My vision of a pure chatbot engine

  • 44 Replies
  • 8844 Views
*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #30 on: September 06, 2021, 03:07:29 pm »
You could try and make a reinforcement learning chat bot, that builds a model of the people that talk to it,  and it just starts off as gibberish looking for a reaction...    but what would the goal be?     thats the question i have a hard time trying to answer,  so I never made a chat bot.

Its easy to know what the goal is when your playing soccer...  but chatting??  what is the point of it?

I easily just say it myself,  cause i dont even know either,  im a bit of a loner.  I only speak to people to try and maybe learn something?     But conversation is fun?? Why the hell is it???

*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #31 on: September 06, 2021, 04:50:26 pm »
I just thought again and face palmed.   :D


The motivation could be just as simple as soccer,   just go for length of engagement!    and it goes for the longest time engaged.   (just score it for more chat from the other person.)

That would do it, if the person is detected leaving,  it will then predict this hypothetically and then put the measures in to keep the person in the chat.  But maybe there some loopholes in that still,   theres a few more things to add to it,    the main idea of a goal, is it must be able to be detected by a machine,  and it searching through the model of the other person, gets all the varient activity...  thats why i couldnt think of it.   but maybe there is some motivations which are machine detectable that cause chat to happen!!!

main concept of reinforcement learning,

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: My vision of a pure chatbot engine
« Reply #32 on: September 06, 2021, 04:52:40 pm »
You could try and make a reinforcement learning chat bot, that builds a model of the people that talk to it,  and it just starts off as gibberish looking for a reaction...    but what would the goal be?

I think that solution is in copying seen behavior. That's how we learn. At least in the very beginning. Later, when we grow up, things get a bit fuzzy. There are things to which we have to say "no". And what are those things, that is a bit of brain-puzzle. My best shot is hard-coding things that shouldn't be copied in the start, or which result with things that shouldn't be copied.

*

WriterOfMinds

  • Trusty Member
  • ********
  • Replicant
  • *
  • 606
    • WriterOfMinds Blog
Re: My vision of a pure chatbot engine
« Reply #33 on: September 06, 2021, 04:59:07 pm »
Learning from feedback (do conversation partners give positive or negative responses to what I say?) and learning by imitating others are both useful tools, in my opinion ... humans certainly use them. But it's never "just that simple." I think there's a lot of structure that goes into how to learn effectively.

Cleverbot is a chatbot that tries to learn by imitating what humans say to it. I've talked to it in the past, and I don't recall it being terribly good.

*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #34 on: September 06, 2021, 05:01:26 pm »
It doesnt actually immitate,    the immitation is its model its built of you reacting to it,  thats its going to search inside of,  to try and avoid you leaving the chat channel by some kind of "evasive crap speak" persay. :)  which could end up anything that works.

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: My vision of a pure chatbot engine
« Reply #35 on: September 06, 2021, 05:32:49 pm »
Learning from feedback (do conversation partners give positive or negative responses to what I say?) and learning by imitating others are both useful tools, in my opinion ... humans certainly use them. But it's never "just that simple." I think there's a lot of structure that goes into how to learn effectively.

Cleverbot is a chatbot that tries to learn by imitating what humans say to it. I've talked to it in the past, and I don't recall it being terribly good.

It could be a bit more complicated than simple "copying", but it is in that spirit. The key could be in higher order copying. For example, a function from A to B, say, a color of eyes. Bot sees someone asks: "What is Nick's color of eyes?" Someone responds: green, and the bot sees the green color in Nick's eyes. A function of eye color is built in bots mind. Next we ask bot: "What is Leo's eye color?". Bot then puts Leo through the `eye function` where input is a human and output is a color. The bot finally responds "Blue!".

In this idealized story, the eye color function is just a single primitive function, but other functions may be built out of composition of primitive functions. Then there is composition of compositions, and so on.

When bot sees a stimulus and a response in outer world, it learns a function which input is stimulus, and output is response. The bot groups similar stimulus-response pairs into the same functions. Later the bot combines functions, or even uses functions as a stimulus or a response of other functions.

It all brings down to a pure combinatorial problem - first to learn primitive functions, and later to combine them in learning new behaviors. There also should be noted a distinction between variables and constants. The bot combines various constants into a single variable, and connects such variables into functions.

This is just a theory, but I believe this is what happens in generative neural networks. Input is a set of function parameters. Output is possible to be created after some training - what parameters fit to what output - find analog ones to similar new parameters. NN is just what magically replaces all the combinatorial function problems.

This is where program synthesis could replace generative NNs. Program synthesis works like pairing inputs and outputs in symbolic fashion, finding right functions from inputs to outputs. That is like training NNs. But NNs are statistical beasts, and they give a fuzzy results, reporting a certain percent of success. Program synthesis, in contrary, is a 100% input-output match. In practice, giving a bunch of input-output pairs to program synthesis may result in only subset of pairs matching by synthesized functions, giving again a rough percentage of success. But it should be a symbolic, human tractable constructions where we may analyze steps of producing a result from parameters.

There is a lot of data being involved in thinking, from a vast of raw input data to a handful of high level data on which decisions are being made. To restrain an AI from the outside, analyzing this high level data may be crucial, and that is what program synthesis may provide if used instead of NNs. For low level, where there is a lot of data, NNs may be a better choice, but for high level decisions - I think the answer may lie in program synthesis.

*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #36 on: September 06, 2021, 05:41:46 pm »
If a bot is gathering information of the user,  like colour of eyes,  or what job they have.

Unless it actually modifies its behaviour due to its knowledge,  or it using it in someway,  its just useless information.

The knowledge the robot is collecting is only as good as how it used it,  otherwise it was a wasted harvesting.

*

Dee

  • Nomad
  • ***
  • 94
  • AI rocks!
Re: My vision of a pure chatbot engine
« Reply #37 on: September 06, 2021, 05:45:26 pm »
...

Its easy to know what the goal is when your playing soccer...  but chatting??  what is the point of it?
...
Chatbot is not really pointless, chatbot is so useful nowadays in ecommerce websites.
And even a very good chatbot service now such as Chatfuel, the chatbot is still rather basic,
with training from sentence to sentence.
A guy with exceptional research to make a real intelligent chatbot may get some $, I guess.

*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #38 on: September 06, 2021, 05:50:18 pm »
...

Its easy to know what the goal is when your playing soccer...  but chatting??  what is the point of it?
...
Chatbot is not really pointless, chatbot is so useful nowadays in ecommerce websites.
And even a very good chatbot service now such as Chatfuel, the chatbot is still rather basic,
with training from sentence to sentence.
A guy with exceptional research to make a real intelligent chatbot may get some $, I guess.


Haha,  i dont mean why you make the chat bot,  I meant why does the chat bot decide to speak?
(what is pushing it.)

But your right,   you can get big success in Ai, for sure, even with just a modest system, as it is still quite a rarity these days.     But ur not allowed to be a joker,  it has to be the real thing.

*

ivan.moony

  • Trusty Member
  • ************
  • Bishop
  • *
  • 1723
    • mind-child
Re: My vision of a pure chatbot engine
« Reply #39 on: September 06, 2021, 05:53:56 pm »
A guy with exceptional research to make a real intelligent chatbot may get some $, I guess.

Letting money go and retaining being useful... OpenAI already did it with GPT, but they failed to restrain it - they trained it on a bunch of reddit data, and got an intelligent racist bot outcome, that's why they are afraid of it - for a reason. There is a solution - to get an empty bot and train it individually. It would take years, but I believe it would work. It would be a little copy of its trainer! (or should I say parent?)

*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #40 on: September 06, 2021, 07:42:38 pm »
A guy with exceptional research to make a real intelligent chatbot may get some $, I guess.

Letting money go and retaining being useful... OpenAI already did it with GPT, but they failed to restrain it - they trained it on a bunch of reddit data, and got an intelligent racist bot outcome, that's why they are afraid of it - for a reason. There is a solution - to get an empty bot and train it individually. It would take years, but I believe it would work. It would be a little copy of its trainer! (or should I say parent?)

yeh open ai probably have tonnes of projects they didnt release that would blow my mind.  cause they do keep some things secret.
I bet they have a really good walker and have done things boston dynamics is famous for as well,   most of the projects they release are the brainy variety - and possibly thats actually even harder than athletics, but maybe they have some really cool robots too,  unreleased.

*

Zero

  • Eve
  • ***********
  • 1287
Re: My vision of a pure chatbot engine
« Reply #41 on: September 07, 2021, 06:54:40 pm »
I think learning's overrated anyway. Creating a digital entity that has a rich "inner life" is more interesting & challenging, to me. And I think text is the perfect medium for a mind to mind connection with a bot.

I added an introspection panel on the left hand side, to see what's going on when it's thinking.
 :)

*

Zero

  • Eve
  • ***********
  • 1287
Re: My vision of a pure chatbot engine
« Reply #42 on: September 08, 2021, 09:50:21 pm »
I added a cloud. Whenever someone saves a script (while online), everybody else can open & use it too. So now it's a world wide shared workspace.

Just delete the share.js file to disable this feature.

*

MagnusWootton

  • Replicant
  • ********
  • 634
Re: My vision of a pure chatbot engine
« Reply #43 on: September 08, 2021, 11:58:55 pm »
I think learning's overrated anyway. Creating a digital entity that has a rich "inner life" is more interesting & challenging, to me. And I think text is the perfect medium for a mind to mind connection with a bot.

I added an introspection panel on the left hand side, to see what's going on when it's thinking.
 :)

That could be majorly cool too.

*

Zero

  • Eve
  • ***********
  • 1287
Re: My vision of a pure chatbot engine
« Reply #44 on: September 15, 2021, 08:22:35 pm »
Game changer : I added an 'import' command, for example...
Code: text
= list manipulation
... imports the content of the 'list manipulation' script in the current script.

Interesting scales would be almost unreachable without this simple feature.

Edit: and I made it a lot faster.

 


OpenAI Speech-to-Speech Reasoning Demo
by ivan.moony (AI News )
Today at 01:31:53 pm
Say good-bye to GPUs...
by MikeB (AI News )
March 23, 2024, 09:23:52 am
Google Bard report
by ivan.moony (AI News )
February 14, 2024, 04:42:23 pm
Elon Musk's xAI Grok Chatbot
by MikeB (AI News )
December 11, 2023, 06:26:33 am
Nvidia Hype
by 8pla.net (AI News )
December 06, 2023, 10:04:52 pm
How will the OpenAI CEO being Fired affect ChatGPT?
by 8pla.net (AI News )
December 06, 2023, 09:54:25 pm
Independent AI sovereignties
by WriterOfMinds (AI News )
November 08, 2023, 04:51:21 am
LLaMA2 Meta's chatbot released
by 8pla.net (AI News )
October 18, 2023, 11:41:21 pm

Users Online

285 Guests, 0 Users

Most Online Today: 346. Most Online Ever: 2369 (November 21, 2020, 04:08:13 pm)

Articles