The GMIS (General Machine Intelligence System) from a new article posted at codeproject.com looks interesting. Do you think that it could be a breakthrough in the field of robotics

  • $\begingroup$ It's far, far too soon to tell. $\endgroup$ Apr 21 '15 at 10:04
  • $\begingroup$ but Is there any theory can rise above it currently? $\endgroup$
    – myrobot
    Apr 21 '15 at 14:20
  • 1
    $\begingroup$ Your question is interesting, but i don't think it is a good fit for this site because it is primarily opinion based and elicits discussion rather than a canonical answer. $\endgroup$
    – Paul
    Jun 6 '15 at 14:13

I did some work on embedding decision trees in a hash table. That worked extremely well. I could embed many facets of the English language into it such as context sensitive spelling, in only a few seconds, using only a few megabytes. You could also embed complex patterns of movement into it very easily. It basically ends up as so called feature hashing. For example in 3 dimensions (x,y,z) you would create a hash function with 3 inputs h(x,y,z). This is a form of dimensional reduction and also data compression. I took it slightly further and feed in only a few bits of information at a time into the hash function for even better results and better generalization.
It is clear to me that you could fit everything the human mind can do into 100 Gigabytes or less. That means that a current mid-scale supercomputer with 300 Tbytes of SDRAM has 3000 times the capacity of the human mind and is much, much faster. A few years ago the general consensus among computer scientists was that you would see human level AI around 2020. I would agree, however their view was based on extreme hardware requirements. My view is that the hardware is already there, it is just a code issue. Hopefully it won't really change anything. That it will only be like adding 1 million people to the human population, in terms of increased ability to do things.


Not the answer you're looking for? Browse other questions tagged or ask your own question.