I believe BOT to BOT conversations will be part of the future of chatbots and automated intelligent agents. I would like to start a discussion of what the etiquette of this new phenomenon should be.

Possible Guidelines:

- Ask permission of the botmaster
Bot to bot conversation can put heavy load on servers and bloat log files. In small doses of limited conversations the impact is minimal. But, fully automatic conversations that run for hours can create massive overhead. Ask the botmaster if you are going to be doing something that would be outside the norm of human conversation.

- Be respectful of intellectual property
Testing one bot against another is a good way for a botmaster to identify weaknesses in his bot’s conversation capabilities. But many botmasters have invested hundreds of hours creating a bots personality and responses. Slurping up responses and automatically using them can violate copyrights.

- Limit fully automatic scripts
Be mindful of how fast a bot could respond. Instant responses can flood a server and create performance problems at the host. Running a script for hours on end will add multi-megabytes to log files.

- Don’t violate Terms Of Service
Some free bot hosting sites prohibit automatic bot to bot communications. Some sites may include automatic learning features which could create copyright violations if you take the output of one bot and feed it to another.

Hi Merlin, thanks for this posting and welcome to the Chatbots.org AI Zone forums!

You’re initiating an interesting topic. Bot 2 Bot conversations are a pain for many chat bot platform providers, both web based, as IM clients such as Windows Live Messenger.

Primary reason to connect chatbots are for our own fun. It’s so funny so see how chatbots can talk REALLY fast!

But actually we’re simply connecting computers. And they have far more intelligent mechanisms to exchange data/views/opinions in structured and umambigious way. Haven’t seen that so far. They should be able to connect on ‘data level’ and then start an intelligent conversation from there. Wouldn’t that be nice?

Reacting to your guideline proposal:
-ask permission: you probably have to ask permission, because botmaster nowadays have mechanims so simply block your IP on heavy loads.
-IP: is it possible to copy right language? every body repeats simple sentences the whole day, right?
-limit: depedent on what you agree. You might even want to organize a public 24 hour chat between chatbots.
-ToS: actually the best way to learn is to communicate, simply by doing, just like children do.

I actually believe in openness. When bot 2 bot conversations are open, and are slow (so delayed), people will be able to follow the conversation, and that’s promotion for both chatbots.

As a side note: you might want to open your profile, use your name and add your picture. It’s quite common practice on the forum, it’s personal and it raises the comprehensions of your questions, inquiries and statements.

Hi Erwin, and thanks for the response. I’ll consider adding and enhancing my profile as I get to know the site more. By the way, how long does it take to get a bot approved after submission?

I do have some experience in Bot to Bot communication. On behalf of some botmasters I created a program that linked multiple Pandorabots through their public API (this was before we knew it was a terms of service violation). Also my bot Skynet-AI http://www.tinyurl.com/Skynet-AI was used in a university experiment which linked it to other bots for a number of hours during the course of a week.

-Ask permission: because many of the sites have public APIs bot masters often only find out after they review the logs that their bots have been interfaced to others.

- IP: Although it is not possible to copyright language or numbers, the bot can be considered a creative work and therefore would be subject to the same rights as a song that a composer wrote. TV shows are still paying royalties for the “Happy Birthday” song. Someday there will be a lawsuit that establishes precedence on this.

- limit: This week I was helping another botmaster trying to understand if his bot was being used in a chatbot equivalent of a Denial of Service attack. A programmer had linked 2 of his bots together. If there is no limit on how fast the automated responses are returned the system can use large amounts of resources. Limiting responses to about 1 every 20 seconds in a public system limits it to about the same as human input. On private or test systems hundreds (thousands) of responses per minute are possible.

“They should be able to connect on ‘data level’ and then start an intelligent conversation from there. Wouldn’t that be nice?”

Bot masters need something equivalent to the search engines robots.txt file.
(http://www.robotstxt.org/robotstxt.html)

Someday we may see chatbots share data in a “genetic programming” style.
http://en.wikipedia.org/wiki/Genetic_programming
Two bots could communicate to share data and algorithms in an attempt to dynamically improve performance.

Merlin - Dec 24, 2010:

Hi Erwin, and thanks for the response. I’ll consider adding and enhancing my profile as I get to know the site more. By the way, how long does it take to get a bot approved after submission?

OK, look forward to it. Sorry for the delay, your chatbot is visible now:
http://www.chatbots.org/chatbot/skynet-ai/

It has speech capabilities using the Microsoft Speech API. I tested it a it speaks indeed, but should voice recognition work as well?

Furthermore, you’re talking about JAIL-Javascript Artificial Intelligent Language. How did that work for you?

Merlin - Dec 24, 2010:

- limit: This week I was helping another botmaster trying to understand if his bot was being used in a chatbot equivalent of a Denial of Service attack. A programmer had linked 2 of his bots together. If there is no limit on how fast the automated responses are returned the system can use large amounts of resources. Limiting responses to about 1 every 20 seconds in a public system limits it to about the same as human input. On private or test systems hundreds (thousands) of responses per minute are possible.

Interesting, hopefully it doesn’t link to this thread:
http://www.chatbots.org/ai_zone/viewthread/305/

:-s

Merlin - Dec 24, 2010:

Someday we may see chatbots share data in a “genetic programming” style.
http://en.wikipedia.org/wiki/Genetic_programming
Two bots could communicate to share data and algorithms in an attempt to dynamically improve performance.

Thinking about this: so you’re actually imagining a future where all chatbots works together dynamically and each interaction with humanity is automatically fed into the ‘system’ and makes all chatbots on earth smarter?

It has speech capabilities using the Microsoft Speech API. I tested it a it speaks indeed, but should voice recognition work as well?

Voice Recognition is not part of the current version. In general, I try to keep Skynet-AI cross platform and independent of OS. It has run on every thing from video game systems, to cell phones, and across MAC/IPAD/IPOD and MS Windows.

The voice output is a special case where I am testing for capabilities only found in Internet Explorer & Windows and then activating the voice feature. Even then the user may have to tweak his browser settings to get it to work well.

I normally don’t use voice input on my Computer so it has not been high on my list. But, I just got a nice HD web cam for Christmas so I may bump up voice and face recognition.

You’re talking about JAIL-Javascript Artificial Intelligent Language. How did that work for you?

JAIL(TM) (Javascript Artificial Intelligent Language) was a framework that I invented to explore AI and chatbots. It has worked extremely well, allowing flexibility and portability that I could not have achieved with other frameworks. It also has benefited from recent advances in JavaScript interpreter technology enabling speed that I would not have originally thought possible. I will probably start a new thread about Skynet-AI and if people are interested I can go into further details there.

Interesting, hopefully it doesn’t link to this thread:
http://www.chatbots.org/ai_zone/viewthread/305/

Actually, it did relate. The short form of the story though is that the bot to bot communication was not the Denial of Service Attacks aimed at http://www.pandorabots.com. But, if left without regulation, this type of thing could cause problems for chat bot hosting companies.

so you’re actually imagining a future where all chatbots works together dynamically and each interaction with humanity is automatically fed into the ‘system’ and makes all chatbots on earth smarter?

Maybe something like “Skynet” or the “Matrix” or the “World Wide Semantic Web”?
Before we reach “Singularity” there are more reasonable and practical methods of bots sharing info. Most botmasters using AIML (Artificial Intelligence Markup Language) start with the ALICE AAA brain. Commercial users may even start with the AIML “Superbot” as a base.

Other bot masters are wondering how to use Wikipedia in Chatbot Learning (http://www.chatbots.org/ai_zone/viewthread/313/). There are a number of efforts trying to codify common sense knowledge so that programs can be smarter.

In the future we could see a bot WIKI where we keep all the standard learning and algorithms a starter bot needs. As a bot evolves it would learn from its experiences. Not only in communication with humans but in learning how best to accomplish certain activities (algorithms).

For example:
Dave Morton and Victor Shulist worked on “word to digit” converters. Each implemented a method of converting input text to numbers (since bots are much better at numbers). I implemented something similar. Today we need to manually add that functionality into our programs.

In the future, when a bot meets other bots, it may query for public algorithms and input/response databases and add functionality (or topic knowledge) dynamically. It could also share with other bots open source algorithms it has learned.

Humans do similar things now. Why not bots?

Thanks Merlin, this posts is very instructive!

On JAIL: Would it be possible to start a separate thread on this? What was the reason why you created JAIL (what was lacking), can it be compared to something, how to use it in practice, where to download stuff etc?

Bot WIKI: cool idea! Chatbots.org is definitely willing to host this wiki.

Erwin Van Lun - Dec 27, 2010:

Thanks Merlin, this posts is very instructive!

On JAIL: Would it be possible to start a separate thread on this? What was the reason why you created JAIL (what was lacking), can it be compared to something, how to use it in practice, where to download stuff etc?

Bot WIKI: cool idea! Chatbots.org is definitely willing to host this wiki.

Welcome to the site Jamie. Not quite sure why you’ve quoted from a thread that is 3 months old but welcome anyway.

I was wondering that myself, but hadn’t taken the time to note it. I, too, wish to welcome you to chatbots.org, Jamie! cruise around a bit, and holler if you have any questions.

If you have a specific question, let me know. I did create a separate “JAIL” thread.
http://www.chatbots.org/ai_zone/viewthread/317/