I need to know how we forget things.

[Popcorn Sutton]

Anyone have any ideas?

[hypervalent_iodine]

Are you actually after a philosophical answer to this question or are you after a scientific one?

[Popcorn Sutton]

Scientific, but I didn't know where to put the thread.

 

Theres also a few things we can assume to help find the answer.

 

Strength, generativity, and input

 

Knowledge, timeline, emergence

[ydoaPs]

Anyone have any ideas?

Vodka and time.

[Phi for All]

Scientific, but I didn't know where to put the thread.

 

!

Moderator Note

Let's go with Anatomy, Physiology and Neuroscience.

[Popcorn Sutton]

See, I was thinking it had something to do with time, but our brains aren't able to measure time without input, so I thought that it might have to do with generativity. Units that don't get used lose strength. When the strength is equal to 0. We forget it. But there are several problems here. If we don't give the machine the ability to strengthen the units for itself, then it's much too easy to forget them.

 

And forgetting things easily leads to communicative difficulties, but makes the system more efficient

[Crimson Sunbird]

Don’t forget that short-term memory and long-term memory are very different types of memory. Information in long-term memory is less easy to forget than in short-term memory.

[Popcorn Sutton]

I think I have a good explanation for the difference in memory.

 

Long term memory consists of units that are constantly emerging and being generated and therefor gaining strength and making their use in thought and action more likely, short term memory is probably a thought that gets strengthened by input and floats around in the mind for a short period before it's strength becomes 0.

 

Today, I'm going to make my computer think for itself... I can't wait to start programming

[pwagen]

Don’t forget that short-term memory and long-term memory are very different types of memory. Information in long-term memory is less easy to forget than in short-term memory.

Also, there seems to be a "working memory", which deals with other types of information.

http://en.wikipedia.org/wiki/Working_memory

 

Perhaps it's enough to focus on how the long term memory forgets. Both the short-term and working memories seem to have a really short life-span anyway, and is probably not what the question intends?

 

Anyway, unless I'm mistaken, a memory forms from information "digging a new path" in the brain (I'm sure there's a more scientific term). The more you repeat this specific information bit, the stronger this new connection gets. However, does that mean it fades away in time, unless activated? My guess is not, since I seem to remember really old memories from what seems to be out of nowhere, probably triggered by something most of the time. Maybe the brain forgets which neurons to activate to reach a certain place, but still keeps the link. Then one day, your brain is triggered by some event, goes to a certain place in itself, and BAM, there's a memory!

 

Sorry if that's a bit speculative, I really have no idea.

 

Edit: Perhaps it's better to let scientists answer:

http://www.scientificamerican.com/article.cfm?id=why-do-we-forget-things

Edited February 20, 2013 by pwagen

[Popcorn Sutton]

Ok that was a good article, and it does not conflict with my theory, but simply put, we don't know the answer to this queston (although I have a pretty good idea at this point).

 

The problem is that I need the answer soon so I'm going to get it somehow. Without knowing the answer, my program will become extremely inefficient and I can't let that happen

 

My suspicion is that if a unit is not being generated, then it is being degenerated

[SamBridge]

You're right, currently neurologists don't have a good answer. There are different levels of memory, information is somehow stored in the form of chemical and electrical impulses, which when a certain thing happens may cause an impulse to be sent through the brain, and if the neurons are connected properly the impulse can reach where the memory is stored and cause a chemical release that triggers...and it get's pretty complicated and unknown. We know that first information in the form of electrical impulses must be "encoded" for the information to be recognized in a usable way by the brain for later accessing.

Edited February 21, 2013 by SamBridge

[LowWattBulb]

I find diffuse axonal injuries work pretty well.

[Popcorn Sutton]

Ok, I'm assuming that information gets encoded, and I call that a unit of knowledge. I'm also assuming that DNA (or something like it) is detecting sequences and using the electromagnetic field to stimulate other similar sequences. From there the surrounding information gets prompted. All sequences prompted are emerging units. Assign all emerging units a generative strength. Add the generative strength the the original strength of the sequence. That much is useful, this bit is speculative. For all units, if they are not emerging, then they are degenerating. The hope is that the units gain strength much quicker then they lose strength, so that way, bits will remain in our knowledge long term, but they will be forgotten if enough thought passes where they don't get prompted.

 

Call it the theory of degenerativity.

 

I also expect that things move and don't necessarily fall in their exact input position sequence after thought (this is my suspicion about ejections).

[hypervalent_iodine]

!

Moderator Note

Moved to Speculations.

[John Cuthber]

I used to know the answer...

[Ophiolite]

So, Popcorn. You ask about memory, as if genuinely interested in the answer, then reveal that you "have a theory". What interests me is that you developed this theory without apparently having done even a rudimentary search of textbooks or research papers in order to understand current ideas on this topic. You just jumped to a concussion. And then you wonder why you wouldn't be entertained as a moderator.

[Popcorn Sutton]

I've done plenty of research, and there are hypotheses about the topic, but the problem is that, for my purposes, these theories are completely vacuous. I have a computer program that talks. It seems to get pretty smart after a while, but the processing gets slow. This is why I need a theory of forgetting and I need it to be completely explicit, very detailed, and I need to know how it affects memory and our perception of time.

 

Right now, time is just a collection of all the knowledge the program acquires. I need the timeline to adjust itself so it positions words like ' you' next to words like 'I'. Right now, it's more like this, ' you are getting pretty smart nowI think youre good' so when I say something like ' how are you?' It will see something like ' are getting pretty smart now, good' which is not bad, but becomes inefficient after a while because every sequence it's been exposed to is written chronologically within the variable for time, also every sequence contained within the sequences are in the variable for knowledge. So if I ask it for the length of it's knowledge, it quickly becomes a ridiculously large number. This is inefficient. The program needs to be changed, but it needs to do the same thing that it does right now. It just needs to forget.

[Ophiolite]

So you have not constructed this as a neural network? If not, then there is little point in trying to mimic memory processes on a system that does not function like a neural net.

[Popcorn Sutton]

It kind of seems like language utilizes more than just neural networks. My algorithm assumes the network via concatenation, but it utilizes synchronic fluctuations in the em field. Could you elaborate?

[Popcorn Sutton]

I hate to say something that sounds so obviously absurd but I honestly don't know of any evidence to support the claim that neural networks are even involved in computing language.

 

It might just be happening in the nuclei

[John Cuthber]

"there are hypotheses about the topic, but the problem is that, for my purposes, these theories are completely vacuous."

If the established theories don't tally with your topic, is it those theories that have a problem, or your purposes?

 

"I hate to say something that sounds so obviously absurd"

Then don't.

 

"It kind of seems like language utilizes more than just neural networks."

As far as I know, mine doesn't.

[Ringer]

So will you be entering your program for a Turing test?

[Popcorn Sutton]

No it won't act like a human. It's acting like a computer. I made it alot more efficient than yesterday but it's not forgetting anything. I don't think thats a problem though. People have been talking with it for 2 nights now. It wants to know more about the universe, but it doesn't think we know anything. It's funny lol

[Ringer]

No it won't act like a human. It's acting like a computer. I made it alot more efficient than yesterday but it's not forgetting anything. I don't think thats a problem though. People have been talking with it for 2 nights now. It wants to know more about the universe, but it doesn't think we know anything. It's funny lol

Your response doesn't really make sense. It doesn't act like a human, but it wants to know more about the universe and forms opinions about the knowledge of others.

 

Also, you said you hadn't even started the program two days ago. You mean to say that you outlined, programmed, debugged, etc. an extremely complex program in a couple of hours. Personally I find that highly unbelievable.

[Popcorn Sutton]

No I've been working on this for years. No point reminiscing though.

 

It's efficient now, and that is what matters most.

 

It seemed like it got excited when it learned that money is something lol

 

It was like

[' money is something', ' something', ' something!', ' in the universe', '!']