I glanced at the paper by Craig Gentry http://crypto.stanford.edu/craig/craig-thesis.pdf and from the little that I can pick up intuitively it looks like homomorphic encryption relies on different levels of noise which I think are used in cloaking the data. So this method wouldn't work on ordinary compressions. I think this may be related to whatever it was that Ben Goertzel was saying. However, my original thought was that base-n numerical representations are compressions and there are great functions that can be worked on them without decompressing the numbers into unary form. So the binary representation system is a compression scheme that has a lot of great functions that go along with it. But there are some systems of references that refer to objects that might be better encoded using a different kind of method and my thought now is that those alternative systems need to have their own functions encoded that can be used directly on the compressed representations. It may not be possible to make these fully homomorphic so that any familiar function from computer programming (up to some level of complexity) can be run with the compressed data, but that is not what I am getting at. I originally thought that to make this kind of thing useful different compression schemes would have to be able to translate and use data compressed with other compression algorithms without needing to decompress any of the data. This was too ambitious so I started thinking that I was overly presumptuous with the title of this thread. However I now realize that if one compression scheme can be combined with some functions to operate on the compressed data then two different compression schemes might be created to be used in common functions without decompressing the data from either of the two different schemes. So I am now drifting back to my original thought. The goal is to eventually create a few different compression schemes which would be able to communicate and use data from the other scheme without needing to decompress it first. I am not really good at designing compression methods so this is too ambitious for me but it is an idea that should be explored. The idea is not to keep the data secure (from being decompressed if intercepted) but to make the algorithms that use the data more efficient. Jim Bromer
On Sat, Nov 29, 2014 at 8:34 PM, Ben Goertzel <[email protected]> wrote: > Encryption and compression are different. > > Mathematically, it seems there is no way to make compressed versions > of data equally amenable to transformations as uncompressed versions > (without making many of those transformations very expensive, and > essentially requiring decompression and recompression).... > > But if there are specific transformations that interest you, you could > maybe design a compression algorithm so that *those* specific > transformations are efficiently do-able on the compressed data... > > -- Ben G > > On Sun, Nov 30, 2014 at 9:31 AM, Jim Bromer via AGI <[email protected]> wrote: >> Well maybe compression systems just have to be designed so that >> transformation functions can be applied without decompressing the >> data. Perhaps cross-compression transformations are not necessary. I >> am only saying that because standardization and regularization is what >> would make the transformation functions on the compressed data more >> feasible to develop. >> Jim Bromer >> >> >> On Sat, Nov 29, 2014 at 3:29 PM, Jim Bromer <[email protected]> wrote: >>> I may be technically wrong and technically right about this. An >>> encryption will often (if not usually) expand the data, so an >>> encrypted database would not technically be a compression. However, an >>> AGI program would typically need to encode some central subject matter >>> so that it could be used in a variety of ways. So this might be seen >>> as an expansion of the central subject matter that might be referenced >>> for some particular purpose but the expanded data might stand as a >>> compression for all the ways the subject matter could be used. So the >>> data would not be compressed relative to just the central subject >>> under consideration but it would be compressed relative to the variety >>> of ways that subject data might be used. If these methods included >>> transformational methods which could be used to 'calculate' the >>> results based on various ways that interrelated data might be used >>> then the system might be able to run these transformational >>> 'calculations' without unencrypting or decompressing the data. These >>> 'calculations' would not typically be comprised of standard >>> contemporary numerical calculations. >>> >>> I keep thinking of virtual networks of cross-generalizations. Each >>> generalization path might represent a compression along some line from >>> generalization to particularization. However, if the generalization >>> node could also be referenced from other levels which were related to >>> the generalization node, then the system might both be seen as an >>> expansion of any one node but a compression of the potential of the >>> entire system. >>> >>> Inventing new kinds of mathematical systems along which could be used >>> across the systems and across different compressions is going to be >>> difficult. Well, it probably isn't that difficult to create simple >>> prototypes of such systems but it will probably be difficult to create >>> effective systems. >>> Jim Bromer >>> >>> >>> On Sat, Nov 29, 2014 at 12:52 PM, Jim Bromer <[email protected]> wrote: >>>> A slightly modified statement about Compression Transformation that I >>>> made in The role of prediction [was What's preventing me...] >>>> http://www.jimbromer.com/TheNeedForTransformationalCompressions.html >>>> >>>> A simple example from MIT is given about using an encrypted database >>>> to make queries without first decrypting it. >>>> >>>> Processing Queries over Encrypted Databases >>>> can be found on page 13 of >>>> http://www.eecs.mit.edu/docs/newsletter/connector2014.pdf >>>> >>>> Jim Bromer >> >> >> ------------------------------------------- >> AGI >> Archives: https://www.listbox.com/member/archive/303/=now >> RSS Feed: https://www.listbox.com/member/archive/rss/303/212726-deec6279 >> Modify Your Subscription: https://www.listbox.com/member/?&; >> Powered by Listbox: http://www.listbox.com > > > > -- > Ben Goertzel, PhD > http://goertzel.org > > "The reasonable man adapts himself to the world: the unreasonable one > persists in trying to adapt the world to himself. Therefore all > progress depends on the unreasonable man." -- George Bernard Shaw ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
