In recent years we’ve come to know AI mainly through its ability to generate: text, images, video, code. This capability has by now become part of our creative and professional processes, and we’re growing accustomed to it in a certain sense.
But today AI is evolving further: from the pure ability to generate content, it’s moving toward the ability to reason, and from reasoning to action, through the activation and intersection of multiple capabilities.
It no longer simply responds to an input, it autonomously finds a way to achieve a given goal.
This shift from simple input-output generation to autonomy is bringing us closer to one of the most discussed and controversial concepts in AI research: the arrival of AGI (Artificial General Intelligence).
There is still no universally accepted definition of AGI, but the term generally refers to an Artificial Intelligence system with cognitive capabilities general and versatile enough to organize itself to carry out a complex task — one whose completion requires a series of interconnected operations — and to solve a wide variety of problems, even very different from those it was originally trained on, without any human intervention.
An AGI is presented with a new problem, never encountered before, and is able to examine it, understand it, identify what it’s missing in order to solve it, carry out the necessary self-training to begin working, plan a strategy, select and use the most suitable tools, verify the result, and correct its own mistakes until it reaches the best possible output.
We don’t yet know when we’ll be able to say we’ve reached AGI, nor is there a precise line beyond which we could declare ourselves in the AGI era, but some of the capabilities we would have associated with AGI just a few years ago are already starting to appear in the systems we use today.
Even though the change we’re witnessing isn’t as dramatic as the arrival and spread of Midjourney or ChatGPT, it’s probably much more profound.
These are roughly the steps we’re seeing unfold, and what lies ahead:
Chatbot → Reasoner → Agent → AGI → Superintelligence
AI has already shown us it’s capable of astonishing things. We’ve seen the generation of content of every kind: images, music, films, websites, video games, apps, 3D models, strategies, research, advertising campaigns, and much more. Today we’re beginning to see AI take control of our computers to interact with the apps on them and carry out even more complex operations, leading to outputs once unimaginable.
Capabilities that, not long ago, we probably would have placed in the description of a future AGI.
But that doesn’t mean AGI has already arrived. One of the main problems with contemporary AI is what’s called “jagged intelligence” — an irregular or uneven intelligence: a system can achieve extraordinary results on extremely complex problems and, shortly after, make surprising mistakes on seemingly much simpler situations.
The distance from AGI, then, doesn’t depend only on how “intelligent” a model is, but above all on how reliable and autonomous that intelligence is. Current systems can still produce incorrect information, misread a situation, or follow the wrong strategy. Their ability to work autonomously over long periods remains limited, and continuous learning from experience is very different from the human kind.
There’s also a huge gap between digital intelligence and the ability to act in the physical world: while AI models have reached impressive levels across numerous cognitive tasks, general-purpose robotics remains far behind.
The most important transformation has already begun: we’re witnessing the gradual shift from systems that produce something on request to systems capable of reasoning about a goal and acting autonomously to achieve it. It won’t be long before we can look toward the next and presumably final evolutionary step: Artificial Superintelligence (ASI).
The question is no longer “what can Artificial Intelligence do?” The more interesting question is “how can it revolutionize our lives?“
ANGELO FERRETTI
My name is Angelo, and for over 20 years, I’ve been passionate about teaching and creating images focused on architecture and interior design.