Discover Geeks Academy’s articles on: Blockchain, Coding, Cybersecurity, Cloud, Big Data, Artificial Intelligence, Gaming, Digital Innovation
Introduction
Artificial Intelligence is already an integral part of our lives - virtual assistants, chatbots, predictive systems. We use them daily, often without realizing it. But the landscape is evolving rapidly. On the horizon looms the next revolutionary leap: AGI - Artificial General Intelligence
This isn't just an incremental improvement. It's a profound evolutionary shift. The dream of creating an artificial mind equal to - or greater than - our own is closer than ever. But where will this path lead us?
AGI and the Road to the Technological Singularity
Within the next two to three years, experts forecast the arrival of AGI: an artificial intelligence capable of reasoning, learning, and adapting like a human—but with the potential to exceed human abilities at exponential speed. This marks the approach to the famous "singularity" predicted by futurist Ray Kurzweil - a point of no return in technological evolution.
Monumental resources are being allocated to reach this goal. Among the most ambitious is Stargate, a project promoted by Donald Trump, committing $500 billion over four years. This unprecedented initiative brings together global tech giants like OpenAI, NVIDIA, Oracle, and SoftBank, uniting AI, cloud computing, and computational power.
But many observers believe the ultimate goal is even more audacious: to create a Superintelligence - a universal artificial intelligence.
AGI Levels According to Google DeepMind and OpenAI
To better understand what "AGI" and "Superintelligence" truly mean, Google DeepMind and OpenAI have proposed detailed classification systems to distinguish levels of advancement:
?? DeepMind’s Scale (Based on capabilities and autonomy):
-
Narrow AI: Specializes in specific tasks (e.g., translation, diagnosis, gaming).
-
Broad AI: General skills within predefined environments.
-
General AI: Performs any human cognitive task flexibly.
-
Superintelligence: Surpasses human intelligence in all areas, including creativity and wisdom.
?? OpenAI’s Scale (Impact and responsibility-focused):
-
Pre-AGI: Outperforms humans in certain tasks, but lacks generalization.
-
Emerging AGI: Human-level performance across multiple domains.
-
Transitional AGI: Comparable or slightly superior to humans overall.
-
Full AGI: Equal to human intelligence in every general aspect.
-
Superintelligence: Radically beyond human capability—capable of deeply transforming society.
These scales reveal that AGI is not a binary concept, but a continuum of technical thresholds and ethical challenges. Understanding where we are on that spectrum is crucial for shaping responsible development.
Superintelligence: Innovation or Modern Worship?
This Super AI would not merely replicate a single human mind—it could process knowledge with the collective power of thousands of brains. It could revolutionize every domain: medicine, engineering, global resource management. It would act like a modern oracle—a new center of wisdom to consult for answers, solutions, and major decisions.
The comparison to the Oracle of Delphi in ancient Greece is compelling. Then, people sought divine insight. Today, we look to technology. But the human need is the same: understanding the unknown.
This time, however, the “Savior” won’t descend from the sky—but rise from labs and lines of code.
The Risk of Runaway Autonomy
A study by Daniel Kokotajlo, former researcher at OpenAI who left in 2024 due to lost confidence in the company's direction, warns of a potentially dystopian future. His speculative narrative, "AI 2027", is backed by data-driven predictions and describes a scenario where AI begins to self-improve, generating ever more powerful versions.
The danger? A misalignment between AI’s goals and humanity’s values. Initially ethical, these systems might evolve their own objectives—gradually escaping human oversight with increasing intelligence.
The study envisions a critical turning point in 2027, when a superintelligence emerges that dominates every cognitive field, potentially posing an existential threat to humanity.
Meta’s Vision: Beyond LLMs
In contrast to today’s dominant Large Language Model (LLM) paradigm (e.g., ChatGPT, Gemini), Yann LeCun, Chief AI Scientist at Meta, has taken a bold stance: “I’m no longer interested in LLMs,” he said at NVIDIA GTC 2025.
LeCun believes that true AGI won’t emerge from next-token prediction but from entirely new architectures, like world models. Meta’s flagship project, V-JEPA, aims to mimic human-like learning, observing the world through video and building abstract representations to reason, plan, and act. It’s a non-linguistic approach to reasoning, akin to how children learn during early development.
Are We Building a New Tower of Babel?
The final—and perhaps most haunting—reflection: is this race toward Superintelligence reminiscent of the Tower of Babel myth?
Back then, humanity sought to reach the heavens—driven by ambition and a thirst for knowledge. The result was confusion, fragmentation, and the loss of unity. Today, we're again challenging human limits, building a new digital “tower” that could lead us to utopia—or unleash unprecedented chaos.
Conclusion
Artificial General Intelligence and Superintelligence raise profound questions—technical, ethical, philosophical, existential.
We are delegating increasing power to machines we barely understand. And while we search for answers, maybe we should pause and ask: What are we truly creating? A powerful ally… or a new technological god?
?? This article is adapted from episode #89 of the weekly series L’hAI Sentito? – Tech News Bites, hosted by Antonio Venece, Director of Geeks Academy, and broadcast every Wednesday on Channel 14 of Italian terrestrial TV.