Posted by:
Henry Bemis
(
)
Date: October 27, 2017 09:22AM
By way of debate for purpose of advancement - The issues you cite appear to be about points or characteristics about which science is not persuaded even exist.
For example consciousness, freewill, and creativity.
COMMENT: Isn't a bit silly to deny the existence of consciousness? I do not know of a single scientist who would take such a bizarre position. Those who do take such a position are hard-core materialist philosophers, such as the Churchlands and Dennett, have been refuted, mostly by common sense.
Regarding freewill, I agree that whether humans have genuine freewill is disputed. However, what cannot be disputed is the fact that no AI system can possibly have genuine freewill. Moreover, without consciousness, there cannot be the illusion of freewill. The point is that strong AI theorists have to address freewill, one way or another. Regarding creativity, I would refer you to the book, "The Neuroscience of Creativity" edited by Vartanian et al., and particularly to the AI theorist, Margaret A. Boden, chapter "Creativity as a Neuroscientific Mystery." The title says it all! And Boden is a renown AI advocate. The short answer is that AI cannot account for human creativity in all its forms.
_________________________________________
Here you cite three attributes that are very much up for debate. When we discussed this last time i asked if you could define what was unique about the meat computer that exists inside the human skull to the point that it cannot be replicated.
COMMENT: And I responded, consciousness, freewill, and human creativity. Please provide me with a computational explanation, or example, of these human attributes. Cite for me a program, or model that addresses any of this in a meaningful way. The only answer available is to simply deny these attributes exist, which in my view is just nonsense.
__________________________________________
It very much appears that you are making an appeal to the metaphysical here as opposed to the physical.
COMMENT: I am not appealing to anything. I am only pointing out that strong AI cannot account for these rather basic human attributes that we all take for granted.
____________________________________________
In terms of performance in every sphere AI is being tested it is out performing humans. It is certainly outperforming us in the realm of games and analysis. Its ability to hold factual data to facilitate decision making and draw upon that data within defined realms far exceeds human capabilities, intact the test data shows it far exceeds the capabilities of not just one human vs a machine but a large collection of the best minds vs a machine.
COMMENT: I have never said that AI did not have some rather marvelous computational and functional advantages over humans with respect to computing raw data. But that fact alone does not make a computer a "mind." Moreover, it does not suggest that an AI system is functionally better than a human in all areas. For example, all scientific achievements, e.g. Einstein's special and general relativity, involved insights that are not computational. In other words, you cannot plug in the equations of Newtonian physics, Maxwell's equations for electromagnetism, and all of the relevant data, e.g. the speed of light, and the gravitational constant, press the "compute" button, and have Einstein's E=mc>2 pop out for free; or Einstein's field equations; or the conceptual equivalent to special or general relativity. This applies to essentially all of scientific insight. Computation is not enough. Creativity and insight is part of being human, and essential to science.
______________________________________________
I'm not persuaded by your arguments that you're following these developments closely.
COMMENT: I understand the developments of AI very well. They are essentially based upon computational models of neural networks by digital computers. They are all by definition computational systems, which use sophisticated learning algorithms. The achievements in this regard have indeed been remarkable. But the fact that a computer can beat a human in chess; or that Watson can beat humans in Jeopardy, does not in any way equate to the idea that humans are essentially poorly designed computers, or that humans do not have something to offer "intelligence" that computers cannot.
_______________________________________
Your assumption in point 2 seems fundamentally off the mark.
HB COMMENT: Here is the evidence: There is no AI program that currently exists, or even that exists in principle, that can model human agency, or human creativity, in any meaningful manner. Both are fundamentally non-computational. Thus, AI must deem such things as illusory. When you can generate both of these from an AI computational program, only then can AI be taken seriously as a substitute for human intelligence.
RESPONSE: Human Agency and Human Creativity? As noted earlier, you'll need to define these attributes and evidence they exist, you'll also need to explain why you feel they are non reducible. I've summated this above in the question about meat computers. Nothing you claim here seems to hold to scrutiny.
COMMENT: Read the book I noted above! Few, if any AI theorists deny the existence of human creativity; and virtually NONE deny consciousness. Tell me. Why is it that I am giving all of the arguments, and you are not providing anything substantive. YOU TELL ME WHY YOU THINK HUMAN BEINGS ARE JUST COMPUTATIONAL MEAT GRINDERS. YOU TELL ME JUST HOW CONSCIOUSNESS, FREEWILL, AND HUMAN CREATIVITY CAN BE AN ILLUSION. YOU EXPLAIN TO ME HOW SCIENTIFIC ACHIEVEMENT OCCURES THROUGH COMPUTATIONAL PROCESSES ALONE. YOU CAN'T. OBVISOUSLY!
_____________________________________________
What is it about the human brain you claim cannot be replicate using other materials on this planet - we do agree i assume that the brain is merely evolved from inorganic matter into a biological machine?
COMMENT: I agree that the human brain is a biological dynamic systems that involves computation of environmental and internal data to produce functional organismic results. What I dispute, on empirical and logical grounds, is that human beings are only the products of the computational properties of brains. That proposition has not been, and in my opinion can't be, supported. If you disagree, SUPPORT YOUR POSITION!
_________________________________________________
AI will be able to do this.
HB COMMENT: How so? How will AI assimilate information that is not fed into its computational structure. It will only assimilate information that it is provided by humans, and will only process such information as directed by humans. Now, granted it will compute much more efficiently, and that will create all sorts of social opportunities and enhancements, including impressive robots, but that is all. It will never tell us what the moral choice is, and it will never create insight the transcends bare facts and rote computations.
RESPONSE: How do Humans acquire information? The supply of information is provisioned by external data coming in via the senses. AI already uses optical recognition and does it faster and more accurately than the human eye - cognition link. Hooking AI up to comparable data sources such as temperature, movement (Air disturbance), visual, Audio, not to mention millions of data feeds coming in from numerous tech sources, would allow it to draw on massive amounts of information and map relationships in the way we data mine. This is what humans already do, but just not as well.
COMMENT: The key phrase here is "data mine." Sure, AI systems can data mine; and compute that data, in many cases, but not all, better than humans. BUT, and this is a big BUT, AI systems are programed by humans beings to produce the results they produce. AI Systems do NOT generate their own algorithms, i.e. their own programs, which is the backbone of "intelligence" in my view. This is another way of saying that they are NOT creative. At the end of the day, they are nothing but sophisticated processing machines. Humans are much more than that, regardless of how humans might be out performed with respect to their computational abilities.
_________________________________________
You argue about transcendence but there is no evidence this exists. Again it falls foul of the meat computer challenge.
COMMENT: If something is know to exist, and that something defies scientific explanation--even in principle, the explanation can be said to be transcendent of science. This does not necessarily imply anything supernatural. What it means that as matters now stand there is something fundamental that science is lacking by way of understanding the phenomenon in question.
______________________________________________
Moral choice is a great example - first we'd need to define morality, then assuming we could examine what the human moral experience is, which we know from psychology is a rule based system of thought linked to emotional states. If you are attempting to argue that there is a grounded objective morality that AI would need to recognise you'd need to explain that and present it, because i'm seeing no such evidence.
All of this is replicable.
COMMENT: No. I do not have a burden here to explain our moral sense. AI has the burden to explain our moral intuitions computationally. But, how is that possible? Where is AI's model of "the rule based system of thought linked to emotional states?" I do not think that there is a grounded objective morality of moral states, but that does NOT mean that moral judgments are computational. Quite the contrary, it means that they are NOT! But, the fact that moral judgments are not objectively computational does not mean that they are meaningless, or even that they have no metaphysical basis.
________________________________________
Critics of AI appear to use the Venn diagram definition of AI. Attempting to argue that whatever AI cannot yet do is what human intelligence actually is.
COMMENT: No! Critics simply point out, as I have, the limits of AI. Numerous people have done this, perhaps most notably John Searle and Roger Penrose. At the end of the day, it is AIs burden to produce a system that models human insight; as I have suggested above that creates scientific discovery!
____________________________________________
I guess this is where we as fellow thinkers need to start, aligning ourselves as colleagues to white board what we agree humans are or are not, and reduce the human experience of intelligence, the mind, emotion, preselection was much as we can to its parts to see if it is replicable. And if it is great, i'll take Aristotle and follow the argument where it leads.
COMMENT: Fair enough, but a reductionist program is a non-starter. Systems theory (outside of AI) acknowledges emergent properties, and there is no question that there are non-computational emergent properties of brains, e.g. consciousness. If AI can produce a system that instantiates such properties, then I will take notice.
___________________________________________
I think from here on, I would be just repeating myself.
Thanks for the dialogue.