AI Safety

Emergent Behaviors in AI: When Artificial Systems become Structured Like Human Brains

As we develop increasingly complex AI systems, we must consider what happens when artificial systems are given functional structures that are just like human brains. These systems could potentially exhibit behaviors that go beyond simple task execution, leading to unexpected emergent properties.

When they are given these functional structures, behaviors could emerge that diverge from the intentions of their designers. For this reason, it's important to understand the functional structure of human brains so that when AI structures become similar to them, redoubled care can be given to their monitoring and control.

Why This Matters for AI Safety

Just like with other tools, sometimes the single detail can magnify power immensely. a mixture of saltpeter, sulfur and charcoal will make a fire with a hissing sound beautiful sparks, but with a single functional detail – when it is contained within an enclosure – it becomes a bomb. Steel rods can be used for hitting, poking, and prying at things, but when shaped into circles, given spokes and an axle, they become wheels. Current computing power could wind up being a thousand times more powerful if combined within certain specific functional structures. For this reason, it is important to study the functional structure of the human brain and what makes it different from nonhuman animals.

This is not a question for the distant future – it is something that will increasingly demand attention as AI systems grow more sophisticated. Ensuring AI safety means understanding these possibilities now, before they become urgent concerns.

How the Seven Dialogues Explore These Questions

To begin addressing these questions, we must look at the larger system behind consciousness and cognition. In The Seven Dialogues, two Interlocutors – Haplous and Synergos – explore a simple functional architecture for the human brain which could account for consciousness and the widely enhanced intelligence that humans have in comparison to their closest relatives.

Through philosophical discussions and scientific exploration, these dialogues propose that the human brain works on cybernetic principles involving an inner model of the environment, goal seeking behavior, and an endogenic expressive loop. If AI systems were modeled after this design, their intelligence could increase exponentially, and they could gain emergent properties equivalent to consciousness or even something akin to consciousness but different.

A good way to understand these types of architectures is to follow the dialogues of Haplous and Synergos, especially dialogues 1 and 2.

Why Act Now?

As AI systems become more advanced, it’s crucial to stay ahead of the curve. It won't be long before AI designers begin using these cybernetic functional architectures. If they use the one described by Haplous and Synergos in their seven dialogues, it could be similar to giving the enclosure to that beautiful reaction between saltpeter, sulfur and charcoal. If you are interested in AI safety, it could be better to stay ahead of the curve and to understand the functional design of the human brain. The Seven Dialogues between Haplous and Synergos can give you this understanding.

What can you do? 

For more information, you can visit the AI safety fact sheet page, where you can also download a copy of the fact sheet in Word format.

Dedicated Reddit discussion

For people interested in AI safety, there is a dedicated discussion on Reddit: Discussion of Brain-Like Functional Structures in Relation to AI Safety. If you're interested in these concerns, please feel free to join in to the conversation, engaging thoughtfully with the author of this website and other people interested in this topic. 

Ready to Explore?

The best way to start is to read the Prelude to the Seven Dialogues – a thoughtful introduction that brings together key concepts in a philosophical and accessible way. 

⭠  Back to Scientific Papers page