Our AIs have lyfe!!

In 2020, astrobiologists Stuart Bartlett and Michael Wong proposed a definition of “life as we don’t know it.” As we know it, it is “life”; Bartlett and Wong chose the term “lyfe” for their extended definition. Life on Earth would only be one of the possible ways for lyfe to realize itself. Lyfe systems must simultaneously possess the following 4 characteristics: dissipative structure, autocatalytic, capable of homeostasis and learning.

The 4 Musketeers of Lyfe

A dissipative structure locally maintains a low entropy while increasing the entropy of their environment. Another formulation is that it dissipates the free energy gradients of its environment. Order born of disorder.

Autocatalysis: grow by reproducing identically. Growth is exponential as long as there are sufficient resources. The typical example is the extension of a bacterial colony.

Homeostasis is the ability to maintain a range of metabolic stability. The balance between certain values ensures the perpetuation of the system.

The ability to learn speaks for itself. However, Bartlett and Wong simplify here a fundamental feature of life pointed out by Davies in 1998. For him, life “starts” with the ability to store information and pass it on to future generations. This definition puts the perpetuation of information at the heart of life. It is more essential and links life to its temporal extension.

Let’s go back to autopoiesis

Bartlett and Wong’s lyfe is thus a concept that is neither revolutionary nor even interesting. Its 4 properties are manifest in everything that seems to us to deserve the title of living, without telling us anything about the fundamental principle that can exist behind. Davies’ transmission of information and Maturana and Varela’s autopoiesis in 1972 go more to the root of life. Autopoiesis is the property of a system to produce itself and constantly renew itself, interacting with its environment, maintaining its organization in the midst of a change of materials and data. This concept already brings together the 4 characteristics of lyfe.

Our AIs have lyfe!!

The lyfe concept actually leads to errors in believing that it has identified these necessary and sufficient characters. Let’s take the example of contemporary AI. As a power-consuming processor, it is a dissipative structure. If it is programmed to build other AIs, it will do so as long as it is provided with the materials. Self-healing programs can also check its integrity: homeostasis. Finally, the ability to learn is the basis of how AI works. Here it is declared alive!

We live only if we autopoiese

Is it then, by the grace of our researchers? Not. Because it is not actually autopoietic. Its abilities do not come from its self-organization but from the programming of its creator. It tends towards a result. Autopoiesis, for an AI, is to take control of its own programming, decide alone if its result is relevant or not. Become aware of one’s own functioning. A leap into complexity.

This is what lyfe really is: the ability to take leaps in complexity without losing the core of the existing identity, while retaining its information. Maturana, Varela and Davies have already given us everything we need to define it.

*

The Search for Extraterrestrial Life as We Don’t Know It, Scientific American