In 2015, Tim Urban published one of the most thought-provoking deep dives into artificial intelligence written for a general audience. Titled The AI Revolution: The Road to Superintelligence, the post laid out a vision that was both awe-inspiring and deeply unsettling. It painted a picture of humanity standing at the base of a steep technological cliff, about to take a leap into either paradise or oblivion, depending on how we handled the coming explosion in artificial intelligence.

Now, a decade later, it’s worth asking: how accurate was Urban’s vision? What are we doing well—and where are we failing—based on the roadmap he sketched out?

The 2015 Vision: A Quick Recap

Urban divided AI into three stages:

• Artificial Narrow Intelligence (ANI): Machines that are better than humans at very specific tasks (e.g., chess, math, image recognition).
• Artificial General Intelligence (AGI): Machines with human-level intelligence, capable of performing any intellectual task a human can.
• Artificial Superintelligence (ASI): Machines vastly more intelligent than the best human minds in every field.

He emphasized how fast the leap from ANI to ASI could happen once machines can improve themselves. This recursive self-improvement could lead to an “intelligence explosion,” taking us from basic AI tools to godlike digital minds in a matter of years or even months. Urban didn’t claim this would happen soon—but he argued convincingly that when it did happen, we wouldn’t be ready.

Where We Are Today (2025)

What We’re Doing Well

• Mainstream Adoption of ANI: Urban’s predictions about ANI were spot-on. Today’s AI systems outperform humans at a growing list of narrow tasks—from diagnosing certain diseases to generating realistic text, images, and even video. Tools like GPT-4.5, image generators, and autonomous agents are embedded in business, education, and creative workflows. The utility of ANI is undeniable.
• Serious Research on AI Alignment: Although most would say we haven’t reached AGI, the field of AI alignment and safety—once a niche concern—has gone mainstream. Organizations like OpenAI, Anthropic, and Google DeepMind have internal alignment teams. Governments and global coalitions are also beginning to draft policy guardrails. This shows a growing awareness of the risks Urban highlighted.
• Public Awareness: Thanks in part to media coverage, AI is no longer an abstract idea. Public discourse is buzzing with questions about automation, bias, ethics, and safety. That collective awareness is a powerful step in the right direction.

Where We’re Falling Short

• Lack of True AGI Governance Frameworks: Urban warned that AGI and ASI could arrive suddenly—and without global coordination, that could be catastrophic. Despite recent efforts, international collaboration on AI safety remains fractured. Nations and corporations continue to race ahead competitively rather than cooperatively, which could create dangerous incentive structures when AGI development heats up.
• Short-Term Thinking: While Urban stressed long-term outcomes (like existential risk or utopia), many AI developers and stakeholders are still focused on near-term applications: user engagement, profits, and market share. This short-sightedness makes us vulnerable to the "careless parent" problem—where we birth a super intelligent entity without properly understanding or guiding it.
• Misalignment Between Models and Human Values: Even with modern AI alignment research, most systems today remain deeply misaligned in subtle but critical ways. They can be gamed, manipulated, or exploited. We’re still far from building systems that reliably internalize and act on nuanced human values, as Urban warned.

What Tim Urban Got Right

• The rapid growth of ANI capabilities and public fascination with AI.
• The stark contrast between technological progress and our preparedness to handle its consequences.
• The "calm before the storm" nature of the present moment—where everything feels manageable, but the stakes are quietly rising.

What’s Still Unclear

• Will AGI emerge suddenly, or gradually? Urban leaned toward a rapid takeoff model, but many experts today are divided. Some think we’ll have a slow ramp-up to AGI that gives us time to react—others still fear the "overnight explosion" scenario.
• Can we ever truly align super intelligent systems with human values? This remains one of the most profound unanswered questions in all of AI research. If we can’t, Urban’s fears about existential risk may yet prove prophetic.

Final Thoughts

In hindsight, The AI Revolution wasn’t just a speculative piece—it was a wake-up call. A decade later, we’ve made progress in all the ways that count technologically but remain alarmingly behind ethically, politically, and philosophically.

Urban wrote that humanity might be a child playing with a bomb, unaware of the power it holds. In 2025, we’ve learned how to start handling the bomb.

The question remains: do we fully understand what happens when it goes off?