With some knowledge of machine learning and deep learning, it seems very unlikely for AI to develop into the consciousness that we imagine.

To me, consciousness requires a new framework that is very different from what we have today, as current forms of learning seem to be very 'shallow'.

What prevents us from developing artificial consciousness?

(Also, is there a difference between artificial consciousness and sentience?)


It may be helpful to think of consciousness, like intelligence, as a spectrum. The Stanford Encyclopedia of Philosophy, under the section "Creature Consciousness" (2.1) defines sentience as:

Sentience. It may be conscious in the generic sense of simply being a sentient creature, one capable of sensing and responding to its world (Armstrong 1981). Being conscious in this sense may admit of degrees, and just what sort of sensory capacities are sufficient may not be sharply defined. Are fish conscious in the relevant respect? And what of shrimp or bees?
Stanford Encyclopedia of Philosophy, 2.1 "Creature Consciousness"

It seems clear that AI possesses the most basic form of consciousness in terms of "sensing and responding to its world". I might liken this level of consciousness to that of microorganisms, able to interact with their environments but possessing none of the higher functions of humans, or even higher-order animals.

We don't really understand the true nature of higher consciousness and human-level sentience, which doesn't necessarily mean that an automata with sufficient complexity couldn't achieve it. But there is a deeper problem relating to validation (i.e. how would be know the system is truly self-aware, and not simulating self-awareness. [See Searle's Chinese Room, but also McCarthy's Refutation]

In his paper Making Robots Conscious of their Mental States McCarthy states:

Conscious knowledge and other information is distinguished from unconscious information by being observable, and its observation results in conscious knowledge about it. We call this introspective knowledge.

A robot will need to use introspective knowledge in order to operate in the common sense world and accomplish the tasks humans will give it.

Many features of human consciousness will be wanted, some will not, and some abilities not possessed by humans have already been found feasible and useful in limited domains.

We give preliminary fragments of a logical language a robot can use to represent information about its own state of mind.

A robot will often have to conclude that it cannot decide a question on the basis of the information in memory and therefore must seek information externally.

Programs with much introspective consciousness do not yet exist.
Making Robots Conscious of their Mental States, Abstract

This still seems to be the case, however, McCarthy seems hopeful about the "computationalist" approach:

Thinking about consciousness with a view to designing it provides a new approach to some of the problems of consciousness studied by philosophers. One advantage is that it focusses on the aspects of consciousness important for intelligent behavior. If the advocates of qualia are right, it looks like robots won’t need them to exhibit any behavior exhibited by humans.

Although current "strong AI" has to be further qualified as "narrow", the continual increase in memory, processing and algorithmic sophistication may well yield Artificial General Intelligence.

Even if this proves to be the case, I suspect the debate over artificial consciousness will rage on.

See also: Artificial Consciousness

| improve this answer | |

By the cognitive science / psychology literature, we don't have a theory of consciousness, or a measure of sentience. By its very nature, "things" like those can only be understood with respect to ones own subjective experience (as in you can only understand consciousness by comparing it to your own conscious experience).

So no, even if we have a device that perfectly replicates the function of consciousness or sentience, we'd still have a hard time saying that it conclusively is indeed consciousness.

| improve this answer | |
  • $\begingroup$ It will be interesting to see the types of tests that come out of the cognitive science field, should such systems emerge. I suspect the question will remain in the area of metaphysics for a good long time, regardless. The better sci-fi authors/philosophers have grappled with this question, Phillip Dick in particular. In some ways, you might cast his view in context of the "Duck Test". (I wouldn't be surprised if there is strong sentiment for the idea that "if it talks like a person, and acts like a person, then it's probably a person";) $\endgroup$ – DukeZhou Nov 20 '17 at 2:01
  • $\begingroup$ it's also quite hard to agree even on living systems. consider the spectrum of lifeforms ranging from humans to bacteria. some animals in-between are probably conscious (monkeys, dogs, dolphins, elephants). but than we'd run into less clear cases (fish, lizard, cicada) or outright unfathomable cases (ant colonies, viruses, internet memes) $\endgroup$ – k.c. sayz 'k.c sayz' Nov 23 '17 at 1:53
  • 1
    $\begingroup$ or more concretely, we don't quite know yet the function of consciousness $\endgroup$ – k.c. sayz 'k.c sayz' Nov 23 '17 at 1:53

I don't know much but the limitations of creating consiousness through the method we are adopting are if consiousness and sentience are a function of programing and whether programing is the precursor in time, a question I asked similarly is which came first. I think the current train of thought is that consiousness is a result of genetic programming passed on by I enteties that were subject to successfull programing which must boil down to initial chance, but thinking that programming is result of structure and order which chance has no place in.

If consiousness is not a result of programming and came first then surely programming will only result in a mechanical simulation of a part of the minds ability to solve physical problems and be unable to be subject of the thoughts that are not down to physical action and reaction where the communication of it can be observed and possibly judged.

If consiousness is a result of the material then it's simulation I think will be achievable but still poses problems that each of us have in determining whether that simulation has been achieved because it's very difficult to determine if someone else shares a similar consiousness and sentience than yourself, because we have no absolute measure in which to judge it by, and are limited by the natural language and it's constraints that we use to communicate to one another and our personal interpretation of them.

| improve this answer | |

By answering this I am hoping to have a discussion on this answer, as the answer is purely opinion based, and as I cant comment on other answers since I am not yet qualified for that, according to the rules

The limitation, in literal sense, are not any. But limitation is quite obvious, if the goal is to build "machine" that "can" "think", they are still just a symbolic machines no matter, whatever the algorithm is. I think, and hence it seems obvious to me, that awareness arise due to level orientation, which was too "low" for GOFAI(Good-old Fashioned Artificial Intelligence),"low" in the sense that "chunks" of "perception" were already stored and processed, or stored, or processed and the upshot was that there was not any leakages, and hence no matter what algorithm one use, computer of any kind, cant just settle down. In the approaches from Connection-ism, the problem(relatively) is the action-domain is too-simple, but I think it would be successful if the domain would be expanded via capsules like approach, by introduction of modularity.

| improve this answer | |

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.