Artificial intelligence is the study of intelligent systems designed by intelligent beings.
Humans use computers and programming languages to model cognitive and neural processes. New work has been done to create real, brain-like, artificial neural networks in hardware form.
[Disclaimer 1]. I enjoy the computational approach and I have been programming, in one form or another, since I was eight years old. I fully accept that we may never replicate the human mind except in replications of the brain and body themselves, ie. in children.
However, to dismiss or diminish this project or technology or understanding is unfounded. The computational approach, like all science, is a tool and not meant to be taken as some metaphysically substantial way of looking at things. Although, it is hard for even the dualist to argue that information is not being processed and manipulated and stored, somewhere, somehow.
[Disclaimer 2]. The dangers of human technology are myriad and clear. And, I do not argue that this technology could, like all technology, be used for military purposes (actually, it is, very clearly, being used for military purposes in the United States) or even destroy us.
If we did not investigate any technology that could destroy us, then we would have little to investigate. And, let us not forget that technology, as much as it may destroy us, may save us from the inevitable struggle with asteroids and the sun's demise amongst infinite possibilities that threaten our existence as a species.
The greatest challenge in modeling the various domains of human intelligence lies in unraveling the complexity and functionality (see functionalism) of the human mind. Though the human mind is very powerful, it is not capable of completely representing the relationship between a single neuronal connection.
This is why, like in the building of microprocessors, in the creations of conceptually connected data structures, we may have to rely on the aid of computer systems themselves. In other words, computer programs and/or robots may have to be the ones to create more complex computer programs and/or robots because of the difficulties and shortcomings of the human mind in representations of the necessary complexity. This would be no different than using robotic arms to weld the circuits and transistor arrays of modern silicon-wafers.
We can approximate, make crude representations, and generalize various neurochemical interactions. But, we can not visualize millions of molecules floating between and affecting the surface of neural cells in intricate ways. Each of these molecules, in turn, is a complex and vast arrangement of atoms, subatomic particles, and their quantum components.
Computer memory, which is limited only to our technological understanding and implementational ability of it, could be capable of both (a) representing the complexity of the human mind at various levels, down to our most basic understanding, and (b) surpass human complexity and create more advanced intelligences.
I have a theoretical article suggesting possible ways to teach computers how to think about space and time.
Also, I have an article looking at whether any artificial system can truly understand anything. Believing that any other intelligent system that behaves intelligently is somehow, less "understanding" than we are is a grand metaphysical assumption based on nothing more than human arrogance about our apparent form and composition.
Pain is not in the hand but in the process of interpreting information from the hand, whether the hand is actually there or, in cases of phantom-limb pain, even when the hand has been removed but the connection and interpretation of the connection has not been severed. All experience, if some form of mind-brain identity is true, happens because of interpretation by the brain.
If we use the computer analogy as I often find useful for studies of artificial intelligence, the brain, then is the processor (an amazingly complex and densely parallel processor) and the senses like the microphone (ears), speakers (voice), monitor (visual representation), taste and smell could be made with finely attuned molecular sensors, and proprioception would require positioning sensors that relay information about heat and pressure as the nerves do.
The computer, like a human, can be processing huge amounts of information in parallel but the attention of a human can only be on one or at most, a few, things at once. We don't think of the body being monitored and regulated by electrochemical signals in the brain but no one can deny that absolute removal of a person's brain will also result in absolute shut down of one's body.
As to people thinking that our potential ability to create artificial minds and bodies is somehow a negative possibility and that it somehow deameans humanity, I say all the contrary is true. Why do we have children but to continue our intelligence and share our existence with those we have created?
So we begin to create without sperm and egg but with our infinitely possible technology that can take any form we can functionally instantiate with some portion of our mental architecture. The possibilities for artificial mind are infinite and the potential understanding we can garner from the addition of conceptual, memorial, or processing power is infinite.
I am not saying we take these projects and creations lightly, however, and the seriousness we apply to the conditions and education of children should be parallel to how we consider artificial beings.
We need not equip them with pain and emotions, though it is hard to imagine how we could make aesthetic judgements in any way other than through the emotional reaction they cause in us.
Those who would claim that nothing other than biological material is capable of consciousness are espousing nothing short of causational voodoo. Unless we have some kind of access to other minds and to their qualitative experience, which is, in principle, impossible, then we can never judge the experience of the android.
I would imagine that an ethics of artificial intelligence would have to be worked out but it should simply be that if we create artificial life we should assume that anything we give it the capacity to understand and/or feel, we should take it at face value because, in such matters, that is all we have to go on.
Substance prejudice, or the claim that functional properties are not sufficient for qualitative experience, is nonsense because even though it could be true we (i) see no reason why it should be (ii) and have no epistemic justification for such claims in terms of scientific testability |
|
|
Generalized Description of Complete AI System:
_________________________________________________
(A) Peripheral mechanisms in the same sense modalities: vision, hearing, smell, taste, touch. (Vision would be a 2D array from one or more sensors on which assumptions are made, using deep calculations to represent the 2D image with assumptions about surface properties and spatial distance. I have an article on a page concerning this process called symbolic vision.) (Difficulty: Relatively easy with the exception of determining the correct assumptions with which to manipulate information to extract the important representations from them).
(B) Linguistic capacities like that of humans (includes logic, mathematics, grammar). (Difficulty: Very Hard, because this also relies on C, D, and F, specifically)
(C) Memory capacity (for words, faces, objects) like that of humans in a conceptual branching or cascading storage form with inter-conceptual links. This particular point can not be overstated: The way in which modern computer memory stores digital information is causing one of the greatest setbacks in programming true AI. (Difficulty: Perhaps the hardest to implement).
The concepts and their referents must have, at least, (i) varying weights, (ii) dynamic status, and (iii) arrays of references based on the internal characteristics of the memorial representation.
Traditional computer architecture makes it difficult, if not nearly impossible, to carry out these qualities unless there is a fundamental rethinking of memory at a higher level than the simple memory addresses where digital information is stored. The human mind, if it is situated in the human brain, is not digital but fuzzy across a wide spectrum.
(D) Imaginative capacity like that of humans using space and time as substrates for imagination and as tools for analogical reasoning. (Difficulty: Very Hard because it would require, at the most crude level, an internal screen and camera or, at least, the capacity to represent internal representations in both deep structure and in awareness).
(E) Sesory-motor capacities and movement capacity like that of humans. (Relatively easy, though not easy at all)
(F) Emotional referent or background so that thoughts in the AI system would be capable of being assumed concerning aesthetic or value judgements. (Relatively easy if we are crude about it, Hard if we are complete about. However, why you would want the system to be anything other than constantly happy is beyond me).
(G) none of this provides for consciousness but neither does anything we can observe in our biogical material either.
From 2000 | Artificial Neural Networks as a Model for Human Inductive Reasoning
______________________________________________________________________________
The question of why humans generalize from specific instances to general principles has been a long-standing philosophical problem. Hume believed that inductive reasoning was merely built-in to the human mind. The central difficulty in inductive reasoning is that there is no logical connection between what has happened in specific instances and what will happen in the future.
Likewise, what properties we find in one instance or location need not occur in another instance or location. Therefore, it is logically and empirically possible for the laws of physics to change at any moment and for the laws of physics to be different in other areas where they have not been tested.
Artificial Intelligence and Human Reason, Emotion, and Memory:
_________________________________________________________________
Creating human psychological simulations in three-dimensional and temporal virtual reality programs.
No human decision can be made without emotion as well as analytical processes such as deduction (logic / reason) and induction (reference to memory / past experience). The more fine-grained these categories are with respect to the diversity of human states, the more closely ai simulation will be concordant with human thought/action.
Decision type: external (motor only) / internal (psychological only) or both (motor)
Emotional (biochemical with subjective effects) environment:
Energy / Metabolic Rate: affects thought rate, motor rate, mood
Mood: happy, content, sad, angry, jealous, elated, awed, depressed, suicidal
There will be both a thought-space as well as a real-space. Both will be in 3D color. Thought space and Real space can sometimes merge.
Start with this situation: Man in park about to eat sandwich but is about to be attacked by a lion.
Environment: Body, Park Bench, Park, Sandwich, Lion
Current Mental Space- “Sandwich will be good” (fulfill hunger, survival instinct level (1.6000 seconds) countdown from levels of hunger to concept (DEADLY) proceeded by sight of lion (immediate memory) – causes reference to concept (LION) based on shape/size/surface features/sounds – causes FEAR because concept LION includes concept (DEADLY) and reference to survival instinct (1) with time (1 sec)
About artificial intelligence sub topics (listed on the left hand of the page):
______________________________________________________________________
Below artificial intelligence in the tree on this site are essay pages on John Searle's Chinese Room Argument that argues against strong A.I. on the grounds that machines do not possess the right causal substantial properties like biological entities do and an essay where I discuss theoretical ways to teach machines about structural, temporal, and relational features in analogical reasoning or, in other words, how to teach computer programs about relative space, time, and causal features between two entities or systems that have similarities.
