Parents, teachers, and employers are rightly concerned about the potential threat that large-scale language models (LLMs) may pose to traditional pedagogy. Are we building a generation of young people who have to look to ChatGPT for guidance in every aspect of their daily lives? But the real culprit is not artificial intelligence. Social media is disrupting our ability to concentrate, and that’s where regulators should focus.
Considering how easy it is to get good answers to your questions from your favorite AI chatbot, it’s tempting to think that ChatGPT 3.5, released in November 2022, marked a turning point in turning humans into brainless zombies who can barely tie their shoelaces without consulting their smartphones. Many educators would reinforce this view by pointing to the torrent of generic, flavorless, AI-generated essays that must be graded, submitted by students who are incapable of or uninterested in answering the most basic questions in the course material when placed in a traditional exam environment.
But in reality, many of these trends existed long before OpenAI turned the world upside down three years ago. In the 1990s, when the Internet was an emerging technology, Google and other traditional search engines faced many of the criticisms we hear today in the LLM era. In other words, it undermined learning and encouraged copy-pasting instead of critical thinking. Similar discussions were held about GPS, calculators, printing presses, and other early technologies.
For a small number of students, the anxieties of the internet age have become a reality, creating the need for plagiarism prevention software like Turnitin. But for most people, the Internet has greatly expanded their productivity and enhanced, rather than diminished, their ability to acquire and retain knowledge. This is most clearly seen in the explosion of research activity facilitated by the World Wide Web over the past three decades, with a surge in international collaboration and an important conduit for the immediate dissemination of new discoveries.
Therefore, we may be more optimistic about how AI can enhance our brains. A particularly notable benefit is customized learning. LLMs are skilled at teaching students one-on-one by identifying their weaknesses and determining the best way to address them, aided by an almost limitless level of patience that even the most forgiving human teacher could never possess. Simply put, LLM complements human cognition when people can maintain deep attention.
The problem is that people are losing their ability to maintain deep attention due to systematic efforts by social media companies. Their revenue model is based on micro-duration engagement and dopamine reinforcement loops that reduce tolerance to cognitive friction and reduce sustained attention. Over time, the benefits of this shallow processing led to people reading and learning less than they used to. This clearly predates the current LLM boom.
Under fragmented attention, LLM switches from a complement to a substitute for human cognition. Use LLM to clarify arguments, generate ideas without thinking rather than stress test them, and use LLM to avoid cognitive strain. One manifestation of this mechanism is the large number of studies that have demonstrated the productivity gap caused by AI. Elite workers are using AI to further separate themselves from less capable colleagues. Some of them outsource their thinking to LLMs and spend the time they save scrolling through their Facebook feeds instead of thinking about harder problems.
Governments must therefore resist the temptation to limit the use of LLMs and instead focus their efforts on the phenomena that are causing fragmentation of attention.
It is no exaggeration to compare social media to an addictive drug.
It is no exaggeration to compare short-form social media to an addictive drug with all its negative psychological and physiological effects. We cannot expect most individuals to take sufficient responsibility to avoid drug consumption as a personal choice and to protect themselves and society as a whole through a combination of legal and educational means. The same argument applies to short form social media.
For those who frequent the airport, moving walkways will be easy to understand. Something that allows you to move faster without expending extra energy, or something that allows you to move slower without expending energy. If too many people are choosing to stop on moving sidewalks, the ideal solution is not to ban belts. It’s so that when people get on it, they can choose to keep walking.
In the age of LLM, if we want to preserve human intelligence, we should fight artificial distractions, not artificial intelligence.
