0

The theory of evolution seems to be intelligent as it creates life

The mechanism of evolutionary theory consists of mutation, recombination, and natural selection like a genetic algorithm.

Isn't this evolutionary mechanism itself the same as the essence of human intelligence?

Neil Slater
  • 28,678
  • 3
  • 38
  • 60
Dimer
  • 331
  • 1
  • 10
  • I think that the question "Isn't this evolutionary mechanism itself the same as the essence of human intelligence?" is a bit ambiguous because you are using the expression "X is the essence of Y". I think it would have been clearer to ask "Do (natural) evolutionary mechanisms _generate_ intelligence?" (actually, this question may not be on-topic, though questions about the definition of intelligence are on-topic here). The answer is clearly "yes", if you consider animals (and, in particular, mammals and humans) intelligent and you assume that evolutionary mechanisms created us. – nbro Nov 05 '20 at 00:15
  • The other question (actually, an assumption) that you can ask is: "Are EMs intelligent themselves?". Artificial evolutionary algorithms can be used to optimize some function (through search and random operations). If intelligence is the process of optimization, then (at least artificial) evolutionary algorithms are intelligent. I would suggest that you edit your post to ask these more explicit questions, but it's also true that an answer was already given, and I don't know if Neil is willing to address your possilby new questions after the edit. – nbro Nov 05 '20 at 00:18
  • Please, next time, before trying to ask a question, think more about your questions. Try to be specific and not ambiguous. Moreover, try to ask _only one question per post_, unless they are very related. Evolutionary algorithms and evolution are clearly a big topic in AI and your question is related to your assumption that "evolution is intelligent", but it suggests that you didn't really know what your question was. Moreover, try to search through the other existing posts/questions. For example, there is already another very similar/related question: https://ai.stackexchange.com/q/28/2444. – nbro Nov 05 '20 at 00:24

1 Answers1

0

The theory of evolution seems to be intelligent as it creates life

When you say "seems to be intelligent" that begs the question: How are you defining "intelligent"? Which of course is still one of the big issues in AI research.

I think there are some flaws with the argument that "creates life" = "intelligent":

  • Evolution does not create life. It operates on entities where there is a copy mechanism which is not 100% reliable, plus a selective environment that also impacts likelihood of further copies being made. Some form of proto-life (an initial Darwinian ancestor or Ida) needed to exist before evolution started.

  • The process of creating the first proto-life capable of undergoing evolution is generally thought to be a large semi-random search through chemical combinations. Random search is sometimes used in optimisation problems, and might be studied as part of AI search topics. However, it would normally be considered something of a baseline algorithm, and definitley not tick the boxes for all the general traits of intelligence.

Isn't this evolutionary mechanism itself the same as the essence of human intelligence?

The Wikipedia article on artificial intellegence lists challenges faced by researchers and developers in AI. The categories chosen there are:

Reasoning, problem solving; Knowledge representation; Planning; Learning; Natural language processing; Perception; Motion and manipulation; Social intelligence; General intelligence

Together, these are mainly traits of mammalian, avian and a few other multicellular species, with a few traits such as language heavily focused on humans.

I think it is important to separate out the mechanism whereby these traits arose naturally - which is generally agreed to be via an evolutionary process - from how those traits function. Artifical intelligence may use a little bit of reverse engineering from the natural traits in order to inspire design, but most AI systems do not use theory of evolution directly.

When used directly, evolutionary algorithms can be used to solve search and optimisation problems. Also they can be used to solve simplified problems in perception and motion/manipulation. However, we are not able to scale up such algorithms to solve all aspects of general intelligence. Instead, systems like machine learning are designed to work from analysis of the problem, inspired in part by working natural systems. These work far more efficiently than evolutionary algorithms. There are no competitive evolutionary variants of AlphaZero, Watson, GPT-3 or neural-networks used in image processing.

Evolutionary algorithms have their place in AI in practice and research. However, they do not define or encapsulate a form of general intelligence.

Neil Slater
  • 28,678
  • 3
  • 38
  • 60
  • Couldn't all known functions of intelligence be made by evolutionary mechanisms? – Dimer Nov 04 '20 at 11:02
  • @Dimer: They could, and apparently this is what has happened in nature. I try to explain why that's not the same thing as evolutionary algorithms "being intelligent" in the answer. If there is a step missing or unclear reasoning in my answer, I could expand it if you could help me find what's wrong – Neil Slater Nov 04 '20 at 12:07
  • @Dimer: As an analogy, a person can build a house or car. However, we would not usually say that the person was themselves a dwelling or a type of transport. And that is even the case when we can host other creatures and carry objects by hand in a more limited sense. Architects and engineers who study or build dwellings or cars would consider studying the biology or social history of humans as being outside of their expertise, even though those things clearly led to architecture and design of modern vehicles. – Neil Slater Nov 04 '20 at 12:13