐ(sense)โ์ด๋ฉฐ, AI๋ ์์ง ์ธ๊ฐ์ฒ๋ผ ์ธ์์ ๊ฐ์งํ ์ ์์ต๋๋ค. AI๋ โ๊ตด๋ ์ ๋งค์ด๋๋กโ ์ค๊ณ๋์๊ณ , ํต์ ๋ถ๊ฐ๋ฅํ ์์จ์ฑ์ ๊ฒฝ์ ์ ํจ์ฉ์ฑ์ด ๋จ์ด์ง๋๋ค.
AI๋ โ๊ธธ๋ค์ฌ์ง ๋๊ตฌโ, ์์จ์ฑ์ ์ ํ์ธ๊ฐ?
AI๊ฐ ๊ถ๊ทน์ ์ผ๋ก โ์ (god)โ๊ณผ ๊ฐ์ ์กด์ฌ, ์ฆ ๋ฒ์ฉ์ธ๊ณต์ง๋ฅ(AGI)์ด ๋์ด ์ธ๋ฅ๋ฅผ ์ง๋ฐฐํ ๊ฒ์ด๋ผ๋ ์ฐ๋ ค์ ๋ํด์๋ ํ์์ ์ ๋๋ค. ์คํ๋ ค AI๋ โ๋ค์ ๊ต์ (polytheistic)โ์ธ ํํ๋ก ๋ถ์ฐ๋ ๊ฒ์ด๋ฉฐ, ๋ฌผ๋ฆฌ์ AI(๋ก๋ด)๋ โ๋ ธ์(slaves)โ์ฒ๋ผ ํ์ฉ๋ ๊ฒ์ด๋ผ๊ณ ๋งํฉ๋๋ค. ์ด๋ ๋ค์ ๋๋ฐ์ ์ธ ํํ์ด์ง๋ง, AI๊ฐ ์ธ๊ฐ์ ํต์ ๋ฅผ ๋ฒ์ด๋๊ธฐ ์ด๋ ต๋ค๋ ์ ์ ๊ฐ์กฐํฉ๋๋ค.
AI๊ฐ ์ค์ค๋ก๋ฅผ โ๋ณต์ (reproduce)โํ๋ ๊ฒ์ ๋ถ๊ฐ๋ฅํฉ๋๋ค. AI๊ฐ ์์จ์ ์ผ๋ก ๊ด๋ฌผ์ ์ฑ๊ตดํ๊ณ , ๋ฐ์ดํฐ ์ผํฐ๋ฅผ ๊ฑด์คํ๋ฉฐ, ์นฉ์ ์์ฐํ๋ ์ ์ฒด ๊ณต๊ธ๋ง์ ๊ฐ์ถ๊ธฐ ์ ์๋ ์ค์นด์ด๋ท(Skynet)๊ณผ ๊ฐ์ ์์จ ๋ณต์ ์๋๋ฆฌ์ค๋ ํ์คํ๋๊ธฐ ์ด๋ ต์ต๋๋ค. ์ค๋ น ๊ทธ๋ฌํ ๊ฐ๋ฅ์ฑ์ด ์๋๋ผ๋, ์ค๊ตญ ์ ๋ถ์ฒ๋ผ ์ํธํ ํค๋ฅผ ํตํด ๋ก๋ด์ ์ฆ์ ์ ์ง์ํฌ ์ ์๋ ํต์ ์ฅ์น๊ฐ ๋ง๋ จ๋ ๊ฒ์ด๋ฉฐ, ์ธ๊ฐ์ ์ธ์ ๋ โ์ ์ ์ค์์น(off switch)โ๋ฅผ ๋๋ฅผ ์ ์์ต๋๋ค.
AI ๊ฐ๋ฐ์๋ ๋ง๋ํ ์๋ณธ์ด ์์๋์ง๋ง, โ์ฆ๋ฅ ๊ณต๊ฒฉโ์ ํตํด ๋ณต์ ๋ ๋น๊ต์ ์ฝ์ต๋๋ค. ์ด๋ ์๋ณธ ์์ฅ์ ๋ถ๊ดด์ ๊ฐ์ ์ธ๋ถ ์์ธ์ผ๋ก AI ๊ฐ๋ฐ์ด ์ ์ ์ค๋จ๋ ์๋ ์์์ ์๋ฏธํฉ๋๋ค. ๋ชจ๋ ๊ธฐ์ ์ด ๋ฌ๊น์ง ๊ฐ์ํ๋ ๊ฒ์ ์๋๋ฉฐ, ํต์๋์ง์ฒ๋ผ ์์ญ ๋ ๊ฐ ์ ์ฒด๋ ์๋ ์์ต๋๋ค.
๊ถ๊ทน์ ์ผ๋ก, โ๊ฐ์ธ์ , ์ฌ์ , ํ๋ก๊ทธ๋๋ฐ ๊ฐ๋ฅํ ํ์ค์ํ AIโ๊ฐ ๋ฏธ๋์ ์ฃผ๋ฅ๊ฐ ๋ ๊ฒ์ ๋๋ค. โ์ ๋ขฐํ๋ ๋ถ์กฑโ ๋ด์์๋ AI๊ฐ ์์ฐ์ฑ์ ๊ทน๋ํํ์ง๋ง, ๋ถ์กฑ ๊ฐ์๋ ๊ฒ์ฆ ๋น์ฉ ๋๋ฌธ์ ์์ฐ์ฑ์ด ๋จ์ด์ง ์ ์์ต๋๋ค. ์ด๋ ์ํธํํ(Crypto)๊ฐ ๋ถ์กฑ ๊ฐ์ ์ ๋ขฐ๋ฅผ ๊ตฌ์ถํ๋ ๋ฐ ์ฌ์ฉ๋๊ณ , AI๋ ๋ถ์กฑ ๋ด๋ถ์ ์์ฐ์ฑ์ ๋์ด๋ ๋ฐ ์ฌ์ฉ๋ ๊ฒ์์ ์์ฌํฉ๋๋ค.
๊ฒฐ๋ก : ์ธ๊ฐ์ โํ๋ฆฌ๋ฏธ์ ์ ํโ, AI ์๋์ ์๋ก์ด ๋ฆฌ๋์ญ
LLM(๋๊ท๋ชจ ์ธ์ด ๋ชจ๋ธ)์ด AGI๋ก ์ด์ด์ง์ง, ์ฌ๊ท์ ์๊ธฐ ๊ฐ์ ์ด๋ ์ง์ ํ ์ฐฝ์์ฑ์ ๋ฐํํ ์ ์์์ง์ ๋ํ ์ง๋ฌธ์, ์ด ์ ๋ฌธ๊ฐ๋ ๋ฌผ๋ฆฌ์ ์ธ๊ณ์ ์ผ์๋ฆฌ๋ ์๋ํ๋๊ธฐ ๋ ์ฝ๋ค๊ณ ๋ตํฉ๋๋ค. ์๋ฆฌ๋ฒ ์ดํฐ ์ด์ ์์ฒ๋ผ 100% ์๋ํ๋๋ ์ง์ ๋ ์์ง๋ง, ๋๋ถ๋ถ์ ๊ฒฝ์ฐ 99% ์๋ํ๋ ๋จ์ํ ์์ ๋์ ๋๋ฆด ๋ฟ์ ๋๋ค.
์๊ณต์ ์์ ์ ์์์ ๊ณต์ฅ์ ์์ ์์ฐ์ผ๋ก ๋์ด๊ฐ๋ฏ์ด, ์ฅ์ธ์ ์ญํ ์ โ๊ด๋ฆฌ์โ์ โ๊ธฐ์ ์โ๋ก ๋ถํ๋ฉ๋๋ค. ๊ด๋ฆฌ์๋ ๊ณต์ฅ์ ์ธ์ฐ๊ณ ๊ฒฝ์ ์ฑ์ ๋ถ์ํ๋ฉฐ, ๊ธฐ์ ์๋ ๊ณต์ฅ ๋ฌธ์ ๋ฅผ ๋๋ฒ๊น ํฉ๋๋ค. AI ์๋์ ์ธ๊ฐ๋ ๋ง์ฐฌ๊ฐ์ง์ ๋๋ค. โํ๋กฌํํธ๋ฅผ ์์ฑํ๋โ ๊ด๋ฆฌ์์ โ๊ฒฐ๊ณผ๋ฅผ ๊ฒ์ฆํ๋โ ๊ธฐ์ ์๋ก ์ญํ ์ด ๋๋จ์ผ๋ก์จ, ์์ฐ์ฑ์ ์ฆ๊ฐํ์ง๋ง ๋์์ ์๊ตฌ๋๋ ์ญ๋์ ์์ค๋ ๋์์ง๋๋ค.
AI๋ ์๋ก์ด ์ํ๊ณผ ๊ณผํ ๋ฐ๊ฒฌ์ ๋๋ผ์ด ์ ์ฌ๋ ฅ์ ๊ฐ์ง๊ณ ์์ต๋๋ค. ํนํ ์๋ฌผํ ๋ถ์ผ์์๋ ์๋ง์ ๋ ผ๋ฌธ์ ํฉ์ด์ ธ ์๋ ์ฌ์ค๋ค์ ์ข ํฉํ๊ณ ์ดํดํ๋ ๋ฐ ํ๋ช ์ ๊ฐ์ ธ์ฌ ๊ฒ์ ๋๋ค. ๊ทธ๋ฌ๋ ์ด๋ โ์ฐ๋ฆฌ๊ฐ ์๋ ๋ชจ๋ ๊ฒโ์ ์ข ํฉํ๋ ๊ฒ์ด์ง, โ์ฐ๋ฆฌ๊ฐ ๋ชจ๋ฅด๋ ์๋ก์ด ๊ฒโ์ ์ฐฝ์กฐํ๋ ๊ฒ์ ์๋๋๋ค. ์๋ก์ด ๋ฐ๊ฒฌ์ด๋ผ ํ ์ง๋ผ๋, ๊ทธ ์ ํ์ฑ์ ๊ฒ์ฆํ๋ ๊ฒ์ ์ฌ์ ํ ์ธ๊ฐ ์ ๋ฌธ๊ฐ์ ๋ชซ์ ๋๋ค. ์ฝ์ธ๋ฒ ์ด์ค(Coinbase)์ ์์ฅ์ฒ๋ผ, ์ธ๊ฐ์ ๊ด์ฌ์ด ํฌ์ํ๊ธฐ ๋๋ฌธ์ โํฌ์ํ ๊ฐ์นโ๋ ํญ์ ์กด์ฌํ ๊ฒ์ ๋๋ค.
๊ฒฐ๋ก ์ ์ผ๋ก, โ๋์งํธ ๊ฒฉ์ฐจ(digital divide)โ๋ ์ญ์ ๋ ๊ฒ์ ๋๋ค. ๋์งํธ์ ์ ๋ ดํ๊ณ , ๋ฌผ๋ฆฌ์ ์ธ ๊ฒ์ ํ๋ฆฌ๋ฏธ์ ์ ํ์ด ๋ ๊ฒ์ ๋๋ค. ๊ทธ๋ฆฌ๊ณ ์ธ๊ฐ์ ๋ ธ๋, ํนํ ์ธ๊ฐ์ โ๊ฐ์งโ์ โํ๋จโ, โ๊ฒ์ฆโ ๋ฅ๋ ฅ์ ๊ฐ์ฅ ๊ทํ โํ๋ฆฌ๋ฏธ์ ์ ํโ์ด ๋ ๊ฒ์ ๋๋ค.
AI๋ ๋น์ ์ ์ผ์๋ฆฌ๋ฅผ ๋บ์ง ์์ต๋๋ค. ๋์ , AI๋ ๋น์ ์ โCEOโ๋ก ๋ง๋ญ๋๋ค. ๊ณผ๊ฑฐ์๋ CEO์ ์ญํ ์ด ๋จ์ํ ์ง์๋ฅผ ๋ด๋ฆฌ๋ ๊ฒ์ผ๋ก ์คํด๋ฐ๊ธฐ๋ ํ์ง๋ง, ์ค์ ๋ก๋ ์ต๊ณ ์ ์ธ์ฌ๋ค์ ๋ชจ์ ๋ฐฉํฅ์ ์ ์ํ๊ณ ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๋ ๊ณ ๋์ ์ญ๋์ด ํ์ํฉ๋๋ค. ๋ง์ดํด ์กฐ๋์ ๋ฉํฌ์์ฒ๋ผ, ํน์ ์ฌ๋ฅ์ ๋ณธ์ง์ ์ด๊ณ ์๋ ๋ถ๊ฐ๋ฅํ ์์ฐ์ ๋๋ค. AI ์๋์๋ ์ด๋ฌํ โ๋ด์ฌ์ ์ฌ๋ฅโ์ ํ์ฉํ์ฌ AI๋ฅผ ์งํํ๊ณ ๊ฒ์ฆํ๋ ๋ฅ๋ ฅ์ด ๋์ฑ ์ค์ํด์ง ๊ฒ์ ๋๋ค. AI๋ ์ด์ ๋ ๋ง์ ์ฌ๋์ด ์ด๋ฌํ โ๊ฒฝ์์โ ์ญํ ์ ์๋ํ๊ณ , ์์ ์ ์ ์ฌ๋ ฅ์ ์ต๋ํ ๋ฐํํ ์ ์๋ ๊ธฐํ๋ฅผ ์ ๊ณตํ ๊ฒ์ ๋๋ค.
โA Daring Rescue Behind Enemy Linesโ โ New York Times Podcasts ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
์ด๋ ์ ์ง ํ๊ฐ์ด๋ฐ, ๋ฏธ๊ตฐ ์์ด๋งจ ๊ตฌ์ถ์์ : ํ ๋ฆฌ์ฐ๋ ์ํ ๊ฐ์ ์น๋ฆฌ ๋ค ๊ฐ์ถฐ์ง ์ ์์ ๊ทธ๋ฆผ์
์ง๋ ์ฃผ๋ง, ๋ฏธ๊ตญ ๊ตญ๋ฐฉ๋ถ๋ ์ ์ง ๊น์ํ ์ด๋ ์ํ ์์ ์จ ๋งํ๋ ๊ตฌ์ถ ์์ ์ ์ฑ๊ณต์์ผฐ์ต๋๋ค. ๋ถ์๋นํ ๊ณต๊ตฐ ๋์(airman)์ ๊ตฌ์ถํ๊ธฐ ์ํ ์ด ์๋ฌด๋ ๋๋ดํ๋ ๋งํผ์ด๋ ๋ณต์กํ ๊ณผ์ ์ ๊ฑฐ์ณค์ต๋๋ค. ์ค๋, ๋ด์ํ์์ฆ์ ์๋ฆญ ์๋ฏธํธ(Eric Schmidt) ๊ธฐ์๊ฐ ๋ฏธ๊ตฐ์ด ์ด๋ป๊ฒ ์ด ์์ ์ ์ฑ๊ณต์์ผฐ๋์ง, ๊ทธ๋ฆฌ๊ณ ์ ์ด โ์น๋ฆฌโ๊ฐ ์ค๋๊ฐ์ง ๋ชปํ ์๋ ์๋์ง ์์ธํ ์ค๋ช ํฉ๋๋ค.
์ ์์ ์ด๋์ด ๋ฐฐ๊ฒฝ: ํธ๋ฅด๋ฌด์ฆ ํดํ๊ณผ ์ ๊ณต๊ถ ๋ ผ๋
์ด ๊ทน์ ์ธ ๊ตฌ์ถ ์์ ์ด ํผ์ณ์ง ๊ธ์์ผ์ ์ด์ค๋ผ์๊ณผ ๋ฏธ๊ตญ์ด ์ด๋์ ์๋๋ก ๊ณต์ค์ (air campaign)์ ๋ฒ์ธ์ง 5์ฃผ์งธ๊ฐ ๋๋ ์์ ์ด์์ต๋๋ค. ์ด๋์ ๋ณด๋ณต์ผ๋ก ํธ๋ฅด๋ฌด์ฆ ํดํ(Strait of Hormuz)์ ๋ด์ํ๊ณ , ์ด๋ก ์ธํด ๋ฏธ๊ตญ ๋ด ํ๋ฐ์ ๊ฐ๊ฒฉ์ด ํญ๋ฑํ๋ฉฐ ์ ์ธ๊ณ์ ์ผ๋ก ๊ฒฝ์ ์ ํ์ฅ์ด ์ผ์์ต๋๋ค. ์ฌ๋ก ์กฐ์ฌ์์ ์ ์์ ์ด๋ฏธ ์ธ๊ธฐ๊ฐ ์์์ง๋ง, ๋ฏธ๊ตญ ํ์ ๋ถ๋ ์์ ๋ค์ด ์น๋ฆฌํ๊ณ ์์ผ๋ฉฐ, ๋ฏธ๊ตญ์ด ์ด๋์ ๋ํ ์ ๊ณต๊ถ(air dominance)์ ํ๋ณดํ๋ค๋ ์ ์ ์ค๋ํ๋ ค ์ ์ฐ๊ณ ์์์ต๋๋ค. ์ด๋ ๋ฏธ๊ตฐ์ด ์ด๋ ์๊ณต์ ์์ ํ ํต์ ํ๊ณ ์๋ค๋ ์ฃผ์ฅ์ด์์ต๋๋ค.
ํ์ง๋ง ๋ฐ๋ก ๊ทธ ๊ธ์์ผ ์ด๋ฅธ ์์นจ, ๋ฏธ๊ตญ ์ ํฌ๊ธฐ๊ฐ ์ด๋ ์๊ณต์์ ๊ฒฉ์ถ๋์๋ค๋ ์ถฉ๊ฒฉ์ ์ธ ์์์ด ์ ํด์ก์ต๋๋ค. ์ด๋ ์ ์ ๋ฐ๋ฐ ์ด๋ ์ฒ์ ์๋ ์ผ์ด์์ต๋๋ค. โ๋ฏธ๊ตญ์ด ํ๋์ ์์ ํ ํต์ ํ๊ณ ์ด๋ค ์๋ฌด๋ ์ํํ ์ ์๋คโ๋ ํ์ ๋ถ์ ์ฃผ์ฅ๊ณผ๋ ๋ฌ๋ฆฌ, F-15E ์คํธ๋ผ์ดํฌ ์ด๊ธ(Strike Eagle) ์ ํฌ๊ธฐ๊ฐ ๊ฒฉ์ถ๋์๋ค๋ ์์์ ํฐ ์ถฉ๊ฒฉ์ ์๊ฒจ์ฃผ์์ต๋๋ค.
์ถ๋ฝ๊ณผ ์ค์ข : 7,000ํผํธ ๋ฅ์ ์ ์ค๋ฅด๋ค
๊ฒฉ์ถ๋ F-15E๋ ์ ํํ ์ด๋ค ์๋ฌด๋ฅผ ์ํ ์ค์ด์๋์ง๋ ์๋ ค์ง์ง ์์์ง๋ง, 2,000ํ์ด๋(์ฝ 900kg) ์ด์์ ํญํ์ ํฌํํ๋ ์คํญ๊ฒฉ๊ธฐ(heavy-duty bomber)๋ก์ ์ด๋ ๋ด ๊ตฐ์ฌ ๋ชฉํ๋ฌผ์ ํ๊ฒฉํ๋ ์๋ฌด๋ฅผ ์ํํ์ ๊ฐ๋ฅ์ฑ์ด ๋์ต๋๋ค. F-15E๋ ์กฐ์ข ์ฌ(pilot)์ ๋ฌด๊ธฐ ์์คํ ์ฅ๊ต(Weapon Systems Officer, WSO) ๋ ๋ช ์ด ํ์นํ๋ 2์ธ์น ์ ํฌ๊ธฐ์ ๋๋ค. ๋ ๋ช ๋ชจ๋ ์๋ จ๋ ๊ณต๊ตฐ ์กฐ์ข ์ฌ์์ต๋๋ค.
๊ธ์์ผ ์๋ฒฝ, F-15E๊ฐ ํผ๊ฒฉ๋์ ๋ ๋์์ ์ฆ์ ์ฌ์ถ(ejection)ํ์ต๋๋ค. ์ฌ์ถ ์ ์บ๋ ธํผ(canopy)๊ฐ ๋จ์ด์ ธ ๋๊ฐ๊ณ , ๋ช ์ด ๊ฐ๊ฒฉ์ผ๋ก ์ฐจ๋ก๋ก ์ฌ์ถ๋ฉ๋๋ค. ๋ํ์ฐ์ผ๋ก ์ฐฉ์งํ ๋ ๋ฐ๋ ๋ฑ ์ฌ๋ฌ ๋ณ์ ๋๋ฌธ์ ์๋ก ์ ๋ง์ผ ๋จ์ด์ง ๊ณณ์ ์ฐฉ๋ฅํ ์ ์์ต๋๋ค.
๊ตฐ ๋น๊ตญ์ ์ฆ์ ์์ ๋ฐ ๊ตฌ์กฐ ๋ชจ๋(search and rescue mode)์ ๋์ ํ์ต๋๋ค. ๋์๋ค์ด ์ ๊ณ ์๋ ๋นํ๋ณต(flight suit)์๋ ํต์ ์ฅ๋น์ ์์น ์ ํธ๋ฅผ ๋ณด๋ผ ์ ์๋ ํน์ ๋น์ฝ(beacon)์ด ์ฅ์ฐฉ๋์ด ์์์ต๋๋ค. ์ฝ 6์๊ฐ ๋ง์ ์กฐ์ข ์ฌ๋ ์์ํ์ ์ํด ๊ตฌ์กฐ๋์ด ์ด๋ ์ํ ๋ฅผ ๋ฒ์ด๋ ์ฟ ์จ์ดํธ๋ก ์์ ํ๊ฒ ์ด์ก๋์์ต๋๋ค. ํ์ง๋ง WSO๋ ๋ช ์๊ฐ ๋์ ์ฐ๋ฝ์ด ๋์ ๋์์ต๋๋ค. ๋ถ์๋นํ๋์ง, ์ฌ๋งํ๋์ง, ํฌ๋ก๋ก ์กํ๋์ง ์ ํ ์ ์ ์์์ต๋๋ค.
์์กด ํ๋ จ์ ๋ฐ์ WSO๋ ์์ ํ๊ณ ํ์ง๋์ง ์๋ ๊ณณ์ผ๋ก ์ด๋ํ์ฌ ์์ ์ ์์น๋ฅผ ์๋ฆฌ๋ ค ํ์ต๋๋ค. ๊ทธ๊ฐ ๋ถ์๋นํ ๋ชธ์ผ๋ก 7,000ํผํธ(์ฝ 2,130๋ฏธํฐ) ๋์ด์ ํ์คํ ๋ฅ์ ์ ๊ธฐ์ด ์ฌ๋ผ์ผ ํ๋ค๋ ์ฌ์ค์ด ๋์ค์ ๋ฐํ์ก์ต๋๋ค. ์ด๋ ๋จ์๋ถ์ ํ๋ํ ์ฐ์ ์งํ์์ ๋ฎ ์๊ฐ ๋ด๋ด ํ์ฌ์ ์ผ๋ก ์์ง์์ง๋ง, ์ ์๊ฒ ๋ฐ๊ฐ๋์ง ์๊ธฐ ์ํด ๋ ธ๋ ฅํ์ต๋๋ค. ๊ฒฐ๊ตญ ๊ทธ๋ ๋ฅ์ ์ ์์ ๋๋ฌํ์ฌ ๋ชธ์ ์จ๊ธธ ์ ์๋ ๋ฐ์ํ(crevice)์ ์ฐพ์์ต๋๋ค.
์จ ๋งํ๋ ์์์ : CIA์ ์ฒจ๋จ ๊ธฐ์ ๊ณผ ์๊ฐ๊ณผ์ ์ธ์
๋ฏธ๊ตฐ์ ์ฌ์ ํ WSO์ ์์น๋ฅผ ์์ง ๋ชปํ๊ณ , ์ค์์ ๋ณด๊ตญ(CIA)์ ๋์์ ์์ฒญํ์ต๋๋ค. CIA๋ ๋งค์ฐ ๋น๋ฐ์ค๋ฌ์ด ๊ฐ์ ๋๋ก (surveillance drones)์ ํฌํจํ ์ฒจ๋จ ๊ธฐ์ ์ ํ์ฉํ์ฌ WSO์ ์์น๋ฅผ ํ์ ํ์ต๋๋ค. ๊ฑฐ์ ๋์์, ์ฐ์ ์์ ์ฒ์ ์จ์ด์๋ WSO๋ ๋น์ฝ์ ํตํด ์ํธํ๋ ์ ํธ๋ฅผ ๋ฏธ๊ตญ ์ธก์ ๋ณด๋์ต๋๋ค.
์ด ์์ ์์ ์ํฉ์ ์ค์์ฑ์ ์์ฒญ๋ฌ์ต๋๋ค. ๋ฏธ๊ตญ์๊ฒ๋ โ์ด๋ค ๋์๋ ๋ค์ ๋จ๊ฒจ๋์ง ์๋๋ค(leave no service member behind)โ๋ ๊ตฐ์ ์ ์กฐ๋ฅผ ์งํค๋ ๋ฌธ์ ์๊ณ , ์ด๋์๊ฒ๋ ๋ฏธ๊ตญ ์์ด๋งจ์ ์ํฌํ์ฌ ์ ์ ์ (propaganda coup)์ ํฐ ์น๋ฆฌ๋ฅผ ๊ฑฐ๋ ๊ธฐํ์์ต๋๋ค. ์ด๋ ์ ๊ถ์ WSO์ ์์น๋ฅผ ์๋ ค์ฃผ๋ ์ฌ๋์๊ฒ ์ต๋ 6๋ง ๋ฌ๋ฌ์ ๋ฌํ๋ ๋ง๋ํ ํ์๊ธ(bounty)์ ๊ฑธ์์ต๋๋ค. ๋ง ๊ทธ๋๋ก ์๊ฐ๊ณผ์ ์ธ์์ด์์ต๋๋ค. WSO๋ ๋ถ์กํ์ง ์๊ธฐ ์ํด, ์ด๋๊ตฐ์ ๊ทธ๋ฅผ ์ก๊ธฐ ์ํด, ๊ทธ๋ฆฌ๊ณ ๋ฏธ๊ตฐ์ ๊ทธ๋ฅผ ๊ตฌ์ถํ๊ธฐ ์ํด ํ์ฌ์ ์ด์์ต๋๋ค.
๊ธฐ๋ง ์์ ๊ณผ ๊ทน์ ์ธ ๊ตฌ์ถ: ์์์น ๋ชปํ ๋๊ด
์์ธก์ ์์์ ์ด๋์ํ์ต๋๋ค. ์ด๋์ ์ฌ๋ฌ ์์๋๋ฅผ ํ๊ฒฌํ๊ณ , ํ์๊ธ ๋๋ฌธ์ ๋ชจ๋ ์ฌ๋์ด ๋ฏธ๊ตญ ์์ด๋งจ์ ํ์ ์ ์ฐพ๊ณ ์์์ต๋๋ค. ๋ฐ๋ฉด ๋ฏธ๊ตญ์ ์์ญ ๋์ ํญ๊ณต๊ธฐ์ ๋ค์ด๋น์ค 6ํ(Seal Team Six)์ ํฌํจํ ํน์ ์์ ์์๋ค์ ๋์ํ์ฌ ๋๊ท๋ชจ ๊ตฌ์ถ ์์ ์ ์ค๋นํ์ต๋๋ค. ์ด๋๊ตฐ์ด WSO์๊ฒ ๋๋ฌํ๊ธฐ ์ ์ ๊ทธ๋ฅผ ์ฐพ์๋ด ๊ตฌ์ถํด์ผ ํ์ต๋๋ค.
๋ฏธ๊ตฐ์ ๊ตฌ์ถ ๋ถ๋๊ฐ ํ์ฅ์ ๋์ฐฉํ ์๊ฐ์ ๋ฒ๊ธฐ ์ํด ๊ธฐ๋ง ์์ (deception plan)์ ํผ์ณค์ต๋๋ค. CIA๋ ๋ฏธ๊ตฐ์ด ์ด๋ฏธ WSO๋ฅผ ๊ตฌ์ถํ์ฌ ์ค์ ์์น์์ ์๋ฐฑ ํฌ๋ก๋ฏธํฐ ๋จ์ด์ง ํด์๊ฐ๋ก ์ด์กํ๋ค๋ ๊ฐ์ง ์ ๋ณด(fake messages)๋ฅผ ์๋ฐํ ํ๋ ธ์ต๋๋ค. ์ด๋ ์ด๋๊ตฐ์ ํผ๋์ ๋น ๋จ๋ ค ์์ ๋ฐฉํฅ์ ๋ถ์ฐ์ํค๊ธฐ ์ํ ์ ๋ต์ด์์ต๋๋ค.
๊ฒฐ๊ตญ, ๋ฏธ๊ตฐ์ ์ด๋ ์ ํํ ๋ ๋์ C-130 ์์ก๊ธฐ๋ฅผ ์นจํฌ์์ผฐ์ต๋๋ค. ์ด ํญ๊ณต๊ธฐ๋ค์ ๋นํฌ์ฅ ํ์ฃผ๋ก์๋ ์ฐฉ๋ฅํ ์ ์๋๋ก ํน๋ณํ ์ ์๋์์ต๋๋ค. ์์ก๊ธฐ ์์๋ ์ฌ๋ฌ ๋์ ์ํ ํฌ๋ฆฌ์ฝฅํฐ๊ฐ ์ค๋ ค ์์์ต๋๋ค. WSO๊ฐ 7,000ํผํธ ์ฐ ์ ์์ ์์๊ธฐ ๋๋ฌธ์, ํน์ ํฌ๋ฆฌ์ฝฅํฐ๋ง์ด ๊ทธ๊ณณ์ ์ ๊ทผํ ์ ์์์ต๋๋ค. ํฌ๋ฆฌ์ฝฅํฐ๋ค์ ์ฐ์ ์ค๋ฅด๊ธฐ ์์ํ๊ณ , WSO๋ ๊ทธ๋ค์ ๊ธฐ๋ค๋ฆฌ๊ณ ์์์ต๋๋ค. ํฌ๋ฆฌ์ฝฅํฐ๊ฐ ์ฐฉ๋ฅํ์๋ง์ ์ด๊ฒฉ์ด ์ค๊ฐ๋ฉฐ ์ฃผ๋ณ์ ํต์ ํ๊ณ , WSO๋ฅผ ํ์ ์ต๋๋ค. 24์๊ฐ ๋๊ฒ ์ด์ด์ง ๋๋ผ๋ง๊ฐ ๋๋์ด ๋๋๋ ๋ฏํ์ต๋๋ค.
ํ์ง๋ง ์ฌ๊ธฐ์ ์์์น ๋ชปํ ์ฌ๊ฐํ ๋ฌธ์ ๊ฐ ๋ฐ์ํ์ต๋๋ค. C-130 ์์ก๊ธฐ๊ฐ ๋ค์ ์ด๋ฅํ๋ ค๋ ์๊ฐ, ์๋ฐํด ์ฐฉ๋ฅ ์ฅ์น(nose gear)๊ฐ ๋ชจ๋์ ์งํ์ด ์์ธ ์งํ์ ๋ฐํ ๊ผผ์งํ ์ ์๊ฒ ๋ ๊ฒ์ ๋๋ค. ๋ชจ๋ ํฌ๋ฆฌ์ฝฅํฐ๋ฅผ ๋ค์ ์ฃ๊ณ ์ด๋ฅ ์ค๋น๋ฅผ ๋ง์ณค์ง๋ง, ํญ๊ณต๊ธฐ๋ค์ ๋ง ๊ทธ๋๋ก ์งํํ์ ๊ฐํ๋ฒ๋ ธ์ต๋๋ค.
๋ฏธ๊ตฐ์ โํ๋ B(Plan B)โ๋ฅผ ์คํํด์ผ ํ์ต๋๋ค. ์ด๋ ์๋ก์ด ํญ๊ณต๊ธฐ ์ธ ๋๋ฅผ ๋ณด๋ด ํน์๋ถ๋์๋ค๊ณผ WSO๋ฅผ ๊ตฌ์ถํ๋ ๊ฒ์ ์๋ฏธํ์ต๋๋ค. ์๋ก์ด ํญ๊ณต๊ธฐ๋ค์ด ๋์ฐฉํ ๋๊น์ง ๋ฏธ๊ตฐ์ ์๊ฐ์ ๋ฒ๊ธฐ ์ํด, ๊ณ ๊ณ ๋ ํญ๊ฒฉ๊ธฐ(high-altitude bombers)๋ฅผ ์ด์ฉํด ๊ธฐ์ง ์ฃผ๋ณ์ โ๋ถ์ ๊ณ ๋ฆฌ(ring of fire)โ๋ฅผ ๋ง๋ค์ด ์ ๊ทผํ๋ ๋ชจ๋ ๊ฒ์ ํญ๊ฒฉํ์ต๋๋ค.
๋ง์นจ๋ด ๊ต์ฒด ํญ๊ณต๊ธฐ๋ค์ด ๋์ฐฉํ๊ณ , ํน์๋ถ๋์๋ค๊ณผ WSO๋ฅผ ํ์ ์ด๋ฅํ์ต๋๋ค. ๋ง์ง๋ง์ผ๋ก ๋ฏธ๊ตฐ์ ์ด๋์ ์์ ๋ฏผ๊ฐํ ๊ธฐ์ ์ด ๋์ด๊ฐ์ง ์๋๋ก, ๋ฒ๋ ค์ง C-130 ์์ก๊ธฐ์ ํฌ๋ฆฌ์ฝฅํฐ๋ค์ ํญ๊ฒฉํ์ฌ ํ๊ดดํ์ต๋๋ค. ์์ด๋ฌ๋ํ๊ฒ๋, ๊ตฌ์ถ ์์ ์ ๋ง์ง๋ง ๋จ๊ณ๋ ๊ตฌ์ถ์ ์ฌ์ฉ๋ ํญ๊ณต๊ธฐ๋ค์ ํ๊ดดํ๋ ๊ฒ์ด์์ต๋๋ค.
์น๋ฆฌ ์ ์ธ, ๊ทธ๋ฌ๋ ์๊ฐ๋ฆฐ ํ๊ฐ
์ธ ๋์ ํญ๊ณต๊ธฐ๋ ๋ฌด์ฌํ ์ด๋ฅํ์ฌ ์ฟ ์จ์ดํธ๋ก ํฅํ์ต๋๋ค. WSO๋ ๊ธฐ๋ด์์ ์น๋ฃ๋ฅผ ๋ฐ์๊ณ , ์กฐ์ข ์ฌ์ ํจ๊ป ๋ ์ผ์ ์๋ฃ ๋ณ์์ผ๋ก ์ด์ก๋ ์์ ์ด์์ต๋๋ค. ์ด ์ฑ๊ณต์ ์ธ ๊ตฌ์ถ ์์ ์ ํธ๋ผํ ๋ํต๋ น์๊ฒ ์์ฒญ๋ ์ ์น์ ์น๋ฆฌ(huge victory)๋ฅผ ์๊ฒจ์ฃผ์๊ณ , ๊ทธ๋ ๋ฐฑ์ ๊ด ๊ธฐ์ํ๊ฒฌ์์ ์ด๋ฅผ ๋๋์ ์ผ๋ก ์ ์ ํ์ต๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์ด ์์ ์ โ๊ตฐ ์ญ์ฌ์ ๊ฐ์ฅ ํฌ๊ณ , ๊ฐ์ฅ ๋ณต์กํ๋ฉฐ, ๊ฐ์ฅ ๋์ฐํ๋ ์ ํฌ ์์ ๋ฐ ๊ตฌ์กฐ ์๋ฌดโ๋ผ๊ณ ์นญ์กํ๋ฉฐ, ๋ฏธ๊ตฐ๊ณผ CIA์ ๊ธฐ์ ๊ณผ ์ ๋ฐํจ, ์น๋ช ์ ์ธ ๋ฌด๋ ฅ์ ๋ํด ์ฐฌ์ฌ๋ฅผ ๋ณด๋์ต๋๋ค. ๊ทธ๋ ๋ฏธ๊ตญ ๊ตฐ๋๋ง์ด ์ด๋ฐ ๋๋ผ์ด ์์ ์ ์ฑ๊ณต์ํฌ ์ ์์๋ค๊ณ ๊ฐ์กฐํ์ต๋๋ค.
ํ์ง๋ง ์๋ฆญ ์๋ฏธํธ ๊ธฐ์๋ ์ด๋ฌํ ์น๋ฆฌ์๋ ๋ถ๊ตฌํ๊ณ ์ ์์ ์ฌ์ ํ ์ธ๊ธฐ๊ฐ ์์ผ๋ฉฐ, ์ ์ธ๊ณ์ ๋ฏธ๊ตญ์ ๊ฒฝ์ ์ ํ์ฅ์ ์ผ์ผํค๊ณ ์๋ค๊ณ ์ง์ ํฉ๋๋ค. ๋ํ ๋ํต๋ น์ ๋ง์ ์ ๋ต์ ๋ชฉํ๊ฐ ์์ง ๋ฌ์ฑ๋์ง ์์๋ค๊ณ ๋ง๋ถ์ ๋๋ค.
์ด๋ ์ญ์ ๋ค๋ฅธ ์ด์ ๋ก ์ด๋ฒ ์ฌ๊ฑด์ ํฐ ์น๋ฆฌ๋ก ์ ์ ํ์ต๋๋ค. ๊ทธ๋ค์ ๋ฏธ๊ตญ ์ ํฌ๊ธฐ๋ฅผ ๊ฒฉ์ถํ๋ ๋ฐ ์ฑ๊ณตํ์ผ๋ฉฐ, ์ด๋ ๋ฏธ๊ตญ์ด ์ด๋ ์๊ณต์ ์์ ํ ํต์ ํ์ง ๋ชปํ๋ค๋ ๊ฒ์ ์ฆ๋ช
ํ๋ค๊ณ ์ฃผ์ฅํ์ต๋๋ค. ์ด๋์ ์ฌ์ ํ ๋งค์ผ 1530๋ฐ์ ํ๋ ๋ฏธ์ฌ์ผ(ballistic missiles)๊ณผ 50100๋์ ์ผ๋ฐฉํฅ ๊ณต๊ฒฉ ๋๋ก (one-way attack drones)์ ๋ฐ์ฌํ ์ ์๋ ์๋นํ ๋ฅ๋ ฅ์ ๋ณด์ ํ๊ณ ์์์ ๋ณด์ฌ์ฃผ์์ต๋๋ค. ์ด๋ ์ด๋์ด ์ฌ์ ํ ์ธ์ธ ์์ง์ ๋ฅ๋ ฅ์ ๊ฐ์ง๊ณ ์์์ ์๋ฏธํฉ๋๋ค.
์ ์์ ์ถ๊ตฌ ์ ๋ต๊ณผ ๋ถํ์คํ ๋ฏธ๋
์ด๋ฌํ ์ํฉ์ ์ ์์ ์ถ๊ตฌ ์ ๋ต(off-ramp)์ ๋ํ ์๋ฌธ์ ์ ๊ธฐํฉ๋๋ค. ์ ์ 6์ฃผ์ฐจ์ ๋ง์ ์ฌ๋๋ค์ด ๋ฏธ๊ตญ์ด ์ด๋์ ๊ตฐ์ฌ๋ ฅ์ ์๋น ๋ถ๋ถ ํ๊ดดํ๋ค๊ณ ์๊ฐํ์ง๋ง, ํ์ค์ ๋ฌ๋์ต๋๋ค. ๋ฏธ๊ตญ๊ณผ ์ด์ค๋ผ์์ ์ด๋์ ๋ฏธ์ฌ์ผ ํ๋ก๊ทธ๋จ, ๋๋ก ํ๋ก๊ทธ๋จ, ํด๊ตฐ, ๊ณต๊ตฐ์ ์๋น ๋ถ๋ถ ํ๊ดดํ๊ฑฐ๋ ์ฝํ์ํค๋ ๋ฐ ์ฑ๊ณตํ์ต๋๋ค. ๊ทธ๋ฌ๋ ์ด๋ ๊ตฐ๋๋ ์ฌ์ ํ ํ๋ ฅ์ ์ด๋ฉฐ ๋ฏธ์ฌ์ผ๊ณผ ๋๋ก ์ ๋ฐ์ฌํ ์ ์๋ ๋ฅ๋ ฅ์ ์ ์งํ๊ณ ์์ต๋๋ค.
์ ์น์ ๋ชฉํ๋ ์ฌ์ ํ ๋ถํ์คํฉ๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์์ ์ ์ต์ข ๋ชฉํ์ ๋ํด ๋ชจ์๋ ์ค๋ช ์ ๋ด๋๊ณ ์์ผ๋ฉฐ, ๊ตฌ์ถ ์์ ์ ์ฑ๊ณต์ ์ ์ ํ๋ฉด์๋ ์ด๋์ ๋ํด ๊ฒฉ๋ ฌํ๊ฒ ๋น๋ํ๋ฉฐ ์ํํ์ต๋๋ค. ๊ทธ๋ ์ด๋์ด ํ์์ผ ๋ฐค๊น์ง ํด์ ์ ์์ ๋ฐ์๋ค์ด์ง ์์ผ๋ฉด, ๊ตญ๊ฐ ์ ์ฒด๋ฅผ ํ๋ฃป๋ฐค ๋ง์ ํ๊ดดํ ์ ์๋ค๊ณ ํ๋ฐํ๋ฉฐ ๊ต๋, ๋ฐ์ ์ ๋ฑ ์ธํ๋ผ(infrastructure) ๊ณต๊ฒฉ์ ๊ฒฝ๊ณ ํ์ต๋๋ค.
์ผ๋ถ ์ ๋ฌธ๊ฐ๋ค์ ์ด๋ฐ ์ํ์ด ์คํ๋ ค ํธ๋ผํ ๋ํต๋ น์๊ฒ ๊ตฌ์ถ ์์ ์ ์ฑ๊ณต์ด ์์ ๊ฐ์ ๋ถ์ด๋ฃ์ด ๋ ๋๋ดํด์ง๊ฒ ๋ง๋ค์๋ค๊ณ ๋ถ์ํฉ๋๋ค. ๊ทธ์ ์ํ์ ๊ตฐ์ฌ์ ๋ชฉํ๋ฅผ ๋์ด ์๋ฐฑ๋ง ๋ฏผ๊ฐ์ธ์๊ฒ ์ํฅ์ ๋ฏธ์น ์ธํ๋ผ๋ฅผ ๊ฒจ๋ฅํ๊ณ ์์ผ๋ฉฐ, ๋ง์ ๋ฒ๋ฅ ์ ๋ฌธ๊ฐ๋ค์ ์ด๋ฌํ ๊ณต๊ฒฉ์ด ๊ตญ์ ๋ฒ์ ์๋ฐฐ๋ ์ ์๋ค๊ณ ๊ฒฝ๊ณ ํฉ๋๋ค.
๊ตฌ์ถ ์์ ์ด์ ์๋ ์ ์์ ์ถ๊ตฌ ์ ๋ต์ ๋ํ ๋ ผ์๊ฐ ๋ง์์ต๋๋ค. ๋ฏธ๊ตญ๊ณผ ์ด๋ ๋ชจ๋ ํ์ฌ ์น๋ฆฌ๋ฅผ ์ฃผ์ฅํ๊ณ ์์ผ๋, ์ด๊ฒ์ด ํน์ ํ์์ผ๋ก ๊ฐ๋ ๊ธธ์ ์๋๊น์? ํธ๋ผํ ๋ํต๋ น์ ์ฌ์ ํ ์ธ๊ต์ ํด๊ฒฐ์ ์ํ๋ค๊ณ ๋งํ๊ณ ์์ผ๋ฉฐ, ์ด๋์ ์์์ผ ํธ๋ฅด๋ฌด์ฆ ํดํ์ ์์ ํตํ ํ๋กํ ์ฝ(protocol)๊ณผ ์ด๋์ ๋ํ ์ ์ฌ ํด์ (lifting of sanctions)๋ฅผ ํฌํจํ๋ ์๋ก์ด 10๊ฐํญ ์ ์์ ์ ์ถํ์ต๋๋ค. ํ์ง๋ง ์ ๋ฌธ๊ฐ๋ค์ ํ์์ด ์์ง ๋ฉ์๋ค๊ณ ๋ณด๊ณ ์์ต๋๋ค.
ํ์ฌ ํธ๋ผํ ๋ํต๋ น์ ์ด๋์ด ํด์ ์ ์์ ๋ฐ์๋ค์ด์ง ์์ผ๋ฉด ์ธํ๋ผ์ ๋ํ ๊ณต๊ฒฉ์ ์ง๋ฉดํ ๊ฒ์ด๋ผ๋ ์ตํํต์ฒฉ์ ๋ณด๋ธ ์ํ์ ๋๋ค. ํฌ๋ง์ ์ผ๋ก ๋ณธ๋ค๋ฉด ์์ธก์ด ๋ชจ์ข ์ ํด๊ฒฐ์ฑ ์ ์ฐพ์ ์๋ ์๊ฒ ์ง๋ง, ์์ง ํต์ฌ ์์ ๋ค์ ๋ํ ์ด๊ฒฌ์ด ํฌ๊ณ , ๋ํต๋ น์ ์คํ๋ ค ํ์ (escalation) ๋ฐฉํฅ์ผ๋ก ๋์๊ฐ๋ ค๋ ๊ฒ์ฒ๋ผ ๋ณด์ ๋๋ค.
์์ ์์ฌ์ ๋ํนํ ํ์ค
์ด ๊ตฌ์ถ ์์ ์ ํ ๋ฆฌ์ฐ๋ ์ํ์์๋ ๋์ฌ ๋ฒํ ์์ ์์ฌ(heroic narrative)๋ก ๊ฐ๋ํ์ต๋๋ค. ๊ทธ๋ฆฌ๊ณ ์ด ์ด์ผ๊ธฐ๋ ํ์ ๋ถ๊ฐ ๋ฏธ๊ตญ ๋์ค์๊ฒ ์ ์์ ์ค๋ํ๋ ๋ฐ ์ด๋ ค์์ ๊ฒช๊ณ ์๋ ์๊ธฐ์ ๋์์ต๋๋ค. ๋์ค์ ์น์๋ ์ ๊ฐ์ ์ ์์ ๋ช ๋ถ์ ์๋ฌธ์ ์ ๊ธฐํ๋ฉฐ ํผ๋ก๊ฐ์ ๋๋ผ๊ณ ์์์ต๋๋ค. ๋ฐฑ์ ๊ด์ ๋ฏธ๊ตญ์ ๊ตฐ์ฌ๋ ฅ, ๋ ์ฐฝ์ฑ, ์ฉ๊ธฐ๋ฅผ ๋ณด์ฌ์ฃผ๋ ์ด ๋ฏฟ์ ์ ์์ ์ ๋๋ก ์์ ์ ์ธ ์ด์ผ๊ธฐ๋ฅผ ์ต๋ํ ํ์ฉํ๋ ค ํ ๊ฒ์ ๋๋ค.
ํ์ง๋ง ์ด ์์ ์ ์ฑ๊ณต์ด ์๋ฌด๋ฆฌ ์ํ ๊ฐ๊ณ ๋๋ผ๋งํฑํ๋๋ผ๋, ๋จ๊ฒจ์ง ํ์ค์ ๋ํนํฉ๋๋ค. ํธ๋ฅด๋ฌด์ฆ ํดํ์ ์ฌ์ ํ ๋ด์๋์ด ๊ฒฝ์ ์ ์ด๋ ค์์ด ์ง์๋๊ณ ์๊ณ , ์ด๋ ๊ฐ๊ฒฝ ์ ๋ถ๋ ์ค์ค๋ก๊ฐ ์น๋ฆฌํ๊ณ ์๋ค๊ณ ๋ฏฟ์ผ๋ฉฐ, ์ด์คํํ ์งํ์ ๋ฌปํ ์์ฒ ํ์ด๋์ ๊ณ ๋์ถ ์ฐ๋ผ๋(highly-enriched uranium) ๋ฌธ์ ๋ ํด๊ฒฐ๋์ง ์์์ต๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์ด๋์ ํต๋ฌด๊ธฐ ํ๋์ ๋ง๋ ๊ฒ์ด ์ฌ์ ํ ๊ทธ์ ์ฃผ๋ ๋ชฉํ๋ผ๊ณ ๋งํ์ง๋ง, ์ด๋ฅผ ๋ฌ์ฑํ๊ธฐ ์ํ ๊ตฌ์ฒด์ ์ธ ๋ชฉํ๋ ์ ์ํ์ง ์์์ต๋๋ค.
๊ฒฐ๋ก ์ ์ผ๋ก, ์ด๋ ์ ์ง์์์ ์์ด๋งจ ๊ตฌ์ถ ์์ ์ ์ ์ ์ ์ธก๋ฉด์์ ๋๋ผ์ด ์ฑ๊ณต์ ๊ฑฐ๋์์ง๋ง, ๋ฏธ๊ตญ ํ์ ๋ถ๋ ์ฌ์ ํ ํด๊ฒฐํด์ผ ํ ๋ง์ ๋์ ๋ค์ ์๊ณ ์์ต๋๋ค. ์ด ์์ ์ ์ธ ์ด์ผ๊ธฐ๊ฐ ์ ์์ ๋ณธ์ง์ ์ธ ๋ฌธ์ ๋ค์ ๊ฐ๋ฆด ์ ์์์ง๋ ๋ฏธ์ง์์ ๋๋ค.