Memory Systems

You have undoubtly heard the

You have undoubtly heard the old saying what came first the Chicken or the Egg. Well in our case what came first the Software or the Hardware. You are correct in saying “Software has infinitely more “moving parts” and creativity” it takes in most cases more to make the software work than the mechanics.

Memory Privacy w/ Ethereum BlockChain

This is another alternative for storing memories, but I thought it deserved its own forum.  I started a new forum on this about Ethereum and Blockchains.

I wrote a long paper last year (for a very large corp) about how to build a SkyNet AI for home robots.  Someone in the audience asked “What about privacy?”   …which I expected to come up, but had no truly good answer.  Any central system built with today’s standards is hackable, it’s just a question of when.  Also, corps are monetizing our privacy.  I basically had to say “Unless someone comes up with a solution, corporations will control our data and our privacy, interrupted by periodic hacking, theft, or extortion.”    …very sad truth, but that IS the world we live in today.

I now believe the idea exists to start solving this fundamental problem.   The idea is the blockchain, more specifically, some variation of the Ethereum blockchain.  I could be wrong, but this feels closer to a solution and a world I would want to live in than any other.  Much research still needs to be done to make this viable.

I think we all collectively have a choice between a diastrophic Orwellian future (like the present day in many ways), or a future where we are not afraid to have robots, TVs, or anything else with video cameras and other sensors ON all the time.  I personally will likely choose to NOT have these devices until they incorporate something like blockchain tech.  

For now, electrical tape is a temporary remedy for cams, unplugging for things like Alexa.  Phones can be put in drawers (for sound and cam proofing) or simply left in a separate room away from family.  Some people think these countermeasures are not necessary, they are naive.  There are so many documented cases of apps abusing consumers.

I hope this stimulates thought and further discussion.

Using files to store memories

I have been working on storing memories in text files. Each file is a different memory type for example:
BlankCanBlank.txt, BlankAreBlank.txt, BlankHaveBlank.txt, BlankHaveBlankBlank.txt. Look familure Martin?
The file is simple, for example the BlankCanBlank file looks like:
The ! before bark and fly is Not Bark and Not fly.
This way it is easy to parse. I have also created  OORelationship and OMRelationship files. these are my one to one and one to many relationship files. The OORelationship file looks like:
The ! not at the end signifies an execption.
The simple brain as I call it uses these files to create assumptions such as (If a bird has wings then a bird can fly). These assumptions are stored in another file Assumptions.txt and are used during Chat. Once an assumption is confirmed or proven false, the assumption is moved to a memory file and deleted from the assumptions file. I also have a user data file to keep track of the different people the robot (UNAS) talks with. When the robot detects a face and does not recognize that face, he asks the person there full name and then assign that user a user ID and logs this information in the file along with date and time last seen/talked and a conversation ID used with ProgramO (Not sure if I will stay with the AIML programO). I am currently working on NLP (using the JavaScript library Compromise) Martin what NLP engine did you use with Lucy?



re: Richard

Hi Richard,

Sorry not to have responded sooner, I don’t get on LMR much.  It sounds like you are doing some cool stuff.  I commend you for jumping in, it is a fun area with many rewards.

If memory serves me, I was using SharpNLP…because it was C#.  I added layers on top of it.  The NLP libs by themselves are only a start.

On your examples for duck, dog, etc…these sound like what I call “triples”.  You might want to consider them generically so you don’t have to worry about organizing them by topic.   Example:  A duck has wings.   This is another triple, but where do I store…ducks or wings.  Maybe it doesn’t matter if you are loading them all into memory.  When the bot learns something new, it will have to decide where to store it though.   It seems like you need a database to me, while I concede that it can be done with files if you have enough memory and have something like hashtables.   Hey, whatever works for you!

I find that in order to deal with logic questions…you need a large number of triples in memory at the same time.  There are some academics that have assembled millions of them I think, too many to cache, a mechanism like SQL would be needed to warm up subsets.

I like the assumptions piece (which could be related to curiosity) and the person stuff you bring up.  Be careful though, if the assumptions are simply derived from the learned knowledge, it would lead to a lot of redundant storage.  Example…if mammals can birth, do I need to bother remembering that people give birth, dogs give birth, cats, etc.  That type of thing can always be figured out on the fly when needed.

It looks like you are well on your way to something very interesting.  Don’t take any of my input too seriously, just opinions based on incomplete knowledge.

I’ love to hear more, as well as your experiences with AIML



Computer memory is any physical device capable of storing information temporarily or permanently.

go on this Linksys Router Support
if you have a router issue.

Have not given it much thought, but if memory serves me well, a large size SSD is becoming less and less expensive, so it’s more of a question of “memories” stored locally or “in the cloud”. Reminds me of the movie NEXT GEN on Netflix.

Very informative thread! Big thanks to everyone who contributed, I greatly appreciated it as a complete newbie.