The New Digital Divide
The first divide was about who could get online. This one is about who can think with the machine.
For two decades, the phrase "digital divide" meant one thing: access. Who had a computer and who did not. Who had an internet connection and who was cut off from it. It was a divide of infrastructure, and we spent an enormous amount of effort, money, and goodwill working to close it, wiring schools, subsidising devices, extending broadband. That divide is not fully gone, but its logic is familiar and its solution is understood. A new divide is now opening, and it follows entirely different rules.
Access is no longer the question
The striking thing about AI is how little the old divide applies to it. The tools are cheap, frequently free, and available to almost anyone with the phone already in their pocket. In terms of raw access, AI may be the most evenly distributed powerful technology in history. If access were the whole story, this would be the most egalitarian moment imaginable. But access was never the whole story. It was simply the part we knew how to measure.
The divide is now skill
The new divide runs between those who learn to use AI well and those who merely use it. Hand the identical tool to two people and you will get radically different results, not because one has better access but because one has better judgment about what to ask, how to verify, when to trust, and when to push.
The first digital divide was about access. This one is about skill, and skill, unlike access, compounds.
| The old divide | The new divide | |
|---|---|---|
| It was about | Access | Skill |
| The fix | Wires, devices, broadband | Teaching judgment |
| How it closes | A one-time step | Slow, deliberate practice |
| Once you have it | You are done | It keeps compounding |
| Over time | Shrinks | Widens, and accelerates |
Why it widens instead of closing
Here is what makes this divide more dangerous than the last. Access is a step you take once: you get connected, and then you are connected. Skill is not like that. The person who learns to use AI well gets more done, learns faster, and reaches further, which gives them still more leverage to learn more. The gap does not stay constant. It accelerates.
Because access looks so equal, we will be tempted to conclude that opportunity is equal too. It is not. Equal access to a tool that rewards skill produces unequal outcomes.
We will mistake equal access for equal opportunity, and miss the divide forming underneath it.
Give two people the same tool and you have not given them the same chance. You have only hidden the gap behind a screen they both happen to own.
This holds between individuals, companies, and nations. And you cannot subsidise judgment or lay fibre to deliver fluency. It has to be taught, slowly and deliberately, which is exactly the unglamorous work most likely to be skipped.
"But AI is the great equalizer"
The strongest counterargument, and it comes with evidence. A widely cited 2023 study of customer-support agents by Erik Brynjolfsson and colleagues found that an AI assistant helped the least experienced, lowest-performing workers the most, pulling them up toward the level of the best and narrowing the gap between them. If AI lifts the bottom fastest, then it closes divides rather than widening them, and this whole essay is backwards.
That finding is real, and it holds where the work is bounded: a defined task, a known good answer, a script the AI can encode and hand to a novice. There, the tool genuinely levels. But most valuable knowledge work is not bounded. It is open-ended, where there is no script to copy, where the entire game is knowing what to ask, what to keep, and what to throw away. In that territory the skilled do not just get a leg up, they compound, because the tool multiplies judgment they already have. So both things are true: AI lifts the floor on routine work and raises the ceiling on everything else, and the ceiling moves faster. The divide narrows where the answer is known and widens where it is not, and the second category is where the real opportunity lives.
The first divide asked who could get online. This one asks who can think with the machine, and that is a much harder question to answer fairly.
I write about AI, data, and learning regularly at pinaldave.com, and I have been teaching this hands-on in my AI workshops.