Market Shock and the Self-Driving Dilemma
The first to react to the AI revolution alongside the Real Camera broadcast was the stock market. The stock market, often regarded as a leading economic indicator, typically moves six to nine months ahead of the current economic conditions. In other words, stock prices rise or fall in anticipation of the economy six months down the line. Because every movement in stock prices generates massive gains or losses, the market fiercely attempts to predict the future. In that sense, Real Camera served as the ultimate promotional showcase for ID Group's future.
The new products, ideas, and Yoo Jae-won's development capabilities and vision that emerged during the broadcast clearly set ID Group apart from other conglomerates. The stock market, which had already stirred after the first episode aired, exploded following the second. The biggest surge came from ID Technology. Though ID Technology formed the backbone of ID Group, its listing had come relatively late. When it finally went public, its market capitalization exceeded two hundred billion dollars from the outset—a mammoth listing so large that it momentarily drained liquidity from the U.S. stock market and caused shares of other companies to dip.
The day after the second episode of Real Camera aired, ID Technology's stock price climbed more than ten percent. In Korean won, that translated to over twenty trillion won added in a single broadcast. Not only ID Technology but also Android saw gains, fueled by Yoo Jae-won's remark that AI Gold would be integrated into the Android operating system. Even Time Warner Nextcom, which had no direct connection to AI, experienced a rise—proof of the tremendous shock Real Camera delivered.
On the other hand, some companies bled because of the program. The day before the second episode aired, Apple took a direct hit to its vital spot. Apple had boldly unveiled its new iPhone 3G, a model that leveraged the company's greatest strength—its hardware craftsmanship—to resolve every criticism it had faced. From the camera module to main memory capacity, storage transfer speeds, and LCD resolution, the upgrades were substantial. Yet Apple had kept the price increase to a minimum, adopting a pricing strategy uncharacteristic of the brand. The top-tier iPhone 3G was priced at 799 dollars—100 dollars cheaper than the Android S2. It was a deliberate choice based on the strategic judgment that the device could not compete on price with Android smartphones.
Launch day brought an overwhelming response. But within twenty-four hours, the mood reversed completely. An AI assistant! An AI lens! Although no AI-related applications had yet been released, users who witnessed the practical applications of AI through Real Camera were simply ecstatic. After all, if the AI could achieve flawless autonomous driving in the F2 electric vehicle, there was no reason to doubt the performance of an app designed for Android smartphones. It was clear that an AI-equipped smartphone would be announced at IDDC 2003 in August, so countless consumers decided they would rather wait for the new Android model than purchase the new iPhone.
The stock market read these consumer shifts and reflected them in prices, resulting in ID Technology's surge and Apple's decline. Apple's atmosphere resembled a funeral. From CEO Steve Jobs down to the lowest-ranking employees, everyone had poured their hearts into the new product, confident of success after achieving the planned specifications and quality. Yet no one had imagined that a single entertainment documentary could derail their flagship device.
Steve Jobs, in particular, was profoundly shaken. While Apple's smartphone leadership lagged a step behind Android, he had always believed he could overtake them at any moment. After all, he had been envisioning smartphones for a very long time. Moreover, with Jony Ive—a designer who perfectly understood his design philosophy—by his side, nothing had seemed daunting. That confidence held until he witnessed the AI demonstrated in the second episode of Real Camera.
Steve Jobs was more shocked by the AI Eyes app than by the F2's autonomous driving. The moment he saw it, he recognized its immense potential for applications. Furthermore, when Yoo Jae-won mentioned the AI assistant, he made it clear that it went far beyond a simple encyclopedia or OCR function.
“Here is the cost estimate for building the AI system.”
Jeff Williams, Apple's Chief Technology Officer, handed a small document to Steve Jobs, who sat with his chin resting on his hand, deep in thought. Steve Jobs silently accepted the papers and began reading. The document outlined everything from the unfamiliar concept of machine learning to a brief summary of the costs involved in constructing a machine-learning AI.
“Hah!” A reaction escaped Steve Jobs when his eyes stopped at one particular section. “Two hundred million dollars?”
Two hundred million dollars—that was the cost required to build a machine-learning AI. Fortunately, Apple already had nearly twenty million iPhone users and possessed a certain amount of big data through its own internet services. However, it lacked the large-scale server systems needed to analyze that data. Jeff Williams had initially planned to begin machine learning with a system capable of one teraflop per second.
“The initial cost alone is two hundred million dollars,” Jeff Williams added, worried that Steve Jobs might misunderstand it as the total budget. Yet his voice lacked conviction. The field of machine learning was entirely unfamiliar even to him as Chief Technology Officer.
After Real Camera: Yoo Jae-won ended, he had frantically sought out experts for consultations and even asked unrelated parties to gather information on ID Group before finally grasping the basics. In other words, the two hundred million dollars covered only the physical infrastructure for the machine-learning system; no one knew how much more it would cost to elevate the AI to a level competitive with rivals. At least Yoo Jae-won's mention of statistics had greatly aided his understanding of machine learning.
“To put it simply, it is statistics on an enormous scale. Take the AI Eyes app, for example. ID Group's data centers likely contain an immense volume of image data. They analyze each image, process it into information, and then display the data most similar to whatever the user points AI Eyes at. This kind of work used to be done by humans, but it appears ID Group has found a way to make computers do it themselves.”
The more Jeff Williams studied machine learning, the more terrifying Yoo Jae-won seemed. “Even more frightening than AI Eyes is the language learning that began in 1992.”
1992! At that time, Apple had barely understood the internet, let alone AI. The company had researched low-power systems for PDAs and tablet computers and invested in ARM, which later helped create the iPhone. Yet their rival had already been thinking about AI back then—it was astonishing.
“If language analysis is applied to computer systems, it means humans and systems will be able to communicate through natural language. For instance, when sending a message, we have to manipulate the touchscreen manually, but over there, a single spoken command would suffice.”
“What I need to know is whether we can develop an AI-related app before August.”
It was now early March—four months and a few days remained to create an AI for the iPhone. “Do you really need an answer right now?” Jeff Williams replied with a cynical question. Having worked at Apple for as long as Steve Jobs, he was in a position to give such a response. Steve Jobs did not look away from him.
“We could make something that looks similar on the surface. We have already dabbled in big data ourselves. But turning that big data into actual intelligence is difficult, and more importantly, it requires a great deal of time.”
Even Jeff Williams could not claim to have fully mastered the theory of machine learning. Still, he could make a reasonable estimate. In four months, they could create something that merely resembled the AI Yoo Jae-won had shown on Real Camera. But they could not guarantee the quality of the result—it would obviously be poor.
“If we include something like that, we will only end up embarrassing ourselves. Besides, we cannot be certain whether the Android phone to be announced at IDDC in August will even feature an AI assistant, let alone AI Eyes.”
“Can you be certain?”
“Certain? It is merely analysis. The core of an AI assistant is natural language processing. Even Yoo Jae-won cannot conquer that so quickly.”
At Jeff Williams's irritated response, Steve Jobs smiled for the first time. It meant he agreed with Jeff's assessment. The shock from Real Camera had been overwhelming, yet recalling Jeff's explanation, he realized there had been no hint of an AI assistant at all.
“Moreover, the controversies that will arise because they moved first are only just beginning.”
From the issue of collecting big data to the ethical problems that would emerge when AI replaces human jobs, Jeff Williams believed the real challenges for AI were only starting now.
“Very well. It is not too late. Proceed with the Siri project immediately.”
With full authority over Apple's operations, Steve Jobs approved the two-hundred-million-dollar expenditure after a single conversation with Jeff Williams. In truth, he had already made up his mind internally. Though it was not too late, the sense of crisis—that falling behind in AI would mean the end—had struck him the moment he watched Real Camera.
Around the same time, in a lecture hall at Harvard University, Professor Michael Sandel was conducting his class “Justice: What's the Right Thing to Do?” Though the title alone carried a distinctly philosophical air, Professor Sandel's lectures were famous for using various examples to explain theories one by one. Because he made even dense philosophy engaging, the class that had begun small grew in enrollment each year. By 2003, it had expanded to fill an entire large lecture hall with hundreds of students attending simultaneously.
One of the famous examples in Professor Sandel's lectures was the trolley problem. Coincidentally, that was the topic for today as well.
“We have covered the trolley problem so many times that it feels slightly tedious. So today, I will modify the scenario slightly.”
A collective sigh of “Ah” rippled through the hall at Professor Sandel's words. Harvard students' academic zeal was world-class; even in liberal arts courses, it was natural for them to prepare in advance. Moreover, “Justice” was a class that had been repeated for several years, so there were plenty of references available. Yet now he was offering a variation on a past question!
“Here we have a self-driving car. Unfortunately, its brakes have failed, and there is no way to stop it through normal means. Tragically, five people are crossing the crosswalk directly in front of the malfunctioning self-driving car. So here is the question: Should the AI run straight through the five people on the crosswalk, or should it swerve and crash into the wall on the opposite side?”
Fortunately, the framework of the problem had not changed much from the classic trolley dilemma—only the trolley driver had been replaced by the self-driving car F2. However, because the entertainment documentary Real Camera had become such a hot topic, the students' concentration rose sharply. The moment Professor Sandel finished speaking, hands shot up throughout the lecture hall as students offered a variety of opinions. This scene was later reported by journalists who had been observing.
For Professor Michael Sandel and his students, it was merely another philosophical thought experiment and discussion. But unlike in the past, this particular debate carried implications far beyond the classroom. Because the topic concerned AI—an issue that would profoundly impact not only the automotive industry but the entire IT sector—the discussion held real-world weight. Given the massive resonance of Real Camera, people were intensely curious about what decision F2 would make in such a situation. Naturally, inquiries flooded in to ID Group and Yoo Jae-won—from Democratic lawmakers preparing autonomous driving legislation and reluctant Republican figures, to academics and the media—all waiting for Yoo Jae-won's response.