|
Post by K. R. Thórisson on Nov 14, 2012 10:01:55 GMT -5
Fyrst of all what is a pan-architecture? Also can you describe further Transversal function. "Pan-architectural functions" and "transversal functions" are synonymous and mean "across the whole architecture or large parts of it". For example, if an architecture represents tasks in a distributed manner – across the entire architecture – and yet each one of them can be learned in virtually the same way, and modified in the same way, that learning is pan-architectural, transversal.
|
|
|
Post by K. R. Thórisson on Nov 14, 2012 10:04:06 GMT -5
Do you think it is better for a robot to start learning about the physical world in a smaller constructed physical form and get to understand it, then be transferred to a larger body? I find it likely for that to be the case, especially when our attempts at constructivist systems are immature and we need to carefully guide their (self-)growth in so as to get a predictable outcome.
|
|
|
Post by K. R. Thórisson on Nov 14, 2012 10:05:00 GMT -5
Challenge: I belief that AGI can adapt if it reaches the point that it can experiment and learn about the world and update itself to function properly within it to achieve its goals. I don´t see any reason that it should be impossible for robots to have these capabilities in the future. I completely agree :-) The AERA system is already showing this to be true.
|
|
|
Post by K. R. Thórisson on Nov 14, 2012 10:11:02 GMT -5
1) Which has received the most attention so far, new development environments, programming languages, or architectural metaconstruction principles and which is the most important of these to achieve an AGI. Actually, programming environments can come later – although of course we need some preliminary environment to get started. The other two are essential. There is no way to create AGIs unless we increase the power of their internal environments. The complexity that arises from having "modules" or operational units larger than a small statement (equivalent to a line of code in e.g. prolog) prevents the creation of systems that can represent the combinatorics of external realities that we ourselves are exposed to when we act in the world. The kind of building bricks that enable a system to assume the necessary number of states to represent the myriad of possible combinations of external reality that is a prerequisite for it to operate in a complex world, such as the real world, can only be achieved by making the code be the level at which the system modifies itself – hence the need for self-programming. And, as is so often the case, it is not simply one or two things or principles that enable AGIs, it is a myriad of small and large things that must be "aligned" or assembled properly. The AERA systems provides one real example of this.
|
|
|
Post by K. R. Thórisson on Nov 14, 2012 10:21:28 GMT -5
Challenge: Let's say we have an input that looks like this: 1 2 3 4 1 + 1 = 2 1 + 2 = 3 We feed it to an AGI that has just been booted and knows nothing, not even numbers, addition nor equality signs but we want it to answer the question 1 + 3 =. For this to work we need it to figure out that there is a sequence, there is addition and how it works. Using a constructionist method we might do something like make a sequence module and an operators module that then figures out an addition module. Making the paradigm shift towards a constructivist approach, then what, where would we begin? The only sensible way is by taking a pattern-identification approach. The system that we build must have the capability of identifying patterns, both temporal and spatial, and the the ability to create models that predict the way the world works. When it then tests these models in reality it gets feedback – like a scientist getting feedback from an experiment he/she makes – that provides input to either modify and improve the model, discard it, or keep it as is. These models, once they describe basic patterns, can then be combined to form hierarchies, which describe patterns based on patterns. And little by little the knowledge of the system grows, based on the system's experience in the world that it lives, to describe ever larger parts of that world. At some point the system will be ready to "see" the patterns that underly mathematics. Then it's time to apply the SAME mechanisms to that particular set of patterns, and you have a system that is learning mathematics. Of course such systems must have the essential prerequisites for this to work: a memory system with efficient read/write capabilities and appropriate access mechanisms based on association (patterns!), perception mechanisms, and inference mechanisms that include deduction, induction and abduction. Deduction is used for inferring what the models predict (simulation), abduction is used for goal regression (finding out how the system can achieve some goal, or finding out that it does NOT know how to achieve it), and induction for generalizing sets of patterns already observed that have significant things in common. Additionally it would not hurt to have powerful analogy-making capabilities – that is, a transversal similarity function that is independent of level of abstraction and level of detail.
|
|
|
Post by K. R. Thórisson on Nov 14, 2012 10:24:13 GMT -5
(1) "I am aware of only one architecture that has actually implemented such an approach, the Loki system, [..]. The system implemented a live virtual performer in the play Roma Amor which ran for a number of months at Cite des Sciences et de L'Industrie in Paris in 2005, proving beyond question that this approach is tractable." Could you elaborate in some way on this? Specifically, how did the implementation prove that the approach was tractable? It's not a mathematical proof, it is a practical proof: The system actually ran and grew in useful ways during its operation in a theatrical performance in Paris. The performance was open to the public and ran for several months; the Loki system played the role of an "actor" in the piece, controlling sensors, lights, projectors and sound systems during the performance, which included a number of human actors to which Loki responded. Rehearsals showed that Loki evolved with the actors over the time of preparation, just like a human actor would.
|
|