The United Nations’ “AI for Good” summit in Geneva feels less like a sterile tech conference and more like a high-stakes, slightly chaotic carnival of humanity attempting to steer a runaway train. Navigating the venue requires dodging live coding demonstrations, sensory-overload gadget displays, and earnest attendees wandering around in glowing headphones, listening intently to UN panel discussions. In the center of it all sits the “Networking Zone,” headlined by a bizarre, rotating seating contraption dubbed “UFOTECH,” which resembles a giant lazy Susan more than a place for serious deliberation. It is an apt metaphor for the event itself: a spinning, dizzying attempt to seat a diverse array of global stakeholders around a table that is constantly shifting under the weight of an unprecedented technological revolution.
Now in its tenth year, this summit serves as a diplomatic counterweight to the frantic, fear-driven legislative sessions happening in Washington or the corporate boardrooms of Silicon Valley. While the Western tech establishment spends its time debating the existential threat of “superintelligence” or navigating trade wars over high-end semiconductors, the ITU’s mission remains rooted in a stubborn, almost audacious idealism. Secretary-General Doreen Bogdan-Martin set the tone early, insisting that when AI is deployed with intention and care, it possesses the unique capacity to tackle our species’ most stubborn afflictions: chronic hunger, untreated disease, and the rapid degradation of our climate. Yet, even as she touted this optimism, there was an underlying tremor of reality; the very tools designed to “fix” the world are simultaneously creating new fractures in our social fabric.
The question of what “good” actually looks like—and who gets to define it—hung heavily over the 106,000-square-meter convention center. It was a friction point that turned the event into a study of contrasts: polished corporate presentations were constantly met with a low, persistent drumbeat of alarm regarding the influence of unchecked monopolies. There is a palpable anxiety that if we allow a handful of massive, opaque tech giants to dictate how AI is integrated into society, we are effectively hardwiring historical inequalities into the future. For many observers, the honeymoon phase of the tech industry’s relationship with international aid is definitively over, and the tone of the conversation has shifted from starry-eyed wonder to wary, protective skepticism.
Some of the most pointed criticism came from civil society organizations like Access Now,, where human rights defenders are calling for a total reevaluation of the public sector’s “best friend” relationship with big tech. The era of blind trust in corporate black boxes is closing, replaced by a demand for radical transparency that the industry is currently ill-equipped to provide. This tension spilled over into the physical realm when protestors interrupted Amazon CTO Werner Vogels mid-keynote, bringing the harsh realities of geopolitical conflict directly into the sanitized, climate-controlled comfort of the UN event. It was a visceral reminder that for “AI for Good” to be more than a slogan, it must address the ways in which these tools are being weaponized in the real world.
Beyond the ethics of deployment, a significant portion of the summit focused on the “how”—the cold, hard reality of the infrastructure required to run these models. Engineering experts, including Harvard’s Vijay Janapa Reddi, pointed out a massive disconnect: we are infatuated by the hype of AI, yet struggling to find ways to make it meaningfully functional in resource-constrained environments. Reddi’s critique is simple but brutal—in engineering terms, “good” is an abstract, useless word. You cannot build a solution if the definition of success is vague. He argues that we are currently designing planes that barely leave the ground, yet we are marketing them as a global solution to human suffering, failing to bridge the gap between “cool prototype” and “essential utility.”
Ultimately, the summit highlighted that AI has become the latest, and perhaps most difficult, hurdle in global development. We are witnessing a hardening of the “compute divide,” where access to the hardware—the physical chips and data centers—is becoming the new currency of power. If artificial intelligence is to truly serve the global population, it cannot remain a luxury good reserved for the wealthiest economies. The participants in Geneva left with a sobering mandate: if we treat compute as mere technology, we will fail. If, however, we treat it as basic development infrastructure—essential, like electricity or the internet—we might just have a chance to build something that actually performs the “good” we claim to seek.
