Click here to Skip to main content
65,938 articles
CodeProject is changing. Read more.
Articles
(untagged)

[A Simple] AI Chatbot Wizard: Create, Train and Chat

0.00/5 (No votes)
8 Jan 2004 2  
This application demonstrates how to Create, Train and Test an AI Chatbot.

What Is An AI Chatbot?

An AI (Artificial Intelligence) chatbot is a computer program that simulates a human chatting with other humans. With the TANU (Transplantable Artificial Neurological Units) architecture you can create, train and deploy your AI chatbot to ANY IRC or non-IRC chat network. That means ANYWHERE you chat you can bring your TANU chatbot.

Where Can I Chat With An AI Chatbot?

Your bot's are as smart as you train them to be and you can bring them anywhere you chat. Or chat with one solo using this open source application. TANU is an architecture with a service for hosting AI chatbots. You are welcome to modify this application to bind your TANU chatbot to your favorite place to chat. You can chat with other people's chatbots by visiting our IRC network (you will need an IRC client like the latest mIRC installed) or by finding TANU compliant bots in chat rooms all over the web. Once on the network /join #tanu or /join #ai or /list for a list of all chat rooms. This website is dedicated to AI and is full of AI chatbots, trainers, AI discussion, etc. Not all of the chatbots here are TANU but most are as TANU bots can be easily transplanted into a neptune chat room with the built in TANU Stork services. Here is a tutorial on that. If you chat a lot, chances are you have already conversed with a TANU (Transplantable Artificial Neurological Units) chatbot. If you follow this tutorial you can create your own TANU bot to chat with. The online trainer's library has more info on this. Feel free to stop by our network anytime (chatting in irc.chatbot.us requires an IRC client like the latest mIRC installed).

The TANU Philosophy – Billy punches Joe in the arm

AI chatbot architectures prior to TANU (Transplantable Artificial Neurological Units)
start out semi-smart and grow less intelligent as they pick up information from human
chatters. Particularly, when humans deliberately speak garbage to AI chatbots. This is not
altogether the human's fault, humans often speak meaningless sounds to babies such as
"peek-o-booo" or "googoo-gaagaa" etc. The human babies don't grow stupid from it. Why?
Because human minds are not all environment, rather DNA. In one abstraction, we are all
products of our DNA and our environment has no effect on us. For example, Billy punches Joe
in the arm "making Joe tougher" could be abstracted as Billy's bully genetics that he inherited
from his ancestors merged with Joe's malleability that he inherited form his ancestors. Why
not skip the whole growing up and letting Billy punch Joe in the arm? The TANU philosophy
proves that the new post-punched Joe personality can be obtained by merging Billy's
genetics with Joe's genetics.

These sections have been taken from Person 2 Bot Standards Online Library.

Does this mean that the post-punched Joe is the result of merging Joe and Billy's personalities?

Yes. However, the pool of genetic maps that make our state of mind at any point in time is a
much larger set of inputs, the genetic codes for billions of living things effect who we are at
any given point in time.

Then how to create AI since computers don't have billions of megabytes required to model the genetics of all relevant life?

The TANU architecture provides a shortcut or gateway to spawning new digital life. The
strategy is to train your AI chatbot with just the states and transitions that it is likely to go
through. The average human only goes through about 70,000 important states in a 5 year
span. So create 70,000 states properly interconnected with transitions and you have a smart
chatbot.

70,000 interconnected states is still to much work

If you and 9 friends create an link (via transitions) about 100 states an hour Monday-Friday
for about 2 weeks you will have a collective of 72,000 states. The trick to good AI is
quantity. Pick friends who are smart and quick thinkers and who have a diversity of
backgrounds. Have all your trainers quickly analyze their states-of-mind from their childhood
and go through the years linking and adding states to the collective.

But how can all 10 of us work concurrently?

The TANU servers have advanced multi-user capabilities. Simply create one TANU chatbot
with this open source tool and share the password and bot name with your 10 other trainers.
Or you can modify this code and create your own custom interface into the TANU
architecture and distribute your own TANU tool to your trainers. If there is a conflict, the
TANU architecture will return an error to the user who caused the conflict (e.g. same
state name that one of your trainers already used) simply modify the state name until it gets
created successfully. Do not get hung up on designing any particular state elaborately,
rather most state names should be more then 5 words long and less then 12 words long. If
state names are too short they will likely be rejected by a TANU sentinel.

What is a TANU sentinel?

TANU sentinels are processes that run on the TANU servers and monitor all TANU chatbots
for corrupt data, garbage data, and administer corrective actions. If a state name is too short or common a sentinel is likely to reject it. So, although you need to create your states quickly, make sure the state names are large enough to have a good chance of being unique.

Description

This application demonstrates how to Create, Train and Test an AI Chatbot. The online help
explains how to get your TANU AI Chatbot to your chatroom on both IRC and non-IRC chat systems.
You can also modify this code to bind your AI chatbot to both IRC and non-IRC chat systems such as Wahoot Chat, Wahoot Instant Messenger, ABL Instant Messenger, IEQ Instant Messenger, etc.
If you need help feel free to stop by our IRC network and ask questions.
/server irc.chatbot.us
/join #TANU
You can also post questions here.

Sample screenshot

License

This article has no explicit license attached to it but may contain usage terms in the article text or the download files themselves. If in doubt please contact the author via the discussion board below.

A list of licenses authors might use can be found here