As artificial intelligence continues to advance at an astonishing pace, even the leaders at its forefront are beginning to voice concern. OpenAI CEO Sam Altman has openly admitted feeling “scared” about the upcoming release of GPT-5, the newest and most powerful version of ChatGPT. Scheduled to launch in August 2025, GPT-5 is expected to significantly surpass its predecessors in speed and capability. Yet it is precisely this rapid evolution that has led Altman to compare its development to the Manhattan Project. His candid remarks, shared during a recent podcast appearance, highlight growing unease about the lack of oversight, ethical uncertainty, and the accelerating race toward Artificial General Intelligence (AGI).
Speaking on Theo Von’s This Past Weekend podcast, Altman confessed that testing GPT-5 left him with a profound sense of unease. “It feels very fast,” he said, referring not just to the model’s impressive processing power, but to the alarming speed at which AI is advancing overall. His remarks echoed the reflective anxiety of scientists involved in the Manhattan Project, who, after creating the atomic bomb, grappled with the unintended consequences of their innovation.
Altman invoked a similar moment of reckoning, asking rhetorically, “What have we done?” The comparison underscores his concern that GPT-5, like nuclear technology before it, may mark a pivotal turning point—one where humanity’s ability to build something outpaces its readiness to understand or control the fallout.One of Altman’s most pointed comments was about the lack of control in AI development. “It feels like there are no adults in the room,” he said, suggesting that regulatory frameworks have not kept up with AI's breakneck speed. While OpenAI advocates for responsible deployment, the pace at which GPT-5 was built may have outstripped safety protocols and human readiness. These remarks have sparked fresh debate about who, if anyone, is truly in charge of monitoring such transformative technologies.
While specifics about GPT-5 remain under wraps, insiders suggest it will feature major upgrades over GPT-4, including improved multi-step reasoning, longer memory, and more advanced multimodal capabilities. It is also expected to reduce model-switching delays, offering more seamless user interactions. But with power comes risk. Altman has openly said that GPT-4 “kind of sucks” compared to what’s coming next. That raises alarms about how much more potent and possibly uncontrollable GPT-5 could be.
