Neural Networks, Artificial Life and Ethics
- Kythe
- abrennan
- tuc
- abrennan
- Mercurian
- 76sig
This archived discussion is "read only".
» Kythe - Hmmm...
A recent breakthrough has been reported in the capability of neural networks regarding speech recognition:http://www.usc.edu/ext-relations/news_se...
Evidently, researchers have found a previously-overlooked feature of how real neurons "talk" to each other, and by implementing this feature in artificial networks, have supposedly developed a machine that outperforms humans in speech recognition.
The announcement is still "premature", in that it hasn't been extensively peer-reviewed.
However, it stands to reason that either this, or some future development will eventually allow us to build machines that can outperform humans at many tasks now considered exclusively "human".
The researchers in this story gave no reason as to why this development could not be applied to artifical neural networks in general, not just those used for speech recognition.
This raises an interesting question: if one makes a neural network (based intrinsically on the way the human brain works), will it begin to "think" at some point, even "outthink" human beings?
And what happens if and when we develop machines that can "think"? Would they become "conscious"? Would they naturally have a sense of self-preservation or a "need for freedom"? Would they "feel"?
How do we deal with the possibility that such machines might see humans as an impediment to themselves -- humans that understandably might view such machines as mere servants? Much of morality is based upon a relative parity between people -- lack of parity has frequently led to domination. How much power should we give to such machines? Are there other ways to prevent them from doing what we don't want them to do?
What of the ethics of developing machines that think as "tools"? On what grounds do we treat these machines as servants or tools? How are these grounds rectified with not breeding humans for slavery or spare parts?
These themes have been done and redone in popular and sci-fi entertainment, of course. But it seems that we might just have to deal with such issues sooner rather than later.
-- posted by Kythe
» abrennan - is the hippocampus a gland?
the article says that the machine is a great pattern matcher, better than a human. I think that is far removed from thinking.
it is my theory that the dictum "have brain will think" never applies. Nor does "build the better brain and the thoughts will come".
I have no evidence anyway but here are my THOUGHTS, hey is that a pun in this context?
it seems as if the thinking thing has developed along with us, part of us yet not easily definable in any seperate way from us. It is a part of what we are as humans. We are not a cake that you can cut up.
We have become built in the chaos of the natural world. I think that our heart is more involved in what we know as thinking than our brain is.
For evidence of that read the Heats code by Paul Pearsall, a american nuerobiologist of note and a very brave man for making public his findings.
so I don't say that we won't build a machine that thinks. But we aren't even close yet. We have a lot more understanding of our selves to do.
the closest machines will come close to think is the machine I heard about that will teach itself to walk. Now that is on the track that will lead us to thinking.
If it teaches itself. look out!
humans do not appear
The human mechanism
-- posted by abrennan
» tuc - Thinking Machines??
All neural net machines have a teacher that altersthe cross linked connection strenghts. Yet the
fact remains that any von Neuwman machine can
do and think all that a neural net machine can, only it's slower serially. The test of machine thinking on a human level is the turing test and
that was defined in the 50's way long ago.
But the questions that are bothering the professionals are termed Quaile or as it was first proposed by T. Nagel "What's it like to be a Bat" The raw feel or sence Data Question.
Searle made it much tougher when he Queried
How would one reduce The fellings and awareness
of understanding or Semantics to Syntax.And he stated you can't teach a machine, it just does as it's told, it can't feel or understand, only work
tautilogical algorithims.
-- posted by tuc
» Mercurian - Robots
In a recent issue of Scientific American, one computer researcher predicted that robotics might advance sufficiently in 50 years to mirror our own intelligence. Now, if a robot passed the Turing test with flying colors, and the Turing test is an objective standard of intelligence, then surely such a machine would be considered to have attained intelligence similar to our own. You might argue that such a machine would not be "human," but at least it would be able to mimic human intelligence such that it could be called a functional zombie - having no consciousness, but, nonetheless, indistinguishable from human intelligence. If such an occurrence were to happen, then such machines should be treated as machines like computers until they prove themselves otherwise. They should be programmed with Asimov's rules of robotics so that they don't harm humans. In short, I would let experience judge whether or not such intelligent robots are worthy of the same laws, governing principles, and ethics that we give to sentient beings. If robots prove their sentience, their capacity to learn and cognize as humans, then we would, therefore, treat them like humans or sentient aliens.-- posted by Mercurian
» 76sig - Re: Hmmm...
In response to message posted by Kythe:I think a good question to ask at this point is whether or not the 'AI' resembles humans in any other way than intelligence. It may or may not resemble the intelligence of a human, but can it reflect the same emotions or spontaneity that humans can? I agree that many films have attempted to display the downfall of human society with the advent of 'AI', but I don't find this to be the slippery slope that everyone imagines. Without emotions, there is no 'AI'. It takes more capacity to understand human emotions than it does to recognize a color or emulate a sound. Just a thought.
-- posted by 76sig
Please follow the guidelines set forth in the Suite101 Posting Etiquette when adding to the discussion.