Super Turing Machine

From Isaac Mao
Jump to: navigation, search

Is there one or more Super Turing Machine(s) running in our universe, or others?

We don't really need to change Von Neumann architecture,but we do need some reinvention on algorithms. The architecture may not support parallel computing easily, but at least we can realize some concepts of “me”, which may lead to the long disputed problem of computerized “self-identity”. In this sense, “Me Neuron” could be identified in biological research as well from human brain.

Both Hava Siegelmann and Juergen Schmidhuber have deep investigation on RNN, which includes many indications on recursive computing from neuron network. But both of them did not consider the high level of “meaning” of algorithm like Stuart Kauffman said in “Reinventing the Sacred”. I believe some multiscale computing in different algorithm could work together and interactively affect each other,though higher level normally can't change the algorithm working in lower level. So actually the requirements of “meaning” in higher level of system will call lower level of algorithm all the time, in purpose.

But Rao Mikkilineni pointed the problem of traditional Turing Machine is on identity of computer(or program). Without identity, any programs can’t really regard itself as a life one, or even not a machine. Nevertheless, his model on DIME yet to be challenged as well. And of course, all the people above should find a mild space that accommodate both Turning computing and cellular computing, e.g. Human Brain.

Today's Artificial Intelligence(AI) efforts seem getting more and more intelligent, but still far away from a simple animal's decision making complex. The key issue is on the computing model defined by Turing Machine. It's the shortcut to put many existing computing logics into programming languages, or modules(like NLP), to try to strengthen the initial condition of a program. However, it may ignored the learning mechanism behind a simple neuronal network. We may have to seek another path to reach future Artificial Intelligence, say via RNN.

blog comments powered by Disqus