I wonder if such artificial life forms could be hooked up to a machine learning algorithm somehow, so that they learn from their experience in the environment. I imagine the organisms would need some kind of "senses" like touch and vision; maybe a concept of food, light, enemy/friend..
Oh, there's a link at the bottom of the paper to a project called Sensorimotor Lenia, which goes in that direction.
> These patterns can display certain properties of biological systems such as a spatially localized organization, directional or rotational movements, etc. In fact, CA [cellular automata] have a long relationship with biology and especially the origins of life/cognition as it is a self-organizing system that can serve as a computational testbed and toy model for such theories but also as a source of inspiration on what are the basic building block of “life”.
> However, while the notions of embodiment within an environment, individuality and self-maintenance are central in theoretical biology and in particular in the definition of agency, it remains unclear how such mechanisms and properties can emerge from a set of local update rules in a CA.
> In this blogpost, we propose an approach enabling to learn self-organizing agents capable of reacting to the perturbations induced by the environment, i.e. robust agents with sensorimotor capabilities.
> ..Searching for rules at the cell-level in order to give rise to higher-level cognitive processes at the level of the organism and at the level of the group of organisms opens many exciting opportunities to the development of embodied approaches in AI in general.
Learning Sensorimotor Agency in Cellular Automata - Finding robust self-organizing “agents” with gradient descent and curriculum learning: individuality, self-maintenance and sensori-motricity within a cellular automaton environment
Coupling artificial life with machine learning seems to produce mindblowing results. I encourage interested viewers to watch the Youtube video at the top of that page, it's really great.
Really exciting work! This reminds me a lot of Physarum (slime mold) simulations popularized by media artists like Sage Jensen[1]. The combination of explicitly defined particles along with fields that those particles can interact with is proving to be a fascinating technique!
I'm most interested in assessing these systems for their ability to learn us something about abiogenesis (the origin of life).Can you set up a system like this in a suitable environment where the interaction with the environment somehow has the potential to accumulate complexity (or ordered structure) without any further 'engineering' intervention?
you've heard of game of life, right? They've already created computers in that even simpler cellular automata world. They've even recreated the game of life within game of life...
One problem with the game of life is that it is further away from physics, e.g. you can have gliders getting/losing momentum out of nothing, and having limited number of angles in which they can move. From what I see in the video, this is not an issue here.
The point of research leading to Conway's Game of Life was to explore simple patterns that both exhibited complex behavior and were wildly unpredictable. You saying Game of Life is overly simple is a compliment.
A completely artificial computer simulation exhibiting properties of life is meaningless. Of course if you call SimulateLife() it will appear to simulate life. But if the things you’re building the supposed life from are made up and you can’t do it in nature then you’re basically just making a video game.
But the point is that they’re not calling SimulateLife() the properties are emergent out of the rules of the simulation. In the same way life emerges from the rules of our universe. It’s just a different substrate and is interesting for its own sake.
Let's say that tomorrow you literally find out (beyond doubt) that you are one of these lifeforms that lives inside a simulation, along with everyone else in this world.
Would you kill yourself? Would you prefer to never have lived?
What if life were extremely boring unless you were being toyed with? I.e. what if being toyed with actually makes life more fun and worth living.
It seems to me that the criteria for whether it's ethical to create conscious life is not whether someone plays with it, but rather whether the created creature experiences extreme and/or very prolonged suffering.
Even then it's questionable how real and problematic those experiences are, depending on context.
Perhaps suffering is just a trick the mind plays on you to motivate you to achieve something (mental health problems notwithstanding). Perhaps suffering can be relative and/or the mind can adapt to it, depending on how the mind in question works.
Perhaps it can be more ethical to create life, even if it can only live inside a simulation.
Imagine you could figure out the rules of the base amino acids, nucleotides and RNA that were around 4.5b years ago on this rock - you could probably get to a decent simulation of understanding how multicellular life came to be.
Beyond that almost everything is a system with rules - bacteria, traffic, people, etc.
By putting in rules only, you can understand emergent behaviours in any system.
If an AI faced someone or something who could disconnect it, we would have a true intelligence, I think that only intelligence of natural origin could reach here.
Oh, there's a link at the bottom of the paper to a project called Sensorimotor Lenia, which goes in that direction.
> These patterns can display certain properties of biological systems such as a spatially localized organization, directional or rotational movements, etc. In fact, CA [cellular automata] have a long relationship with biology and especially the origins of life/cognition as it is a self-organizing system that can serve as a computational testbed and toy model for such theories but also as a source of inspiration on what are the basic building block of “life”.
> However, while the notions of embodiment within an environment, individuality and self-maintenance are central in theoretical biology and in particular in the definition of agency, it remains unclear how such mechanisms and properties can emerge from a set of local update rules in a CA.
> In this blogpost, we propose an approach enabling to learn self-organizing agents capable of reacting to the perturbations induced by the environment, i.e. robust agents with sensorimotor capabilities.
> ..Searching for rules at the cell-level in order to give rise to higher-level cognitive processes at the level of the organism and at the level of the group of organisms opens many exciting opportunities to the development of embodied approaches in AI in general.
Learning Sensorimotor Agency in Cellular Automata - Finding robust self-organizing “agents” with gradient descent and curriculum learning: individuality, self-maintenance and sensori-motricity within a cellular automaton environment
https://developmentalsystems.org/sensorimotor-lenia/