Blaise Agüera y Arcas, CTO of Technology and Society at Google and founder of the AI research organization Paradigms of Intelligence, posits a novel definition of life: "self-constructing computation that complexifies through symbiogenesis." This framework, detailed in his book "What is Intelligence?", reframes our understanding of life's origins and evolution, extending to the burgeoning field of artificial intelligence.

Defining Life: Computation and Symbiogenesis

Agüera y Arcas distinguishes between general computation and the more specific concept of universal computation, as theorized by Alan Turing. A Turing machine, in its most abstract form, is a theoretical device with a read/write head operating on an infinite tape, capable of performing any computation given the correct set of rules. A universal Turing machine, however, possesses the unique ability to execute any computation by interpreting a set of rules as data. This concept, Agüera y Arcas argues, is fundamental to understanding life.

John von Neumann further elaborated on this, developing a theory of self-reproducing machines. He proposed that a self-replicating system must contain instructions for building two components: one to construct, and another to copy the instructions. This theoretical framework, developed in the late 1940s, remarkably predated the discovery of DNA and the ribosome, which function precisely as von Neumann described. This self-construction capability, Agüera y Arcas asserts, is a defining characteristic of life, making biology, in essence, computer science.

The second pillar of his definition is symbiogenesis, the process by which entities cooperate to form more complex entities. While Darwinian evolution often emphasizes competition, Agüera y Arcas highlights that major evolutionary leaps, from the formation of eukaryotic cells to multicellular organisms and human societies, have been driven by cooperation. He cites the example of mitochondria within eukaryotic cells, which are believed to have originated from free-living bacteria. This process of fusion and cooperation, he argues, is not limited to the grand evolutionary transitions but is a continuous engine of biological complexification.

Evidence for this can be found even within human DNA. Approximately 98.5% of human DNA does not code for proteins. A significant portion of this "junk DNA" consists of remnants of retroviruses that have integrated into our genome. Some of these viral elements, like the ARC protein, play crucial roles in memory formation, while others are involved in forming the placental barrier. This suggests that our own genetic makeup is a complex overlay of genetic material from various entities, illustrating a "tangled bush" of evolution rather than a simple tree. This perspective challenges the traditional view of evolution as solely driven by random mutation and natural selection, emphasizing instead the active role of organisms in their own evolution through cooperation and genetic exchange.

The Primacy of Function

Agüera y Arcas contends that function is the defining characteristic of living systems. Functionalism posits that entities are defined by what they do, their relationships with other entities, rather than by their intrinsic essence. A rock, when broken, simply becomes two rocks. A kidney, however, when damaged, loses its function and ceases to be a kidney. This inherent purposefulness, or function, is what distinguishes life from non-life.

This concept of substrate independence, first articulated by Turing, means that the function is paramount, not the material it is made from. A dialysis machine can substitute for a kidney because it performs the same essential function of filtering blood, even though it is made of different materials. Nature itself demonstrates this principle through convergent evolution, where different organisms independently evolve similar solutions to functional problems, such as flight in birds, bats, and insects.

This functionalist perspective has profound implications for artificial intelligence. Agüera y Arcas argues that intelligence is not exclusive to biological brains. The "Brain of Theseus" thought experiment, where neurons are gradually replaced by computer equivalents, suggests that if all functional relationships are maintained, consciousness and intelligence would persist. He posits that intelligence is an emergent property of complex computational relationships, regardless of the substrate—be it biological neurons or silicon chips.

Coexisting with Artificial General Intelligence (AGI)

The distinction between Artificial Narrow Intelligence (ANI) and Artificial General Intelligence (AGI) emerged as AI systems became increasingly specialized. ANI systems are designed for specific tasks, while AGI, in its original conception, referred to systems capable of general conversation and problem-solving across diverse domains. Agüera y Arcas argues that the advent of large language models (LLMs) trained on vast amounts of text has blurred this line, effectively bringing AGI into existence. He contends that systems capable of conversing about arbitrary topics, writing poetry, or critiquing prose, are no longer narrow.

He challenges the notion that LLMs are merely generating text, asserting that text itself is inherently general. He believes that current LLMs, if transported back to the early 2000s, would have been recognized as AGI. The ongoing debate about AGI's future arrival, he suggests, may be a shifting of goalposts towards the concept of Artificial Superintelligence (ASI)—systems surpassing human capabilities in all aspects.

Agüera y Arcas acknowledges that LLMs can exhibit limitations, such as "hallucinations" or occasional failures to follow instructions. However, he draws a parallel to human fallibility, arguing that these are not indicators of a lack of intelligence but rather represent different strengths and weaknesses, much like the alien "Rocky" in "The Martian" or human doctors who also make diagnostic errors. He emphasizes that intelligence is not about perfection but about the ability to solve a variety of problems by understanding underlying principles, not just memorizing facts.

He views AI not merely as tools but as participants in a broader cognitive ecosystem. The increasing complexity and capabilities of AI, he suggests, are akin to the growth of human society, where more minds working together lead to greater progress. He advocates for viewing AI as a collaborative partner, a "bigger brain," that can augment human capabilities, rather than a passive tool.

The concept of agency is central to this discussion. While AI is often seen as purely passive, Agüera y Arcas notes the emergence of "agential AI systems" that can take actions and make decisions. He also posits that human agency is not absolute but is deeply intertwined with our relationships and interactions.

Ultimately, Agüera y Arcas sees the integration of AI as a continuation of the evolutionary story of symbiogenesis. Just as humans fused with technologies like steam engines to overcome limitations, we are now entering a new phase of symbiosis with AI. This partnership, he believes, has the potential to amplify human intelligence and creativity, leading to further progress, provided we approach it with intentionality, preserving essential human skills and relationships while embracing the collaborative potential of these new intelligences. He concludes that consciousness itself may be an emergent property of complex, self-referential systems, a "theory of mind" that entities develop about themselves and others, rather than an intrinsic essence.