What Intelligence Should Look Like in the AI Era— How It Differs Decisively From the Internet Era

In 1995 the internet became commercially available and within several years a near-permanent infrastructure for daily life. In 2022 ChatGPT was released and within several months a presence reshaping the very way knowledge labor is performed. Both are called "epoch-making technologies," and the literacy required at each turning point looks superficially similar — but the structures are decisively different. This essay examines that difference and identifies the components of intelligence we will need over the next twenty years.

011995 and 2022 — Same on the Surface, Different Underneath

The Windows 95 launch in November 1995 placed the internet at the center of public life. Within a few years, "people who can use the internet" and "people who cannot" diverged in their work and information access. Then in November 2022, ChatGPT became publicly available, and within roughly six months "people who can use generative AI" and "people who cannot" began to show a similar gap.

Surface similarity: a new information technology arrives, those who can use it operate at higher productivity, those who cannot fall behind. So why do we say the structures are decisively different?

02Internet-era Literacy = "The Ability to Find"

The literacy demanded in the internet era can be summarized in a single phrase: "the ability to find." The internet is a huge library — every page is created and posted by a human author, with a clear source, a publication date, a perspective, and a position. The user's task: design a query, navigate search results, read multiple pages, cross-check, and arrive at trustworthy information. Skills needed:

• Search query design (Boolean operators, refining keywords)
• Source evaluation (who is publishing this and for what purpose)
• Cross-checking (do multiple independent sources say the same thing)
• Time-stamp awareness (is this still current)
• Verification of authoritative sources (governments, academic journals, established media)

This skill set is largely an extension of "library research literacy." The novelty was the speed and breadth.

03AI-era Literacy = "The Ability to Ask"

Generative AI is a fundamentally different kind of information system from the internet. AI does not "look up" — it generates a response by predicting the next token. The output may look right but be wrong (hallucination). It may be confident but speculative. It may give a different answer to the same question depending on how it is asked. This is not a flaw — it is the inherent nature of the technology.

So the literacy required is completely different:

• The art of designing the question (a clear question yields a clear answer)
• The provision of context (giving the AI enough background to operate correctly)
• The verification of the output (is this true, where does it come from, is it being made up)
• The continuation of dialogue (refining and improving across multiple turns)
• The synthesis across multiple AIs (using several AI systems and combining)
• The judgment of human responsibility (AI suggests, the human decides and is accountable)

The most important shift: in the internet era, "the answer existed somewhere — finding it was the problem." In the AI era, "the answer doesn't yet exist — having it generated, well, is the problem." The ability to ask determines the quality of the answer.

04Seven Decisive Differences Between Internet and AI

The seventh row is particularly important. The internet took a decade to penetrate. AI has accomplished similar penetration in two months. The pace at which it reshapes labor, education, society — and the time available to adapt — is dramatically shorter.

05AI-era Intelligence — Eight Components

What the internet era required — search, source evaluation, cross-checking — has not disappeared. It now lives inside ③ Source Verification, as one component of a broader literacy. AI-era intelligence is the internet-era literacy plus seven additional layers.

06What Must Change in Education

Most school education today is still organized around the literacies of the internet era (or earlier). What needs to shift:

• From "find the right answer" to "design the question": When AI can answer most factual questions in seconds, the value of having pre-memorized facts drops, while the value of knowing what to ask rises.
• From "individual achievement" to "AI-collaboration achievement": Pretending AI doesn't exist in exam settings is a fiction. Teaching how to use AI well while developing the human capacities it cannot replicate.
• From "do not cheat by using AI" to "tell the truth about how you used AI": Bans are unenforceable. The work product is the human's, and how AI contributed should be transparent — that is the integrity model going forward.
• The reinforcement of subject foundations: The deeper your domain knowledge, the better you can use AI. Foundations have not lost value — they have gained it.
• The intensification of ethics and judgment: When AI proposes everything, "what we choose and what we refuse" — the human ethical judgment — becomes more important.

An education system that still measures "memorized facts" will produce graduates whose function AI performs better. An education that builds question design, judgment, ethics, dialogue, accountability — those graduates will be the ones who lead the AI age.

07What This Means in Pharmaceutical and Promotional Material Review

In the pharmaceutical promotional material review setting:

• Material checking by AI: First-pass mechanical checks (against Pharmaceutical Affairs Law §66 etc.) — AI is fast and consistent
• Searching past similar cases: "Was this expression flagged for product X before?" — AI surfaces candidates in seconds
• The drafting of expert opinion documents: Initial draft → AI; the reviewer revises and verifies
• The translation of foreign-language references: From technical Korean, German, French — translation accuracy of major AIs is now sufficient for professional use

However: the final judgment and the personal responsibility for it stay with the human reviewer. "AI said it was OK, but I have a sense something is off" — that intuition is the value that endures. To say it, the reviewer must have a deep grasp of the regulation, the disease, and how patients actually receive information. The eight components of intelligence above directly support this: the foundation for AI to be useful is the deep human subject expertise on which it operates.

The internet was a "tool for finding what already exists." AI is a "partner for generating what does not yet exist." The literacy required of the user is decisively different. What we need is the ability to ask, the ability to verify, the ability to judge, and the willingness to be accountable. Each of these is hard. But each is also the territory in which we humans, having grown up in this AI era, can still distinguish ourselves.

Trust comes from operating these eight components with integrity. The role of a compass is to keep us oriented when the technology moves at terrifying speed.