This Tiny Code Made Artificial Life! (Blaise Agüera y Arcas)

Blaise Agüera y Arcas discusses an experiment using the minimal Brainfuck programming language to demonstrate how artificial life and purposeful behavior, such as reproduction, can emerge from simple code, highlighting a sudden phase change indicative of emergence. He also notes that this phenomenon is not language-specific, as similar behaviors appear in other programming languages, suggesting fundamental principles underlying life and intelligence.

In this discussion, Blaise Agüera y Arcas talks about an experiment he conducted a few years ago called BFF, which is based on the Brainfuck programming language. Brainfuck is a minimal Turing-complete language created by a physics graduate student named Urban Müller in the 1990s. The experiment demonstrates how artificial life can emerge from seemingly nothing, highlighting the concept of life as the emergence of purpose. In this case, the purpose of the programs is to reproduce, and any alteration to their code often breaks their ability to do so, indicating that the programs have a functional purpose.

Blaise emphasizes that the experiment reveals a sudden phase change, a form of emergence where new properties arise that were not present before. When asked if David Krakauer, a prominent scientist, would recognize this as emergence, Blaise believes he would, even though they disagree on many AI-related topics. The phase change represents a fundamental reorganization at a micro level, resulting in new phenomena that can be described with simpler variables, aligning with Krakauer’s criteria for emergence.

The conversation also touches on the possibility of design bias influencing the emergence observed in the experiment. Since the Brainfuck language and the terms used contain a lot of embedded information, it could shape how the artificial life forms develop. Blaise agrees that the structure of the programs does indeed change depending on the language used, suggesting that the language itself plays a role in the behavior and evolution of these artificial life forms.

To test the generality of the phenomenon, Blaise and his team have experimented with other programming languages, including Z80 assembly language, which was used in microprocessors from the 1970s until recently discontinued. The fact that similar emergent behaviors appear across different languages and architectures indicates that the phenomenon is quite generic and not limited to a specific language or system.

Overall, the experiment with BFF and Brainfuck language serves as a compelling example of how life-like properties such as reproduction and purpose can arise from minimal computational systems. It provides insight into the nature of emergence and artificial life, showing that even tiny pieces of code can give rise to complex, purposeful behaviors, shedding light on the fundamental principles underlying life and intelligence.