English
Based on โClaude Code Can Be Your Second Brainโ from Every Watch the original video
The AI Amplifier: Inside the Ultimate Second Brain for Deep Work
In an age where digital tools promise productivity but often deliver distraction, Noah Brier has engineered a system that transcends the typical. Co-founder of successful startups like Percolate and Variance, and now leading AI strategy consultancy Alfik, Brier has cultivated a unique approach to harnessing artificial intelligence. His setup, which he affectionately refers to as a โheavy duty Claude Code setupโ rather than just for coding, transforms an Obsidian note-taking vault into a true second brain โ a relentless thinking partner, researcher, and even a logkeeper of his evolving ideas, all accessible from anywhere.
This isnโt just about using AI for quick answers or content generation; itโs about deep, sustained intellectual work, reimagined. Brierโs innovation lies in his ability to leverage AI not as a replacement for human thought, but as a profound amplifier, enabling him to think, research, write, and even ship code from the most unexpected of places: his phone.
The Second Brain Unleashed: Noahโs Obsidian-Claude Code Nexus
At the heart of Brierโs system is Obsidian, a popular note-taking application that stores notes as plain markdown files in a userโs local file system. This choice is deliberate, offering flexibility and ownership that traditional cloud-based note apps often lack. โOne of the big advantages with Obsidian as a note-taking platform is that itโs a bunch of markdown files and a bunch of folders,โ Brier explains, โand they can then be synced with Git and you can do lots of other fun kinds of things.โ
On top of this Obsidian vault, Brier runs Claude Code, Anthropicโs AI assistant designed for coding and text processing. But Brierโs primary use isnโt for writing code; itโs for interacting with his vast archive of notes. His most recent โobsession,โ he shares, has been standing up a home server in his basement to host this entire setup, making his Claude Code-powered Obsidian vault accessible and fully functional on his phone. This seemingly simple step has been transformative, turning his phone from a device primarily for consumption into a powerful tool for deep intellectual engagement.
โI found whether itโs like Claude Code and Obsidian orโฆ even Claude Code and code right,โ Brier notes, โbeing able to sign in on your phone and have Claude Code push a small update to something because you just realize it while youโre out is amazing.โ This capability extends the reach of his second brain, allowing him to seamlessly transition between desktop and mobile environments for complex tasks.
Beyond Generative: The Power of AIโs โReadingโ
A core philosophy underpinning Brierโs workflow is the distinction between โthinking modeโ and โwriting mode.โ He observes that most AI models, often labeled โgenerative,โ are overly eager to produce artifacts โ outlines, drafts, or completed pieces. This, he argues, misses the AIโs most powerful, yet often overlooked, capability: its ability to โread.โ
โI think partially because we call it generative, thereโs entirely too much focus on its ability to write and not enough focus on its ability to read,โ Brier states. โIts ability to read is incredible, right? And I think, arguably, sort of like much more useful on a day-to-day basis. Like we produce artifacts far less frequently than we just like think about things.โ
To enforce this distinction, Brier employs a specific โthinking partnerโ sub-agent within Claude Code. This agent is meticulously prompted with instructions like: โYour role is to facilitate thinkingโฆ donโt try to write the thing.โ He even includes explicit directives: โI do not under any circumstances want you to try to write it. Take this literally. Do not create outlines, drafts, or any versions of talks/writing. Only gather and organize the requested materials.โ This ensures the AI acts as a Socratic guide, asking sharp questions and helping him explore complex problems, rather than prematurely drafting solutions.
When embarking on a new project, such as preparing for a conference talk, Brierโs process is meticulous:
- Project Initialization: He creates a new folder (following the PARA method for organization) and starts Claude Code in his root Obsidian directory, giving it access to his entire vault.
- Contextualization: He feeds the AI past talks and general ideas to establish style and scope.
- Vault-Wide Research: Crucially, he instructs Claude Code to search his entire 1500+ note archive for existing research relevant to the new project and pull it into a dedicated โresearchโ folder. This jumpstarts the process with his own prior knowledge.
- External Chats: Transcripts from conversations with other AI models (ChatGPT, Grok, Claude) are also clipped and stored in a โchatsโ folder, becoming part of the research corpus.
- Daily Progress Log: At the end of each day, Brier has Claude Code review all new notes and chats, then summarize what heโs learned and the best ideas that emerged, helping to push the project forward.
Perhaps one of the most powerful features for maintaining deep work is the โcatch me upโ command. After interruptions, Brier can simply ask Claude Code, โCan you catch me up on the last three days of research?โ The AI then sifts through all new material in the project folder, providing a concise summary of progress and insights. โItโs pretty amazing also to just be able to kind of revisit deep work like this,โ Brier marvels, acknowledging that picking up work after a break is often the hardest part.
The Voice of Genius: Grok in the Car and Beyond
While Claude Code powers his Obsidian vault, Brier also champions Grokโs voice mode as a superior tool for on-the-go deep thinking. He singles out Grok 2, 3, or 4 for its intelligence and โtool callingโ capabilities โ its ability to integrate with other functions or search the web effectively โ which he finds significantly better than competitors like ChatGPT or Gemini.
โI will fight anybody who says anything different,โ he declares, adding, โI just found Grokโs voice mode to be significantly smarter than anybody elseโs.โ He describes it as โa podcast made specifically for you about whatever youโre curious about.โ
Brier shares several anecdotes illustrating Grokโs utility:
- A 5-hour drive to New Hampshire for summer camp drop-off turned into a 2-hour research session, working through a complex piece just by connecting his phone to Bluetooth.
- A โmind-blowing sessionโ on self-attention mechanisms, where Grok delivered the best explanation heโd ever heard.
- Using Grok in his Tesla (where itโs now integrated) to research Walter Benjaminโs โThe Work of Art in the Age of Mechanical Reproduction,โ exploring themes of elitist critiques against new technologies.
This ability to engage in profound intellectual exploration while commuting or driving represents a paradigm shift. โItโs just Iโฆ and you know obviously all the sort of other ChatGPT and Claude and all these things of just being able to sort of like go and do research and really think and and explore things in this device thatโs always been useful but like not useful for deep work,โ he explains. โI feel like it it itโs really changed my ability to do that.โ
AI as Bureaucracyโs Bane: The โThomasโs English Muffinโ Theory
Brierโs current talk preparation, a detailed example he walks us through, ties together historical insights with his vision for AIโs impact on organizations. The talk explores the OSSโs (Office of Strategic Services, precursor to the CIA) โSimple Sabotage Field Manualโ โ a guide for citizen saboteurs during WWII. He highlights the manualโs recommendations for white-collar workers: โalways refer things to committee,โ โalways revisit previously made decisions,โ โdonโt act with too much haste.โ
Brier connects this historical bureaucracy to modern organizational challenges, proposing that AI can โsidestep a lot of the bureaucracy that exists inside large organizations because it sort of has thisโฆ goo-like effect where it can kind of fit into any crevice or crack.โ He calls this his โThomasโs English muffin theory of AI,โ where AI gets into โthe nooks and crannies.โ
This theory posits that unlike past technologies, which often forced companies to adopt new structures or centralize tools (e.g., everyone must use Jira, not Asana), AI can act as a โfuzzy interface.โ It doesnโt care what tools or data structures teams use because โitโs all just data structures to them.โ This allows companies to maintain diverse, decentralized workflows while AI models in the middle translate and integrate information.
He draws a parallel to Wild Bill Donovan, who started the OSS, and his emphasis on โempowering individualsโ and โautonomy at the edgesโ โ a philosophy he sees echoed in AIโs potential to enable distributed, yet coordinated, work. Brier even floats a nascent idea: โbureaucracy as positional encoding,โ suggesting that bureaucracy, while often negative, was an innovation for large-scale operations. AI, through its ability to parallelize work (like transformers displacing sequential models), might offer a new way to achieve structure and hierarchy without the traditional rigidities.
This isnโt just theoretical. The host, Dan, offers a real-world example from Every, the company behind the video. With 15 people running six different products, each with its own tech stack, AI is enabling โtacit code sharing.โ Instead of abstracting common functionalities (like fast file search) into modular libraries โ a โheavy liftโ โ developers can simply use Claude Code to understand and adapt code from other product repos. โEveryone gets more productive because AI can kind of translate,โ Dan explains.
Noah Brierโs journey into building an AI-powered second brain reveals a profound shift in how we can engage with information and deepen our thinking. By treating AI as a collaborative partner rather than just a generative tool, and by leveraging its unique capabilities to read, organize, and stimulate thought across all devices, he demonstrates a future where deep work is not only more accessible but also more profound. This isnโt just about personal productivity; itโs a blueprint for how AI can fundamentally reshape our intellectual processes and organizational structures, one thought, one research query, one โnook and crannyโ at a time.
Based on โTwo Superpowers Across the Tableโ from New York Times Podcasts Watch the original video
The Uneasy Summit: Trumpโs Beijing Trip Under the Shadow of Iran
For the first time in nearly a decade, President Donald Trump was set to meet with Chinese President Xi Jinping in Beijing. But this wasnโt the triumphant return of a newly elected leader, basking in the glow of global pageantry. As Rachel Abrams from The Daily noted, the meeting on Wednesday, May 13th, of Trumpโs hypothetical second term, came at a fraught moment. Trump was struggling to extricate the United States from a war with Iran, forcing him to face off against China, arguably the biggest long-term threat to American dominance across technology, trade, and military power.
My colleague, David Sanger, reporting live from his Beijing hotel, painted a vivid picture of the high stakes and low expectations surrounding this critical summit. What would come of it, and more pressingly, what wouldnโt?
A Weakened Hand in the Dragonโs Lair
The contrast with President Trumpโs first visit to China in 2017 couldnโt have been starker. Then, he was a freshly minted president, treated to the full spectacle of Chinese hospitality. โPresident Xi, I want to thank you for that incredible welcoming ceremony,โ Trump had declared, addressing the โchronic imbalanceโ in trade and pledging to protect American intellectual property. He was just beginning to shape his China policy, focusing on the nationโs rise and envisioning a stronger bilateral relationship.
This time, however, Trump arrived under a significant cloud: the ongoing conflict with Iran. The summit, originally planned for April, had been postponed with the expectation that by mid-May, Iran would have capitulated. Instead, Iran was resisting, with other world leaders like Chancellor Mertz of Germany suggesting the United States had been humiliated. The Chinese themselves were reportedly โa bit mystifiedโ by the USโs struggles to secure the Strait of Hormuz or defeat what they perceived as a โsecond or third rate powerโ like Iran.
โAll summits are about optics,โ Sanger observed, โand while Iโm sure there will be all the pomp and ceremony that goes with these, the fact of the matter is he comes into this summit looking a bit weakened.โ This shift in perceived strength was a critical backdrop to any negotiations.
The Low-Hanging Fruit: Beef, Beans, and Boeing
With Trumpโs arrival, discussions inevitably began with trade and economic relationships. While the fundamental divides between the two superpowers loomed large, the immediate focus would likely be on what David Sanger termed โthe low-hanging fruit.โ
โThis is Donald Trump and his idea of a summit is to emerge with a bunch of business deals, even if they donโt fundamentally change the nature of the relationship,โ Sanger explained. The predictable announcements would revolve around โthe three Bs: beef, beans, and Boeing.โ These distinctly American exportsโsoybeans, aircraft, and specialty beefโwere items China often purchased anyway. They offered President Trump tangible โdeliverablesโ to present to the American people, allowing him to declare โthe biggest purchases ever,โ even if they were essentially commodities.
The true test, however, lay in the more contentious โTโ: tariffs. China had been a primary target of Trumpโs tariff regime, making it a dominant discussion point. Yet, Xi Jinping likely felt he held a strong hand. China had previously demonstrated leverage by cutting off rare earths and magnets to the US in retaliation. Moreover, recent US Supreme Court and trade court decisions had forced the repayment of some collected tariffs and challenged the presidentโs ability to impose sweeping duties.
One specific tariff issue of significant concern to China was the 100% tariff on Chinese cars entering the US, notably imposed by Joe Biden during his tenure. While China had globally exported 7 million cars last year, a million of which were destined for the US in Trumpโs first term, the American market was now largely closed. Despite Chinaโs push, a deal on this front was unlikely. The threat Chinese car production posed to American automakers mirrored the challenge Japan presented decades ago, making a significant concession politically unfeasible for the US.
These immediate trade tensions, while important, often overshadow the โmuch more fundamental dividesโ that truly define the US-China relationship.
The Core Divides: A Battle for Global Dominance
Beneath the surface of trade deals and tariff disputes lies a profound โdeath struggleโ for global dominance. President Xi Jinping has openly declared his ambition for China to be the worldโs number one military, economic, political, and even cultural power by 2049. The central question for the United States, as Sanger articulated, is โhow do you deal with a country thatโs trying to displace you as the worldโs number one power?โ
The Nuclear Arms Race
One of the most concerning areas of this competition is nuclear weapons. For decades, China maintained a โminimum nuclear deterrentโ of 100-200 weapons. However, under Xi, a massive buildup began, largely unnoticed by the US until years later. Today, China possesses approximately 600 nuclear weapons, with Pentagon estimates projecting 1,000 by 2030 and parity with the US and Russia by 2035.
This rapid expansion takes on critical significance following the expiration of the last remaining arms control treaty between Russia and the United States. President Trump, rightly, insisted that any new treaty must include China. Yet, the Chinese have firmly stated their disinterest in arms control discussions until they achieve an arsenal comparable in size to the US and Russia, refusing to negotiate from a perceived disadvantage. Trumpโs stated intention to raise nuclear arms control with Xi was therefore unlikely to yield much progress.
The Taiwan Question
Taiwan remains a perennial flashpoint. President Xiโs objective is to subtly shift US policy, making Taiwan more doubtful of American aid in the event of a Chinese takeover. This often manifests in seemingly minor โwording changesโ in diplomatic language. For instance, American officials typically state they โwould not support Taiwan declaring its own independence.โ The Chinese, however, desire a shift to โoppose,โ a word carrying immense diplomatic weight, implying US recognition of the Peopleโs Republic as the sole legitimate China and opposition to any challenge to that claim.
Despite the critical importance of Taiwanโa major producer of advanced semiconductors through Taiwan Semiconductor Corporation (TSMC), essential for the AI revolutionโPresident Trump has historically shown reluctance to fully engage on the issue. This creates a potential opening for Xi, who might offer guarantees on chip supply in exchange for a less robust US stance on the islandโs broader autonomy.
The AI Arms Race: Uncharted Territory
Perhaps the most urgent and least understood challenge is the race for artificial intelligence dominance. The current discussions between world leaders on AI are โpretty limited and pretty superficial,โ Sanger noted. Even basic agreements, like ensuring humans, not AI, control nuclear weapons, took months to negotiate during the Biden administration. The goal is to establish new forms of arms control for AI, recognizing that traditional methods of counting and inspecting physical weapons donโt apply to algorithms and code. These would largely be โcodes of conduct,โ inherently difficult to enforce.
Both the US and China are locked in a fierce competition to โwin the AI battle,โ making them hesitant to implement restrictions that might hinder their own growth. Yet, the need for guardrails is undeniable. This urgency was underscored by the recent emergence of โMythos,โ Anthropicโs advanced AI product, which was deemed so effective at conducting offensive cyberattacksโinstantly finding vulnerabilities in utility grids and infrastructureโthat it was withheld from public release. While Mythos could also be used defensively, China is undoubtedly developing similar large language models, likely just months away from their own version.
This prospect deeply worries the administration, but a significant concern among American China experts is that top officials, distracted by the Middle East conflict, are โill-preparedโ to tackle these complex AI issues.
The Shadow of Iran
The ongoing war in Iran fundamentally altered the conditions for this summit. President Trump had envisioned a swift resolution, with Iran capitulating and the Strait of Hormuz reopening. None of these outcomes materialized.
The Chinese have a profound economic interest in resolving the conflict, with over 30% of their oil and gas transiting the Strait of Hormuz. The soaring energy prices exacerbated by the war were devastating many Chinese enterprises, already struggling from a domestic economic slowdown. Trump would undoubtedly press Xi to stop supplying targeting data and technology to Iran for their missiles and interceptors. He would also urge China, as a major purchaser of Iranian goods, to use its influence to persuade Iran to reopen the strait. It remained to be seen whether the cautious President Xi, despite his strong interest in ending the war, would quietly intervene on the USโs behalf.
Defining a โWinโ in a High-Stakes Game
Given the complexity and urgency of these issues, what would constitute a โwinโ for each leader at the end of the day?
For President Trump, a win would likely be measured in tangible business deals, allowing him to return home and declare success as the โdealmaker-in-chief.โ This narrative, however, would sidestep the deeper, more fundamental questions of power and influence. It would also fail to address the perception, particularly after the Iran conflict, that the United States might not be as invincible as some leaders once believed.
For President Xi, the game is a longer one. He has little interest in individual deals beyond their utility in โgreasing the wheels of diplomacy.โ Xiโs primary objective is to demonstrate that China is the more stable, reliable power on the global stage. As he recently said in a veiled swipe at Trump, China is โnot invading other countries or following the law of the jungle.โ His long-term plan is to show that the world will increasingly depend on Chinese power and capital to rebuild and establish new trading relationships. In this context, Xi might also seek to highlight any perceived lack of a coherent, long-term strategy from the United States.
Ultimately, this summit was less about immediate breakthroughs on the thorniest issues and more about a demonstration of strength, or perceived weakness, on a global stage where two superpowers are engaged in a relentless struggle for the future. The real outcomes would likely unfold not in the press conferences, but in the quiet shifts of diplomatic language and the strategic maneuvering that continues long after the leaders have departed Beijing.
ํ๊ตญ์ด
โClaude Code Can Be Your Second Brainโ โ Every ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
ํด๋ก๋ ์ฝ๋๋ก ๊ตฌ์ถํ โ๋ชจ๋ฐ์ผ ์ 2์ ๋โ: AI ๊ธฐ๋ฐ ์ฌ๊ณ ํ์ฅ๋ฒ
๋ ธ์ ๋ธ๋ผ์ด์ด(Noah Brier)๋ AI ์ ๋ต ์ปจ์คํ ๊ธฐ์ ์ํฝ(Alfik)์ ์ด์ํ๋ฉฐ, ํด๋ก๋ ์ฝ๋(Claude Code)๋ฅผ ํ์ฉํด ์์ ๋ง์ ๋ ํนํ โ์ 2์ ๋(Second Brain)โ ์์คํ ์ ๊ตฌ์ถํ์ต๋๋ค. ๊ทธ์ ์์คํ ์ ๋จ์ํ ๋ฉ๋ชจ๋ฅผ ์ ๋ฆฌํ๋ ๊ฒ์ ๋์ด, ๊น์ด ์๋ ์ฌ๊ณ ์ ์ฐ๊ตฌ, ๊ธ์ฐ๊ธฐ, ์ฌ์ง์ด ์ฝ๋ฉ๊น์ง ์ค๋งํธํฐ ํ๋๋ก ๊ฐ๋ฅํ๊ฒ ๋ง๋๋ ํ์ ์ ์ธ ์ ๊ทผ ๋ฐฉ์์ ๋ณด์ฌ์ค๋๋ค. ์ด ๊ธฐ์ฌ๋ ๋ ธ์ ๋ธ๋ผ์ด์ด๊ฐ ์ด๋ป๊ฒ ํด๋ก๋ ์ฝ๋๋ฅผ ์ง์ ํ โ์ฌ๊ณ ํํธ๋(Thinking Partner)โ๋ก ํ์ฉํ๋ฉฐ, ๊ฐ์ธ๊ณผ ์กฐ์ง์ ์์ฐ์ฑ์ ๊ทน๋ํํ๋์ง ๊ทธ์ ํต์ฐฐ์ ๊น์ด ์๊ฒ ํ๊ตฌํฉ๋๋ค.
์๋ฒ๋ ธํธ์์ ์ต์๋์ธ, ๊ทธ๋ฆฌ๊ณ ํด๋ก๋ ์ฝ๋๋ก: ์ฌ๊ณ ๋๊ตฌ์ ์งํ
๋ ธ์ ๋ธ๋ผ์ด์ด๋ ๊ณผ๊ฑฐ ์๋ฒ๋ ธํธ(Evernote)๋ฅผ ๋ ์ฐฝ์ ์ผ๋ก ํ์ฉํ๋ฉฐ โ์ฌ๊ณ ๋๊ตฌ(Tools for Thought)โ ๋ถ์ผ์์ ์ ๊ตฌ์ ์ธ ์๋ชฉ์ ๋ณด์ฌ์คฌ์ต๋๋ค. ํ์ง๋ง ๊ทธ๋ ์๋ง์ ์ฌ๋๋ค๊ณผ ๋ง์ฐฌ๊ฐ์ง๋ก ์๋ฒ๋ ธํธ๋ฅผ ๋ ๋ ์ต์๋์ธ(Obsidian)์ผ๋ก ์ ํํ์ต๋๋ค. ์ต์๋์ธ์ ๋ชจ๋ ๋ ธํธ๊ฐ ๋งํฌ๋ค์ด(Markdown) ํ์ผ ํํ๋ก ๋ก์ปฌ์ ์ ์ฅ๋๋ฉฐ, ๊น(Git)๊ณผ ๊ฐ์ ๋ฒ์ ๊ด๋ฆฌ ์์คํ ๊ณผ ์ฝ๊ฒ ์ฐ๋๋๋ ์ฅ์ ์ด ์์ต๋๋ค.
์ด๋ฌํ ์ต์๋์ธ์ ์ ์ฐ์ฑ์ ๋ ธ์์ ํด๋ก๋ ์ฝ๋ ํ์ฉ์ ํต์ฌ์ ์ธ ๊ธฐ๋ฐ์ด ๋ฉ๋๋ค. ๊ทธ๋ ์ง ์งํ์ค์ ํ ์๋ฒ(Home Server)๋ฅผ ๊ตฌ์ถํ๊ณ , ๊ทธ ์์ ์ต์๋์ธ ๋ณผํธ(Obsidian Vault, ๋ ธํธ ์ ์ฅ์)๋ฅผ ๋ฃ์ด๋์์ต๋๋ค. ๊ทธ๋ฆฌ๊ณ ์ด ์ต์๋์ธ ์์ ํด๋ก๋ ์ฝ๋๋ฅผ ์คํ์์ผ, ์ธ์ ์ด๋์๋ ์์ ์ ๋ชจ๋ ๋ ธํธ์ ์ ๊ทผํ๊ณ AI์ ์ํธ์์ฉํ ์ ์๋ ํ๊ฒฝ์ ๋ง๋ค์์ต๋๋ค. ํด๋ก๋ ์ฝ๋์ ๊ฐ์ฅ ์ค์ํ ์ฉ๋๋ ๋ฐ๋ก ์ด ๋ ธํธ๋ค๊ณผ์ ์ํธ์์ฉ์ ์์ต๋๋ค. ์ด๋ฅผ ํตํด ๋ ธ์๋ ์ค๋งํธํฐ๋ง์ผ๋ก๋ ์๊ฐํ๊ณ , ์ฐ๊ตฌํ๊ณ , ๊ธ์ ์ฐ๊ณ , ์ฝ๋๋ฅผ ๋ฐฐํฌํ๋ ๋ฑ ์ฌ์ธต์ ์ธ ์์ ์ ์ํํฉ๋๋ค.
์์์ โ์๊ฐํ๋ ํํธ๋โ: AI ๊ธฐ๋ฐ ๋ชจ๋ฐ์ผ ์์ฐ์ฑ์ ์ฌ์ ์
๋ ธ์๋ AI๊ฐ ๋ชจ๋ฐ์ผ ํ๊ฒฝ์์์ ์์ฐ์ฑ์ ํ์ ์ ์ผ๋ก ๋ณํ์ํค๊ณ ์๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค. ๊ณผ๊ฑฐ ์ค๋งํธํฐ์ ๊ธ์ฐ๊ธฐ๋ ์ฝ๋ฉ ๊ฐ์ โ์ฌ์ธต ์์ (Deep Work)โ์๋ ์ ํฉํ์ง ์์ ๊ธฐ๊ธฐ๋ก ์ฌ๊ฒจ์ก์ต๋๋ค. ํ์ง๋ง AI๋ ์ด๋ฌํ ํ๊ณ๋ฅผ ๊ทน๋ณตํ๊ณ ํด๋ํฐ์ ์ง์ ํ โ์์ธต ์์ ๋๊ตฌโ๋ก ํ๋ฐ๊ฟ์์ผฐ์ต๋๋ค.
ํนํ ๊ทธ๋ ๊ทธ๋ก ์์ฑ ๋ชจ๋(Grock Voice Mode)์ ํ์ํจ์ ๊น์ ์ธ์์ ๋ฐ์์ต๋๋ค. ๊ทธ๋ ๊ทธ๋ก์ ์์ฑ ๋ชจ๋๊ฐ ๋ค๋ฅธ AI ๋ชจ๋ธ(์ฑGPT, ์ ๋ฏธ๋์ด ๋ฑ)์ ์์ฑ ๋ชจ๋๋ณด๋ค โํจ์ฌ ๋ฐ์ด๋๊ณ , ํจ์ฌ ์ค๋งํธํ๋ฉฐ, ๋๊ตฌ ํธ์ถ(Tool Calling)๊ณผ ์ฐ๊ตฌ ์ํ ๋ฅ๋ ฅ์ด ์๋ฑํ๋คโ๊ณ ๋จ์ธํฉ๋๋ค. ํ ์ฌ๋ผ(Tesla) ์ฐจ๋์ ๊ทธ๋ก์ด ๋ด์ฅ๋ ์ดํ ๋์ฑ ํธ๋ฆฌํ๊ฒ ์ฌ์ฉํ๊ณ ์์ผ๋ฉฐ, ์ฌ์ง์ด ํ ์ฌ๋ผ์ ํ์ฌ๋๊ธฐ ์ ์๋ 5์๊ฐ ์ด์ ์ค 2์๊ฐ ๋์ ๊ทธ๋ก ์์ฑ ๋ชจ๋๋ฅผ ํตํด ๊ธ์ ์ฃผ์ ๋ฅผ ์ฐ๊ตฌํ๊ณ ์์ด๋์ด๋ฅผ ๋ฐ์ ์์ผฐ๋ ๊ฒฝํ์ ๊ณต์ ํฉ๋๋ค.
โ๋ง์น ๋๋ง์ ์ํ ํ์บ์คํธ๋ฅผ ๋ฃ๋ ๊ฒ ๊ฐ์์. ๊ถ๊ธํ ๋ชจ๋ ๊ฒ์ ๋ํด ๋ง์ถคํ ์ค๋ช ์ ์ ๊ณตํด์ฃผ์ฃ .โ ๋ ธ์๋ ๊ทธ๋ก ์์ฑ ๋ชจ๋๊ฐ ๋ณต์กํ ๊ฐ๋ (์: ์ ํ ์ดํ ์ (Self-Attention))์ ์ดํดํ๋ ๋ฐ ์ต๊ณ ์ ์ค๋ช ์ ์ ๊ณตํ๋ค๊ณ ๊ทน์ฐฌํฉ๋๋ค. ๊ทธ๋ ์ด์ ์ค์๋ ๋ฐํฐ ๋ฒค์ผ๋ฏผ(Walter Benjamin)์ โ์ด๋ฏธ์ง์ ๋๋ ์์ฐ(Mass Production of Images)โ์ ๋ํ ์์ด๋์ด๋ฅผ ํ์ํ๊ณ , ๊ทธ์ ๋์๋ ์ฌ์๊ฐ๋ค์ ์ฐ๊ตฌํ๋ ๋ฑ ๊น์ด ์๋ ์ฌ๊ณ ์ ์ฐ๊ตฌ๋ฅผ ๋์์์ด ์ด์ด๊ฐ ์ ์๊ฒ ๋์์ต๋๋ค.
โ์๊ฐ ๋ชจ๋โ์ โ์์ฑ ๋ชจ๋โ์ ๋ถ๋ฆฌ: AI๋ฅผ ํ์ฉํ ์ฌ์ธต ์ฌ๊ณ ๋ฒ
๋ ธ์๋ AI ๋ชจ๋ธ์ ํ์ฉํ ๋ ๊ฐ์ฅ ์ค์ํ ํต์ฐฐ ์ค ํ๋๋ก โ์๊ฐ ๋ชจ๋โ์ โ์์ฑ ๋ชจ๋โ๋ฅผ ๋ช ํํ ๋ถ๋ฆฌํด์ผ ํ๋ค๊ณ ๊ฐ์กฐํฉ๋๋ค. ๋ง์ ์ฌ๋๋ค์ด AI๋ฅผ โ์์ฑํ(Generative)โ์ด๋ผ๊ณ ๋ถ๋ฅด๊ธฐ ๋๋ฌธ์, AI๊ฐ ์ฆ์ โ๊ฒฐ๊ณผ๋ฌผ(Artifact)โ์ ๋ง๋ค์ด๋ด๋ ๋ฅ๋ ฅ์๋ง ์ด์ ์ ๋ง์ถฅ๋๋ค. ํ์ง๋ง ๋ ธ์๋ AI์ ์ง์ ํ ๊ฐ์น๊ฐ โ์ฝ๋(Reading)โ ๋ฅ๋ ฅ์ ์๋ค๊ณ ์ฃผ์ฅํฉ๋๋ค.
โAI์ ์ฝ๋ ๋ฅ๋ ฅ์ ๋๋๊ณ , ์ผ์์ ์ธ ์ธก๋ฉด์์ ํจ์ฌ ๋ ์ ์ฉํฉ๋๋ค. ์ฐ๋ฆฌ๋ ๊ฒฐ๊ณผ๋ฌผ์ ๋ง๋ค์ด๋ด๋ ๊ฒ๋ณด๋ค ์๊ฐํ๋ ์๊ฐ์ด ํจ์ฌ ๋ง๊ธฐ ๋๋ฌธ์ด์ฃ .โ
๊ทธ๋ AI๊ฐ ๋๋ฌด ๋นจ๋ฆฌ ๊ธ์ฐ๊ธฐ๋ ์์ฝ ๊ฐ์ ๊ฒฐ๊ณผ๋ฌผ ์์ฑ์ ๋ฐ์ด๋ค๋ ค๋ ๊ฒฝํฅ์ด ์๋ค๊ณ ์ง์ ํ๋ฉฐ, ์ด๋ฅผ ์ ์ดํ๊ธฐ ์ํด ๋งค์ฐ ๊ตฌ์ฒด์ ์ธ ์ง์๋ฅผ ๋ด๋ฆฝ๋๋ค. ์๋ฅผ ๋ค์ด, ํด๋ก๋ ์ฝ๋์๊ฒ โ์ง๊ธ์ ๊ธ์ฐ๊ธฐ๋ฅผ ๋์์ฃผ์ง ๋ง๋ผ. ์ค์ง ์๊ฐํ๊ณ ์ง๋ฌธ์ ๋์ ธ๋ผ. ๊ฐ์๋ ์ด์์ ๋ง๋ค์ง ๋ง๋ผ. ์์ฒญ๋ ์๋ฃ๋ฅผ ์์งํ๊ณ ์ ๋ฆฌํ๋ ๋ฐ๋ง ์ง์คํ๋ผโ๊ณ ๋ช ์์ ์ผ๋ก ๋ช ๋ นํฉ๋๋ค.
๊ทธ์ ์์ ํ๋ฆ์ ๋ค์๊ณผ ๊ฐ์ต๋๋ค:
- ํ๋ก์ ํธ ํด๋ ์์ฑ: ์๋ก์ด ๊ฐ์ฐ์ด๋ ๊ธ์ฐ๊ธฐ ํ๋ก์ ํธ๊ฐ ์์๋๋ฉด, ์ต์๋์ธ์ ์ ์ฉ ํด๋๋ฅผ ๋ง๋ญ๋๋ค.
- ์ด๊ธฐ ํ๋กฌํํธ: ํด๋ก๋ ์ฝ๋์๊ฒ ์์ ์ด โ์๊ฐ ๋ชจ๋โ์ ์์ผ๋ฉฐ, ๊ณผ๊ฑฐ ๊ฐ์ฐ ์๋ฃ๋ฅผ ์ ๊ณตํ์ฌ ์์ ์ ์คํ์ผ์ ์ดํด์ํค๊ณ , ๊ฐ์ฐ์ ํต์ฌ ์์ด๋์ด๋ฅผ ๊ณต์ ํฉ๋๋ค.
- ๊ธฐ์กด ์ง์ ํ์ฉ: ํด๋ก๋ ์ฝ๋์๊ฒ ์ต์๋์ธ ๋ณผํธ ์ ์ฒด(์ฝ 1,500๊ฐ์ ๋ ธํธ)๋ฅผ ๊ฒ์ํ์ฌ ํ๋ก์ ํธ์ ๊ด๋ จ๋ ๊ธฐ์กด ์ฐ๊ตฌ ์๋ฃ๋ ๋ ธํธ๋ฅผ โ์ฐ๊ตฌ(Research)โ ํด๋๋ก ๊ฐ์ ธ์ค๋๋ก ์ง์ํฉ๋๋ค. ์ด๋ AI๋ ๋จ์ํ ํค์๋ ๊ฒ์์ ๋์ด, ๋ ธ์์ ๊ธฐ์กด ์ฌ๊ณ ํ๋ฆ๊ณผ ๊ด๋ จ๋ ์๋ฃ๋ฅผ ์ฐพ์๋ด๋ ๋ฐ ๋์์ ์ค๋๋ค.
- ์ฌ๊ณ ํํธ๋ ์์ด์ ํธ ํ์ฉ: ํด๋ก๋ ์ฝ๋ ๋ด์ โ๋ณต์กํ ๋ฌธ์ ๋ฅผ ํ์ํ๋ ๋ฐ ํนํ๋ ํ๋ ฅ์ ์ฌ๊ณ ํํธ๋โ ์ญํ ์ ํ๋ ์๋ธ ์์ด์ ํธ(Sub-agent)๋ฅผ ์์ฑํฉ๋๋ค. ์ด ์์ด์ ํธ๋ ๋ ธ์์๊ฒ ๋ ์นด๋ก์ด ์ง๋ฌธ์ ๋์ง๊ณ , ๊ทธ์ ์๊ฐ์ ์ ๋ฆฌํ๋ฉฐ, ํ์ ๊ณผ์ ์์ ์ป์ ํต์ฐฐ์ ์ง์์ ์ผ๋ก ๊ธฐ๋กํฉ๋๋ค.
- ์ธ๋ถ ๋ํ ํตํฉ: ์ฑGPT, ํด๋ก๋, ๊ทธ๋ก ๋ฑ ๋ค๋ฅธ AI์ ๋๋ด๋ ๋ํ ๋ด์ฉ(์ ์ฒด ์คํฌ๋ฆฝํธ)์ ์ต์๋์ธ ๋ด โ๋ํ(Chats)โ ํด๋์ ํด๋ฆฌํํ์ฌ ํด๋ก๋ ์ฝ๋๊ฐ ์ฐธ๊ณ ํ๋๋ก ํฉ๋๋ค. ์ด๋ฅผ ํตํด AI๋ ๋ ธ์์ ๋ค์ํ ์ฌ๊ณ ๊ณผ์ ์ ์ข ํฉ์ ์ผ๋ก ์ดํดํ ์ ์์ต๋๋ค.
- ๋ฐ์ผ๋ฆฌ ์งํ ์ํฉ ๊ธฐ๋ก: ๋งค์ผ AI๊ฐ ๊ทธ๋ ์์ฑ๋ ๋ชจ๋ ๋ ธํธ๋ฅผ ๊ฒํ ํ๊ณ , ๊ฐ์ฐ ์งํ์ ๋์์ด ๋๋ ํ์ต ๋ด์ฉ๊ณผ ์์ด๋์ด๋ฅผ ์ ๋ฆฌํ์ฌ ๊ธฐ๋กํ๋๋ก ํฉ๋๋ค.
- ํ๋ฆ ๋๊น ๋ฐฉ์ง: ์ ๋ฌด ๋ฑ์ผ๋ก ์ธํด ์์ ํ๋ฆ์ด ๋๊ฒผ์ ๋, โ์ง๋ ๋ฉฐ์น ๊ฐ์ ์ฐ๊ตฌ ๋ด์ฉ์ ์์ฝํด๋ฌ๋ผโ๊ณ ์์ฒญํ๋ฉด, AI๋ ํด๋น ๊ธฐ๊ฐ ๋์์ ๋ชจ๋ ๋ ธํธ๋ฅผ ์ฝ๊ณ ์ฃผ์ ํต์ฐฐ๊ณผ ์งํ ์ํฉ์ ์ ๋ฆฌํด์ค๋๋ค. ์ด๋ ๋ค์ ์์ ์ ๋ชฐ์ ํ๋ ๋ฐ ํฐ ๋์์ ์ค๋๋ค.
์ด๋ฌํ ๊ณผ์ ์ ํตํด ๋ ธ์๋ โํธ๋์คํฌ๋จธ๊ฐ ์ธ์์ ์ง์ด์ผํค๊ณ ์๋ค(Transformers are eating the world)โ๋ ๊ทธ์ ํต์ฌ ์์ด๋์ด(ํน์ ๋ชฉ์ ์ฝ๋๋ฅผ ๋์ฒดํ๋ ํธ๋์คํฌ๋จธ ๋ชจ๋ธ, ํ ์ฌ๋ผ๊ฐ ์ ๊ฒฝ๋ง์ผ๋ก 30๋ง ์ค์ ์ฝ๋๋ฅผ ์ ๊ฑฐํ ์ฌ๋ก ๋ฑ)์ โ๊ด๋ฃ์ฃผ์(Bureaucracy)โ์ ๋ณธ์ง, ๊ทธ๋ฆฌ๊ณ OSS(์ ๋ต์ฌ๋ฌด๊ตญ) ์ค๋ฆฝ์ ์์ผ๋ ๋น ๋๋ ธ๋ฐ(Wild Bill Donovan)์ ๋ฆฌ๋์ญ ์ฒ ํ์ ์ฎ๋ ๋ณต์กํ ๊ฐ์ฐ ์์ด๋์ด๋ฅผ ๊ตฌ์ฒดํํ๊ณ ์์ต๋๋ค.
AI, ์กฐ์ง ๊ด๋ฃ์ฃผ์๋ฅผ ํ๋ฌด๋ โํ ๋ง์ค ์๊ธ๋ฆฌ์ ๋จธํโ ์ด๋ก
๋ ธ์๋ AI๊ฐ ์กฐ์ง์ ๊ด๋ฃ์ฃผ์๋ฅผ ํด๊ฒฐํ๋ ๋ฐ ์ค์ํ ์ญํ ์ ํ ๊ฒ์ด๋ผ๋ ํฅ๋ฏธ๋ก์ด ์ฒ ํ์ ์ ์ํฉ๋๋ค. ๊ทธ๋ ๊ด๋ฃ์ฃผ์๊ฐ ํ๋ ๊ธฐ์ ์ด์ ๋ฐฉ์์ ํ์ ์ด์์ผ๋ฉฐ, ๋๊ท๋ชจ ์กฐ์ง ์ด์์ ํ์ํ ๊ณ์ธต๊ณผ ๊ตฌ์กฐ๋ฅผ ์ ๊ณตํ๋ค๊ณ ์ค๋ช ํฉ๋๋ค. ๊ทธ๋ฌ๋ ํ๋์ ์์๋ ๋นํจ์จ๊ณผ ๊ฒฝ์ง์ฑ์ ์์ง์ด ๋์์ต๋๋ค.
๊ณผ๊ฑฐ์ ๊ธฐ์ ๋ค์ ์กฐ์ง์๊ฒ ๊ธฐ์กด ๊ตฌ์กฐ๋ฅผ ์ ์งํ๋ฉฐ ๊ธฐ์ ์ ํตํฉํ ์ง, ์๋๋ฉด ์๋ก์ด ๊ตฌ์กฐ๋ฅผ ์ฑํํ ์ง ๊ฒฐ์ ํ๋๋ก ๊ฐ์ํ์ต๋๋ค. ๋๋ถ๋ถ์ ๊ฒฝ์ฐ ์๋ก์ด ์ํํธ์จ์ด๋ ์๋ก์ด ๊ตฌ์กฐ๋ฅผ ์๊ตฌํ๊ณ , ์ด๋ก ์ธํด ๋ง์ ์ํํธ์จ์ด ํ๋ก์ ํธ๊ฐ ์คํจํ์ต๋๋ค. ํ์ง๋ง AI๋ ๋ค๋ฆ ๋๋ค.
๋ ธ์๋ AI๋ฅผ โํ ๋ง์ค ์๊ธ๋ฆฌ์ ๋จธํ(Thomasโs English Muffin) ์ด๋ก โ์ ๋น์ ํฉ๋๋ค. ์๊ธ๋ฆฌ์ ๋จธํ์ โ๋ํฌ ์ค ํฌ๋๋(nooks and crannies)โ์ฒ๋ผ, AI๋ ์กฐ์ง์ โํ์์ ๊ท ์ดโ ์์ผ๋ก ํ๊ณ ๋ค์ด ๊ธฐ์กด์ ๋ค์ํ ์์ ๋ฐฉ์์ ๊ทธ๋๋ก ์ ์งํ๋ฉด์๋ ์ ์ฒด๋ฅผ ์ฐ๊ฒฐํ๋ โํผ์ง ์ธํฐํ์ด์ค(Fuzzy Interface)โ ์ญํ ์ ํ ์ ์๋ค๋ ๊ฒ์ ๋๋ค.
์๋ฅผ ๋ค์ด, ํ ํ์ ์์ฌ๋(Asana)๋ฅผ, ๋ค๋ฅธ ํ์ ์ง๋ผ(Jira)๋ฅผ, ๋ ๋ค๋ฅธ ํ์ ๋ฆฌ๋์ด(Linear)๋ฅผ ์ฌ์ฉํ๋ ํ์ฌ์์ ๊ธฐ์กด์๋ ๋ชจ๋ ํ์ด ํ๋์ ๋๊ตฌ๋ก ํตํฉ๋๋๋ก ๊ฐ์๋ฐ์์ต๋๋ค. ์ด ๊ณผ์ ์์ ๋ง์ ํ์ด ๋ถ๋ง์ ๊ฐ์ก์ต๋๋ค. ๊ทธ๋ฌ๋ AI๋ ๊ฐ ํ์ด ์ ํธํ๋ ๋๊ตฌ๋ฅผ ๊ณ์ ์ฌ์ฉํ๋๋ก ํ์ฉํ๋ฉด์๋, ๋ชจ๋ ๋ฐ์ดํฐ๋ฅผ AI ๋ชจ๋ธ์ด ์ดํดํ ์ ์๋ ๊ณตํต ๊ตฌ์กฐ๋ก ๋ณํํ์ฌ ์ค์์์ ๊ด๋ฆฌํ ์ ์๊ฒ ํฉ๋๋ค. AI๋ ์ด๋ค ๋๊ตฌ๋ฅผ ์ฌ์ฉํ๋ ์๊ดํ์ง ์๊ธฐ ๋๋ฌธ์ ๋๋ค. ์ด๋ ๊ธฐ์ ์ฉ ์ํํธ์จ์ด ๋์ ์ ๊ทผ๋ณธ์ ์ธ ๋ฌธ์ ์๋ โ๋์ ๋ฐ ๋ณํ ๊ด๋ฆฌ(Adoption and Change Management)โ์ ๋ถ๋ด์ ํฌ๊ฒ ์ค์ฌ์ค ์ ์์ต๋๋ค.
์๋ธ๋ฆฌ(Every)์ฌ ๋ด๋ถ์ ์ฌ๋ก๋ ์ด๋ฌํ ๋ ธ์์ ์ด๋ก ์ ๋ท๋ฐ์นจํฉ๋๋ค. ์๋ธ๋ฆฌ๋ 6๊ฐ์ ์ ํ์ 15๋ช ์ ์ธ์์ผ๋ก ์ด์ํ๋ฉฐ, ๊ฐ ์ ํ๋ง๋ค ์๋ก ๋ค๋ฅธ ๊ธฐ์ ์คํ(๋ ์ผ์ฆ, ํ์ ์คํฌ๋ฆฝํธ ๋ฑ)์ ์ฌ์ฉํฉ๋๋ค. ์ต๊ทผ โ์คํํด(Sparkle)โ์ด๋ผ๋ ํ์ผ ๊ฒ์ ๋๊ตฌ์ โํ๋ผ(Parah)โ๋ผ๋ ๋ฒ๋ฅ ๋ณด์กฐ AI ์ ํ์ ๊ฐ๋ฐํ๋ฉด์, ๋ ์ ํ ๋ชจ๋ โ๋น ๋ฅธ ํ์ผ ๊ฒ์โ ๊ธฐ๋ฅ์ด ํ์ํ์ต๋๋ค.
๊ณผ๊ฑฐ๋ผ๋ฉด ์คํํด์ ํ์ผ ๊ฒ์ ๊ธฐ๋ฅ์ ์ถ์ํํ๊ณ ๋ชจ๋ํํ์ฌ ์ฌ์ฌ์ฉ ๊ฐ๋ฅํ ๋ผ์ด๋ธ๋ฌ๋ฆฌ๋ก ๋ง๋ค์ด์ผ ํ์ ๊ฒ์ ๋๋ค. ์ด๋ ์๋นํ ์๊ฐ๊ณผ ๋ ธ๋ ฅ์ด ๋๋ ์์ ์ ๋๋ค. ํ์ง๋ง ์๋ธ๋ฆฌ๋ ๋์ ํ๋ผ ๊ฐ๋ฐ์๋ฅผ ์คํํด ๋ ํฌ(Repo, ์ ์ฅ์)์ ์ถ๊ฐํ๊ณ , ํด๋ก๋ ์ฝ๋์๊ฒ ์คํํด์ ์๋ ๋ฐฉ์์ ํ์ ํ ํ ํ๋ผ์ ๋ง๋ ์์ฒด ๋ฒ์ ์ ๋ง๋ค๋๋ก ์ง์ํ์ต๋๋ค.
์ด์ฒ๋ผ AI๋ ์๋ก ๋ค๋ฅธ ํ๊ฒฝ๊ณผ ๋ฐฉ์์ผ๋ก ์ ์ฌํ ๋ฌธ์ ๋ฅผ ํด๊ฒฐํ๋ ์ฌ๋ฌ ๋ ํฌ ์ฌ์ด์์ โ์๋ฌต์ ์ฝ๋ ๊ณต์ (Tacit Code Sharing)โ๋ฅผ ๊ฐ๋ฅํ๊ฒ ํฉ๋๋ค. AI๊ฐ ์ฝ๋๋ฅผ ๋ฒ์ญํ๊ณ ์ดํดํจ์ผ๋ก์จ, ๊ฐ๋ฐ์๋ค์ ๋ณต์กํ ์ถ์ํ๋ ๋ชจ๋ํ ์์ ์์ด๋ ๋ค๋ฅธ ํ์ ์ง์๊ณผ ๊ฒฝํ์ ํ์ฉํ์ฌ ์์ฐ์ฑ์ ๋์ผ ์ ์๊ฒ ๋ ๊ฒ์ ๋๋ค.
๊ฒฐ๋ก : AI๋ฅผ ํตํ ์ง์ ๊ด๋ฆฌ ๋ฐ ์์ฐ์ฑ์ ๋ฏธ๋
๋ ธ์ ๋ธ๋ผ์ด์ด์ ์ฌ๋ก๋ AI๊ฐ ๋จ์ํ ๋๊ตฌ๋ฅผ ๋์ด ๊ฐ์ธ์ ์ฌ๊ณ ์ ํ์ต, ๊ทธ๋ฆฌ๊ณ ์กฐ์ง์ ์๋ ๋ฐฉ์๊น์ง ํ์ ํ ์ ์๋ ์ ์ฌ๋ ฅ์ ๋ช ํํ ๋ณด์ฌ์ค๋๋ค. ํด๋ก๋ ์ฝ๋๋ฅผ โ์ 2์ ๋โ์ด์ โ์ฌ๊ณ ํํธ๋โ๋ก ํ์ฉํ๋ ๊ทธ์ ๋ฐฉ์์ ์ฐ๋ฆฌ๊ฐ AI๋ฅผ ์ด๋ป๊ฒ ํ์ฉํด์ผ ํ๋์ง์ ๋ํ ์ค์ํ ์ฒญ์ฌ์ง์ ์ ์ํฉ๋๋ค.
ํต์ฌ์ AI์ โ์์ฑโ ๋ฅ๋ ฅ๋ณด๋ค โ์ดํดโ ๋ฅ๋ ฅ์ ์ง์คํ๊ณ , AI์๊ฒ ๋ช ํํ ์ญํ ์ ๋ถ์ฌํ์ฌ ์ฐ๋ฆฌ์ ์ฌ๊ณ ๊ณผ์ ์ ๋ณด์ํ๋๋ก ๋ง๋๋ ๊ฒ์ ๋๋ค. ๋ํ, AI๋ ์กฐ์ง ๋ด ๋ค์ํ ์์คํ ๊ณผ ์ธ๊ฐ์ ์์ ๋ฐฉ์์ ์ฐ๊ฒฐํ๋ ์ ์ฐํ ์ธํฐํ์ด์ค๊ฐ ๋์ด, ๋นํจ์จ์ ์ธ ๊ด๋ฃ์ฃผ์๋ฅผ ์ํํ๊ณ โ๋ณํ ๊ด๋ฆฌโ์ ๋ถ๋ด ์์ด ์์ฐ์ฑ์ ํฅ์์ํฌ ์ ์์ต๋๋ค.
AI๋ ๋ ์ด์ ๋ฏธ๋์ ๊ธฐ์ ์ด ์๋๋ผ, ์ง๊ธ ๋ฐ๋ก ์ฐ๋ฆฌ์ ์ง์ ๊ด๋ฆฌ์ ์์ฐ์ฑ์ ์ฌ์ ์ํ๋ ๊ฐ๋ ฅํ ๋๋ฐ์์ ๋๋ค. ๋ ธ์ ๋ธ๋ผ์ด์ด์ ๊ฒฝํ์ ์ด๋ฌํ AI ์๋๋ฅผ ์ด์๊ฐ๋ ์ฐ๋ฆฌ์๊ฒ ๊น์ ์๊ฐ๊ณผ ์ค์ฉ์ ์ธ ์ง์นจ์ ์ ๊ณตํ๊ณ ์์ต๋๋ค.
โTwo Superpowers Across the Tableโ โ New York Times Podcasts ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
๊ฒฉ๋์ ์๋, ๋ฏธ์ค ์ ์ํ๋ด: ์ด๋ ๊ฐ๋ฑ๊ณผ ํจ๊ถ ๊ฒฝ์ ์ ํธ๋ผํ์ ์์งํ์ ๋์ข
๋ด์ํ์์ฆ ํ์บ์คํธ โ๋ ๋ฐ์ผ๋ฆฌ(The Daily)โ๋ ์ต๊ทผ ๊ทผ 10๋ ๋ง์ ๋ฒ ์ด์ง์์ ์ด๋ฆฐ ๋๋๋ ํธ๋ผํ ๋ฏธ๊ตญ ๋ํต๋ น๊ณผ ์์งํ ์ค๊ตญ ๊ตญ๊ฐ์ฃผ์์ ์ ์ํ๋ด์ ์ฌ์ธต ๋ถ์ํ๋ฉฐ, ์ด ํ๋ด์ด ๋จ์ํ ์ธ๊ต์ ๋ง๋จ์ ๋์ด์ ๋ณต์กํ ์ญํ ๊ด๊ณ์ ๋ฏธ๋ ํจ๊ถ ๊ฒฝ์์ ์ ์ด์ ์์ ์กฐ๋ช ํ์ต๋๋ค. ์ด๋๊ณผ์ ๊ฐ๋ฑ์ ๋ฐ๋ชฉ ์กํ ํธ๋ผํ ๋ํต๋ น์ด ๊ณผ๊ฑฐ์๋ ํ์ฐํ ๋ฌ๋ผ์ง ์์์ผ๋ก ์ค๊ตญ์ ์ฐพ์ผ๋ฉด์, ๊ธฐ์ ๋ถํฐ ๋ฌด์ญ์ ์ด๋ฅด๊ธฐ๊น์ง ๋ชจ๋ ๋ถ์ผ์์ ๋ฏธ๊ตญ์ ์ง๋ฐฐ๋ ฅ์ ๊ฐ์ฅ ํฐ ์ํ์ด ๋๋ ์ค๊ตญ๊ณผ์ ๋๊ฒฐ์ ์ง๋ฉดํ์ต๋๋ค. ๋ด์ํ์์ฆ์ ๋ฐ์ด๋น๋ ์์ด(David Sanger) ๊ธฐ์๋ ์ด๋ฒ ํ๋ด์์ ๋ฌด์์ด ๋ ผ์๋ ๊ฐ๋ฅ์ฑ์ด ๋๊ณ , ๋ ์ค์ํ๊ฒ๋ ๋ฌด์์ด ๋ ผ์๋์ง ์์ ๊ฒ์ธ์ง์ ๋ํ ํต์ฐฐ์ ์ ๊ณตํ์ต๋๋ค.
์ด๋ ๊ฐ๋ฑ์ ๊ทธ๋ฆผ์ ์๋, ์ฝํ๋ ํธ๋ผํ์ ์ ์ฅ
2017๋ ๊ณผ ๋ฌ๋ผ์ง ์์
2017๋ , ํธ๋ผํ ๋ํต๋ น์ ์ค๊ตญ๊ณผ ์ ์ธ๊ณ๋ฅผ ๋๋ผ๊ฒ ํ๋ฉฐ ๋ํต๋ น์ ๋น์ ๋ ์งํ ๋ฒ ์ด์ง์ ๋ฐฉ๋ฌธํ์ต๋๋ค. ๋น์ ๊ทธ๋ ์ค๊ตญ์ด ๋งค์ฐ ๋ฅ์ํ๊ฒ ํผ์น๋ ๋ชจ๋ ํ๋ คํ ์์ ๊ณผ ํ์์ ๋ฐ์์ต๋๋ค. ์ ์ฃผ์์ โํ์์ ์ธ ํ์์์ ๊ฐ์ฌ๋๋ฆฐ๋คโ๋ ํธ๋ผํ ๋ํต๋ น์ ๋ง์์ ์ ์ ์๋ฏ, ์ต๊ณ ์ ์์ฐ๋ฅผ ๋ฒ ํ์์ต๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ๋น์ ์ค๊ตญ์ด ์์ ์ ๊ฒฝ์ ๋ฐ ๊ตญ๋ฐฉ ์ ์ฑ ๊ณผ ์ด๋ป๊ฒ ์ฐ๊ด๋ ์ง ๊ณ ๋ฏผํ๊ธฐ ์์ํ์ผ๋ฉฐ, ์ง์ ์ฌ์ฐ๊ถ ๋ณดํธ์ ๊ณต์ ํ ๊ฒฝ์ ํ๊ฒฝ ์กฐ์ฑ์ ๋ํ ๋ฏธ๊ตญ์ ์์ง๋ฅผ ๊ฐ์กฐํ๋ฉด์ ์ค๊ตญ์ ๋ถ์์ ์ฃผ์ ๊ฐ๋๊ตญ์ผ๋ก ์ธ์ ํ๋ ๋ถ์๊ธฐ์์ต๋๋ค.
๊ทธ๋ฌ๋ ์ด๋ฒ ๋ฐฉ๋ฌธ์ ์ํฉ์ด ์์ ํ ๋ฌ๋์ต๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ๋ณธ๋ 4์์ ์ค๊ตญ์ ๋ฐฉ๋ฌธํ ์์ ์ด์์ผ๋, ์ด๋๊ณผ์ ๊ฐ๋ฑ์ด ์ค์๊น์ง๋ ์ด๋์ ํญ๋ณต์ผ๋ก ๋๋ ๊ฒ์ด๋ผ๋ ํ์ ์๋ 6์ฃผ ์ฐ๊ธฐํ์ต๋๋ค. ํ์ง๋ง ๊ทธ์ ์์์ ๋น๋๊ฐ๊ณ , ์ด๋์ ์ฌ์ ํ ์ ํญํ๊ณ ์์ผ๋ฉฐ ๋ ์ผ์ ๋ฉ๋ฅด์ธ ์ด๋ฆฌ๋ฅผ ๋น๋กฏํ ๋ค๋ฅธ ์ธ๊ณ ์ง๋์๋ค์ ์ด๋์ด ๋ฏธ๊ตญ์ ๊ตด์์์ผฐ๋ค๊ณ ํ๊ฐํ๋ ์ํฉ์ ๋๋ค. ์ค๊ตญ ๋ด๋ถ์์๋ ๋ฏธ๊ตญ์ด ํธ๋ฅด๋ฌด์ฆ ํดํ๊ณผ ๊ฐ์ ์์ญ์ ๊ฐ๋ฐฉํ๊ฑฐ๋ ์ด๋๊ณผ ๊ฐ์ 2, 3๋ฅ ๊ตญ๊ฐ๋ฅผ ์ ์ํ๋ ๋ฐ ์ ๊ทธ๋ ๊ฒ ์ด๋ ค์์ ๊ฒช๋์ง ์์ํดํ๋ ๋ถ์๊ธฐ์ ๋๋ค. ๋ชจ๋ ์ ์ํ๋ด์ด ๊ทธ๋ ๋ฏ โ์ธ๊ด(optics)โ์ด ์ค์ํ์ง๋ง, ํธ๋ผํ ๋ํต๋ น์ ์ด๋ฒ ํ๋ด์ ๋ค์ ์ฝํ๋ ๋ชจ์ต์ผ๋ก ์ํ๊ฒ ๋์์ต๋๋ค.
ํธ๋ฅด๋ฌด์ฆ ํดํ๊ณผ ์ค๊ตญ์ ์๋ฌธ
ํธ๋ฅด๋ฌด์ฆ ํดํ ๋ฌธ์ ๋ ํนํ ์ค๊ตญ์๊ฒ ํฐ ์๋ฌธ์ผ๋ก ๋ค๊ฐ์์ต๋๋ค. ์ค๊ตญ์ ์ด๋์ด ๋ฏธ๊ตญ์ ์ ์ฌ์ ๊ตฐ์ฌ์ ์๋ ฅ์๋ ๋ถ๊ตฌํ๊ณ ์ฌ์ ํ ํธ๋ฅด๋ฌด์ฆ ํดํ์ ๋ด์ํ๋ฉฐ ์ ํญํ๊ณ ์๋ ์ํฉ์ ์ดํดํ๊ธฐ ์ด๋ ต๊ฒ ์ฌ๊น๋๋ค. ์๋ฐฉ์ ์๊ฐ์์๋ ์ด๋์ด โ2๋ฅ ๋๋ 3๋ฅโ ๊ตญ๊ฐ๋ก ๋ถ๋ฅ๋ ์ ์์์๋ ๋ถ๊ตฌํ๊ณ , ๋ฏธ๊ตญ์ด ์ด๋์ ์๋๋ก ๋ช ํํ ์ฐ์๋ฅผ ์ ํ์ง ๋ชปํ๋ ๋ชจ์ต์ ๋ฏธ๊ตญ์ ๊ตญ์ ์ ์์๊ณผ ๊ตฐ์ฌ๋ ฅ์ ๋ํ ์๊ตฌ์ฌ์ ์์๋์ต๋๋ค. ์ด๋ ํธ๋ผํ ๋ํต๋ น์ ์ธ๊ต์ ์ ์ง๋ฅผ ๋์ฑ ์ฝํ์ํค๋ ์์ธ์ผ๋ก ์์ฉํ์ต๋๋ค.
์ ์ํ๋ด์ โ์ฌ์ด ๊ณผ์ โ: 3B ๋ฌด์ญ ํ์
์๊ณ ๊ธฐ, ์ฝฉ, ๋ณด์ โ ๋ฏธ๊ตญ์ฐ ์์ถํ
ํธ๋ผํ ๋ํต๋ น์ด ์ฐธ์ฌํ๋ ์ ์ํ๋ด์ ๊ฐ์ฅ ํ์คํ ์์ ๋ ํญ์ ๋ฌด์ญ๊ณผ ๊ฒฝ์ ๊ด๊ณ์ ๋๋ค. ์ด๋ฒ ํ๋ด์์๋ ์ค๊ตญ์ ๋๊ท๋ชจ ๊ตฌ๋งค ๋ฐํ์ ์ ์ฌ์ ์ธ ์ฌ์ ๊ฑฐ๋ ๋ฐ ์ํด๊ฐ์(MOU)๊ฐ ๊ธฐ๋๋ฉ๋๋ค. ๋ฐ์ด๋น๋ ์์ด ๊ธฐ์๋ ์ด๋ฅผ โ๋ฎ๊ฒ ๋งค๋ฌ๋ฆฐ ๊ณผ์ผ(low-hanging fruit)โ์ด๋ผ๊ณ ํํํ๋ฉฐ, ์ด๋ฌํ ์๊ฑด๋ค์ด ํธ๋ผํ ๋ํต๋ น์ด ์ด๋ค ์ ์ํ๋ด์์๋ ์ถ๊ตฌํ๋ ์์ฌ์ด ์ฑ๊ณผ์์ ๊ฐ์กฐํ์ต๋๋ค.
ํนํ ์ฃผ๋ชฉ๋ฐ๋ ๊ฒ์ ์ด๋ฅธ๋ฐ โ3Bโ์ ๋๋ค: ์๊ณ ๊ธฐ(Beef), ์ฝฉ(Beans), ๋ณด์(Boeing). ์ด๋ค์ ์ค๊ตญ์ด ์ด์ฐจํผ ๊ตฌ๋งคํด์ผ ํ๊ฑฐ๋, ๋ฏธ๊ตญ์ด ์ฆ๊ฐ์ ์ธ ์ฑ๊ณผ๋ก ์ ์ํ๊ธฐ ์ข์ ๋ํ์ ์ธ ๋ฏธ๊ตญ์ฐ ์์ถํ์ ๋๋ค.
- ์ฝฉ(Soybeans): ์ค๊ตญ์ ์ค๋ซ๋์ ๋ฏธ๊ตญ์ฐ ์ฝฉ์ ๊ตฌ๋งคํด์์ง๋ง, ๊ฐ๊ฒฉ์ด ๋น์ธ๋ค๊ณ ์ฌ๊น๋๋ค.
- ๋ณด์(Boeing): ์ค๊ตญ์ ๋ณด์ ํญ๊ณต๊ธฐ์ ์์กดํด์์ง๋ง, ์ด์ ๋ ์์ฒด์ ์ผ๋ก ํ๋ฅญํ ํญ๊ณต๊ธฐ๋ฅผ ์์ฐํ๊ณ ์์ต๋๋ค.
- ์๊ณ ๊ธฐ(Beef): ์ค๊ตญ๋ ์์ฒด ์๊ณ ๊ธฐ ์ฐ์ ์ ์ก์ฑํ๊ณ ์์ง๋ง, ์ฌ์ ํ ํน์ ๋ฏธ๊ตญ์ฐ ๊ณ ๊ธ ์๊ณ ๊ธฐ๋ฅผ ๊ตฌ๋งคํ๊ณ ์์ต๋๋ค.
์ด๋ฌํ ํ๋ชฉ๋ค์ ์ค๊ตญ์ด ์ด๋ ์ ๋ ํ์๋ก ํ๋ ๊ฒ๋ค์ด๊ธฐ์, ํธ๋ผํ ๋ํต๋ น์ ์ด๋ฅผ ํตํด ๋ฏธ๊ตญ ๊ตญ๋ฏผ์๊ฒ ์ฆ๊ฐ์ ์ธ ์ฑ๊ณผ๋ฅผ ๋ณด์ฌ์ค ์ ์์ต๋๋ค. ๊ทธ๋ โ์ญ๋ ์ต๋ ๊ตฌ๋งค๋โ์ด๋ผ๊ณ ์ ์ธํ ํ ์ง๋ง, ๋ณธ์ง์ ์ผ๋ก ์ด๋ค์ ์๊ตญ์ด ๊ฐ์ฅ ์ฝ๊ฒ ๊ฑฐ๋ํ ์ ์๋ ์ํ๋ค์ ๋๋ค.
๊ด์ธ ๋ฌธ์ ์ ์ค๊ตญ ์๋์ฐจ
๋ฐ์ด๋น๋ ์์ด ๊ธฐ์๋ โ3Bโ ๋ ผ์์์ ๊ด์ธ(Tariffs)๊ฐ ์ธ๊ธ๋์ง ์์ ์ ์ ์ง์ ํ์ต๋๋ค. ๊ด์ธ๋ ์ค๊ตญ์๊ฒ ๋งค์ฐ ์ค์ํ ๋ฌธ์ ์ ๋๋ค. ์ค๊ตญ์ ํธ๋ผํ ๋ํต๋ น์ ๊ด์ธ ์ ์ฑ ์ ์ฃผ์ ํ์ ์ด์์ผ๋ฉฐ, ์ด๋ ์ด๋ฒ ํ๋ด ๋ ผ์์ ์๋น ๋ถ๋ถ์ ์ฐจ์งํ ๊ฒ์ ๋๋ค. ๊ทธ๋ฌ๋ ์ค๊ตญ์ ์ด๋ฏธ ์๋ ์ ํฌํ ๋ฅ์ ์์ ์์ถ์ ์ค๋จํ๋ ๋ณด๋ณต ์กฐ์น๋ฅผ ํตํด ๋ฏธ๊ตญ์ ์๋นํ ์๋ ฅ์ ๊ฐํ์ฌ ํธ๋ผํ ๋ํต๋ น์ ์ผ๋ถ ์ ์ฑ ์ ํํด์ํจ ๊ฒฝํ์ด ์์ต๋๋ค.
๋ํ, ๋ฏธ๊ตญ ๋๋ฒ์์ ํ๊ฒฐ๋ก ์ด๋ฏธ ์ง์๋ ์ผ๋ถ ๊ด์ธ๋ฅผ ํ๊ธํด์ผ ํ๊ณ , ์ต๊ทผ ๋ฌด์ญ ๋ฒ์์ ํ๊ฒฐ์ ํธ๋ผํ ๋ํต๋ น์ด ๋ชจ๋ ๊ตญ๊ฐ์ 10%์ ๊ด์ธ๋ฅผ ๋ถ๊ณผํ ์ ์๋ ๋ฅ๋ ฅ์ ์ ๋์ ๊ฑธ์์ต๋๋ค. ์์งํ ์ฃผ์์ ์ด๋ฏธ ์ธ์์ ์ ๋ฐ์ ์ด๊ธฐ๊ณ ์์ํ๋ค๊ณ ๋๋ ์ ์์ต๋๋ค.
ํนํ ์ค๊ตญ์ ํน์ ํ๋ชฉ์ ๋ํ ๊ด์ธ, ๊ทธ์ค์์๋ ์ค๊ตญ์ฐ ์๋์ฐจ์ ๋ํ ๊ด์ธ๋ฅผ ํฌ๊ฒ ์ฐ๋ คํฉ๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์ฒซ ์๊ธฐ ๋์ ์ค๊ตญ์ ์ฐ๊ฐ ์ฝ 100๋ง ๋์ ์๋์ฐจ๋ฅผ ์์ถํ์ง๋ง, ์๋ ์๋ ์ ์ธ๊ณ์ ์ผ๋ก 700๋ง ๋๋ฅผ ์์ถํ์ต๋๋ค. ๊ทธ๋ฌ๋ ๋ฏธ๊ตญ ์์ฅ์์๋ ์กฐ ๋ฐ์ด๋ ๋ํต๋ น์ด 100%์ ๊ด์ธ๋ฅผ ๋ถ๊ณผํ์ฌ ์ฌ์ค์ ์์ถ์ด ์ฐจ๋จ๋ ์ํ์ ๋๋ค. ์ค๊ตญ์ ์ด๋ฌํ ์ํฉ์ ๋ํด ๊ฐ๋ ฅํ ๋ถ๋ง์ ์ ๊ธฐํ๊ณ ์์ผ๋ฉฐ, ๋ฏธ๊ตญ ์์ฅ์ผ๋ก์ ์๋์ฐจ ์์ถ ์ฌ๊ฐ๋ฅผ ๊ฐ๋ ฅํ ์ํ ๊ฒ์ ๋๋ค.
ํ์ง๋ง ๋ฏธ๊ตญ์ ์ด ๋ฌธ์ ์์ ์๋ณดํ์ง ์์ ๊ฒ์ผ๋ก ์์๋ฉ๋๋ค. ์์ญ ๋ ์ ์ผ๋ณธ ์๋์ฐจ๊ฐ ๊ทธ๋ฌ๋ฏ์ด, ์ค๊ตญ์ ์๋์ฐจ ์์ฐ ๋ฅ๋ ฅ์ ๋ฏธ๊ตญ ์๋์ฐจ ์ ์กฐ์ ์ฒด์ ์๋นํ ์ํ์ด ๋๊ธฐ ๋๋ฌธ์ ๋๋ค. ์ด๋ฌํ ๊ด์ธ ๊ด๋ จ ๋ ผ์๋ ๊ตญ๊ฐ ๊ฐ์ ์ผ์์ ์ธ ๋ฌด์ญ ๊ฐ๋ฑ์ ์ผ๋ถ์ด๋ฉฐ, ์ด๋ฒ ํ๋ด์์ ํด๊ฒฐ๋ ์ ์๋ โ์ฌ์ด ๊ณผ์ โ๋ก ๋ถ๋ฅ๋ ์ ์์ต๋๋ค.
ํ ์ด๋ธ ์๋ ๋์ธ ์ง์ง ์์ : ํจ๊ถ ๊ฒฝ์์ ์ฌ์ฅ๋ถ
์ ์ํ๋ด์์ ๋ค๋ค์ง๋ ์ฆ๊ฐ์ ์ธ ๋ฌด์ญ ๋ฌธ์ ์ธ์๋, ์๊ตญ ๊ฐ์๋ ํฅํ ์์ญ ๋ ๊ฐ ๋๊ฐ ์ง๋ฐฐ์ ์ธ ๊ฒฝ์ , ๊ตฐ์ฌ, ๊ธฐ์ ๊ฐ๊ตญ์ด ๋ ๊ฒ์ธ๊ฐ ํ๋ ๊ทผ๋ณธ์ ์ธ ์ง๋ฌธ๊ณผ ๊ด๋ จ๋ ํจ์ฌ ๋ ๊น์ ๋ถ์ด์ด ์กด์ฌํฉ๋๋ค. ์ด๋ฌํ ํต์ฌ ์์ ๋ค์ ์ง์ ์ ์ธ ํํ๋ก ๋ ผ์๋์ง ์์ ๊ฐ๋ฅ์ฑ์ด ๋์ต๋๋ค.
ํต๋ฌด๊ธฐ ๊ฒฝ์๊ณผ ๊ตฐ๋น ํต์
์ง๋ 10๋ ๊ฐ ๋ฏธ๊ตญ๊ณผ ์ค๊ตญ์ โ์น์ดํ ์ฌํฌโ๋ฅผ ๋ฒ์ฌ์์ต๋๋ค. ์์งํ ์ฃผ์์ 2049๋ ๊น์ง ์ค๊ตญ์ ์ธ๊ณ ์ต๊ณ ์ ๊ตฐ์ฌ, ๊ฒฝ์ , ์ ์น, ์ฌ์ง์ด ๋ฌธํ ๊ฐ๊ตญ์ผ๋ก ๋ง๋ค๊ฒ ๋ค๋ ๋ชฉํ๋ฅผ ์จ๊ธฐ์ง ์์์ต๋๋ค. ์ด๋ฌํ ๋งฅ๋ฝ์์ ์ค๊ตญ์ ํต๋ฌด๊ธฐ ์ฆ๊ฐ์ ๋ฏธ๊ตญ์ ๊ทผ๋ณธ์ ์ธ ์ฐ๋ ค ์ฌํญ์ ๋๋ค.
๋ง์ค์ฉ๋ฅ ์๋ ์ดํ ์์ญ ๋ ๋์ ์ค๊ตญ์ ์ต์ํ์ ํต ์ต์ง๋ ฅ์ ์ ์งํ๋ฉฐ 100~200๊ธฐ์ ํต๋ฌด๊ธฐ๋ฅผ ๋ณด์ ํ์ต๋๋ค. ๊ทธ๋ฌ๋ ์์งํ ์ฃผ์ ์ง๊ถ ์ดํ ์ค๊ตญ์ ๋๊ท๋ชจ ํต๋ฌด๊ธฐ ์ฆ๊ฐ์ ์์ํ์ผ๋ฉฐ, ๋ฏธ๊ตญ์ ์๋ ์ด ์ง๋์์ผ ์ด๋ฅผ ์์ ํ ์ธ์งํ์ต๋๋ค. ํ์ฌ ์ค๊ตญ์ ์ฝ 600๊ธฐ์ ํต๋ฌด๊ธฐ๋ฅผ ๋ณด์ ํ๊ณ ์์ผ๋ฉฐ, ๋ฏธ๊ตญ ๊ตญ๋ฐฉ๋ถ(Pentagon)๋ 2030๋ ๊น์ง ์ฝ 1,000๊ธฐ, 2035๋ ๊น์ง๋ ๋ฏธ๊ตญ๊ณผ ๋ฌ์์ ์์ค์ ํต๋ฌด๊ธฐ๋ฅผ ๋ณด์ ํ ๊ฒ์ผ๋ก ์ถ์ ํฉ๋๋ค.
์ด๊ฒ์ด ์ค์ํ ์ด์ ๋ ๋ช ๋ฌ ์ ๋ฌ์์์ ๋ฏธ๊ตญ ๊ฐ์ ๋ง์ง๋ง ๊ตฐ๋น ํต์ ์กฐ์ฝ์ด ๋ง๋ฃ๋์๊ธฐ ๋๋ฌธ์ ๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์ค๊ตญ์ด ํฉ์์ ์ฐธ์ฌํ์ง ์๋ ํ ์๋ก์ด ์กฐ์ฝ์ ํ์ํ๋ ๊ฒ์ ๋ฌด์๋ฏธํ๋ค๊ณ ์ฃผ์ฅํ์ต๋๋ค. ๊ทธ๋ฌ๋ ์ค๊ตญ์ ์์ ๋ค์ด ๋ฏธ๊ตญ๊ณผ ๋๋ฑํ ์์ค์ ํต๋ฌด๊ธฐ๋ฅผ ๋ณด์ ํ๊ธฐ ์ ๊น์ง๋ ๊ตฐ๋น ํต์ ์ ๊ด์ฌ์ ๋์ง ์๊ฒ ๋ค๋ ์ ์ฅ์ ๊ณ ์ํ๊ณ ์์ต๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์ ์ฃผ์์๊ฒ ํต ๊ตฐ๋น ํต์ ๋ฌธ์ ๋ฅผ ์ ๊ธฐํ ๊ณํ์ด์ง๋ง, ์ค๊ตญ์ด ๋ ผ์ํ ์ค๋น๊ฐ ๋์ด ์์ง ์์ ํฐ ์ง์ ์ ์ด๋ฃจ๊ธฐ๋ ์ด๋ ค์ธ ๊ฒ์ ๋๋ค.
๋๋ง์ ๋ฏธ๋์ ๋ฏธ๊ตญ์ โ๋ง ๋ฐ๊พธ๊ธฐโ
์ค๊ตญ์ด ๋ ผ์ํ ์ค๋น๊ฐ ๋์ด ์๋ ์ฃผ์ ๋ ๋ฐ๋ก ๋๋ง(Taiwan)๊ณผ ๋๋ง์ ์กด์ฌ๊ถ์ ๋๋ค. ์์งํ ์ฃผ์์ ๋ฏธ๊ตญ์ด ์๊ฐ์ด ์ง๋จ์ ๋ฐ๋ผ ์ค๊ตญ์ด ๋๋ง์ ์ฅ์ ํ๋ ๊ฒ์ ๋ํด ๋ ๊ฐ๋ฐฉ์ ์ธ ์ ์ฅ์ ์ทจํ๊ณ ์๋ค๋ ๊ฒ์ ๋ณด์ฌ์ฃผ๊ธฐ ์ํด ๋ชจ๋ ๋ ธ๋ ฅ์ ๋คํ ๊ฒ์ ๋๋ค. ๊ทธ๋ ์ฆ๊ฐ์ ์ธ ๋๋ง ์นจ๊ณต์ ์ํ ์ด๋๋ฅผ ๊ธฐ๋ํ์ง ์์ง๋ง, ๋๋ง์ธ๋ค์ด ๋ฏธ๊ตญ์ ์ง์์ ๋ํด ๋ ์๊ตฌ์ฌ์ ๊ฐ๊ฒ ํ ์ ์๋ โ๋ฏธ๋ฌํ ๋ฌธ๊ตฌ ๋ณ๊ฒฝโ์ ์ถ๊ตฌํฉ๋๋ค.
๊ฐ์ฅ ํํ ์ธ๊ธ๋๋ ์์๋ ๋ฏธ๊ตญ ๊ด๋ฆฌ๋ค์ด ๋๋ง์ ๋ ๋ฆฝ ์ ์ธ์ โ์ง์งํ์ง ์๋๋ค(not support)โ๊ณ ๋งํ๋ ๊ฒ์ ์ค๊ตญ์ด โ๋ฐ๋ํ๋ค(oppose)โ๋ก ๋ฐ๊พธ๊ธฐ๋ฅผ ์ํ๋ ๊ฒ์ ๋๋ค. ์ด๋ ์์ ๋ณํ์ฒ๋ผ ๋ค๋ฆฌ์ง๋ง, ์ธ๊ต์ ์ผ๋ก๋ ์์ฒญ๋ ์๋ฏธ๋ฅผ ๊ฐ์ง๋๋ค. โ๋ฐ๋ํ๋คโ๋ ๊ฒ์ ๋ฏธ๊ตญ์ด ์ ์ผํ ์ง์ ํ ์ค๊ตญ์ ์คํ์ธ๋ฏผ๊ณตํ๊ตญ์ด๋ฉฐ, ์ด์ ๋์ ํ๋ ค๋ ์ด๋ค ๋ ธ๋ ฅ๋ ๋ฐ๋ํ๋ค๋ ํน์ ๊ฒฌํด๋ฅผ ์ทจํ ๊ฒ์์ ์๋ฏธํ๊ธฐ ๋๋ฌธ์ ๋๋ค. ์ด๋ฌํ โ๋ง ๋ฐ๊พธ๊ธฐโ๋ ์ค์ ์ ์ธ ํ๋์ด ์๋๋๋ผ๋ ๋๋ง์๊ฒ๋ ํฐ ์ํ์ผ๋ก ๋ค๊ฐ์ฌ ์ ์์ต๋๋ค. ํธ๋ผํ ๋ํต๋ น์ ์ฆํฅ์ ์ธ(freelance) ์ฑํฅ์ ๊ณ ๋ คํ ๋, ์ค๊ตญ์ ๊ทธ๊ฐ ์ด ๋ฌธ์ ์์ ์์์น ๋ชปํ ๋ฐ์ธ์ ํ๊ธฐ๋ฅผ ๋ฐ๋ ๊ฒ์ ๋๋ค.
๋๋ง์ ๋ํ ์ธ๊ณต์ง๋ฅ(AI) ๊ด๋ จ ํต์ฌ ๊ธฐ์ ์ธ ๋ฐ๋์ฒด ์นฉ์ ์ฃผ์ ์์ฐ๊ตญ์ ๋๋ค. ์ธ๊ณ์์ ๊ฐ์ฅ ๋ฐ์ ๋ ๋ฐ๋์ฒด ์ ์กฐ์ ์ฒด์ธ ๋๋ง ๋ฐ๋์ฒด ํ์ฌ(Taiwan Semiconductor Corporation, TSMC)์ ์์ค์ ๋๋ง์ ์ง์ค๋์ด ์์ต๋๋ค. AI ํ๋ช ์ ํต์ฌ์ธ AI ์นฉ ์์ฐ์ ์ํด ๋ฏธ๊ตญ์ TSMC์ ์์ฐ์ ์์กดํ๊ณ ์์ต๋๋ค. ์์งํ ์ฃผ์์ ํธ๋ผํ ๋ํต๋ น์๊ฒ ์นฉ ๊ณต๊ธ์ ๋ณด์ฅํ๋ฉฐ ๋๋ง ์ฌ์ ๋๋จธ์ง ๋ถ๋ถ์ ๋ํด์๋ ์ ๊ฒฝ ์ฐ์ง ์๋๋ก ์ ๋ํ ์๋ ์์ต๋๋ค.
AI ํจ๊ถ ์ ์๊ณผ โ๋ฏธํ ์คโ์ ๊ฒฝ๊ณ
AI๋ ํต๋ฌด๊ธฐ๋ ๋๋ง ๋ฌธ์ ๋ณด๋ค ํจ์ฌ ๋ ์๊ธํ๊ฒ ๋๊ปด์ง๋ ์ ์ธ๊ณ์ ์ธ ๊ด์ฌ์ฌ์ ๋๋ค. ์ด ๊ธฐ์ ์ ์ฐ๋ฆฌ์ ์๋ณด, ์ฌ์ง์ด ์ดํด๋ฅผ ๋์ด์ ๋น ๋ฅด๊ฒ ๋ฐ์ ํ๊ณ ์์ต๋๋ค. ๊ทธ๋ฌ๋ ์๊ตญ ์ ์ ๊ฐ์ AI ๊ด๋ จ ๋ ผ์๋ ๋งค์ฐ ์ ํ์ ์ด๊ณ ํผ์์ ์ผ ๊ฐ๋ฅ์ฑ์ด ๋์ต๋๋ค. ๋ฐ์ด๋ ํ์ ๋ถ ์์ ์๋ ๋ฏธ๊ตญ๊ณผ ์ค๊ตญ์ด ํต๋ฌด๊ธฐ๊ฐ ์ธ๊ณต์ง๋ฅ ํ๋ก๊ทธ๋จ์ ์ํด ์ง์๋์ง ์๋๋ก ํฉ์ํ๋ ๋ฐ ์๊ฐ์์ด ๊ฑธ๋ ธ์ต๋๋ค.
AI ์๋์๋ ํต ์๋์ ๊ตฐ๋น ํต์ ๋ฐฉ์์ด ์ ์ ์ฉ๋์ง ์์ต๋๋ค. AI๋ ์ฝ๊ฒ ๊ฒ์ฌํ๊ฑฐ๋ ์๋ฅผ ์ธ๊ฑฐ๋ ํต์ ํ ์ ์๋ ํน์ ๋ฌด๊ธฐ๊ฐ ์๋๊ธฐ ๋๋ฌธ์ ๋๋ค. ๋ฐ๋ผ์ AI์ ๋ํ ํต์ ๋ ์ฃผ๋ก ๊ฐ์ ํ๊ธฐ ์ด๋ ค์ด ํ๋ ๊ฐ๋ น(codes of conduct)์ ํํ๊ฐ ๋ ๊ฒ์ ๋๋ค. ์๊ตญ ์ง๋์๋ค์ด ์ด ์ฃผ์ ๋ฅผ ์ธ๊ธํ๊ฑฐ๋ ์ค๋ฌด์ง์๊ฒ ๋ ผ์๋ฅผ ์ง์ํ ์ง์กฐ์ฐจ ๋ถํ์คํฉ๋๋ค.
๋ฏธ๊ตญ๊ณผ ์ค๊ตญ ๋ชจ๋ AI ํจ๊ถ ๊ฒฝ์์์ ์น๋ฆฌํ๊ธฐ ์ํด ์ ๋ ฅํฌ๊ตฌํ๊ณ ์์ต๋๋ค. ์ด๋ฌํ ๊ฒฝ์ ๋๋ฌธ์ ์๊ตญ ๋ชจ๋ ์ฑ์ฅ์ ์ ํํ ์ด๋ค ์กฐ์น๋ ์ทจํ๊ณ ์ถ์ด ํ์ง ์์ ์ ์์ต๋๋ค. ๊ทธ๋ฌ๋ ์๊ตญ ๋ชจ๋ ์๋๋ฐฉ์ ๋ํ ์ผ์ข ์ ์์ ์ฅ์น(guardrails)๋ฅผ ๋ง๋ จํ๋ ๋ฐ ์ดํด๊ด๊ณ๊ฐ ์๋ค๋ ๊ฒ์ ๋ถ๋ช ํด ๋ณด์ ๋๋ค.
ํธ๋ผํ ํ์ ๋ถ๋ ์ฒ์์๋ โํ ๊ตญ๊ฐ๊ฐ ์ด ๊ฒฝ์์์ ์น๋ฆฌํ ๊ฒ์ด๋ฉฐ, ์ด๋ ์ฐ์ ์ด ๊ท์ ์์ด ์์ ๋กญ๊ฒ ํ๋ํ๋๋ก ํ์ฉํ๋ ๊ตญ๊ฐ๊ฐ ๋ ๊ฒโ์ด๋ผ๋ ์ ์ฅ์ด์์ต๋๋ค. ๊ทธ๋ฌ๋ ์ต๊ทผ ๋ช ์ฃผ ๋์ ์ด๋ฌํ ํ๋๋ ๋ณํํ์ต๋๋ค. ๋ณํ์ ๊ณ๊ธฐ๋ ์ธ๊ณต์ง๋ฅ ๊ธฐ์ ์คํธ๋กํฝ(Anthropic)์ด ๊ฐ๋ฐํ โ๋ฏธํ ์ค(Mythos)โ๋ผ๋ ๊ฐ์์ ์ ํ ๋๋ฌธ์ ๋๋ค. ์ด ์ ํ์ ๊ณต๊ฒฉ์ ์ธ ์ฌ์ด๋ฒ ๊ณต๊ฒฉ์ ์ํํ๋ ๋ฐ ๋งค์ฐ ํจ๊ณผ์ ์ด์ด์ ๋์ค์๊ฒ ๊ณต๊ฐ๋์ง ์์์ต๋๋ค. ๋ฏธํ ์ค๋ ์ ํธ๋ฆฌํฐ ๊ทธ๋ฆฌ๋๋ ๋ค๋ฅธ ์ธํ๋ผ ์ฝ๋์ ์ทจ์ฝ์ ์ ๊ฑฐ์ ์ฆ์ ์ฐพ์๋ด ๊ณต๊ฒฉ์ ๋ช ๋ นํ ์ ์์ต๋๋ค. ๋ฌผ๋ก ์ด๋ ๋ฐฉ์ด์๋ ์ ์ฉํ ์ ์์ง๋ง, ์ค๊ตญ ์ญ์ ๋ฏธํ ์ค์ ์ ์ฌํ ๋๊ท๋ชจ ์ธ์ด ๋ชจ๋ธ์ ๊ฐ๋ฐ ์ค์ด๋ฉฐ, 6๊ฐ์์์ 1๋ ์์ ์์ฒด์ ์ธ ๋ฏธํ ์ค๋ฅผ ๊ฐ๊ฒ ๋ ๊ฒ์ ๋๋ค. ์ด๋ ๋ฏธ๊ตญ ํ์ ๋ถ๋ฅผ ์ฌ๊ฐํ๊ฒ ์ฐ๋ คํ๊ฒ ๋ง๋ค ์ฌ์์ ๋๋ค. ๋ง์ ๋ฏธ๊ตญ ๋ด ์ค๊ตญ ์ ๋ฌธ๊ฐ๋ค์ ํธ๋ผํ ํ์ ๋ถ ๊ณ ์ ๊ด๋ฆฌ๋ค์ด ์ค๋ ์ฌํ์ ๋๋ฌด ๋ชฐ๋ํ์ฌ ์ด๋ฒ ํ๋ด์ ์ ๋๋ก ์ค๋น๋์ง ์์ ์ํ๋ก ์ํ๊ณ ์๋ค๋ ์ ์ ์ฐ๋ คํ๊ณ ์์ต๋๋ค.
์ด๋ ์ฌํ, ์ ์ํ๋ด์ ์์์น ๋ชปํ ๋ณ์
ํธ๋ฅด๋ฌด์ฆ ํดํ ๊ฐ๋ฐฉ์ ์ ๋ฐํจ
ํธ๋ผํ ๋ํต๋ น์ ํ๋ด์ 6์ฃผ ์ฐ๊ธฐํ ๋, ์ ์์ด ๋๋๊ณ ์ด๋์ด ํญ๋ณตํ๋ฉฐ ํธ๋ฅด๋ฌด์ฆ ํดํ์ด ๊ฐ๋ฐฉ๋ ๊ฒ์ด๋ผ๊ณ ์์ํ์ต๋๋ค. ํ์ง๋ง ์ด ์ธ ๊ฐ์ง ๋ชจ๋ ํ์ค์ด ๋์ง ์์์ต๋๋ค. ์ด๋ ์ ์์ ์ด๋ฒ ํ๋ด์์ ํต๋ฌด๊ธฐ๋ AI ๊ฐ์ ๋ค๋ฅธ ์๊ธํ ์ฃผ์ ๋ณด๋ค ๋ ๋ง์ ๋ ผ์ ์๊ฐ์ ์ฐจ์งํ ๊ฐ๋ฅ์ฑ์ด ๋์ต๋๋ค.
์ค๊ตญ์ ํธ๋ฅด๋ฌด์ฆ ํดํ์ด ๊ฐ๋ฐฉ๋๊ธฐ๋ฅผ ๊ฐ๋ ฅํ ์ํฉ๋๋ค. ์ค๊ตญ์ ์์ ๋ฐ ๊ฐ์ค์ 30% ์ด์์ ์ด ํดํ์ ํตํด ์ด์กํ๋ฉฐ, ์ด๋ ์ ์ ์ด์ ์๋ ์๊ตญ ๊ฒฝ์ ์ํฉ์ด ์ข์ง ์์๋ ํฐ๋ผ ์๋์ง ๊ฐ๊ฒฉ ์์น์ด ๋ง์ ์ค๊ตญ ๊ธฐ์ ๋ค์ ๊ณ ์ฌ์ํค๊ณ ์๊ธฐ ๋๋ฌธ์ ๋๋ค. ๊ทธ๋ฌ๋ ์ค๊ตญ์ ์ง์ ๊ฐ์ ํ ๋งํ ์ญ๋์ด ํฌ์ง ์์ต๋๋ค.
์ค๊ตญ์ ์ญํ ๊ณผ ๋๋ ๋ง
ํธ๋ผํ ๋ํต๋ น์ ์ค๊ตญ์๊ฒ ์ด๋ ๋ฏธ์ฌ์ผ, ์๊ฒฉ๊ธฐ ๋ฐ ์ ๋ฐ์ ์ธ ์ ์ ์ํ์ ํ์ํ ํ์ ๋ฐ์ดํฐ์ ๊ธฐ์ ๊ณต๊ธ์ ์ค๋จํด ์ค ๊ฒ์ ์์ฒญํ ๊ฒ์ ๋๋ค. ๋ํ, ๋ฏธ๊ตญ์ ์ค๊ตญ์ด ์ด๋ ์ํ์ ์ฃผ์ ๊ตฌ๋งค์๋ก์ ์ด๋์ ๋ํ ์ํฅ๋ ฅ์ ํ์ฌํ์ฌ ์ด๋์ด ํดํ์ ๊ฐ๋ฐฉํ๋๋ก ์ค๋ํด ์ค ๊ฒ์ ์๊ตฌํ ๊ฒ์ ๋๋ค.
์์งํ ์ฃผ์์ ๋งค์ฐ ์ ์คํ ์ธ๋ฌผ์ด์ง๋ง, ์ ์ ์ข ์์ ๋ํ ๊ฐ๋ ฅํ ์ดํด๊ด๊ณ๋ฅผ ๊ฐ์ง๊ณ ์์ต๋๋ค. ๋ฐ๋ผ์ ๊ทธ๊ฐ ๋ฏธ๊ตญ์ ํธ์์ ์กฐ์ฉํ ์ด๋์ ๊ฐ์ ํ ์ง ์ฌ๋ถ๊ฐ ์ฃผ๋ชฉ๋ฉ๋๋ค. ์ด๋ ์ค๊ตญ์๊ฒ ๋ณต์กํ ๋๋ ๋ง๊ฐ ๋ ๊ฒ์ ๋๋ค.
๊ฐ์์ โ์น๋ฆฌโ ๋ฐฉ์ ์: ํธ๋ผํ์ ๋จ๊ธฐ ์ฑ๊ณผ, ์์งํ์ ์ฅ๊ธฐ ์ ๋ต
์ด๋ฒ ์ ์ํ๋ด์์ ๋ ์ง๋์๊ฐ ๊ฐ๊ฐ โ์น๋ฆฌโ๋ผ๊ณ ๋๋ ๋งํ ๊ฒ์ ๋ฌด์์ผ๊น์? ๊ทธ๋ฆฌ๊ณ ๊ทธ๋ค์ด ๋์ค์๊ฒ โ์น๋ฆฌโ๋ก ํฌ์ฅํ ์ ์๋ ๊ฒ์ ๋ฌด์์ผ๊น์?
ํธ๋ผํ ๋ํต๋ น์๊ฒ๋ ๋ถ๋ช ํฉ๋๋ค. ๊ทธ๋ ํญ์ ๋๊ท๋ชจ ์ฌ์ ๊ฑฐ๋๋ฅผ ํตํด ํ๋ด์์ ๋ฒ์ด๋๋ฉฐ, ์์ ์ โ์ต๊ณ ์ ํ์๊ฐ(dealmaker-in-chief)โ๋ก ์ ์ธํ๊ณ ์ถ์ด ํ ๊ฒ์ ๋๋ค. ๊ทธ๋ฌ๋ ๋ฐ์ด๋น๋ ์์ด ๊ธฐ์๊ฐ ์ง์ ํ๋ฏ์ด, ์ด๋ฌํ ๋จ๊ธฐ์ ์ธ ์ฑ๊ณผ๋ ์ด๋ ์ฌํ ์ดํ ์ค๊ตญ์ด ๋ฏธ๊ตญ์ ๊ณผ๊ฑฐ์ฒ๋ผ ๋ฌด์ ์ ์กด์ฌ๋ก ๋ณด์ง ์์ ์๋ ์๋ค๋ ๊ทผ๋ณธ์ ์ธ ์ง๋ฌธ์๋ ๋ตํ์ง ๋ชปํฉ๋๋ค.
์์งํ ์ฃผ์์๊ฒ๋ โ์ฅ๊ธฐ์ ์ธ ๊ฒ์โ์ ๋๋ค. ๊ทธ๋ ๊ฐ๋ณ์ ์ธ ๊ฑฐ๋ ์์ฒด์๋ ํฌ๊ฒ ๊ด์ฌ์ด ์์ต๋๋ค. ๋ง์ฝ ๊ทธ๋ฐ ๊ฑฐ๋๋ค์ด ์ธ๊ต์ ์คํ์ ์ญํ ์ ํ๋ค๋ฉด ๊ด์ฐฎ์ต๋๋ค. ์ ์ฃผ์์ ์์ ์ด ํ์ฌ ๋ ์์ ์ ์ธ ๊ฐ๋๊ตญ์ ์ด์ํ๊ณ ์์์ ๋ณด์ฌ์ฃผ๊ณ ์ถ์ด ํ ๊ฒ์ ๋๋ค. ๊ทธ๋ ๋ค๋ฅธ ๋๋ผ๋ฅผ ์นจ๋ตํ๊ฑฐ๋ โ์ ๊ธ์ ๋ฒ์น(law of the jungle)โ์ ๋ฐ๋ฅด์ง ์๋๋ค๊ณ ์ต๊ทผ ํธ๋ผํ ๋ํต๋ น์ ๊ฒจ๋ฅํ ๋ฏํ ๋ฐ์ธ์ ํ์ต๋๋ค. ๊ทธ๋ ์๊ฐ์ด ์ง๋จ์ ๋ฐ๋ผ ์ ์ธ๊ณ๊ฐ ์ค๊ตญ์ ํ๊ณผ ์๋ณธ์ ์์กดํ์ฌ ์ฌ๊ฑดํ๊ณ ์๋ก์ด ๋ฌด์ญ ๊ด๊ณ๋ฅผ ๊ตฌ์ถํ๊ฒ ๋ ๊ฒ์ด๋ผ๊ณ ๋ฏฟ๊ณ ์์ต๋๋ค. ์ด๊ฒ์ด ์ ์ฃผ์์ ์ฅ๊ธฐ์ ์ธ ๊ณํ์ด๋ฉฐ, ๊ทธ๋ ๋ฏธ๊ตญ์ด ์์ฒด์ ์ธ ์ฅ๊ธฐ ๊ณํ์ ๊ฐ์ง๊ณ ์๋์ง ์์ฌํ ์๋ ์์ต๋๋ค.
์ด๋ฒ ๋ฏธ์ค ์ ์ํ๋ด์ ๋จ์ํ ๋ ์ ์์ ๋ง๋จ์ด ์๋, ๊ฒฉ๋ณํ๋ ๊ตญ์ ์ง์ ์์์ ๋ ์ด๊ฐ๋๊ตญ์ด ํผ์น๋ ์น์ดํ ํจ๊ถ ๊ฒฝ์์ ํ ์ฅ๋ฉด์ ๋๋ค. ์ด๋ ์ฌํ๋ผ๋ ์์์น ๋ชปํ ๋ณ์ ์์์ ํธ๋ผํ ๋ํต๋ น์ ์์ ์ ์ ์ง๋ฅผ ํ๋ณตํ๋ ค ์ ์ฐ๊ณ , ์์งํ ์ฃผ์์ ์ฅ๊ธฐ์ ์ธ ๋น์ ์ ๊ฐ์ง๊ณ ์ค๊ตญ์ ์์์ ๊ฐํํ๋ ค ํ ๊ฒ์ ๋๋ค. ์ด๋ฒ ํ๋ด์ ๊ฒฐ๊ณผ๋ ๋จ๊ธฐ์ ์ธ ์ฌ์ ์ฑ๊ณผ๋ฅผ ๋์ด, ํฅํ ์์ญ ๋ ๊ฐ ์ธ๊ณ์ ๊ถ๋ ฅ ์งํ์ ๊ฒฐ์ ํ๋ ๋ฐ ์ค์ํ ์ด์ ํ๊ฐ ๋ ๊ฒ์ ๋๋ค.