SoC #6 - AI Isn’t the Enemy—Our Fear Is

Last Thursday night, I found myself at a small restaurant with nine friends—a cross section of the AI world. I was there from OpenAI; a few were from Anthropic and a couple from Google DeepMind. One friend had just quit his job at an AI lab to pursue one of his oldest passions: drawing manga.

Our conversation floated, as it usually does, between our slightly obsessive interests—mine happen to be literature and classical music—and our excitement about the new generation of agentic coding models that might soon liberate us from excruciating tasks (no one should ever have to manually review a 9,000-line PR). At that table, we were simply ten friends who love AI and who work, in our own ways, to ensure its impact on the world is as positive as possible. There were no “OpenAI vs. Anthropic” conversations. We were unified—even if there is little intellectual overlap between Dostoevsky, Formula 1, and geology.

A few hours later that same night, the news broke that Sam Altman’s home had been attacked. I was shocked and deeply unsettled. Sam has a family; he has the most unbelievably adorable baby boy. To many of us, Sam is an emblem of technological progress. He is one of the main reasons most AI labs exist today: he took an ex-ante improbable bet on a technology that most of the world assumed would never work—and he happened to be right. Yet an unfortunate cocktail of incendiary headlines and inaccurate narratives has turned him, and at times AI more broadly, into a kind of scapegoat for every existing human fear: unemployment, doom, war.

Humans need explanations. When we can’t make sense of something complex, we tell a simpler story. We point at someone to blame; it lightens the cognitive load. Much of my academic research before joining OpenAI revolved around scapegoating. Historically, humans start searching for scapegoats in moments of great uncertainty and economic or political crisis—the Black Death in the 14th century, the persecution of “witches” in the 1600s, or the constant scapegoating of immigrants in moments of economic fragility.

We do not have to repeat those patterns. We need to do a better job at communicating; we need to participate, more genuinely and more frequently, in the economic and political debates around the technology. We also need to be both honest and humble about the fact that none of us know exactly how this will play out.

We do however have history on our side. Over the last two centuries, scientific and technological innovation has been the primary driver of human progress. The printing press democratized knowledge and paved the way for the Reformation, the Enlightenment, and eventually the modern world. The steam engine and electricity generated the industrial revolutions that drastically expanded the size, complexity, and wealth of human economies. The discovery of antibiotics added decades to our life expectancy. Some economists even argue that the washing machine—because of the dramatic decrease in time dedicated to domestic labor—might be the single most important event for the emancipation of women in the 20th century.

The numbers are staggering: 200 years ago, 90% of the world’s population lived in extreme poverty; today, it’s about 8%. Global average life expectancy jumped from approximately 30 years in 1800 to 73 years in 2023. It seems almost self-evident that we feel happier, more fulfilled, and more hopeful when we are not sick, when we have enough food, when we have additional free time to spend on what we find meaningful and fulfilling.

We therefore have a duty to be optimistic about the future—otherwise why would we be working this hard to build it? But optimism is not enough. We need to be extremely intentional about making sure we optimize for that exact future. We’ve all witnessed first-hand the effects of social media on our collective well-being. I don’t believe the people who built it were evil; I don’t think they were even wrong. They just didn’t know that when you design an algorithmic ranking system for the first time in history, you’re accidentally creating the largest addiction machine the world has ever seen.

With AI, we have a chance to do things differently—to build a technology that meaningfully frees people from the excruciating, menial parts of work and allows them to invest their attention and energy in what is more interesting, creative, and fulfilling.

Technology is always double edged. The very thing that can make us more fulfilled and free is the same thing that can make us lose ourselves. That’s why we need to spend time and research funds verifying—through long-term, rigorous studies and carefully crafted experiments—that the technology we’re building is indeed increasing humans’ sense of agency and autonomy. That is the bar; I think it’s an achievable one.

I worry that AI might become a new source of political polarization and of irrational collective fear—just as almost every great technological revolution has in its early days. When the printing press emerged, many believed it would create an “information overload” and erode our memory (Plato wrote the same thing about writing). Fears about the introduction of the railroad, the telephone, the radio, and the Internet were all, at some point, as intense as some of the fears we are hearing today about AI.

Yet, sitting around a dinner table, all of us—OpenAI, Anthropic, Google DeepMind—are just a bunch of nerds who get abnormally excited about technology. Everyone is working extremely hard to make sure the best possible world materializes—one in which there is an abundance of economic resources and, most importantly, where humans have more agency and autonomy.

I really believe in that. And I hope that our work in the next few years can prove it.

SoC (Stream of Consciousness) are short musings on AI, philosophy, and the directions intelligence is starting to take in our life.

I also love that stream of consciousness abbreviates to SoC: a system-on-a-chip compresses an entire architecture into something small and self-contained.

That’s what I want these pieces to be too — compact architectures of thought.

— With Love, Houda