[flr]

Yes, these are the reasons why both the religious and eliminitivist views to consciousness are to me the most uninteresting: they sidestep and deny from the beginning the questions.

[flr]

We know how computers do multiplications of 2 integers: the flow of electricity though circuit elements is pattern isomorphic with what we call multiplication in base 2.

However, did the scientists figured out how the neurons in our brain do the multiplication (or summation) of 2 integers? Are we so advanced yet ? Anyone can point to a link (somehow) related to this?

[evan_au]

only a living being could possibly solve... using imagination

As I understand it, most living beings do not use imagination.

For most small organisms, their largest data store is in their genome, which has pre-programmed responses to the world as they find it. "Learning" at the DNA level occurs through genetic drift within a population.

Some degree of genetic learning occurs within an individual, at the epigenetic level, which may lead a malnourished  organism to behave slightly differently than its well-nourished clone with the same DNA.

Those with more data storage outside their genome (for example in a brain) can store more temporary information (like where is food today vs where was food on other days) and modify their behavior based on the results of trial-and-error experiences during their lives. Pattern matching then retrieves the most likely solution.

Much of that brain-based data storage is used for functions like movement, finding food and reproduction - the same things that occupy a single-celled organism. It is just that they are far more flexible about it. They are in some ways more helpless while they populate this storage with useful experiences, gained through trial & error.

I agree that imagination is powerful - it allows one to use the model of the real world which has been stored in the brain, and try out various strategies in imagination without the danger of actually trying them (the "inner movie"). It is trial and error, with reduced risk. This is one aspect of Planning.

Animal psychologists look for species which display the ability to Plan actions, coming up with novel actions in novel situations that work first time (or almost work). This is one indicator of Planning and Intelligence.

Mental rehearsal is also one aspect of dreaming - the ability to carry out actions in the brain while the body is paralyzed and unable to act on it immediately. Perhaps dreaming is an example of imagination in action?

I am interested to see your definition of imagination, and how it differs from planning.

If that computer is programmed with binary logic, then it must use mathematics to solve the problem.

Yes, Wolfram Alpha has a large database of mathematical solutions, which the computer uses to seek the best method, apply it to solve equations and graph the solutions.

It even is able to determine when one approach is not working, and try a different method, effectively learning by trial and error, when normal methods fail.

Because it is doing this internally, before changing the physical world, you could say that it is Planning.

Mathematics has powerful tools when dealing with mathematical problems. But Artificial Neural Networks (ANN) also use mathematics to train the network on real-world data, extracting the patterns of interest, even outside mathematical domains. In a sense, backpropagation is like dreaming: the result is to modify the weights of actions based on the outcomes of those actions.

I guess we have to ask what is the ratio of data in a computer system which is "hard coded" (like DNA) vs learned on-the-fly (like a brain). When more is learned than provided as initial data, we could say that a computer system is Intelligent.

[dlorde]

If that computer is programmed with binary logic, then it must use mathematics to solve the problem. Mathematics however cannot solve problems dealing with the metaphysics of reality. Thus only a living being could possibly solve 2+2=5 using imagination, a property of the mind which allow one to bend the physical laws
of sentience.

'to bend the physical laws of sentience' sounds like pseudo-profound BS - a Chopra-esque deepity - unless, of course, you can explain what these physical laws are (why not show the maths while you're at it), and how they can be 'bent'  []

A digital computer programmed with binary logic can be programmed to emulate the relevant functions of a biological neuron - in fact, of a network of biological neurons - in all it's noisy, lossy, flexible, resilience. Such a network, appropriately structured, can function just like it's biological equivalent. It functions at a level of abstraction above the digital substrate, so it doesn't have the algorithmic limitations of that substrate. 

For example, an interesting recently published demonstration is a virtual (emulated) neural network, nicknamed 'Annabell' (Artificial Neural Network with Adaptive Behavior Exploited for Language Learning), which is "capable of learning to communicate through natural language starting from tabula rasa [blank slate], without any a-priori knowledge of the structure of phrases, meaning of words, role of the different classes of words...". In other words, it can learn a language (e.g. English) well enough to give sensible answers to simple questions, without having any pre-programmed dictionary or syntax. It learns semantic processing by association, from a training dataset of sentences based on the language experience of 3-5 year-olds.

Annabell emulates 2.1 million neurons, interconnected through 33 billion virtual connections, is written in C++, and you can compile and run it on a desktop computer.

You can read the published paper here: A Cognitive Neural Architecture Able to Learn and Communicate through Natural Language, and you can download the open-source software and documentation from GitHub here: annabell.

It seems to me that the question is not whether a conscious AI could be made, but whether we really want to make one. There are at least two major government-sponsored projects to produce human brain emulations (e.g. the EU's Blue Brain Project), but they're focusing more on medical and computing research objectives than consciousness per-se. It's possible, perhaps even likely, that a low level consciousness will result from a well-trained emulation of similar structure and complexity, but I suspect a lot of dedicated time and effort would be required to get something approaching human-style consciousness, and that would come with all kinds of metaphysical (and legal?) problems about its moral and ethical status.

[smart (OP)]

'to bend the physical laws of sentience' sounds like pseudo-profound BS - a Chopra-esque deepity - unless, of course, you can explain what these physical laws are (why not show the maths while you're at it), and how they can be 'bent'  []

By that expression I meant that imagination allows one to resolve the ubiquity of consciousness using metaphysical freedom. A computer based on algorithmics have no imagination, no emotions, and no consciousness.

The pseudo-profound BS in my humble opinion is that artificial intelligence could ever create from dead molecules a conscious being.

[cheryl j]

Dead molecules?

[RD]

... dead molecules ...

At what point do "dead molecules" become alive ?.
Are the the sugar-molecules in my cup-of-coffee alive ? .
If that sugar is "dead molecules" do they magically become alive when I drink the coffee ?.

https://en.wikipedia.org/wiki/Vitalism

[smart (OP)]

"Purpose and meaning are inseparable aspects of life, similarly as consciousness. We cannot expect those in dead molecules. We do not give any moral and ethical importance to an accumulation of dead molecules, but such a
consideration is a must for the life principle. Hence, abiogenesis is an insult to the life force."

A machine, as far I understand, is mechanically assembled from inanimate matter (i.e: dead molecules) thus it have no internal purposes like consciousness or imagination.

Whether the Vedantic view is a vitalistic philosophy makes me uncomfortable... I prefer believing that the Vedantic view is a philosophy of science because its concepts are still prevalent in modern biology.  []

http://www.tandfonline.com/doi/pdf/10.1080/19420889.2015.1085138

[puppypower]

The human psyche is composed of many layers. Conscious and unconscious is only the beginning of this differentiation; differentiation 1.0.  One of the problems with the science of the unconscious mind, is very few scientists use first hand data. Most depend on second and third hand data, which does not form a complete picture resulting in rational polytheism.

Going to Hawaii, in the first person, will tell you more than listening to 100 people tell you about their vacation in Hawaii? The reason is, there are parts of the brain that will react differently. Imagine doing astronomy using only the testimony of others, most of which will be from laymen; patients and test subjects. Having a tooth ache is different from being told about a tooth ache. In all cases, theory will not be totally objective to first hand experience.

Based on first hand experience, collecting data from my own psyche I found that the personality is based on layers. On the surface of the psyche is the persona or the mask of the ego. The persona is easy to see. The persona is connected to our choices of dress, style, tattoos, slang, make-up, attitude, etc. The persona is the mask we show the world and is how strangers see us.

Below the persona is the ego, proper, which contains our personal working memories. The ego proper contains things friends and family are aware of that the world may not see or be allowed to see. The tough guy may not want his friends to know he cries while listening to certain music. He prefers they see him as his tough guy persona. But his mothers sees more than his mask.

Below the ego is the personal unconscious, which contains memories of things we were once aware of, such as the 90% of the facts we learned in school, but forgot after final exams. These memories can be retrieved with hypnosis since they were conscious to us at one time.

Below the personal unconscious is the shadow. The shadow is the bridge between the personal unconscious and a deeper layer of the psyche called the collective unconscious. The collective unconscious is connected to the DNA and contains firmware which is not personal but collective in nature, and defines humans as a species; human nature.

The shadow is the bridge between the personal and collective unconscious and contains traces of both. Like our shadow in the sun, the shadow follows us around. The shadow often contains things the ego is not aware of, which others might see. It contains. among other things, our quirks. For example, in gambling some people have tells, which are unconscious quirks that can betray the quality of your cards; nose scratch.

Below the shadow is the collective unconscious. These firmware define human collective human propensities which define humans as a species. The lion and dog have their own collective unconscious firmware. The first layers of the collective unconscious are firmware connected to natural human instinct.

The next layer, in men, are firmware of relationship, which help us as humans form object, idea and human relationships. For example, the subjectivity of prestige comes from this layer of the firmware. Prestige tells us how a object, idea or person stacks up in the hierarchy of social order. The pet rock can be way up there in prestige, if the firmware is active.

The next deeper level are firmware connected to meaning. These are more concerned to the naked truth beyond the prestige of consensus. The pet rock might be popular in terms of prestige, but the firmware of meaning is more objective, and uses a different method to stack up reality. When Einstein developed relativity this layer was active. Below this is the inner self, which controls the collective unconscious. The firmware of the collective unconscious are the many masks of the inner self.

The metaphysical nature of reality is connected to being unconscious of the entire unconscious. The various unconscious layers will react to reality and with most people being unaware, this will add something extra, via projection, that we can't see with science. We sense it is out there, but in reality it is inside of us, shining out.   

Exploring the unconscious mind can be done in the second and third person, but this does not filter out all the metaphysic layers which makes psychology a soft science and makes theory diverge into rational polytheism. Rational monotheism will require first hand data, which will be very similar for all.

[dlorde]

... I meant that imagination allows one to resolve the ubiquity of consciousness using metaphysical freedom.

Riiiight... more of the same  []

A computer based on algorithmics have no imagination, no emotions, and no consciousness.

Debatable, even today (see computer imagination) - but, as I said, it makes mores sense to use an algorithmic system to emulate a non-algorithmic system like a neural network.

The pseudo-profound BS in my humble opinion is that artificial intelligence could ever create from dead molecules a conscious being.

Nature creates conscious beings from 'dead'(!) molecules all the time - but it's not the substrate that matters, consciousness is computational, information processing, so can - in principle - be emulated by any universal Turing machine; in practice, neuromorphic systems are likely to be most suitable, but, as previously mentioned, these can virtualised on a digital algorithmic system. Modern digital systems are fast enough to emulate a large network of neurons and their connectivity with a single microprocessor, and still operate faster than the biological equivalent. The main technical problem for a large-scale emulation is power consumption and heat dissipation, though these are being addressed.

[flr]

these can virtualised on a digital algorithmic ...

I believe John Searle said something like:

One will never get wet from a simulation of rain.
To get wet you still need real rain.

===
As for the rich and fascinating topic consciousness, my belief is that:
1. A Turning Machine will never generate consciousnesses, because it is an insufficient physical state/structure for such task. Similar with "China Nation" experiment.
 2. Forget about eliminitivists and religious views as they fall in the same category: laziness to even look into the issue.

[RD]

One will never get wet from a simulation of rain.
To get wet you still need real rain.

Real ants are not required for formication.

My point being, disease & drugs are capable of distorting perceptions of reality,
demonstrating that if it were possible to hack-into someone's nervous-system it would be possible to accurately simulate any experience , (including wetness).

[Ethos_]

A few thoughts about AI:

Could an AI computer feel sorrow or regret when faced with an error of it's own making?

Could an AI computer fall in love with another AI computer without being instructed to do so?

Could an AI computer appreciate art to the extent that it could distinguish between beauty and ugliness also without instruction?

And lastly, could an AI computer enjoy the activity of "playing" even though the "playing" had no specific profit or progress as it's goal?

I frankly don't know the answers to these questions myself and I would hazard a guess that it's highly unlikely that definitive answers to these questions will ever be answered with any degree of certainty.

[flr]

... it were possible to hack into someone's nervous-system it would be possible to accurately simulate any experience , (including wetness).

If hack into someone nervous system to induce to that someone the feel of rain one is no longer simulating anything because he is working the real thing (the right physical substrate) that generate aware perceptions.

Note that you already assumed an "someone" so the right physical structure to generate that whole process/thing labeled as "someone"  existed already and it was not generated from scratch by codes run by some Turning machine.

I also can  hack inside my TV to make the image more red-ish or whatever, but that does not mean I understand the basic  and the ultimate physical laws that holds together that piece of matter(called TV). Similarly, figuring out the neuronal correlated of consciousness (at the level of inter-neurons connections) may leave untouched the question "Why that neuronal state is conscious in the first place?"

[flr]

A few thoughts about AI:

Could an AI computer feel sorrow or regret when faced with an error of it's own making?

Could an AI computer fall in love with another AI computer without being instructed to do so?

Could an AI computer appreciate art to the extent that it could distinguish between beauty and ugliness also without instruction?

And lastly, could an AI computer enjoy the activity of "playing" even though the "playing" had no specific profit or progress as it's goal?

I frankly don't know the answers to these questions myself and I would hazard a guess that it's highly unlikely that definitive answers to these questions will ever be answered with any degree of certainty.

If by AI you mean Turing Machine, then my answer is NO. More than just Turning Machine is needed to get "feel", just like more than Turning Machine is needed to solve Halting or tilling problems.

My suspicion is that our minds/thinking rely on physical processes that cannot be completely translated to a Turning Machine, in other words our minds may rely on a computing model superior to Turning machines.

Also, the raw conscious "feel" of something may have nothing to do with any kind of computing model.

[evan_au]

Could an AI computer appreciate art to the extent that it could distinguish between beauty and ugliness also without instruction?

A parallel question: Could a human appreciate art to the extent that it could distinguish between beauty and ugliness also without instruction?
I think not.

Indications are that things like music, art and beauty are basically fashions which are cultural artifacts of our own creation.
People growing up in one musical culture have trouble appreciating very different musical styles and even musical scales that are popular in other cultures.
In many respects, I think our appreciation of art and beauty is trained via subconscious cultural cues; some experiments have shown that a baby relates more to the music, rhythm and language to which a baby was exposed before birth.

One area that seems to be a universal measure of human (and biological) beauty is symmetry. When the shape of an organism is controlled by the outworking of a genetic algorithm, major deviations from symmetry indicate a major genetic or environmental crisis. We regard this as a blight, even when the asymmetry is too small to register consciously.

Could an AI computer appreciate art?

Art is very much in the eye of the beholder. But when some apps recommend similar images or similar music, part of their determination comes from algorithms that study the content, not just what other people also consumed.

Could an AI computer feel sorrow or regret when faced with an error of it's own making?

There are people who think that AI can only work well when computers have emotional states that reinforce learning or cause it to change its behavior.
Current AI learning algorithms (like backpropagation) do change behavior, but don't rely on emotional states.

could an AI computer enjoy the activity of "playing" even though the "playing" had no specific profit or progress as it's goal?

I think this is closely linked to the previous question. A human (or an AI) starts with a low level of experience.
"Play" is a safe way of gaining experience, and it must be pleasurable and self-rewarding, or we would not do it, and we would end up naive and unready to face the world.

If we punished and shamed every child (or AI) for every time they failed to make an adult decision, we would end up with insecure children (or AIs) who were unwilling to try anything.
 
The same goes for organisations - an environment which encourages innovation (which will often fail in small or big ways) is more successful than an organization that punishes even a hint of failure; the latter just produces bureaucracies where everyone just focuses on "protecting their butt" (like politics).
 
A rich and welcoming play environment produces more innovated and interested adults who are self-motivated to learn new things. We need to approach this with AIs that are continually self-motivated to learn.

[Ethos_]

 
A rich and welcoming play environment produces more innovated and interested adults who are self-motivated to learn new things. We need to approach this with AIs that are continually self-motivated to learn.

Interesting thoughts Evan and this last observation is the most interesting one in my opinion. This approach may be the key to achieving a successfully self aware AI. After all, it's probably the most significant means by which very young humans initiate many of their first experiences in the
learning process. I must confess that I agree with your observations. Nevertheless, a truly self aware AI leaves me with a certain degree of trepidation. And there are several top computer scientists that share this fear as well.

[flr]

Current AI learning algorithms (like backpropagation) do change behavior, but don't rely on emotional states.

Essentially they are about how to do a fit.
One can get as much emotions and (self) awarness from these ANNs as from fitting a and b from y=a*x+b then use a and b to estimate y from any other x.

[dlorde]

I believe John Searle said something like:

One will never get wet from a simulation of rain.
To get wet you still need real rain.

He was quite right. A simulated tornado won't blow your house down.

However, you can calculate with a simulated (or, more precisely, emulated) calculator (I have one on my phone), and you can run real Windows programs on an emulated Windows operating system (e.g. on a Mac). The point being that computation (information processing) is substrate-neutral; you can do it with analogue computers, digital computers, neural networks - any capable real-world information processor (a Universal Turing Machine equivalent) can - in principle - compute what any other can (though some are impractical).

If the biological neural networks in brain are computational in function (and the evidence is very strongly in favour of this) - although not conventionally algorithmic - then their computational functions can be emulated by non-biological information processors, such as digital microprocessors programmed to emulate those networks (assuming all relevant aspects are emulated).

Whether this means that an artificial consciousness is possible depends on whether you think consciousness is a computational process of the brain or not.

my belief is that:
1. A Turning Machine will never generate consciousnesses, because it is an insufficient physical state/structure for such task. Similar with "China Nation" experiment.

Can you explain this? It seems to me that the 'China Nation' thought experiment simply describes a human brain on a large scale, so would have all it's relevant properties (we assume the people behave like neurons in all relevant respects, are organised and inter-connected in the same way, and all other relevant influences are suitably accounted for, e.g. blood dynamics, neurotransmitters, hormones, etc).

[evan_au]

A Turing Machine will never generate consciousnesses, because it is an insufficient physical state/structure for such task.

"Never" is a long time.

Join together 100 billion neurons—with 100 trillion connections—and you have yourself a human brain

The state of the art in 2014 was 1 chip that holds 1 million neurons and 256 million connections, or about 1/400,000 of a brain.
Cobble together 10,000 of these chips in a data warehouse, and you have 1/40 of a brain.

Moore's law doubles about every 2 years, so if this trend continues, in about 15 years (2030), we could produce something in a research lab with "sufficient physical state/structure" to be comparable with a human brain. In 20 years (2035), many companies could afford them. Make it 35 years (2050), and such a thing could be affordable in every home. In 15 years (2030) it could be ready for fielding by the military, which has a totally different cost structure.

Could consciousness emerge from this mass of chips? Nobody knows. But it is conceivable.

Elon Musk, Steven Hawking and others have warned about the dangers of autonomous military robots. As I see it, we have about 10 years to rediscover and start applying something like Asimov's "Three Laws of Robotics" - a code of robot ethics. Unfortunately, the military is the group least likely to appreciate an advance like this.

PS: If you use the historical 18 months for the doubling time of Moore's Law, the above dates come scarily closer.