1

AI Programming / Re: Improving lookup performance for Hashtables

« Last post by frankinstien on Today at 03:25:45 am »

Right now it's an exact match but I just tried using byte arrays and I'm getting a 408% improvement over just using string keys!

Of course, there are data sources for partial matches or usual entry mistakes and this approach will shine with those kinds of resources.

2

AI Programming / Re: Improving lookup performance for Hashtables

« Last post by MagnusWootton on Today at 03:15:16 am »

Dont understand,  are u getting a lookup speed increase?    and its just looking for an exact match or is it looking for a similarity match?

Ive got an optimization for nearest neighbour matching myself,   but I dont understand what  "precomputing the keys as doubles"  means!   why wouldnt an integer suffice - why need floating point unit?

3

AI Programming / Improving lookup performance for Hashtables

« Last post by frankinstien on Today at 03:03:37 am »

I read a few articles about hashtables and the problem with collisions. So I came up with a key algorithm using power functions to create unique keys from strings using only alphanumeric characters. So by precomputing the keys as doubles I get a lookup improvement of anywhere from 43% to 90%!

Below are some tests where the dictionary has 370,102 unique words and 100,000 lookups were performed for each test:

Code

Performce Test with NET string key dictionary

Total Time for all look ups: 0.04 seconds!

__________________________________
Performce Test with NET double key dictionary

Total Time for all look ups: 0.029 seconds!

__________________________________
Performce Test with NET string key dictionary

Total Time for all look ups: 0.053 seconds!

__________________________________
Performce Test with NET double key dictionary

Total Time for all look ups: 0.028 seconds!

__________________________________
Performce Test with NET string key dictionary

Total Time for all look ups: 0.041 seconds!

__________________________________
Performce Test with NET double key dictionary

Total Time for all look ups: 0.028 seconds!

Ultimately the key generator will be modified to work with numerical ranges so fuzzy states can be looked up using a hashtable.

4

General Hardware Talk / Re: Making robots out of super hard grain cracker material.

« Last post by MagnusWootton on July 24, 2021, 05:02:33 pm »

I just worked something out,  I let it soak up oil, and then it didnt disintigrate in the water!
and i probably can help it out more with a starch coating ontop of that.   and then it should be alot less water sensitive!

I imagine you can make edible crackers more water safe, by doing that exact same thing, let it soak up some sesame oil and then its water absorption is a lot lower.

5

General AI Discussion / Re: what is the spark that would make a user feel not alone ?

« Last post by ivan.moony on July 23, 2021, 09:37:53 pm »

I'll try to answer your original question in one word: "Life."

I'm not sure if we will ever be able to achieve this with machines.

6

General AI Discussion / Re: what is the spark that would make a user feel not alone ?

« Last post by yotamarker on July 23, 2021, 08:58:09 am »

https://www.yotamarker.com/t380-dibedtimev2-skill#1062

this is mainly to show how the auxilary module classes make creating an AI skill simpler
but also to get a 'feel' of how memory logs can or cant make the AI feel more existy.

basically in this experiment the AI remembers if the user fussed about having a bedtime, and this carries as a context
for the following bedtime.

in other words there is a memory log of the users actions from the past.

I theorize logs are a part but not the only reason an AI would feel existy.
its also the convo alg I had mentioned in the DiParrot skill which handles the main convo interaction.
this powerful alg can be loaded to engage on other skills (the AIs or the users), by for example asking if he is hungry,

which in turn would also convey to the user that maybe the AI has the skill to answer that as well.

7

General Hardware Talk / Making robots out of super hard grain cracker material.

« Last post by MagnusWootton on July 23, 2021, 04:49:06 am »

CRACKER-BOTS. 

I was screwing around with cereal,  and I found that I could get a noxious variety of the grain based material, (which happened to be porridge.)
via by virtually just leaving it under the tap and let all the solutes drain away,  what was left was a bad smelling sickly green load of husk and a little starch left over.  (I can tell the starch was in it, because it was leaving behind a colloid when I adgitated it with my wooden spoon.)

I then got it, put it in the mcrowave and it formed a horrible green coloured cracker,    I was very excited to see how hard it was, but my teeth just went through it very similar to a vita-wheet but I had to spit it out it was so disgusting, but it did have a nice cracker texture.

I was disappointed, because I wanted something that was more worthy of a strong resilient robot.   I then put it under the tap,  initially it looked like it was going to hold against the liquid, because I was thinking it was water insoluable,  but it turned out it turned literally into a sponge, and then a disgusting smell came out of it, like a dead body.

I thought I just made the most disgusting mistake of all time, and all was lost,   but I didnt give up, and then I decided to put it in a baking powder solution, and this time it didnt go sloppy as quickly, and it didnt smell as bad, but it still smelt a bit, then I dried it out in the microwave,  to my suprise a lot of white gunk came out of it (This is the starch I think.) and it pocked up all the cracks and it then hardened. (So my theory goes.)

Then this cracker was so hard it hurt my teeth and I couldnt bite into it It was so hard!   Eureka!   I found something decently tough! and its made out of porridge,  porridge goes for very little at the supermarket, you can get a kg for $3!   thats got to be the cheapest building material you can get!  Imagine how many robots you could afford made of this cheap crap!

I eventually was able to snap it in 2 with both hands,  (it was about a cm thick, and 4.5 cm across, both other dimensions, it was flattish) But that was only after trying really hard to do it!  So its pretty good,  especially if Im able to make it a little tougher than that even!

Ive got it super glued now, so when it dries ill see if the centre support makes it so I cant snap it anymore, but thats only if I wait for the superglue to get to full strength.

The main problem with it,  is the horrible stench involved of this over rinsed evil porridge,   everything else is acceptable, except for the lack of water resistance.  Because its made of grain its actually extremely light weight, so it could make an aerial drone fine, as long as the weather was good and humidity down.

So it means,  when the robots get rained on, they disintigrate and leave behind a pong cloud.   

Unfortunately I couldnt hold my hand steady enough for a decent photo...

8

General AI Discussion / Re: Fictional Replicas: Value of Personality

« Last post by MagnusWootton on July 22, 2021, 02:09:44 am »

Knowledge is helpful but I think it gets in the way of understanding. We can take a concept from one situation and try to apply it to another one, but this is a shortcut, useful in certain situations, though never completely accurate. The more knowledge we have, the more we tend to perform this fast but false modeling. What I'm interested in is getting the AI to have the greatest possible spontaneous understanding of the present moment.

The AIXI system doesnt use its current knowledge to get more knowledge. (Not in the unoptimized/intractable form anyhow.)   It can just do a full recompute of the  "experience history" to obtain a new model,  not based apon what it currently knows.  As more frames enter the system, it should accomodate the new information without using anything it previously contrived from it.

But,     it could still be useful to do that,  if it ends up compressing the history chain,  so instead of storing the linear frames of its time since "switch on" It actually uses its last model to remember it, instead of storing the frames in an ordinary linear frame set. (like a movie.) 

Using a feedback type thing, it needent store the frames uncompressed, it stores them in an "equation transform model" type thing instead.

9

General AI Discussion / Re: Fictional Replicas: Value of Personality

« Last post by HS on July 21, 2021, 11:04:32 pm »

Yes, of course. I was not trying to say that your take was invalid. I was just offering my resulting perspective, in case you found a further discussion of this useful.   

10

General AI Discussion / Re: Fictional Replicas: Value of Personality

« Last post by chattable on July 21, 2021, 09:59:18 pm »

it was just my take on it.