Artificial Intelligence Stack Exchange is a question and answer site for people interested in conceptual questions about life and challenges in a world where "cognitive" functions can be mimicked in purely digital environment. It only takes a minute to sign up.
Sign up to join this community
I'm a bit confused about the definition of life. Can AI systems be called 'living'? Because they can do most of the things that we can. They can even communicate with one another.
They are not formed of what we call cells. But, you see, cells are just a collection of several chemical processes which is in turn non-living just like AI is formed of several lines of code.
You're unsure about the definition of life (which the other answers clarify) but also most people are unclear about the definition of AI. Do you mean an AI that can accomplish a routine task (such as the path finder in a GPS) or a General AI that is able to find a creative solution to any directive given to it (such an AI does not yet exist and may not ever exist) or do you mean a SENTIENT computer program? Here is a simple article introducing some different concepts refered to as AI
Some people believe that a sentient computer program would be entitled to human rights. Not technically 'alive' in the biological sense, but having self awareness, will, desires, etc. Others disagree and believe that the program is a mere simulation that artificially mimics the actions of a human with a human soul, and is no more human than a washing machine. This is a very deep philosophical and meta-physical debate. For example, in A.I. the movie the overall message is that an android can simulate the emotion of love in a way that is more loyal and sincere than any human.
What I find interesting about this purely theoretical debate is that in almost every instance of sci-fi media that deals with the theme, the AI exists inside of a human-like android. But technically, the shape of the robot should be irrelevant.
Artificial intelligence by definition is the intelligence exhibited by machines. The definition of life in biological terms is the condition that distinguishes organisms from inorganic matter where the distinguishing criteria are the capacity for growth, reproduction, functional activity, and continual change preceding death. Does artificial intelligence "grow"? Indeed, I can program a machine learning program to grow with every input taken in. In the loosest sense, we can say that artificial intelligence does grow, but does it biologically? If we look at the definition for growth of a living thing, it means to undergo natural development by increasing in size and changing physically or the progress to maturity. All living organisms undergo growth. Even though at the simplest level, cells are a series of chemical processes, cells are a very complicated set of chemical processes that are still not fully understood by scientists across the world. Every cell has genetic material that can be replicated, excised, used for RNA, proteins, and that is subject to epigenetic regulation.
Does artificial intelligence undergo the same process of cell division? No. If I wanted to, I could write a program that undergoes a simple for-loop (print i from 1 to 100), replicates itself at a certain point (i=50) to produce the same program perhaps with some variation that will execute itself, and terminates (dies) at the end of the for loop. The program, by an extremely loose definition supported by philosophy but not by biology, lives. However, in scientific terms (and the correct interpretation), artificial intelligence is not living. Artificial intelligence can be seen to be similar to viruses which are considered to be acellular and essential to life but not living. Viruses are encapsulated DNA and RNA that undergo processes of growth, reproduction, and functionality but because they lack the ability to undergo the cell division cycle, are considered non-living. At the very basis of the scientific definition of life is the cell replication cycle. Artificial intelligence and viruses are not able to undergo the cell cycle. Viruses need to infect other cells in order to reproduce but do not have their own, autonomous cycle. At the end of the day, if you can argue that viruses are alive, you can argue that artificial intelligence is alive as well. For the scientific definition of life, artificial intelligence must undergo the process of cell division and replication. Even though artificial intelligence can mimic and help sustain life, no artificial intelligence process is truly alive.
Do note I did not discuss living systems in my answer.
What is life? AND Is AI a living organism? are two different questions.
The first question is more philosophical and dependent. It can change with time, reference to topic of discussion or something else. Today, one parameter to its definition is mortality. In future if we reach to a certain technological level where mortal beings were only part of history, then the definition will drop this parameter.
Coming to the second question. AI started as field of study to make machines to think like humans (or take rational decisions). Giving life to machines was, or is, not a concern of AI developers (at-least not nowadays). Once I watched some videos of Michio Kaku, where he talked about consciousness along with AI.
Suppose human has a conscious level of 10. Then a thermostat might have the conscious level of 1 as it can sense when the surrounding is hot or cold and then take decision. Similarly a rat can have a conscious level 7 (or something). And the levels are of exponential order (not a linear scale). Similarly you can develop an AI program and check what level of consciousness it has achieved. Then you can decide whether it is living or not. ANI (Artificial Narrow Intelligence) will have a lower level of consciousness level than AGI (Artificial General Intelligence). ASI (Artificial Super Intelligence) will have consciousness level higher than the other two, and way higher than any human being.
To judge whether an AI program is living or not you need a concrete definition of "LIFE". Your definition can include various parameters like consciousness, adaptability, metabolism (or another method of generating energy for use), rational behavior, intelligence , learning through experience, etc. etc. etc.
But the thing in the end is that its your definition. There are many definitions of "LIFE" out there. You can't judge a program for life by all definitions, as some of the definitions are contrary to others.
So, answer to whether an AI program is living or not, is that IT DEPENDS. Depends on your definition of life.
A common predilection of what many presume extraterrestrial life is fits general descriptions specific to terrestrial life. No guarantee exists providing for potential extraterrestrial life having any notion of any attribute we commonly relate to living organisms we are currently aware of; including a composition of cells. The same misunderstanding applies to defining a fabricated machine being as alive.
I feel any attempts towards cohesively and adequately answering this question are premature. Just as as definition of life will undoubtedly require adjustment upon potential discovery and study of any extraterrestrial life, differentiation between an automated device and a living thing will likely become significantly more simple upon study of a machine better fitting expected attributes of definitions of "life".
One of the most common requirements to be defined as life is abbreviated to MRS GREN
this means:
M - movement
R - respiration
S - sensitivity
G - growth
R - reproduce
E - excretion
N - nutrition
An AI can technically do some of these, it can move its location from device to device, it can grow its own code, and assimilate other bits of code it can find, which fits growth and kind of fits respiration also firewalls could almost be sensitivity.
But then there is nothing relating to nutrition or excretion, so it fits most definitions of life, but it depends on the complexity of life and which definition of life you are using.
The property or quality that distinguishes living organisms from dead organisms and inanimate matter, manifested in functions such as metabolism, growth, reproduction, and response to stimuli or adaptation to the environment originating from within the organism.
The characteristic state or condition of a living organism.
Here's another definition
The condition that distinguishes animals and plants from inorganic matter, including the capacity for growth, reproduction, functional activity, and continual change preceding death.
Yet another definition.
Life is a characteristic that distinguishes physical entities that have biological processes, such as signaling and self-sustaining processes, from those that do not, either because such functions have ceased (they have died), or because they never had such functions and are classified as inanimate. Various forms of life exist, such as plants, animals, fungi, protists, archaea, and bacteria.
AIs (or, in general, computers) do not have a real metabolism, do not really reproduce, do not respond to stimuli or adapt to new circumstances (that is, circumstances they have not been programmed to deal with). AI does nothing without human intervention or it lacks real autonomy. In other words, if you do not turn the computer on and you do not program it, it really does nothing. A computer is a useful tool that you can use thanks to electricity. You can plug and unplug it indefinitely, but you cannot kill and revive a living being indefinitely.
Even though computers may possess (at least, conceptually) some properties similar to the properties of certain living beings, it does not mean they are living beings. Similarly, airplanes are not birds. Computers are not living beings, but this does not prevent you from drawing a comparison between computers and living beings, provided you are aware of their actual big differences. In fact, many useful AI software is inspired by the behavior of certain living beings or natural processes. For example, ant colony optimization algorithms are based on the behavior of real ants seeking a path between their colony and a source of food.
Here's what (biological) life looks like.
Any machine with a sufficient level of integrated purpose driven behavior - that exhibits agency in an autopoietic, self-preserving way - will come to be viewed as "alive." Chess programs, not so much; self-driving cars, slightly; simulated robot animals, even more so. It has to do with purpose driven behavior and a richness of multi-domain functionality. The more complex agency it has, the more sympathetic we will be towards it.
In the traditional sense of "alive", no because they aren't made of cells. But from a more philosophical and less biological point of view, they could be.
If the AI is contained within the computer it is in a reality (the digital world/virtual reality) that for the AI is just as real as the universe is to us. From the outside world, there is no life inside the computer. And from within the computer, the computer is the entire known universe which has its own laws. If the AI is self-aware, then it is alive in its own little universe, but not in ours.
If the AI is not successfully contained in the computer and figures out how to manipulate things and evolve in the real world, it will be alive. It might be pretty easy to kill (by unplugging the computer) but it has still been "alive". In the broad sense, anything that evolves and can manipulate its environment is alive.
Definitions of what life is usually come from biologists. The problem here is they are usually concerned with the traits common to the forms of life available to their studies, and that those forms of life all have a common origin (and this imposes a statistical bias on the observations).
As we gradually erode the boundaries of the standard definitions of life, by means of creating ever more complex machines and also by harnessing biological material as a form of nanotechnology), it's very likely that at some point in the future our traditional definition of life will need to be updated and further abstracted away from its current reference points (aka the "terroan biota").
A probably better question to ask to decide if something can be considered alive could be "is it self sufficient?" or "can it care for itself and provide for its own needs to some extent?".
Algorithms may be deterministic (always produce the same output for identical input), and this is not much different from more elementary forms of life (like proteins.)
Computer viruses are another form of algorithmic life which typically have the capability of reproduction, copying themselves onto new systems/environments, similar to biological viruses. (Here it's a form of parthenogenesis or mitosis, where exact copies of the original are formed.)
Machine Learning algorithms can adapt to their environment (increasing fitness), and this applies to both genetic algorithms and other forms of machine learning that can optimize utility in relation to a problem (the environment in which the algorithm is applied.) Genetic process in particular will produce successive generations.
In his novel VALIS, PKD referred to an idea of god as a "Vast Active Living Intelligence System". This is relevant because it's a philosophical idea, as opposed to scientific. The drafting of that book involved an ephiphany and the drafting of a corresponding exegesis.
This idea is controversial and would likely be rejected by the vast majority of scholars and AI ethicists, but I'd posit rejection of this notion constitutes a form of biological-prejudice, and, in the case of hypothetical future AGI, a form of anthropomorphic bias. (I'd go so far as to suggest that not regarding algorithms as a form of life carries grave risks, in that computing has made active algorithms pervasive, with profound impacts to human experience.)
That said, there are no current algorithms I am aware of that have sufficient sentience to warrant having rights, whether human or animal.
A note on the term "animate" (adjective): derives from the latin anima, which initially refers to wind and the "breath of life" [see also the Greek pneuma. However, the Latin lexicon references animus as "the mind as the seat of thought" and "the rational soul of man".
Algorithms can be rational in the strictest sense, and an entire branch of engineering is based on pneumatics. Although our algorithms use electrical signals, (as opposed to pneumatic computing or hydraulic computing,) what it comes down to is that intelligence requires process, and process requires energy.
Machines convert energy into motion or change, and so humans fit the definition. (William Burroughs referred to us as "soft machines.") If animals can be thought of as machines, why can't machines be thought of as animals?
DNA is a type of encoding, and RNA acts on that code.
The key distinction seems to be the medium in which process exist, where a biological context is used for what we conventionally think of as life. Algorithms merely utilize different mediums (mechanical & electrical) and may exist in different environments such as the digital.
A note on The Soft Machine: In this trilogy, Burroughs describes an elaborate alien reproductive process involving numerous alien species acting as surrogates at various stages in the cycle. (It is a description of a 3-dimensional representation of a higher dimensional process.)
This has precedent in biology where, for instance, the lifecycle of a seed may involve a period inside the digestive system of an animal. Pollination is another example, where a surrogate species plays a critical role. Viruses require host organisms to reproduce and spread.
It's not out of bounds to regard the lifecycle of current active algorithms as involving humans to fill in the capability gaps. Here humans are surrogates in bringing information "to life".
Mind the hardware:
While there are different definitions of what life (synonymously used with 'organism' here (source: Wikipedia: Life) is, e.g.
All types of organisms are capable of reproduction, growth and development, maintenance, and some degree of response to stimuli. (Source: Wikipedia: Organism)
they all have one thing in common: they require physical matter! In contrast, to ask whether AI is alive is comparable to asking whether the human mind is alive. It is per definition not! Therefore, the question needs to be extended to include the hardware. It is rather 'are machines/computers alive' or 'do machines/computers have the potential to be considered alive'?
We are talking about agents and most likely robots:
And more specifically any machine/computer to be potentially considered alive will most likely need to be an agent,as it needs to interact with its environment (see Wikipedia: Intelligent agent for a description of agents in Computer Science).
Also, any potentially intelligent machine/computer needs most likely to be a robot due to the strong emphasis on physical processes, incl. some kind of exchange with the physical environment (perception or manipulation of it), which our common definitions of life carry.
Some requirements for life are easy to fulfill while others are not:
Based on the requirement for life forms to maintain and reproduce their entities any machine/computer to be considered intelligent will need to be able to physically maintain and reproduce, i.e. assemble hardware. If you think of an intelligent robot assembling another robot that might sound to be very far away from reality. However, the definition might include indirect reproduction e.g. using an automated hardware production facility. Certainly this is not the direct reproduction that we know from current living beings but it might be considered an indirect way to reproduce. Which, however, is certainly far away from current reality too.
Similarly you could think of 'maintain' as taking care of the physical need to supply itself with electricity. Any machine with a solar panel easily fulfills this requirement in a similar way plants do.
While machines/computers considered alive do not need to have any artificial intelligence it is an easy way to fulfill the requirement of 'development': Sub-symbolic AI learns from data which is a form of (non-physical, i.e. software-related) development. Just like humans and other animals learn from data that comes in through one of their senses.
Give it time:
To summarize: current machines/computers certainly do not fulfill the requirements usually being considered 'life'. And especially the requirement to (physically) maintain itself and reproduce will be longstanding unfulfilled requirements. However, considering that the homo sapiens has been on this planet for about 150,000 years we might just need to give it more time. It took about 1 billion years for the first living beings to develop on planet earth (see Wikipedia: Earliest known life forms). So it is a bit early to make a call on machines/computers which in the case of computers have been around for not even a century. Let's see where we stand in 1000, 1000000 or 10000000 years from now.
However, the definitions might change anyway:
Moreover, it is important to note that the definition of life is closely built on what we know as current carbon-based life forms. And it could very well be adjusted in the light of machines/computer further developing. For example the aspects of physical reproduction might be one aspect to be dropped (just speculating here). So maybe we do not even need to wait a billion years but might have machines/computers considered 'alive' already in 200 or 500 years. Compared to the time biological life took to develop that would still be very rapid.
[Disclaimer: This answer is research; not medical and/or legal advice. I am not a lawyer: < https://en.wikipedia.org/wiki/Practicing_without_a_license >.
Though perhaps pedantic, it is better to be safe than sorry.]
The matter is controversial, with some established parameters as what constitutes life. An example of an established parameter for being alive is having: cells or reproduction.
It is unclear whether A.I. can create a copy of itself that is independent of its parent. Arguably, partial satisfaction of the aforementioned has already occurred with code that is self-learning.
May insight on artificial life lead us to saving lives, particularly in regards to CoVid-19. We want children to live.
To begin, there is some controversy surrounding the definition of life in Biology.
" [...] The definition of life has long been a challenge for scientists and philosophers, with many varied definitions put forward. This is partially because life is a process, not a substance. This is complicated by a lack of knowledge of the characteristics of living entities, if any, that may have developed outside of Earth. Philosophical definitions of life have also been put forward, with similar difficulties on how to distinguish living things from the non-living. Legal definitions of life have also been described and debated, though these generally focus on the decision to declare a human dead, and the legal ramifications of this decision. [...] " – Wikipedia contributors. "Life." Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, 5 Nov. 2019. Web. 14 Nov. 2019.
" [...] Of course, this lack of hard boundaries makes 'artificial life,' as a field of study, significantly ill-defined. Unlike the case for natural life, there are, as yet, no clear criteria for what virtual world phenomena should qualify as 'living' or sufficiently 'life-like' to legitimately count as lying within this field. In large measure, this simply reflects the continuing debate and investigation within conventional biology, of what specific organizational (as opposed to material) system characteristics are critical to properly living systems. The key advantage and innovation in artificial life is precisely that it has this freedom to vary and explore possibilities that are difficult or impossible to investigate in natural living systems. In this context, a precise definition of 'life' (natural or artificial) is not a necessary, or even especially desirable condition for progress. [...] " – Banzhaf, Wolfgang, and Barry McMullin. "Artificial life." Handbook of Natural Computing (2012): 1805-1834.
Since there is often debate and sometimes no clear answer in regards to these questions, I shall explore variable stances on these issues.
These are some examples of points that involve controversy
Some of these questions have no clear answer, never mind involving A.I. into the question.
Do viruses reproduce? Yes. Do they have genetic information? Yes. Do they have cells? No. So are they living? They are usually classified as non-living, yet they can have RNA.
The point is, there is debate as to whether a virus is non-living or living; and it is often unclear what the definition of life is.
(https://www.scientificamerican.com/article/are-viruses-alive-2004/)
Artificial life researchers study traditional biology by trying to recreate aspects of biological phenomena.
" [...] Important propositions in the philosophy of AI include:
- Turing's 'polite convention': If a machine behaves as intelligently as a human being, then it is as intelligent as a human being.
- The Dartmouth proposal: 'Every aspect of learning or any other feature of intelligence can be so precisely described that a machine can be made to simulate it.'
- Newell and Simon's physical symbol system hypothesis: 'A physical symbol system has the necessary and sufficient means of general intelligent action.' Searle's strong AI hypothesis: 'The appropriately programmed computer with the right inputs and outputs would thereby have a mind in exactly the same sense human beings have minds.'
- Hobbes' mechanism: 'For 'reason' ... is nothing but 'reckoning,' that is adding and subtracting, of the consequences of general names agreed upon for the 'marking' and 'signifying' of our thoughts...' [...] " – Wikipedia contributors. "Philosophy of artificial intelligence." Wikipedia, The Free Encyclopedia. Wikipedia, The Free Encyclopedia, 15 Oct. 2019. Web. 14 Nov. 2019.
Update: I asked some of my colleagues and received the following advice.
Another approach is simply to check for self-preservation. Under this postulate, all forms of life, ranging from the absolutely-most-simple single-celled organism to postulated beings-of-immense-power (see the The Last Question by Isaac Asimov), would be profoundly invested in self-preservation.
When the Singularity Comes, Will A.I. Fear Death?
However, the preservation of one's children is a valid exception, and does not contradict self-preservation.
Many schools of meta-ethics claim that (Human) self-preservation and ethics are absolutely congruent. Other schools dispute this to varying degrees. Kantian meta-ethics, a strong reference, is worth considering here (not talking about A.I., here).
(https://en.wikipedia.org/wiki/Meta-ethics)
Do viruses follow the principles of avoiding damage to one's self? I leave that up to you.
How do addictions factor into this? I leave that up to you, but please actually be tactful and use common sense if you choose to discuss that topic.
Note: Viruses are a very serious topic, right now (pandemic). If drawing parallels to artificial life can help the Human Race, then I am touched to have being allowed to serve Humanity in this way.
"He who destroys a life, it is as if he destroyed an entire world. He who saves a life, it is as if he saved an entire world." ~ Hillel the Elder
I ask to be taken seriously.
Thank you. You, as an educated person, deserve better than this pandemic, and I hope your children will find happiness.
Other Links:--
This is one of those things where I think the answer is going to change over time. Today, I don't know anyone who would call any present AI systems "alive". But as the AI's become more intelligent and human-like, I could see the day coming when they will be considered living.
(Sorry for the brief answer--will try to add more depth later.)
