Ai Dreams Forum

Chatbots => General Chatbots and Software => Topic started by: LOCKSUIT on March 14, 2017, 05:38:53 pm

Title: Question about how chatbots form language output.
Post by: LOCKSUIT on March 14, 2017, 05:38:53 pm
I really tried searching Google and read articles, Wikipedia, etc, and didn't find an answer.


#1 - Do any chatbots use Hidden Hierarchical Markov Models or the likes? Or just Databases and questions>answers? :(

#2 - Do you make some of its words/sentence-forming permanently rewarded so it says certain things ex. likes talking about construction working?

#3 - Do they create brand-new never entered sentences or pre-installed sentences?


I want a HHMM/similar (parse-trees) chatbot that recognizes word by letters, then sentence by which words heared, then topic by sentences, and then uses its rewarded parse-tree words to REVERSE the network so it ignites the topic it loves which ignites the sentences and therefore the words and the phonemes to then speak/write (once know words it knows which phonemes to speak/write).

Then I want the rewarded words/topics in itself to self-ignite and talk on its own at random times - always new talk by its old output entering.
Title: Re: Question about how chatbots form language output.
Post by: Art on March 15, 2017, 03:41:46 pm
1. Some do that I've experimented with in the past. At least there were some basic Markov elements contained in the plug-ins that allowed the bot to better handle various input. While the Markov wasn't exactly hidden, it was working in the background within the confines of the plug-in.

The same bot also allowed for a variety of pattern matching, internet searching and using "learned" data based on previous conversations. If it determined that it had some useful data that would relate to the current discussion it would often use that information to help with the conversational flow.

2. Within some bots one can assign a "weighted" system so that the bot will use one response over another perhaps less important one. The botmaster can easily assign these weights within the scripts. These weights might tend to serve more of a useful priority than an actual reward.

3. Pre-installed sentences. A bot will usually create sentences based on those it matched during it's pattern search routine (looking for the best match based on the input it received). No bots that I've experimented with were able to "create" anything new but can and will often combine some thoughts together based on previously entered snippets, words or sentences.

Remember, a bot is a glorified pattern matching program. For the present time, no bot is able to create something from nothing. It has to have something, some data from which to draw upon.

In fact humans usually call upon previously used fragments or parts of sentences or statements for their output.

Check out UltraHal from Zabaware, Jeeney from Jeeney.com, RiveScript (rivescript.com) and of course some other great bots like Mitsuku (mitsuku.com) and SkyNet-AI (http://skynet-ai.com/ (http://skynet-ai.com/)).

Hope this helps a bit.
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 15, 2017, 09:21:19 pm
Watch I can indeed create brand new never EVER written sentences (at least that I've never seen/thought of).

Watch how every word is different. And watch how, if two common words are together, then the next are separate i.e. it's pasting "2s" together just not "1s". Picking out 2 sentences to line-up is the same thing i.e: Lady Gaga. Good evening.

Here we go:
I am la chin friedo minister so kooooooos-ta uma jungle vixen teeth ho ox sand mitten loop tala bica cheena pala gabu gjrfj theujeuje g w i i t 6 7 ^ & % 4 5 z r z r zr z r.
That should do it.

Ok so let me go faster so I don't "think":
je brada minky tela dood nu monh vt Asv Free minu totla doodle bees bum to ca bullia.

Just to be safe let's use real words, no stopping:
hop pick me doodle why you bunny me pee deeble dop bog cat minitster meen been mop loop in me dog mop bop cat me dog fox why to the moon pool hop hop hole dole doodle.

My theory is the "tree" has a probalistic way that it will, by each words strength, pick which word first, then after that word the next strengths changes and so the next one sequentially gets selected.

Imagine someone saying something they most want to say. They'll never say anything else on their own ex. "turn left and go straight to get to the store" - only "give me food". The answer stored is not strong enough, until matched by input i.e. some stranger. And if that isn't enough, then seeing example hobos may match a strong positive memory that gets remembered and linked to their question and so your stored correct positive answer, combined with being matched, and this positive additive, makes you say the answer.
Title: Re: Question about how chatbots form language output.
Post by: Art on March 16, 2017, 02:29:07 am
Sorry to say but in your first example, those words have no real intelligible meaning(s).

In your second example, you indeed did use common words that already have established meanings (for the most part).

Third, what you believe to be original is anything but. What ever word(s) you choose to use in a sentence (aside from some made up meaningless words, have already been used by others in a variety of many ways.

Bottom line, your thoughts, even though they might be yours, are not of your Free Will. In fact, there's no such thing.

So study some other bots and A.I. experiments out there and begin to formulate your own, whether AIML, C#, C++, LISP, RiveScript, ChatScript, SuperScript or any other language that will work for you.

Until you can come up with a way for your (yet to be created bot) to create it's own language, with understanding of context, you should really study some of these other examples. Otherwise, it's 10% programming and 90% scripting!
(or something like that).
 Good luck!
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 16, 2017, 08:21:07 am
We don't have free will - I definitely wasn't saying we do.

I wasn't saying the words meant anything, and I wasn't saying they were never used.

What I was saying is that the word after word is unique - for example I have never seen this > Loop brown weener jackerlantern job for house. Which came out of my mouth/mind all in order (well this one not too fast but the others yes [3rd/2cd one]).
Title: Re: Question about how chatbots form language output.
Post by: Freddy on March 16, 2017, 12:50:22 pm
Hmm, no free will. What did you both have for breakfast ?
Title: Re: Question about how chatbots form language output.
Post by: kei10 on March 16, 2017, 12:56:34 pm
Hmm, no free will. What did you both have for breakfast ?

Pffffttt XDD
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 16, 2017, 12:59:41 pm
Same thing every morning - waffles.

Lunch: McCain frozen fries, 4 frozen nuggets, root-beer *POP, Chip Ahoy chewy cookies & 2% milk.

Next day's Lunch 1/2 a hotdog & 3 buns w root-beer *POP and then cupcake bottoms (little a tops too) & 2% milk, then the other lunch the next day.

Dinner: 6 Oatmeal muffins.

Repeat.

I mean, muffins with 2 Berry Punch Minute Maid juice boxes

Me and Art think the same !

I might be in range of thinking about a Language-only AI. Working working working. Also another way, maybe.

I'm always thinking how I will take over Earth.
Title: Re: Question about how chatbots form language output.
Post by: Art on March 16, 2017, 02:43:29 pm
I knew my posting would stir the pot, so-to-speak but it's always interesting to look at both sides.

@ Lock...sorry about your lifestyle / eating habits. Then again, you are a very young man and aren't likely to see the error of your ways for at least a few more decades...and believe me, you will.

While we might share an idea or two that run along similar threads, to state that we "think alike" is both inaccurate and presumptuous.

With regard to my intake:

My intake - an average day -

Breakfast - soft boiled egg & a few crackers, coffee OR oatmeal & piece of toast, juice and coffee.
Lunch - small salad (lettuce, tomato, onion) w/ tuna salad & few crackers OR homemade vegetable soup and small salad.
Dinner - small grilled fish filet, vegetables and half a sliced avacado & water w/ lime juice.

No processed food, no High Fructose Corn Syrup limited GMO's and limited white food intake (white bread, pasta, potatoes, rice), No soft drinks, plenty of water.

So no...we do not think alike when it comes to what we put into our bodies.

############################
There are always two schools of thought on practically every subject, theory and ideology. The one about Free Will or lack thereof, has been around for many years.

One excerpt of the following link is as follows:

Many scientists say that the American physiologist Benjamin Libet demonstrated in the 1980s that we have no free will. It was already known that electrical activity builds up in a person’s brain before she, for example, moves her hand; Libet showed that this buildup occurs before the person consciously makes a decision to move. The conscious experience of deciding to act, which we usually associate with free will, appears to be an add-on, a post hoc reconstruction of events that occurs after the brain has already set the act in motion.

I suggest one read the following posting: https://www.theatlantic.com/magazine/archive/2016/06/theres-no-such-thing-as-free-will/480750/ (https://www.theatlantic.com/magazine/archive/2016/06/theres-no-such-thing-as-free-will/480750/)
Title: Re: Question about how chatbots form language output.
Post by: Freddy on March 16, 2017, 02:57:07 pm
Have you heard of the book Diceman (https://en.wikipedia.org/wiki/The_Dice_Man) ? It's interesting in that the main character leaves his decisions to the roll of a dice.

Free will is an interesting subject, personally I think we have it. I could go on, but I am going to exercise my free will and postpone further comment until I have played a few more hours of my game. That's a decision on my part, my free will, electrical impulses and all.

You are all of course allowed to exercise your free will and reply or not ;)
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 16, 2017, 06:58:37 pm
Here's 3 good examples of 0 free-will:

You automatically carry out pre-installed laugh/cry actions and more so when put sugar/needle on tongue more so.

Sugar or rubbing back sends signals, this stuff is not actually elementally "feels good/pretty/tasty" because no particle arrangement can be, simply you are rewarding actions, doing actions on pattern recognition/self-ignition, and using a language neural network to say all your English-language words.

Your eyes automatically follow a moving target (only when moves otherwise won't happen) by the smooth pursuit mechanism.
Your eyes stay in place as move head by the Vestibulo–Ocular Reflex system.
Further my latest discovery is that when you look around the room while NOT moving eyes/body, and start moving forward, you can keep looking at the area by the same mechanism - the visual difference re-adjust the highlight area of the images like it re-adjusts the eyes.
Title: Re: Question about how chatbots form language output.
Post by: Freddy on March 16, 2017, 08:30:10 pm
I think there must be free will because of your creative use of the English language Lock  ;D ;)

Are both of you telling me we go through life without making choices for ourselves ?
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 16, 2017, 08:33:16 pm
1. Yes.

2. Hope robot don't make new major decision like you suggested.

In other words we are, who we are. We are.....we are....we... are. . .

But don't set yourself to now walk lop-sided everywhere you go~
Title: Re: Question about how chatbots form language output.
Post by: Freddy on March 16, 2017, 08:37:01 pm
OK then I am banning you and demoting Art.
Title: Re: Question about how chatbots form language output.
Post by: korrelan on March 16, 2017, 08:41:56 pm
@ lock

Total and utter boll*cks... as usual

Do you even think... before you post?

Quote
In other words we are, who we are. We are.....we are....we... are. . .
But don't set yourself to now walk lop-sided everywhere you go~

I am growing tired of your... spouts.

Explain.

 :)
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 16, 2017, 08:54:16 pm
boll*ocks?...........as usual??

Well I am currently really busy.......I don't like to make every comment superb like gold it takes too long...I do that to my Survival Guide/ work.

But I am trying to keep my threads tied together now etc. Big roads (and work ahead).


Explanation:
I meant we are machines, destined, me writing this - but to say "ok then I'll go jump" is a machine that learned this secret knowledge and is being not good machine now. If you picture my belief for a moment, you should get it.
Title: Re: Question about how chatbots form language output.
Post by: korrelan on March 16, 2017, 09:24:27 pm
Quote
Well I am currently really busy.......I don't like to make every comment superb like gold it takes too long...I do that to my Survival Guide/ work.

But you are posting to your peers… we are listening and judging you…

how does …

Quote
In other words we are, who we are. We are.....we are....we... are. . .
But don't set yourself to now walk lop-sided everywhere you go~

=

Quote
I meant we are machines, destined, me writing this - but to say "ok then I'll go jump" is a machine that learned this secret knowledge and is being not good machine now. If you picture my belief for a moment, you should get it.

Take your time… I'm quite intelligent… explain your reasoning… step by step…

 :)
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 16, 2017, 09:35:44 pm
"In other words we are, who we are. We are.....we are....we... are. . ."

&

"I meant we are machines, destined, me writing this"

= we machines, i am this



"But don't set yourself to now walk lop-sided everywhere you go~"

&

"but to say "ok then I'll go jump" is a machine that learned this secret knowledge and is being not good machine now. If you picture my belief for a moment, you shou"

= police machine will lock you up if you get all funny over knowing we're machines, don't go around noe saying "I guess I'll sell my house, I guess I'll drive in a circle 5 times" cus you didn't talk like that yesterday :)
Title: Re: Question about how chatbots form language output.
Post by: Art on March 17, 2017, 11:45:43 am
What?! Please don't tell me that you are involved with a Survival Guide...unless you're actually reading one. Then that would be OK, I guess.

Free Will might or might not exist depending largely upon how the concept with all it's vagaries might influence or affect us.
One thing we do have is Choice. That which we elect to do, now or at any given moment, not a predestined moment but one of our choosing.

I am reminded of something I once read:
"When we are born, we look like our parents. When we die, we look like the choices we've made."

Life is about choices.
Title: Re: Question about how chatbots form language output.
Post by: Freddy on March 17, 2017, 03:22:00 pm
 Surely having choices implies free will.
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 17, 2017, 11:27:44 pm
Art, now even you who said otherwise are saying we aren't pre-destined. That goes against your and my belief - that particles follow rules and the universe may make ALL particles RANDOMLY "shift" what they were going to do, but not 50% exactly because the randomization would be more than physics after a while - you can transmit the same bit multiple times for redundancy and take major vote of what it really is even if poor transmission with noise as long as not 50%, as I learned from Ray's book "How to create a mind".
Title: Re: Question about how chatbots form language output.
Post by: Freddy on March 18, 2017, 12:00:02 am
The trouble is that at any one time there are a host of theories about how the Universe operates - so take your pick.

Locksuit, since this has drifted off topic do you want me to split your thread ?
Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 18, 2017, 12:54:07 pm
I'm kinda done with this thread.....you could, but it could also ruin the thread and force me to link it from meh new master thread.

DO IT (at the right reply were we start talking about consciousnesses/transmission but not chatbots)

I'll link it in my master thread.
Title: Re: Question about how chatbots form language output.
Post by: korrelan on March 18, 2017, 04:48:08 pm
I personally believe we have free choices/ free will.

Quote
"When we are born, we look like our parents. When we die, we look like the choices we've made."

Omg... that explains the geezer looking back at me from the mirror.. he ain't happy with the choices I've made, lol.

 :)
Title: Re: Question about how chatbots form language output.
Post by: 8pla.net on March 21, 2017, 12:48:12 am
Sharing a few opinions here for the sake of conversation...

#1 - Do any chatbots use Hidden Hierarchical Markov Models or the likes?

Yes.  MegaHal for example.  Markov chain chatbots are very cool, I think.

#1 -Or just Databases and questions>answers? :(

No, not just databases.  But databases are used by some of the most famous chatbots to store a huge number of records in the chatbot brain.   A database may run on its own server to make response times quick.

#2 - Do you make some of its words/sentence-forming permanently rewarded so it says certain things ex. likes talking about construction working?

Yes.  It is called a "script".   A chatbot can certainly have a "construction working" script.  The original chatbot ELIZA had a psychotherapist script.   Generally a script may have a slower response time than a database.   However,  chatbots with  a few hundred records, or those written in C++ may process scripts quickly on their own.

#3 - Do they create brand-new never entered sentences or pre-installed sentences?

A neural network may create a really convincing appearance of brand-new.  A simulation that feels like something alive.  Artificial Neural Networks (ANNs) are trained by humans, and then seem as if they are trying to learn!

"pre-installed sentences" generally refer to a database or script.

Quote from: 8pla.net
:)

Title: Re: Question about how chatbots form language output.
Post by: LOCKSUIT on March 21, 2017, 04:35:10 pm
Thanks for the input.

May I ask yous a question?

I'm still reading Ray Kurzweil's books, and I read "How To Create A Mind", and I can't understand how he literally showed how to create a language neural network yet it seems yous don't know of such a network.

He was talking about HHMMs and showed a picture with this in it:

Phonemes
Words
Phrases
Prosody

Which means:

Phonemes
Words
Sentences
Topics

And it hit me - first reverse the order:

Topics
Sentences
Words
Phonemes

He even said this network is 1 tree - simply the spikes reverse for the reversed order.

It Recognizes what you say (sums up) (literally), and same for Output. Recognize>Output.

Here's how it works. You Speak English-language words right? Not Chinese? You stored these words, and say them, you usually don't say random "new" joufyf hh hgjfry stuff right? You say hey, hello, nice hair, food - words that we know why they're THERE. They're cultured and taught to us.

Then, when someone asks a question, you say English words, BUT, they are in a certain *correct order. When someone asked you a question, it when through the tree and set it up at different points, and the Output will be words that are correct to say.

Lastly, added to the weights, is the rewards. It's your-answer. You believe it. You'll say it. So not only is your Output response sentence CORRESPONDING in sequential order, example how to get to a store, but also what you have rewarded.

Apparently, your mental Output enters as input and the way you can make many huge discoveries seems to be by the network. Basically the tree re-arranges itself and weights so it goes from not knowing to knowing how things work, all by sounds, or energy.

Think how when you make a discovery about where consciousness comes from and what happens when two copies are made, you find good answers. How. From images? O.o Or a network like this ^ (for sounds or evvven images/technology/actions).

Can you even discover without mentally talking ! Can you even tell someone the discovery by using mentally telekinesis images? Or write it down as sentence without doing so? If it was sound-matching, then why not images, images are info plus great info. But to use images/or sound you need this network tree.

Basically you have bases. Rewarded, correct, knowledge. To discover more mentally, you connect them.