[AI Error Note] "Artificial Intelligence, That's Totally a Bubble"

③Conditions for Innovation Seen in the Case of Electricity: Time
The Light Bulb Invented in 1879 Took 70 Years to Become Widespread
Electricity Enabled Changes in Factory Structures
Computers Also Faced Criticism Asking "What Has Really Changed?"

Editor's NoteThe atmosphere has become such that nothing seems possible without artificial intelligence (AI). AI is almost always attached to the promotion of various products and services.
Office workers are also struggling. This is because of the easy demands at work like, "Hey, let's make something with AI too."
Does having AI guarantee a big hit? Not really.
In fact, the world of AI has been a continuous series of failures.
The so-called AI success stories represent only 0.1% of cases. 99.9% of companies and services have failed.
However, one of the fastest paths to success is studying failures.
The same applies to AI. Examining 99.9% of failures is a shortcut to the 0.1% of successes.
The 'AI Wrong Answer Note' series explores failure cases related to AI products, services, companies, and individuals.

Image generated by ChatGPT (DALL E·3)

Although there is a commotion as if the world will be turned upside down immediately by artificial intelligence (AI), not many people feel any immediate changes. Like the past 'dot-com bubble,' the 'AI bubble theory'?that AI is just an illusion?often resurfaces.

The bubble theory is not entirely baseless. Since the spectacular debut of ChatGPT in November 2022, numerous AI startups have emerged, and massive investments have been made. Developing AI technology requires astronomical amounts of money. The UK major investment bank Barclays predicted that big tech companies would spend $60 billion annually on AI model development by 2026. However, the resulting revenue is expected to be only about $20 billion.

OpenAI Forecasts Losses Until 2029... Can It Endure?

Like the bubbles before, there is quite a bit of talk that the AI bubble will burst soon. Will that really happen?

The situation is not easy even for OpenAI, arguably the most successful AI company. The IT media The Information reported on October 10, 2024 (local time) that OpenAI, the developer of ChatGPT, is expected to continue operating at a loss until 2029.

According to a report obtained by the media, OpenAI's revenue is expected to reach $100 billion (135 trillion KRW) by 2029. This is about 30 times the estimated revenue of $3.7 billion this year.

Will it finally escape the deficit situation then? No. Despite the enormous revenue scale of $100 billion, it is expected to remain unprofitable. Although revenue is increasing every year, costs are also rising significantly. A loss of $5 billion is expected this year alone, and losses are projected to increase even more sharply until 2028.

"AI is a revolution? Isn't that a scam?"

This is the background of such critical voices. The expected productivity revolution has not appeared. AI technology is advancing rapidly every day, but productivity-related indicators show little movement.

Nevertheless, we cannot label the AI revolution as merely a bubble or failure. The history of technology tells us that failures experienced by new technologies in their early stages are part of the process toward success.

Reflecting on Innovation Conditions Through Electricity: 'Time'

Even after the invention of electricity, it took a long time for it to become a common power source.

At 3 p.m. on September 4, 1882, electric lights lit up the financial district of Wall Street, New York. It was a historic moment when the world's first commercial power plant, Pearl Street Station, built by Edison, began operation. However, the dazzling start of history was faint. Although the power plant could light 400 bulbs, the initial customer base was only 85.

"Electric lights will explode and turn the whole neighborhood into ashes.
Electricity is too dangerous and an unnecessary technology."

Even the banks that provided funding shook their heads. The main power source at the time was the steam engine. It was literally the technology that drove the 'Industrial Revolution.' Investors turned away, saying, "Steam engines are enough."

The light bulb Edison invented earlier in 1879 also had a humiliating history. Edison and his supporters expected the light bulb to change the world immediately. Their dreams and the market's response were very different.

By 1907, only about 8% of American households used electricity. It was not until 1925 that more than half of urban households used electricity. Rural areas were even less electrified. Even in the early 1930s, the electrification rate was below 10%.

Through President Franklin Roosevelt's Rural Electrification Administration, nationwide electrification was completed only in the 1950s. It took nearly 70 years for electricity to become a universal infrastructure.

Electricity Replaced Steam Engines... Even Changed Factory Structures

The steam engine developed and unveiled by Corliss in 1876. New England Wireless & Steam Museum website

The changes in industrial sites are even more interesting. Although it was first realized in the 1880s that electricity had great potential to improve factory operations, it took about 40 years to understand factory systems that could effectively use electricity.

Factories in the steam engine era had to endure serious inefficiencies. Power from a huge central steam engine was transmitted through long shafts and belts, during which 20-40% of the power was lost. This was inevitable due to friction and vibration energy losses in shafts and belts. Moreover, steam engines needed 1-2 hours to build up sufficient steam pressure after starting in the morning. It's hard to imagine now when motors start immediately by just plugging in a cord.

Factory structures were inevitably dependent on steam engines. Huge boilers and steam engines occupied the center of buildings, and all machines had to be arranged around these power shafts. Factories at the time were mainly multi-story buildings of 4-5 floors. Using gravity to connect belts from power shafts on upper floors to lower floors was efficient. The building structure itself was designed for power transmission.

Modern automated factory. Photo by Getty Images Bank

The advent of electric motors freed all these constraints. It was realized that power could be used even far from the power source by introducing electricity. Installing individual electric motors on each machine solved the power loss problem, and machines could start immediately just by connecting to the power supply.

The philosophy of factory design also changed. It shifted from 'efficiency of power transmission' to 'efficiency of production processes.' Machines could now be freely arranged according to workflow.

Freed from the structural constraints of steam engines, innovative ideas to improve factory and workplace design and workflows began to emerge one after another. By 1920, everyone understood that the true purpose of electricity introduction was not simply 'fuel cost reduction' but 'dramatic productivity improvement' through innovative factory design.

If it had been the steam-powered era, Fordism would not have existed. Using electric motors, flexible assembly lines could be built. The famous conveyor belt could also be controlled at a constant speed.

'Bubble Theory' That Computers Could Not Escape Either

Computers were introduced into the workplace in the 1970s, but from 1976 to 1990, productivity in the U.S. economy actually declined. Large-scale investments in new equipment, along with technology, organizational changes, and employee retraining, required time before leading to productivity improvements.

The history of personal computers (PCs) shows a similar pattern. Although PC adoption expanded in the 1980s, productivity improvement was minimal. Companies invested huge amounts to introduce computers, but most employees used them only as simple typewriters. The dominant criticism at the time can be summarized as follows.

"You can see the computer age everywhere but in the productivity statistics."

This famous quote was made by Robert Solow, Nobel laureate in economics in 1987.

The change began in the mid-1990s with the advent of the 'Internet.' With the emergence of email, intranets, and e-commerce, computers evolved from simple document creation tools to core platforms for communication and business. The value of networked computers increased exponentially.

Essential Condition for New Technologies to Overcome Failures: Time

Andrew Ng, one of the 'four great AI scholars' and one of the most influential figures in the AI market, compared AI to 'electricity.'

Historical cases such as electricity and computers teach us an important lesson. For revolutionary technologies to realize their potential, the element of 'time' is essential. Not just vague waiting time, but the time society needs to accept and adapt to new technologies.

The AI revolution also requires fundamental reforms of existing production systems and structures, as seen in the cases of computers and electricity. This includes infrastructure construction, development of complementary technologies, changes in organizations and institutions, and learning and adaptation by users.

Currently, AI can be seen as going through this process. The emergence of generative AI like ChatGPT may be at a stage similar to the early inventions of the light bulb or PC. The AI revolution is still in its infancy, and it will take more time for our society to fully embrace and utilize AI. Just as electricity became more than a simple lighting tool and became a driving force for industrial innovation, AI will fundamentally change the way we work and the structure of society beyond being a mere tool for task automation.

Gutenberg invented the printing press and went bankrupt within a few years, passing away quietly. However, his printing press lifted the dark veil of the Middle Ages and triggered a knowledge revolution. We are enjoying its benefits today. Judging the AI revolution solely by its current limited impact may be premature. History teaches us that 'the true value of revolutionary technology gradually but surely reveals itself over time.'

Next Series Preview

④ AI: The belief that everything will work just by introduction (December 1)
⑤ Why do we waste time at airports? (December 7)
⑥ "Let's make something with that AI too," said the executive (December 8)