AI Agent Shock, AlphaGo Duel Ends, Economic Power Shift

● Agentic AI reshapes the economy

Lee Sedol and AlphaGo 10th Anniversary: “The duel is over, collaboration begins” — the moment AI agents, explainable AI, and people literacy move from Go to the economy

In the AI era, the perspective that sees it only in terms of “model performance” is changing.

In this article, the key takeaway is exactly three things.

First, what Lee Sedol emphasized at the AlphaGo 10th anniversary event wasn’t ‘a duel’ but ‘collaboration (work with an agent)’.
Second, the point of making a Go AI in 18 minutes—it appears not as a simple event, but as a symbol of an “everyone can build and use it” era.
Third, as explainable AI (interpretation models) and an education/review (Instruction·Interpretation·Feedback) structure are attached, it’s becoming clear that the way AI enters real work and learning is changing.

I’ll summarize this flow from the viewpoint of the economy and AI trends—“like news.”

1) 10th anniversary event: Why did Lee Sedol bring up ‘collaboration with an AI agent’ instead of ‘a duel with AI’?

Event background: Lee Sedol (9-dan, professor at Ulsan National Institute of Science and Technology) held an event with the Korean AI startup ‘Inhance’ to commemorate the 10th anniversary of his match against AlphaGo
Message shift: Reinterpreting the symbolism of a duel remembered as “humanity lost, but Lee Sedol won’t be blamed” as agent collaboration
Key observation: AI is now at a stage where it becomes a fellow who helps you at your side (Agentic Work), rather than an entity that exists to defeat the opponent

Economically important here is that this change quickly spreads into industrial structure.

As differentiation becomes harder with just competing large models,
the competitive axis is shifting to how companies embed and operate AI in work (workflow·roles·feedback).

2) What “building a Go AI in 18 minutes” means: AI build cost and time are shrinking fast

Based on the article/dialogue content, the points emphasized at the event are these.

Not building the model itself, but implementing a system that “plays Go, checks, and executes” in a short time
Infrastructure change: not massive compute infrastructure for matches, but a combination of specific hardware (DGX Spark support) + cloud (AI cloud)
As a result: moving from “a technology only a tiny few can do” to “a technology that iterates on prototypes quickly”

At this point, readers may have one question.

“Aren’t we basically shifting the fight to who can build AI better and who can use it better now?”

Yes.

So the SEO keywords that go into this can be naturally organized as well.

AI agent
Generative AI
Explainable AI
AI workflow
AI digital transformation

These five words bundle the situation where the competition shifts from “performance competition → application competition” in one go.

3) Not “a match,” but an “agent demonstration”: they created and checked a Go program with voice commands

At this event, Lee Sedol wasn’t simply placed in a role of “going up against AI once.”

Voice commands via an AI agent → generate a Go program
Directly playing/inspecting → confirming “how well it’s been made”
Shift in viewpoint: it’s no longer that humans “receive” the AI’s answers; it’s a structure where humans “set goals” and the AI “implements” them

The reason this matters is simple.

Now, AI’s value is moving from
“Does it get the right answer?” to
“Does it complete a portion of the work (with quality/completeness)?”

In other words, it’s agentic systems that get completed by work units, not the model itself, that become the competitive advantage.

4) From B2B to B2C: Lee Sedol acknowledged the possibility of “personal expansion,” but also pointed out the reality—“we still need to see the use cases”

The nuance in the conversation was like this.

Right now it’s mainly being used in a B2B format
After that, expansion to B2C is also possible
But there’s caution that “there may not be many use cases for individuals to use right away”

Translating this into economic trends:

Companies adopt B2B first because they can quickly verify ROI (return on investment)
B2C depends on when it “enters daily life,” so user experience (UX), trust, and the cost structure are key

So the next market stage won’t open because “the technology works.”
It opens when it’s clear why I’m spending money/time.

5) The difference between AlphaGo and 10 years later: evolving not from “the scale of computation,” but from “exclusion/learning method” and the “review/interpretation structure”

This is the theme that came up often in the conversation.

The value of review (replay/commentary/verification) has grown
Go is closer to “finishing a work” than winning or losing, so review is essential
After AlphaGo, the way pros used to ‘do research’ shifted to a way of ‘studying by looking at AI’
And AlphaGo’s characteristic—its method of excluding irrelevant things—was also mentioned

This is where explainable AI fits in.

Even in the event demonstration, they didn’t just have a “playing model.”
They explain that there were multiple components like an interpretation model, a teaching model, and a match-specific form.

This structure will be extremely important in education and corporate operations going forward.

For trust and an improvement loop to form, people must be able to understand the choices AI makes
If you only provide “black-box results,” organizations end up with higher operating costs anyway

So explainable AI is becoming both a technology trend and a “condition for real-world adoption.”

6) The “Divine Move in 78,” but the core of the actual game flow was in 68 moves

What Lee Sedol emphasized at the event goes beyond the commentary flow about the ‘78 moves.’

When commentators talk about bugs after the 78th move, from the perspective of someone who knows Go, they look more closely at the earlier flow (especially the 68th move)
The 68th move is the moment of “creating an irreversible gap”
From there, the outcome converges sharply on “whether the bug appears or not”

Interpreting this from an AI/economic viewpoint:

More often than a one-shot correct move (78 moves),
irreversible structural changes (68 moves) determine the result.

It’s the same in AI.

Rather than an event where the model gets it right once
It’s the initial conditions/workflow/constraints/training data/interpretation loop that “fixes the game”

So success in an AI project is decided not by late scoring, but by early-stage design.

7) A revolution in how to study Go: starting to follow “positioning (early design)” with an AI sense, not “battles”

The change Lee Sedol described is quite intuitive.

In the past, AI had to do ‘massive calculations’ to play
Now, they’ve started making it so AI follows positioning from the early game in a sense-based way
Pros now “learn the sense of how AI plays,” not just “finding the number of correct moves”

One-line summary here:

Go became a game that absorbs AI’s thinking patterns from the opening, and that changed the way people study itself.

This flow applies almost identically to other industries.

The more a field requires strategy (investing, logistics, games, manufacturing quality, etc.),
the more early-stage design determines success or failure.

8) When will “AI teaches” become possible: performance is ahead, but “human touch” remains

There were questions about education in the conversation, and the answers were realistic too.

In some areas like math and English, AI already does better than humans
But to “replace cram schools,” the human touch in the teaching method remains
AI is becoming quickly more natural in tone and explanation style, but measuring educational effectiveness is complex
Abstract strategies with entry barriers like Go are especially promising for AI-based learning

This also connects to corporate HR/education (reskilling).

That’s why the ‘People Literacy’ concept introduced at the event is important.

AI literacy (tool understanding) alone is not enough, and
the ability to braid AI into work through collaboration with colleagues ultimately creates performance.

9) The socio-economic conclusion of 10 years of AlphaGo: even a 78th-move miracle and even a 10-year build are ultimately completed by “human-expert intervention”

This is the part from Lee Sedol’s AlphaGo dialogue that comes across most strongly from an “industry perspective.”

The AlphaGo incident was a catalyst that massively attracted talent to enter the AI industry
However, in the next stage (practical application, domain expansion), the involvement of domain experts is essential
Problem definition, feedback, and verification made by humans create the next level of the industry

In other words, AlphaGo wasn’t “the end”—it was the starting point where developer interaction attaches, and it accumulated over 10 years, leading to today’s AI digital transformation (AX) trend.

Main content to convey (the ‘conclusion less often found elsewhere’ that I pulled out)

The AI era has moved from competing for ‘stronger models’ to competing for ‘the ability to complete work with agents.’
It’s not about the performance of a single shot (78 moves), but about how irreversible structural changes (68 moves) make or break early-stage design and workflow.
With explainable AI + interpretation/education models + a review loop, enterprise adoption speeds up. You can’t just “show and stop the result”; you have to “understand and improve.”
People literacy (collaboration ability) is ultimately the gate to organizational performance. Using technology may be easy, but organizations must change collaboration practices between people.

< Summary >

At the AlphaGo 10th anniversary event, Lee Sedol and Inhance showed the flow of the era not as “a duel” but as “collaboration with AI agents.”
The 18-minute Go AI demonstration suggests that the time required to build AI is being shortened and that infrastructure is also being configured more efficiently.
In addition, as the system adds an interpretation/teaching/review structure (explainable AI) beyond simple matches, the way AI is evolving to enter education and practical work is changing.
The way people study Go also shifted from “calculating correct answers” to “following AI sense to design openings,” and the conclusion in the industry is that domain expert involvement and feedback loops create the next step.

[Two related article links]

*Source: [ 티타임즈TV ]

– 이세돌-알파고 10주년 행사, 이세돌은 왜 바둑에이전트 만들었나?

NextGenInsight.Net