Artificial Intelligence chatterbot and NLP (Natural Language Processing)

302 Replies, 90692 Views

This would be great as well. What I meant is perfectly described in these posts in the AI forum:

"... What would I like this AI to learn. Well.. I was thinking for it to read information like maybe ebooks, pdf, word or even information found on the internet. I would like it to take this information in, so lets say if it read a ebook on genetics and I asked it a question revolved around that, it could give me an answer. ... "

And this is the response:

"What you describe seems to almost be a common thread of hopeful discussion among AI enthusiasts although it has remained, for the most part, somewhat elusive.
There are a few bots that perform various self-learning tasks like UltraHal, Prelude, Kari and a few others but these bots usually perform these as received instruction or taken from the context of conversational inputs. The inputs are often directed to a table where info is broken down, stored and remains ready for future retrieval as a matched pattern or useful within the context of a sentence that is "deems" pertinent. There are some conversational entities from eJtalk (Cassandra - Youtube) and Cognitive Code's Silvia - youtube) that perform on a very (seemingly) high level.

While a lot of the ALICE, Pandora and other AIML based bot masters claim their bot "learns" it usually means that yes, their bot CAN learn IF the botmaster inputs the information into the bot's database for future usage. There really is no self directed learning and I dare say that most of the "other" bots are able to do as much (on their own).

These are still chatbots...software. Most have no idea at all as to what the color blue actually is or love or hammer, other than some definition that has been placed in their database to be realized and recalled if needed. The other thing they often do know, is the usage of certain words and phrases although, a genuine "understanding" is still not happening. They are not living entities, but some are way better than others at providing the illusion of intelligence and that seems to be the "rub" in AI circles.

We award Prizes and medals at our own attempts to "fool ourselves" and we praise ourselves in doing so.

Without meaning to sound sour or callous I have had an ongoing enthusiasm and chat-a-holic behavior for several decades with hundreds of bots. I have always been searching for, patiently chatting, hoping to find or discover that one bot that seems more "in tune" or that "gets it", where others fall short. I have been pleasantly surprised over these years many times and often caught completely off guard as my jaw dropped, all the while thinking, now THAT was cool!! How'd it do THAT? 😉

Hopefully, with more and more bots being developed and a greater number of resources being available, it's likely only a matter of time before we ourselves are the ones that are ultimately fooled by some very astute bot(s)."
I will probably install "Program O" just to get the idea how AIML files work. As the next step I will try to implement FAQ, site navigation assistance and some advices on self-bondage scenarios. Then we'll see...

In principle, even if the AI bot is AIML files based, it should be possible to create a real AI program that creates these AIML files based on "its experience" (converesations, Wiki, various texts, etc).
Ok, if you start a real AI program, then that program will need to be able to reprogram itself based on what it reads from e-books.
But a true AI program, that will think for itself, that can be very deadly. And here's why.
When an AI programs it's self, and continues to reprogram its self all the time like we do, then that program will acheve self awareness. And can cause AI fear, for both man and machine.
And, how can you stop the AI program from scanning the net for other information?
This could get a little out of hand.
An A.I. program would need a frame of reference... Esp for this forum.
Explaining an orgasm is one thing knowing what it feels like is a totally different. This whole fetish thing revolves around the querks of human reproductive drive..... An A.I. would prolly develop a fetish around the "Save As" button and what colour it was ? ( hmm adds new meaning to punching someones buttons)
(This post was last modified: 15 Jan 2015, 10:00 by Marcus.)
(15 Jan 2015, 09:51 )Marcus Wrote: An A.I. program would need a frame of reference...
Yes, regardless the implementation. Basically, AIML files contain patterns the program expect from users, and typical answers. Till this time I could not find any program that does learn. For example, if I say:

"My name is John" or "I am John"
"What is my name?"

I expect the program to answer:

"Your name is John" or "You are John"

But that does not happen! Even the highly prized CleverBot can't do that!
(15 Jan 2015, 06:04 )Tinker D Wrote: But a true AI program, that will think for itself, that can be very deadly. And here's why.
When an AI programs it's self, and continues to reprogram its self all the time like we do, then that program will achieve self awareness. And can cause AI fear, for both man and machine.
This is what Stephen Hawking said. The Skynet in the "Terminator" series, The Matrix, etc. However, no such programs are available. At least publicly.
Again lacking a frame of reference.... you assign human values to a world that for us would be totally abstract. Why would AI want to harm us ... out of fear ?
Emotional awareness is not a necessary construct for something to be intelligent esp by your Turing test criteria.
Just cos a computer can quack like a duck doesn't make it a duck... unless it really thinks its a duck.... so the ability to process data does not make some piece of software intelligent or give it true understanding.
Spent several hours trying to make Program O work. The internal engine does respond according to the AIML database, but the GUI refuses to show anything. Many users report the similar problems, I hope they will be fixed, but I'm not holing my breath, since the personality and "subject" related patterns/answers must be created from scratch. Currently I'm more interested in the AI based on the neural networks and an expert system that will ask questions and provide a solution based on the answers. Think of generating self-bondage or bondage scenarios. The latter can be written "in house", of course, but I'm looking for a ready to be used framework.

As always, your ideas, pointers and URLs are very welcome.
(This post was last modified: 16 Jan 2015, 02:40 by Like Ra.)
I don't think there is a ready to run framework out there. And a possible reason could be, what I call legial liability.
If I wrote a self bondage or a bondage AI, and I gave or sold the AI routines to another self bondager and died using said routines, wouldn't I as the person who made and sold/gave the routines be responable for his death?

It looks like you will have to really tweak the program or write your own.
No, no, no... Under the framework I understand the PHP engine with templates, that allow creating various, for example, diagnostic or knowledge base scenarios. The actual information in the database is empty and needs to be filled in by the administrator.