I propose that the *only* way ahead is via a fundamentally new representational formalism that would clearly reveal what critical information has been missing from the conventional forms of data representation. That is, we need to get to the form of data representation that captures completely new side of objects we have been missing so far, which can only be achieved by relying on the entirely new representational 'tools'.
I'd like to defend two points:
1. Because AI is a fundamentally new kind of science, which should be dealing with the information processing in Nature, even *the birth* of AI as science, of necessity, must reveal entirely new scientific vistas.
2. I can see only one possible way this could happen: as mentioned above, it is via a fundamentally new representation formalism capturing previously inaccessible view of objects / processes.
If you can see a different way such (non-incremental) scientific birth could happen, I would like to here about it. Note that any incremental way will not do, since it cannot lead to a new kind of science.
depends on how u see the cognitive science and AI. modelling brain will change the fate of AI, many are trying to model it. more specific cognitive neuroscience
Cognitive science, including cognitive neuroscience, have been *talking* about the brain and touching almost every nanometer of it for ages. The issue is not to talk about the brain but to unlock the secrets of the representational formalism of *the mind*, which must be the same as that of Nature in general.
, Henry Markram. is the director of the blue brain project
http://www.technologyreview.com/news/409107/a-working-brain-model/
you can see it for clarity how cog neuroscience can help
cognitive science is very new field and from ages we are not studying mind/brain with multidisciplinary perspective. cognitive science will work as a basic science for the development of technology of AI. grand aim of AI is to build human mind/brain. untill you know what mind/brain is, how you will model it? .
Indrajeet, you seem to be confusing the mind and the brain all the time. ;--)
Indrajeet, may I also suggest to forget about the brain, since it is just one of the possible evolutionary / biological 'implementations' of the 'hardware' aspects of *the Mind*. Think of another galaxy and a planet with advanced forms of life in it, where many aspects of the 'implementation' are, most likely, quite different from the familiar to us..
If that exists still you will need to know how that alien brain/mind is build/works to model that. Still need of study of mind/brain
Finally, can you understand even computer software by studying its hardware?
yes lev i do agree that AI can develop without brain/mind study. but i think by getting some knowledge from mind/brain studies will speed its process
Indrajeet , i want to assure you that I do appreciate the unprecedented difficulties all of us have faced in trying to understand the mind phenomenon, possibly the last and most important scientific frontier. In that connection, I do understand why one would be looking for some sure ground one can stand on. But this is exactly the main difficulty in addressing the mind phenomenon: What is the ground we should be looking for?
Note the above two-step logic I'm proposing. First, since we are, for the first time, confronted with a new kind of science, unrelated to what science has been studying so far, it is quite prudent to begin completely anew. And second, the most *direct* way to open up entirely new scientific vistas is to generalize our numeric representational formalism (to get to a structural representation). Of course there are many other reasons why the road leads to structural representation. Moreover, in order not to full ourselves, any proposal of structural representation must *suggest* immediately what the new, radical, possibilities are being opened up.
However, the first point, although quite 'natural', is crucial here from a strategic point of view: without a radical break with our known formalisms we are doomed to go in circles (as AI has been doing now for half a century).
But who has such unusual scientific courage so needed for the development of AI?
Interesting question about how different representation can make difference in AI research! Different applications, information demand different type of representation. The challenge lies in selecting the right representation for the task in hand. Before proceeding further, I would ask myself these two questions.
Are humans the right person to decide which representation is best suited for the task?
Should AI algorithm be empowered to select the representation which will enable it to do its task in optimal way?
I think a good approach will be to have the reasoning or AI algorithm choose one or more representation for task at hand. This can be the representation 'tool' that you are talking about. Representation tool can select a representation based on some constraints. It should be possible to explore different representation and select the one which is most suitable.
We need of pool of representations for the representation tool to select from. This is something which should be left to the AI algorithm to decide rather than us, humans selecting based on our assumptions. We could start with simplest forms of representation like string, integer, raw binary data, and let the representation tool abstract more complex representation from them.
Subhash, the whole point is that there are reasons to believe there exists a universal form of representation.
Unfortunately, our present experience with various "representations" has led most, including yourself, to believe that the situation with multiple representations is inevitable and will persist.
- Badges
- Science topic
- Similar topics
- Computer Science
- Data Mining