Remember job applications? You know, you write a resume and a cover letter. Then someone will read them. You'll want to dig out a tasteful button-down shirt and try to impress the interviewer. Maybe a quip or a question about your amateur bowling career will land you the job. Maybe you'll never hear back again.
This is all changing. When Christian Vinson applied for a job in finance this year, he knew he was actually applying to artificial intelligence. Many companies now use screening software to sift through mountains of job applications, so he spent nearly 10 hours letting two artificial intelligence chatbots learn about his work experience. He filled in details about his educational background, a banking internship and five years of military service. He then gave instructions to the bots to tailor his cover letter to hundreds of job postings.
An AI wrote his job applications, and it was an AI that read them—like a dog chasing its own tail. “This is the new normal,” said Vinson, 30, who eventually found a job at an investment bank in New York.
We live in a world where AI tools have penetrated the core of a staggering number of professional, academic, and personal interactions. Lately, these chatbots and intelligent agents have done more than just talk to us. Students use ChatGPT to write papers, and teachers use artificial intelligence to grade papers. Employees send lengthy AI-generated emails to colleagues, who respond with lengthy chatbot texts they generate themselves. AI Podcast hosts and AI Podcast guests dish on topics like gas prices, the war in Iran, and artificial intelligence. The tech world wants to take this further: In March, Meta acquired a social network designed for bots to talk to each other.
There is a popular concept in the technology circle called "dead network theory", which believes that the vast majority of digital traffic will eventually be generated by software, and human voices will be reduced to a minority. We may now have a corollary to this prediction—call it the “robot loop”—where humans on both sides of the interaction relinquish their roles to artificial intelligence. This kind of recursive communication can be creepy and error-prone, and a simple mistake can get passed back and forth uncorrected. Even stranger, they turn our conversational dynamics on their head, turning chatbots into interlocutors and humans into their facilitators.
Depending on who you ask, robot loops can be banal and dystopian, or they can be genuinely useful. Regardless of how people feel about this, researchers say we will see robot loops more and more frequently as we increasingly entrust artificial intelligence systems to do our bidding.
Someday in the not-too-distant future, a water pipe might burst and you'll be in dire need of a plumber. Maybe you'd call 10 plumbers yourself to get a price, but maybe you'd hand all those dials over to a personal digital assistant that can perform tasks like sending emails, buying plane tickets, and writing code. Meanwhile, Jiaxin Pei, an assistant professor at the School of Information at the University of Texas at Austin, said plumbers may struggle to cope with the influx of calls from tireless artificial intelligence assistants. Maybe they’ll assign their own smart agent to act as a virtual receptionist. “This is the endless cycle we are creating,” Pei Jiaxin said. How will all this affect the way we interact with each other?
In some ways, human history is a long history of panic over the potential of new technologies to change the way we communicate. Socrates was worried that writing would destroy our memory—we only know this because Plato wrote it down. Critics in the early 20th century believed that the telephone posed a threat to civilization. Critics have predicted time and time again that the conveniences of modern life may lead to a decline in the quality of human interactions. The author of a 1994 New York Times article argued that e-mail is "so compelling and convenient that its sheer volume of information and lack of consideration may undermine its usefulness," although he acknowledged that e-mail does have its benefits: "No more stamps to lick."
None of these innovations ended a productive discussion, although they each changed the texture of the discussion. Today, we can communicate with more people than ever before without being in the same physical space. As a result, researchers believe we may be actually talking to each other less. Many of our conversations take place on platforms that encourage certain types of communication, especially those that are intense, quirky, or otherwise have the potential to retain our attention for the financial benefit of the platform.
Still, AI is different from these previous advances. While email transfers information between people, chatbots imitate humans and may even replace them. Suddenly, humans are left to decide for themselves whether they would rather talk to a real person or a program that approximates a real person. Many people have already decided that they would rather turn to a chatbot for financial advice, medical diagnosis, or companionship.
People are already using AI-generated avatars to speak for them. When Bay Area pastor Justin Lester isn't busy, people can talk to his "digital twin." The next step is likely to be a future where people's own digital twins might talk to Leicester's digital twin. Our digital middlemen will increasingly collide. We talk to our own chatbots, and our chatbots talk to each other. CEOs have created avatars powered by artificial intelligence that can hold meetings with their employees. How long will it be until their subordinates send avatars to attend these meetings?
Many people are already living in this wonderful new world. Over the past few months, I've interviewed them. A graduate student used artificial intelligence to complete his homework, and the professor's virtual assistant graded him. One telecommunications executive noticed a customer sending an AI voice agent to argue with his company’s AI customer service system. Another man used a chatbot on a dating app to draft a message to a woman who appeared to be doing the exact same thing. "How long are we going to chat GPT with each other?" he finally wrote. But she never replied.
Most of them admitted that there was a hollow feeling to the interactions, that while they followed the basic framework of human conversation, the texture was completely wrong. A single chatbot can generate a deluge of text that is as fake as cheap jewelry. Put two robots together and have a conversation, and they'll produce a steady stream of conversation that, while following all the rules, is jarring because it's missing certain key elements: curiosity, empathy, and real experience.
