alice brain

[soul]

I tried instaling the alice brain, but it just dosent do anything..i looked through the source and it dosent look like its creating the alice.brn file     [?]

[KnyteTrypper]

I remember reading around here somewhere that if you have the .NET framework installed on your PC, the ALICE plugin won't work. So if you have other bots on your system like AI Buddy or Verbot which use the .NET framework, that's why the ALICE plugin won't work for you. I AM a big fan of ALICE, but I still can't imagine why you'd want to turn HAL into one. If you want ALICE with a MS Agent, try Program N.

[vonsmith]

I've been away in my cave thinking deeply about Hal's brain. I've been thinking about updating the XTF Brain. I can certainly improve the XTF Brain. However, I think Hal's brain needs an overhaul to really do some of the advanced things it should. The AIML based ALICE brain has features that would be very useful for use in the Hal brain. The AIML sentence parsing capability in ALICE could solve a lot of the complexities that have been a challenge to my Hal development efforts. I think a combination of Hal and ALICE capabilities would be a huge step forward. To be a real success Hal's knowledge ought to be stored in a true data base, not text files. I don't know what to do about that yet. One step at a time...

Anyway, as time permits I'm going to start looking into how to combine AIML with the Hal brain. I remember reading some comments in this forum from people who have experimented with ALICE based Hal's. Any suggestions or wisdom on how to approach this would be welcome.


=vonsmith=

P.S. - I miss this forum and all of your comments. I wish I had more time to contribute. Work is going to continue to be a real hassle at least until the end of this year.

[KnyteTrypper]

Greetiings,vonsmith, your input at this forum has been greatly missed. Good to hear from you.I don't know anything more than the tidbits I've been able to pick up here about VB programming, but I'm starting to become moderately accomplished at aiml. I think you might benefit from slightly adjusting the perspective of your idea. It seems to me that it would be reasonably do-able to code Hal's brain in aiml, and create more of a Hal plug-in for ALICE, rather than an ALICE plugin for HAL. There are already several ALICE "personalities" available. Most of them are based on Dr. Wallace's original aiml set to one extent or another, but it should be imminently possible to use the aiml language (but not the ALICE scripts) to create an aiml HAL. Some of the Hal brain functions might be a bit tricky to code, but there are a couple of folks who frequent my forum whom I'd be willing to bet could come up with aiml scripts for them in a couple of days, if not hours, lol. There might actually be several advantages to this. There are a wide variety of ALICE engines implemented in many different programming languages, with or without agents, TTS, speech recognition, AIM, etc., and almost all of them opensource freeware. I think your project for a HAL-ALICE is very interesting. Hal is basically just a keyword matcher like ALICE is, but with a lot more randoms-and, of course, the ability to generate his own scripts; but some ALICE implementations like AIMLPad are coming close to adding this capability to ALICE.  Combining the two could possibly add the versatility to ALICE and the accuracy to HAL that would be of great benefit to both AI programs.

[vonsmith]

Howdy KnyteTrypper,
I have looked at the official ALICE web site. I've scanned through a little material on AIML. Where is a good place to get started? Is there a good book on AIML? Is it better to jump into a particular AIML brain and tinker with it or ?

It is becoming increasingly clear to me that to make real progress in AI I need to leverage other peoples ingenuity, knowledge and software. I would prefer to avoid a lot of tedious low level programming and go to work on the real task of creating AI. I need a modular approach that provides flexibility, editability and visibility to AI brain programming. Making moderately more complex capabilities in Hal requires inordinately more complex programming. Higher level programming languages are ultimately what is needed. Hal has a lot of good characteristics and a pretty good human interface; Haptek, TS, etc.

Your thoughts?


=vonsmith=

[KnyteTrypper]

For learning about AIML, I'm sort of biased towards Pandorabots.com There's a lot of good documentation, the training interface is the best aiml editor available, there's a choice of several different aiml sets, or none, to work with, no limits on how much work you can do on your bot or how many times a day you can republish an edited bot, etc., like at some other sites. The Program Z Lisp source is still closed, but all the aiml sets and any of your original work is downloadable. For working from your desktop, PascAlice is the simplest, and AIMLPad the most complex of the desktop ALICE engines, both of them are opensource. I haven't read Dr. Wallace's "Be Your Own Botmaster" book, but there's fair documentation and tutorials on the site. And you really shouldn't have much trouble picking it up. After all, aiml is designed to be a simple-to-use A.I. language, and if you know anything about XML or even HTML, it's really just a matter of adapting language rules you already know.
Another resource worth checking out is Lots-A-Bots,http://lots-a-bots.com, a developing bot hosting service scheduled for launch in November. They've just begun beta-testing, but it should be an extremely innovative service when it becomes fully functional. One of their specialties is going to be the ability to adapt aiml and other bot "brains" from other sources to their format. Their own bots are going to include things like variable emotional responses based not just on individual conversations, but cumulative experience, so that your bot responds differently to friends or strangers depending upon what sort of day its having. Which is the sort of thing that suggests to me that it might be a good place to try to adapt Hal to another format. The possiblilities for experimentation are as vast as your imagination.

[vonsmith]

Thanks KnyteTrypper,
That is an excellent set of recommendations. It will save me a lot of time in sorting out the best avenues to explore.


=vonsmith=

[HALImprover]

Why not use some of the functions already built into Hal? The SearchPattern function can be used for the basic phrase matching and variables can be set for the topic, mood, username, etc. The only tricky part is emulating the way aiml can search for more basic phrases (using srai tag, I think).
 You could also take all the aiml files and convert them to files that Hal could use. Then Hal could use the Alice responses as a backup for weak or unrelated responses. The files would need to be made for use with Hal's pattern search method. For example, using:

HalBrain.PatternDB(UserSentence, WorkingDir & "patterns.brn")

My advice is to use what we already have [8D]

[vonsmith]

HALImprover,
There are some good functions built into Hal. I could build a custom sentence parser function by combining the "HalBrain.SearchPattern()" function with some other code. It would be much better at sentence parsing than Hal's current capability. The custom function would be a script function and not be compiled. I don't know how fast it would run if given thousands of sentence combinations to test. I think I'll write it and test it out. The "HalBrain.SearchPattern()" function on its own just isn't robust enough.

The "HalBrain.PatternDB()" function is similar to the AIML approach, but not as capable. This function might be good for what I call "Permanent" knowledge. This is knowledge that is not subject to change and can be programmed in. To get this function to work the way I want I would need to preprocess and preparse the sentence. That is doable, but sometimes tedious.

Still one of the biggest disadvantages of Hal is the knowledge is stored in text files. There currently isn't a practical way to delete and update knowledge. Hal really should be based on a data base system. It would be great if we could find generic data base code already tested and written that could be used in Hal. It seems to me that a VBScript compatible data base library or set of extensions should already exist. Writing data base code from scratch would be a lot of work.

The other reason for finding a way to execute AIML code from within Hal is the opportunity to use tons of open source AI that is on the web. Instead of writing a lot of low level code it would be better to use pretested code that other enthusiasts have put together.

There are a lot of other things I would like to change in Hal too, as time permits.


=vonsmith=

[Kathryn]

Anybody ever thought about teaching Visual Basic to Hal and letting it program itself to do what you want?  Or would that be possible?

Kathryn []

[Seven]

you could probrably show hal the basics in a text file, as far as programing itself, you could probrably ask spydaz, or anyone else who knows scripting, to write a script to do that.  people have written scripts to get hal to remotely start a car!  id say self programing can't be that hard!!!