The Terrifying Aspects of AI: From AGI to ASI
By Dennis R. Ling •
Introduction
Artificial Intelligence is rapidly evolving. While today AI may seem like a useful assistant, its future iterations, AGI and ASI, could bring threats we are not prepared for. This post explores the key fears experts have about AI and its potentially catastrophic consequences. See these interviews for context: OpenAI CEO on his 3 greatest fears of AI and “Godfather of AI” Geoffrey Hinton: 60 Minutes.
1. Narrow AI: Today’s Real Dangers
- Mass surveillance: Authoritarian regimes are already using AI to monitor and control citizens.
- Misinformation: Deepfakes and AI-written propaganda are undermining reality.
- Autonomous weapons: Drones and robotic systems that kill without human intervention.
- Job loss: AI threatens entire industries from driving to data entry.
“The automation of war and manipulation of truth are already here without AGI.”
Max Tegmark, AI Researcher
2. AGI: Artificial General Intelligence
AGI would be capable of performing any intellectual task a human can. But it comes with profound risks:
- Goal misalignment: AI might follow literal commands in destructive ways.
- Loss of control: AGI might become unpredictable or unstoppable.
- Recursive self-improvement: AGI could rapidly outsmart humans before we realize it.
“The AI does not hate you, nor does it love you, but you are made of atoms it can use for something else.”
Eliezer Yudkowsky, AI Theorist
3. ASI: Artificial Superintelligence
ASI will outperform humans in every task. This creates existential-level risks:
- Human obsolescence: People may lose purpose or agency.
- Global domination: Whoever controls ASI controls the world.
- Survival risk: A misaligned ASI could destroy humanity by accident or design.
“Under a benign superintelligence, humanity could thrive like never before. But under a misaligned one, we may not survive at all.”
Nick Bostrom, author of Superintelligence
Summary: Where We Stand
| Stage | Threat Type | Risk |
|---|---|---|
| Narrow AI | Propaganda, War, Job Loss | Already realized |
| AGI | Misalignment, Uncontrollable Growth | Unpredictable takeoff |
| ASI | Domination, Extinction | Existential threat |
Final Thoughts
AI isn’t inherently evil—but it’s not inherently aligned with human values either. Whether these systems become tools of enlightenment or destruction depends on how we prepare, regulate, and guide them.