7

I am interested in studying AI, and I thought it would be a good idea to study the nature of intelligence before stepping into the field. I googled "books to read about intelligence", but it gave me a useless list of books that makes people intelligent.

I don't have a background in philosophy, and I would appreciate if you can refer to any introductory/intermediate books that cover the general subject of intelligence in philosophical point of view.

Thank you.

edit: After some research and thoughts, intelligence is nothing but an ability to process information. Then, intelligence can be viewed as:

F: World -> Meaning

F(information) = interpertation

and intelligence is the function F that maps the information to an abstract world in which our consciousness and ideas lives in.

1
  • 2
    For definitional questions of this sort online encyclopedias are a better source, see Intelligence and references there. This site takes more focused questions.
    – Conifold
    Nov 26 '20 at 11:23
3

One of the most important causes of frequent confusion, is the difference between Artificial Intelligence and Artificial General Intelligence. Yes AI is pretty much just processing information but because intelligence has been redefined in this field. It's like how physicists use the term information in the context of entropy, it's a crucial, but somewhat slippery, distinction that contradicts everyday usage of the word. AI was originally intended to be used like AGI is now, but as people worked on it & we have started to understand how sophisticated things our brains do that we take for granted, the goalposts have moved. And AI has come to mean any little step on the journey to more sophisticated computers.

AGI is it's own big topic. The Human Brain Project aims to find out more. OrchOR theory suggests there may be quantum processes at work in the brain or at least in memory, that put the task several extra orders of magnitude away from the scope of that project - studying connectomes of simple organisms is key to this, & I would say currently supportive of OrchOR. Not much is known about quantum computing in practice (there are only 2 quantum algorithms), but they are simulatable by classical computers, which themselves given enough time could all simulate each other, because they are Turing Machines (or more correctly, TM-equivalent). Integrated Information Theory is an important proposal in how systems might become 'more intelligent, and it helps explain dfferent states of human consciousness & arousal, a previously neglected topic in intelligence.

I heartily recommend John Vervaeke's Awakening From The Meaning Crisis lectures, just as a short introduction to philosophy, but more for framework he builds of terms like relevance-realisation, cognitive grip, and salience landscapes. This gets at the fundamental difference between information and meaning - the sifting of information for what is relevant, and assembling the useful bits, in a task-related way.

The computer intelligence bible, that you will find a LOT more people in computing talk about than read, is Douglas Hofstadter's Godel Escher Bach: an Eternal Golden Braid. It's not that it's technical, it really isn't, it's written a little like Alice In Wonderland, & a little like early Greek philosophy, and is full of thought-experiments (Gedankenexperiments) which people outside of physics always underestimate. But, the book takes patience & tenacity to get through it's doorstop size, while it keeps stretches your brain. Some chapters take chewing over for ages to absorb. It's an amazing work.

Hofstadter proposed that the defining quality of consciousness is what he called being a 'strange loop'. This comes out of a context of Godel's Incompleteness & the halting problem in relation to Turing Machines. So it takes a little background to appreciate. He argues that something happens when you relate lots of sets of data together, that goes beyond circular reasoning, or using axioms, or infinite regress (Munchausen's Trilemma, a funny way of referencing 'impossible' things in the Baron Munchausen story), because you get tangled hierarchies, with self-reference, feedbacks, and moving around in our hierarchy of methods of knowing in a way that is not quite machine-logical, and that the origins of consciousness are there. This article has some nice examples of using it in philosophy - it's like a way to 'shew the fly out of the bottle', to use logic to go 'beyond' logic.

Here's a relevent short post: How is knowledge possible?

And a longer post working from the top-down philosophical side on meaning, and covering intersubjectivity, an important idea in peer-to-peer reality building: According to the major theories of concepts, where do meanings come from?

There is a powerful tendency for people in science and computing to think there is nothing very interesting or special about human minds. And unfortunately, a powerful strand in philosopy which says there is something so soecial about them, scientists aren't on track to figuring them out - the 'qualia' idea and the so called Hard Problem Of Consciousness. I strongly recommend not joining either camp. The story of physics has been from thinking we were a few results away from explaining everything in 1900, and now we don't know what 95% of the universe is made of - our greatest progress has been to begin understanding the scope of our ignorance. I feel strongly we are on a similar trajectory about intelligence.

3

Intelligence is related sometimes (heard it in a philosophy meeting, Toulouse, France) to the ability to think and act coherently towards our deepest and most prioritary goal: survival. So, it is not really intelligent for humans to make money by cutting trees or killing animals, we're just digging our own grave.

If an A-I entity would be able to survive and persist in time (autorepair, reproduce, enter into the ecological balance, counteract the 2nd law of thermodynamics), independently of us, with no help, perhaps it could be called intelligent.

Surviving along time is not simple. The entity should adapt to changing conditions. So, the entity would need to evolve. This is perhaps the most relevant element of any surviving entity: the ability to learn, about itself and the environment, and to act improving the interactions to keep an internal order.

Regarding entropy: it is a feature of a system, therefore the entity regarding its environment. The 2nd law states essentially that energy spreads, that is, the entity will tend to reach the same state as the environment: dispersal. So, the tendency to keep integrity and functionality should be embedded in the system.

An example:

It is easy to program a c++ object that keeps a variable "energy" which decreases with any interaction with the environment (e.g. on each method), and increases with some specific actions. But as long as such "energy" is not the electrical energy that keeps it existing, it is not valid. A robot perhaps would do the task: run to the wall plug and connect itself to recharge batteries. In such case, it would already need to recognize and find wallplugs.

But what if the entity would be built in France, and moved to England? Wallplugs are not compatible! The entity should have a mechanism that would allow it to find a solution.

Keep in mind that a simple battery discharge is the equivalent to the death of the entity. So, it should have a mechanism to prevent that. Now, that would be intelligent. And since then, since the moment robots adapt to a changing environment, and learn new solutions that increase their probabilities of survival, we should be careful of them.

Update: following the question, is it necessary for the entity to have a physical body? No, according to this definition. Perhaps if we would be able to develop an abstract entity simulating its survival, the same rules can be used to build it physically.

8
  • So, an amoeba gulping its food for survival is "smarter" than a sophisticated computer program that is designed to do a specific task. Also, self-learning programs wouldn't be an instance of intelligent being since it wouldn't struggle for its own survival. \\ In order for a program to survive, it needs a body to contain the content (semantics) of the program. So does it imply that physical body is a prerequisite condition for a being to possess intelligence?
    – Paul Pogba
    Nov 27 '20 at 3:00
  • Following this concept of intelligence, 1. yes, an amoeba is more intelligent than a mathematician working for a polluting company. 2. A lot of self learning apps try to survive in its context, but that's not real survival, although, it is an advance. 3. No, I don't think a body is mandatory, but if you can develop a game where entities find creative solutions to survive, that is be perhaps the most successful step towards a real intelligent entity.
    – RodolfoAP
    Nov 27 '20 at 3:05
  • "an amoeba gulping it's food for survival" is really a poor characterization. An amoeba is smarter than a supercomputer because it adapts to the environment, it keeps energy against the 2nd law of thermodynamics, it reproduces to avoid exhaustion of it... etc. Gulping is just a task, and is not what I would give credit an amoeba for.
    – RodolfoAP
    Nov 27 '20 at 3:11
  • 3. If we pulverize all memory cards of every electronic device in this world, then no program would survive. So if a machine has an instinct to survive, it will employ all tactics possible to somehow preserve its body. In this sense, even though body is not a prerequisite of an intelligent being, an intelligent being will try to preserve its body for survival. Does this sound plausible?
    – Paul Pogba
    Nov 27 '20 at 3:18
  • You are mixing concepts. A body is not necessary if the entity is meant to survive in a video game. But if it is meant to survive in real conditions, yes, it would perhaps need a body (unless it exists and reproduces as an electronic virus in neural systems of animals). And in this case, it would fight to death for survival, possibly.
    – RodolfoAP
    Nov 27 '20 at 3:24
2

I can't answer this question directly because the question "what is intelligence?" is deeply philosophical and might not have an easy answer.

But since you are interested in this because of AI:

The field of AI is not, for the most part, concerned with making machines "intelligent" or "sentient" because these attributes can't even be well defined.

Once you leave this behind, what stays is that what we want is rational agents.

Agent: Senses its environment and can act to change it. (could be a physical but also a software environment)

Rational: Given its sensor inputs, chooses the action which will maximize its goal function. (example: minimize distance to target destination, avoid violating traffic laws, etc.)

I recommend Russel and Norvigs excellent "Artificial Intelligence: A modern approach". It tackles the philosophical questions in the introduction, gets rid of misconceptions, and then shows the awesomeness that is AI (from rule-based AIs, to planning, to learning).

2

The nature of intelligence is a topic of much debate. We argue endlessly about cognitive behaviours in animals and whether some behaviour implies intelligence or not. Funnily enough, environmentalists and vivisectionists tend to disagree with each other, who'd have thought it. Try asking a fisherman and a vegan whether fish feel pain.

As far as AI goes, back in the day it was imagined that a sufficiently large computer would be intelligent. When serious number-crunchers appeared, they patently weren't. GIGO - Garbage In, Garbage Out became a fundamental rule of programming. The gold-standard for AI instead became the ability to play chess. But when chess-playing toys appeared in the 1980s, the goalposts were moved again, to things like fuzzy logic and handwriting recognition. When that left the things as dumb as ever, the post holes were dug in for machine learning, soon to be joined by Big Data. All that has yielded little more than mindless chatbots, targeted advertising and socially-biased facial recognition. The next destination for the goal posts appears to be autonomous vehicles, but frankly I am not waiting up.

One school of thought suggests that true intelligence might be necessarily accompanied by sentience and cognition. Other commentators have suggested that intelligence is not an on/off thing. It builds by degrees from a simple worm shying away from heat to fully sentient cognition. The best modern systems are probably comparable to the more sophisticated insects, such as bees. (Don't knock bees, twenty years or so ago the yardstick was ants).

One thing history does teach us: by studying AI, you will learn what intelligence is not. Maybe you will be the one to demonstrate what it actually is - have fun anyway!

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.