Exploring the Origins of the "Self": Insights from Biology to AI

???? Recently, I delved into a fascinating paper titled "Reafference and the origin of the self in early nervous system evolution" by Gáspár Jékely, Peter Godfrey-Smith, and Fred Keijzer. It offers thought-provoking insights into how the 'self' emerged in biological systems, particularly during the early stages of nervous system evolution.

The study focuses on the concept of 'reafference' - the impact of an organism's actions on its sensory mechanisms. This concept is pivotal in understanding the interplay between an organism’s sensory experiences and its actions, marking the beginning of what we might call a primitive sense of self or 'body-self' in early multicellular organisms.

Early in the history of life on earth, a fundamental concept of 'body-self' evolved, most likely as early as unicellular organisms. 'Body-self' emerged as an essential feature of life during the emergence of simple multi-cellular animal life, without needing a highly organized organ like a brain. Such an organism has a 'body-self' if it has a particular form of organization. That form of an organization includes motility and sensing, where action and sensing are tied together through 'reafference'; which is defined as any effect on an organism’s sensory mechanisms due to the organism’s own actions. The 'body-self' enables the organism to sense and act as a single unit. It possesses a self that separates itself from the world.

???? Key Aspects of the Biological 'Self':

1. Reafference Principle: A groundbreaking idea suggesting that actions initiated by an organism lead to sensory changes, integral for interacting with its environment.

2. Body-Self: An organizational form where organisms act and sense as cohesive units, integrating motility and reafferent sensing.

3. Reafference Types: The paper makes a key distinction between sensing internal body changes (deformational reafference) and sensing environmental movement (translocational reafference).

4. Corollary Discharge: This mechanism tracks the sensory consequences of an organism's actions, distinguishing between self-induced changes and external stimuli.

???? From Biology to AI: This research has implications for the field of Artificial Intelligence, particularly Large Language Models (LLMs) and Artificial General Intelligence (AGI). Here are some parallels and insights:

1. Feedback Loops: Like biological systems, LLMs and AGIs can benefit from incorporating feedback mechanisms, although the 'self' in AI lacks physical and sensory experiences.

2. Integration and Coordination: The integration of sensing and action, crucial in early nervous systems, parallels the need for coherent integration of various modules in AGI.

3. Predictive Understanding: Just as biological systems understand the consequences of their actions, AGI should be capable of predicting and understanding the implications of its actions.

4. Adaptive Learning: Learning from sensory input in biology is akin to adaptive learning from actions or outputs in AI systems.

5. Self-Concept in AI: While biological evolution led to a body-self, the challenge in AI is to develop a coherent, persistent identity over time.

6. Ethical Considerations: The development of a 'self' or consciousness in AI, akin to natural forms in biology, raises ethical and philosophical questions.

?

If you are a proponent of the view arguing that for LLMs to achieve a higher level of functioning, akin to human reasoning, they would require more than unsupervised learning, the concept of ‘self’ may just be one possible route. While parallels exist in the development of a sense of 'self' in both biological systems and AGIs, the contexts and manifestations are significantly different. The concept of self in AI is perhaps more about creating systems capable of adaptation, learning, and maintaining a consistent identity, rather than mirroring the self-awareness found in biological organisms. There is still much to explore!