From Electricity to Artificial Intelligence: How a General-Purpose Technology Discovers and Prices the Future World

What a technology can do does not determine what humans will use it for. There is a key concept in between: the imaginable value that can be priced.

Capital never lacks imagination, but it is far more willing to place bets on imaginations that come with measurable metrics. Edison's electric light could be valued against existing gas lighting, and it quickly secured funding from Wall Street; Morse's telegraph had no ready market reference point, and in the end, it only moved forward with a $30,000 allocation from Congress. The market prices the future every single day, but it excels most at valuing those futures that can already be quantified into a clear number and compared with others.

From the very moment artificial intelligence emerged, it has been faced with a massive bill: programmers, lawyers, customer service representatives, teachers, doctors, analysts, and all forms of cognitive labor across the entire modern economy. This bill is so enormous that AI can thrive perfectly well even without looking outward for new use cases.

This is both its convenience and its danger.

A large number of current AI applications are essentially equipping old job positions with "electric motors": lawyers review contracts faster, programmers write code faster, doctors draft medical records faster. The history of electricity reminds us that the truly massive gains did not come from replacing steam engines with electric motors, but from factories eventually removing the overhead line shafts and redesigning their entire production model around electricity.

Factory buildings grow old, and people will defend their own factories.

At the same time, hundreds of millions of ordinary end users are constantly testing out a huge number of unnamed "electrical appliances" at the grassroots level: a doctor's private medical record sorting workflow, a teacher's self-developed teaching method, a small shop owner's custom email processing routine. As generation costs keep plummeting, the new bottleneck has shifted to screening: who can identify the truly universally valuable insights from the massive pool of private experiments?

This also leads to a clear policy recommendation.

The government can openly acknowledge that it does not know where new use cases will emerge, so it does not pick winning enterprises, industries, or problems, but instead distributes a small amount of exploration rights to every individual. A Universal Token for all residents is exactly the public power socket of the artificial intelligence era.

Issue every resident a non-transferable quota for model usage, so that more people can experiment with those unnamed, yet-to-be-valued use cases. There is no penalty for trial and error; once a valid use case is identified, a private workflow can be verified, disseminated, and eventually grow into an entirely new market.

The future of artificial intelligence does not only depend on how much more powerful the models can become.

It also depends on whether we are willing to reserve a little search space for unknown use cases long before the market can put a price tag on them.

Part One How a New Technology Finds Its Use Cases

The laws of nature determine what a technology is capable of, but they never determine what humans will ultimately use it for.

1. The 80-Year Gap

In 1800, Volta stacked alternating zinc and copper plates separated by brine-soaked cloth, and obtained a continuous electric current for the first time. In 1831, Faraday inserted a magnet into a coil and pulled it out, generating induced current. 30 years separated these two milestones. It took another 50 years from Faraday's discovery to 1882, when Edison built the landmark commercial central power station at Pearl Street in New York.

Electricity was never idle during those 80 years. Around 1808, Davy publicly demonstrated the arc lamp at the Royal Institution of Great Britain. In 1805, people already used electric current to deposit one metal on the surface of another; by around 1840, process improvements in Birmingham turned electroplating into a profitable business. Telegraph lines began to be laid in the 1840s. Lighthouses were equipped with arc lamps. Doctors used electricity as a trendy therapy, and street fairs treated it as a novelty trick. Use cases emerged one after another.

But these use cases were far apart from each other, each completely isolated. The telegraph had its own dedicated batteries, electroplating used separate batteries, and arc lamps were powered by independent generators. There was no shared infrastructure, no daily, standardized supply that people could access easily. 80 years passed between the discovery of continuous current and electricity becoming a readily available daily utility, and it took several more decades for electricity to be delivered to every household just like running water. The "gap" I refer to is exactly this period of time.

This piece of history reminds us to separate two distinct things: one is discovering a new capability, and the other is inventing new activities built around that capability. The former is given by nature, the latter is not. Electromagnetic induction is written into the laws of physics, but electric lights, elevators, refrigerators, and trams are not. They are human inventions, and they were invented very slowly.

Looking back today, we tend to assume that the full list of electricity use cases was pre-written, waiting for technology to mature to be fulfilled one by one. People back then had no such list. They had no idea why households needed electricity, why factories needed electricity, or why cities needed electricity. Electricity was a powerful answer sitting right there, but people had to find the corresponding questions one by one.

2. The Telegraph: The First Major Use Case, and the Most Unfamiliar One

The first major use case of electricity was the telegraph. This story is worth elaborating on, because it is fundamentally different from the later electric light.

Morse was a painter, not an electrician. In 1832, on a ship returning to the US from Europe, he heard people discussing electromagnets and got the idea. By 1838, he and his collaborators had proven that encoded signals could be transmitted over considerable distances of wire. He spent the next five years looking for funding. He approached private investors, traveled to Europe to seek patents and support, but still could not raise enough money to build a long-distance demonstration line. In 1843, the US Congress allocated $30,000, and the first line between Washington and Baltimore was completed the following year. As soon as the line went live, private capital flooded into the telegraph industry almost immediately.

Why could he not find private funding for the first five years? It was not that no one wanted faster news. Merchants had always wanted to learn price updates earlier, newspapers had always wanted to get news faster, and governments had always wanted to transmit orders more rapidly. The demand was there, but no one could quantify how large that demand was. The only comparable reference points at the time were postal services and horse couriers. If you multiplied the cost of a postage stamp, the telegraph seemed barely worth anything: a letter only cost a few cents, who would pay a huge premium to get it a few days earlier? But the telegraph later achieved things that the postal service never could. It spawned news agencies, aligned grain and cotton prices across different regions on the exact same day, allowed railways to schedule trains down to the minute, and together with the railway network, it drove the adoption of unified time zones. None of these things existed in 1843, and no one could calculate their potential revenue.

Of course, there are other explanations for Morse's difficulties. The technology had not yet proven itself over long distances, the first demonstration line had the properties of a public good where the builder would bear all the losses, patents were uncertain, and Morse himself had no reputation in the business world. All these are plausible factors, and I do not intend to attribute the financing difficulty to a single cause. The value of this story is that it reveals a problem, rather than proving a single answer: a future with no existing market metrics is far harder to secure the first round of verification funding. Once the first funding is in place, the demonstration is completed, the metrics emerge, and private capital will follow.

Morse's story separates two distinct scenarios: a use case with no demand, and a use case whose demand cannot be valued. The telegraph falls into the latter category. The imagination was there: Morse himself envisioned a nationwide network. Capital did not lack confidence, it lacked a measuring stick. What reference can I use to estimate how much this is worth? Without comparable benchmarks, investors cannot put it side by side with other investment opportunities. The market can hardly price things that cannot be compared.

3. Edison: Imagination Gets Its Measuring Stick

Four decades later, Edison was doing something completely different.

When he announced he would develop the electric light in 1878, his goal was crystal clear from day one: to replace gas lighting. He designed not just a single light bulb, but an entire integrated system: generators, wires, electricity meters, lamp holders. The entire system was designed around the price point of gas lighting. The brightness of the bulb should be comparable to gas lamps, the electricity tariff should be similar to gas bills, and the wiring process should be as convenient as laying gas pipes. Funding from Morgan and Vanderbilt arrived very quickly.

Edison was of course also imagining the future. He envisioned a world where every household had electric lights, a world far larger than the existing gas lighting market at the time. But his imagination had a solid anchor. The annual revenue of gas companies was public record, the number of households in a city could be counted, and the share of lighting in household expenditure was a well-documented figure. He could say: if my light is slightly cheaper than gas, all that money will flow to me. Investors understood this statement perfectly.

Put Morse and Edison side by side, and the difference is obvious. Both were imagining a future that did not exist yet. Morse's future had nothing to multiply against, while Edison's future did. The former spent five years securing $30,000 from Congress, while the latter got Wall Street funding in just a few months.

I call Edison's type of imagination "pricable imagination". It is still imagination, and it can still be wrong — Edison famously made the wrong call on direct current vs alternating current. But it is a mistake that can be financed. Capital can compare it with other potential mistakes, assign it a valuation, and spread out its risks. Morse's imagination could not do any of these things.

4. Existing Categories as Pricing References for Imagination

We can extract a general conclusion from these two stories.

The capital market never lacks imagination. It values the future every day, putting price tags on mines that have not started production, drugs that have not hit the market, and railways that are not yet completed. The market never only looks at the status quo. The real difference is that different futures have different degrees of valuability.

An existing product category provides three things for imagination: first, an observable expenditure scale — how much revenue the gas company made this year. Second, a pre-existing group of customers who are already paying for the service, so you do not need to convince them that they need lighting. Third, a shared set of language that allows you to say "10% of this market". With these three elements, imagination has a clear reference point. A person can say: the current market is this large, I can capture a portion of it, and if we assume the market will continue to grow, the future value will be this number. This calculation can be wildly inaccurate, but it is a calculation that can be written down on paper, and compared against other calculations in a meeting room.

A truly new activity often cannot even articulate what "this market" is. It has no existing expenditure scale, because no one is spending money on it yet. It has no existing customers, because customers do not even know they want it. It has no shared descriptive language, because the vocabulary to describe this demand has not been invented. It can only rely on one person's personal belief. Beliefs cannot be aggregated, compared, or risk-spread.

So the role of existing categories is far greater than just "they are already there". They are the valuation reference points for imagination. Any future that can borrow these references can more easily secure funding. Those that cannot have to rely on congressional allocations, sponsors, the inventor's own savings, or simply wait for time to pass.

This is far more precise than the usual claim of "market failure". The market does not fail. It excels at allocating capital across comparable futures, and it does this job extremely well. It is simply not good at pricing futures that have no valuation anchors. These two things are often lumped together as "market discovers demand". What the market discovers is "valuable demand".

5. Use Cases Beget More Use Cases

The subsequent history of electricity can be retold following this exact logic.

The electric light came first, because gas lighting provided its measuring stick. Electric motors entered factories smoothly, because steam engines provided its measuring stick: a motor could be priced against the steam horsepower it replaced. Trams spread very quickly, because horse-drawn streetcars had already been running in cities for decades. Every early use case gained a foothold by borrowing from an old, existing category.

Truly new things came much later, and they emerged in a very different way from electric lights. Elevators already existed in the steam and hydraulic era, and New York and Chicago were already using them to build 10-story buildings in the 1880s. After electric motors matured, elevators became faster, more reliable, and could easily serve far more floors, making the business case for high-rise buildings far more viable. Refrigeration followed the same pattern: it first gained a foothold in breweries, meat storage facilities, and cold chain transportation, where people were already paying for ice and freshness preservation. As refrigeration kept getting cheaper, household refrigerators became a conceivable product; after household refrigerators became widespread, supermarkets and more complete cold chains gained new room to expand, eventually completely transforming the way people eat today.

There are three distinct mechanisms here that need to be separated. The first is the substitution anchor: gas lamps for electric lights, where the old category provides an observable market size for the new technology, making it easy to secure financing. The second is feasibility expansion: refrigeration for the cold chain, where a key input drops in price, making previously uneconomical activities economically viable. The value of the cold chain cannot be calculated by multiplying the market size of refrigeration — refrigeration only makes it possible. The third is use case accumulation: existing applications leave behind equipment, standards, skills, and supply chains, lowering the cost of developing the next generation of applications. All three are important, but only the first one provides a valuation reference for imagination via existing categories. The latter two do not provide measuring sticks, they provide possibilities.

As a result, new categories rarely emerge out of thin air. Early use cases first gain valuation by borrowing from old categories, and after they are established, they open up space for the next batch of use cases through price reduction, standardization, and supporting inventions. The largest batch of new activities in the electricity era were built up layer by layer in this way. Each layer had to stabilize first, get its own price, and build out supporting equipment, before the next layer could have something to build on.

This also explains why the electricity revolution took so long. The 80 years without a complete list of use cases existed because no existing large enough market could absorb electricity all at once. The lighting market was not small at the time, but it was very limited compared to the entire economy. The industrial power market was already occupied by steam engines, so electric motors had to replace them one by one. There was no unified, massive existing category that could provide a universal measuring stick for electricity. Electricity had to find use cases one by one, each new use case adding a new measuring stick that could be used to imagine the next one.

6. Before We Move to Artificial Intelligence

We can now bring the discussion back to the present.

When a general-purpose technology emerges, it is not only faced with its own capabilities, but also with a pre-existing economy. Some things in this economy have clear prices, others do not. The direction the technology searches for use cases is largely determined by what measuring sticks it can borrow at the moment of its birth. Electricity could barely borrow any measuring sticks when it emerged, so it was forced to search outward for 80 years.

The situation for artificial intelligence at birth is completely different. The thing it most closely resembles happens to be one of the most expensive, widespread, and clearly accounted inputs in the modern economy: human cognitive labor. Lawyers, programmers, accountants, customer service representatives, designers, doctors, teachers, analysts — every single one of these roles has clear salary data, industry scale, profit margin, and statistical records. From the moment AI was born, it was faced with the largest bill in the world, and every single line item on that bill can be directly multiplied to calculate potential value.

This is exactly what we will discuss in the next section.

Part Two From the Day It Was Born, Artificial Intelligence Saw a Massive Pre-Existing World

7. The First World Artificial Intelligence Saw

At the end of 2022, the public first began to use large-scale models capable of natural language conversation. Within a few months, the list of use cases emerged. Programmers write code, so code-writing assistants were built. Customer service representatives answer questions, so customer service chatbots were built. Lawyers review contracts, doctors draft medical records, accountants do bookkeeping, teachers prepare lessons, analysts generate reports — every single profession quickly got its corresponding AI product.

These use cases all share one common feature: every single item on the list is an activity that already existed in the existing economy, and every single one has a clear salary attached to it. This is no coincidence. The people building these products first saw a payroll sheet. The question they asked was: which of these expensive labor costs can be taken over by machines?

Some people say this is a lack of imagination. I do not agree. The people building code assistants are imagining a world with so much software that it is unthinkable today, and writing a few lines of code for programmers is just