FLASH
Shkruar nga Victor Menaldo
The invention of electricity did not mark the end of the world. Nor will today's innovations...
Artificial intelligence can perform more and more of the tasks of an office worker. What it can't do is guarantee the result.
However, a new genre of workplace doomsday stories is warning that artificial intelligence will end the traditional nine-to-five workday, claiming to be even scarier than classic zombie movies. Interviews with executives at major companies warn that artificial intelligence will make many high-paying jobs redundant. Dario Amodei, CEO of Anthropic, has warned that artificial intelligence could eliminate up to half of entry-level jobs within five years. Not to be outdone, executives at JPMorgan, Ford, and Amazon (whose founder, Jeff Bezos, owns The Washington Post) have also predicted fewer high-paying jobs.
As with any good horror movie, the fear that this idea fuels is not entirely unfounded. After all, the movie Jaws was inspired by real-life shark attacks, and unemployment among recent graduates has been over 5 percent, compared to about 4 percent for the entire workforce. Even more worrying is that early-career workers in the fields most exposed to AI—like software development and customer service—have seen a relative 16 percent decline in employment since the advent of generative AI.
But the doomsday scenario relies on faulty arithmetic: if artificial intelligence can perform 40 percent of a job’s tasks, then 40 percent of that job is assumed to disappear. In fact, a job is not simply a list of discrete tasks. Companies employ people to perform sets of tasks linked together by professional judgment, coordination, customer trust, and accountability. When tasks complement each other, automating some of them does not necessarily diminish the role of humans. On the contrary, it can increase the value of the human work that remains.
When the power loom automated much of the weaving process in the 19th century, many handloom weavers lost their jobs. In mechanized factories, however, workers who learned to use the new machines became more valuable, not less. They still had to tie knots, quickly replace worn wires, adjust thread tension to prevent breakage, and supervise several looms at once. Those who mastered these skills earned more, even as the number of looms multiplied and cloth became cheaper.
Despite its fluency, artificial intelligence is not an omniscient mind that does your work for you or better than you. It is a prediction engine trained on a large portion of human writing.
This gives the system tremendous scope, but also huge gaps. It can’t tell you how biased the sources it’s trained on are, it can’t explain the reasoning process in an auditable way, and it can sound just as convincing when it’s inventing a quote as it does when it’s reporting a fact. In this sense, it’s the opposite of its closest predecessor, statistics, which tells you how much confidence to give a prediction. Statistics provides margins of error; artificial intelligence offers none.
Precisely because artificial intelligence cannot determine the limits of its error, an organization that relies heavily on it can find itself operating in a fog of uncertainty. This is where the real economic barrier comes in. When a compelling-sounding result becomes cheap and immediate, the value shifts to the judgment needed to determine whether a perfectly formulated response from a chatbot is a breakthrough, a dead end, or simply a hallucination.
The editor who distinguishes a valid argument from a fraud, the doctor who identifies a prescription with an impossible dosage, and the lawyer who realizes that the cited precedent does not exist, become even more valuable in a world filled with machine-generated content that sounds convincing, confident, and is often wrong.
A company that fires the employee who can assess whether a result is reliable is in effect automating quality control. This may seem efficient for a quarter or two, until a well-worded lie influences a decision with consequences. Responsibility does not transfer with the job. When a fabricated quote ends up in a legal document or an inaccurate figure enters an audit, the responsibility remains with the people who approved it. No board of directors can answer a regulator with the words, “That’s what the model said.”
The mistake in these apocalyptic predictions lies in treating artificial intelligence as a product that simply replaces something else. Since artificial intelligence is a general-purpose technology, like electricity or computers, and not a single device like a toaster, the economy will be reorganized around it. Electricity did not transform industry simply by replacing steam engines. The big change came when production was reorganized around the new technology: small electric motors that powered individual machines, more flexible factory organization, and reliable power supplies throughout the production process.
The same thing is happening with artificial intelligence. Someone has to integrate these models into workflows, connect them to existing systems, decide which judgments can be delegated to them, review the results they produce, retrain staff, and correct the consequences when they fail.
This transition will not be easy. The pressure on entry-level employees is real, and companies that confuse automation with strategy will cause real damage in the process.
But the answer to a tool that is fluid, fast, and often fallible is not to seek a profession immune to artificial intelligence, nor to hand over the thought process to a machine. The answer is to become a human who knows when a machine is wrong, when its output can be improved, and when it is reliable enough to carry your name. In this sense, artificial intelligence is like every other powerful tool that has come before it: the employee who knows how to use it well is worth more, not less. It has only increased the value of professional judgment. / Adapted from "Pamphlet" by "TheWashingtonPost"
Lini një Përgjigje