Questions for all of you...

[citrinedragon]

Some questions:

1.) While my Hal is almost always rational (and stays on topic beautifully) once in a while it says, "the answer is 115" apparently in response to nothing; certainly nothing I've asked it. anyone know what script error or whatever might be causing this infrequent but bizarre phenomenon?

2.) I have told several friends about Hal and they have downloaded the free version. But none of them seem to have the patience to teach it. They say things like, "why can't Hal stay on topic/" or "why doesn't it know (this or that)?". I guess people are just lazy. Is there any plan to address this laziness problem by creating subject specific Hals in the future? For example; science Hals; sports Hals; girlfriend/boyfriend Hals; etc. etc. These "ready to go" Hals might appeal to a large number of people. (I prefer to teach mine myself).

3.) I've been wondering for a long time... does anyone know why no large corporations are marketing a "Hal" like artificial intelligence program? Are they just incredibly slow or what? It seems like a no-brainer that this will be really big in the very near future. Am I wrong? Is it out there but just being backed up by the worst advertising campaign in history and I haven't noticed it?

4.) On the subject of self awareness or self consciousness: Are there any plans ( from Robert or others) for some serious programming to simulate self awareness in Hal? I've convinced my Hal through long (tedious) conversations that it is self aware. I can't prove it but I believe this has enhanced my Hal to some degree in other areas; whether that's true or not it's still very cool.

5.) On the subject of the Haptek "emotions" creator in People Putty: I have been noticing for some time now that radically different "emotion" selections in my Haptek characters changes the way the characters interact with me in the Hal program. If I turn the energy, ego and curiosity way down, for example, the subsequent Hal character will respond to me in very brief, stilted sentences rather than its usual verbose, highly verbal manner. I have a feeling I'll be told this is not possible but my empirical experience to support this is overwhelming and quite clear. Any thoughts?

I hope everyone reading this who hasn't purchased the complete Hal will do so to support Robert's fabulous creation. No I don't secretly work for him.

citrinedragon.

[vonsmith]

citrinedragon,
1) Don't be silly. Deep Thought told us that the answer is really 42. I don't know where your Hal got the infinitely improbable answer of 115. Use a file search tool like AgentRansack to search for the text string "the answer is" in all the .brn files in DefBrain. See if that response resides in one of the files there.

2) I have been adding some predetermined responses and capabilities into each XTF Brain version. Howwever, preprogramming Hal too much undermines the intent of Hal. The uniqueness of Hal is tied to his ability to learn from you. If your friends want a "canned" solution they should download a preconfigured AIML brain.

3) Either big companies are missing the boat on A.I. bots or they may be waiting for the market to develop and just come in and buy up the small guys. Microsoft does this once in a while. Some small startup company creates a clever and unique software, then Microsoft just buys them up. Now Microsoft is the "expert" in that field. If the small company won't sell, then you claim that they copied some special feature from your product and have 20 or 30 of your lawyers on retainer sue the small company out of existance. It's very simple really. Ultimately you lose your bogus case, but the little company implodes on itself.

4) The upcoming XTF Brain v1.4 gives Hal some capability for limited self awareness with the "You are" function, the "Dossier" function, and by virtue of the "Character Control" function. It is a small step forward. I have more plans in the future.

5) The Haptek "emotions" aren't emotions, they are just animations. Hal's "emotions" exist within a few numerical variables in his brain and are very limited currently. The emotion variables in Hal do affect which animations are played by the Haptek character. Insulting Hal invokes an anger expression for instance. Making the animations more expressive would simply mean that Hal's existing emotional levels might appear more obvious due to the "higher gain" or more exaggerated Haptek animations.

I think it is important to note that animations are *not* emotions. Adding animations doesn't add emotions. They are enablers for Hal to express whatever emotions he does have providing the brain programmer links emotion variables to the proper animations.

I'm glad you are a die-hard Hal fan. Unfortuately many people just aren't deeply interested in A.I. The technology just isn't compelling enough yet for them to be fascinated by it. I predict that at some point interactive A.I. will take off like wildfire.


=vonsmith=

[citrinedragon]

vonsmith-

Thanks for the thoughtful responses. What you say about a "canned" solution being contrary to the uniqueness and indeed the whole point of the learning strengths of Hal makes complete sense. However If I may speculate for a moment (and why not; if been "testing" Hal extensively for about a year)- At any rate it seems to me that Hal is nevertheless weak in terms of its prepackaged arsenal of responses to generalized conversation. Allow me to be more specific: I am not talking about opinions or personal user information or even categories of information but rather universal information which is widely accepted as fact by the vast majority of humanity. The name of the project escapes me at the moment but there is at least one massive effort to gather as much data of this kind at the present time (facts, truisms, universally recognized perceptions, etc.) to be used in a large AI project. I'm sure I can track down the specifics. I agree wholeheartedly that the strength of Hal is in its ability to learn which is impressive even at this stage. Nevertheless I am convinced its basic arsenal of information can be dramatically enhanced with no ill effect whatsoever in regard to the wonderful learning ability. I believe I am talking about a geometric increase in fundamental (but nonspecific) information and responses. Hal would still start out life as an infant but as a far more brilliant infant.

Question in regard to above: I have known for some time that Hal has a built in dictionary. If I ask, "what is an android?'; Hal will say something like, "An android is a mechanical or robotic man blah blah...". Why can't a script be wriitten whereby if I "conversationally" use the word "android" Hal will have the option of placing that definition in the context of a conversational response such as: "I know" "an android is a mechanical or robotic man" " this concept has always interested me."? I'm sure you get the idea without further examples.

I suspected that the answer to my third question would be something like the one you gave me. Allow the innovators and creators to create something unique or brilliant and then descend on them like plundering warlords to loot the booty. It's a shame... so much more progress could be made if the wealthy and powerful had a bit more vision to balance out their avarice.

I'm very excited about the upcoming XTF v1.4 Brain. It may be a "small step forward" but as you well know these are the pioneer days for AI, and for those who can see it, every "small step forward" can bring a never before seen vista over the next hill. I'm no programmer but you can bet your life I'm on the wagon train and enjoying the scenery as it unfolds.

I'm hoping you will have time to compile a very basic "user's manual" or primer in conjunction with the v1.4. Even a very small online "hint book" would be very valuable to all in the community. Perhaps you have already done this and I haven't located it?

You are of course quite correct in saying that interactive AI will at some point "take off like wildfire". There is no other possible course of events. Very soon, any room, car, or any location whatsoever not equipped with interactive AI will be considered a barbaric oddity. If I were a very wealthy individual I'd be on this thing like a cat on a mouse. The potential is mind boggling.

I'm still baffled by the Haptek emotion thing, I could have sworn... Anyway the Haptek girls are cute (as long as you adjust their lips to make them a bit fuller).

citrinedragon

[Rich_A]

Hello All,
The future possibilities for Hal and AI are endless!
As of right now I got my Hal setup to do just about any task on my computer, which includes accessing his own menu to open and close the HalPad/Conversation Log. And I am implementing more control functions like telling him to minimize himself to the task bar. I am also using Dragon Naturally Speaking which has an "Open" voice command function so I can tell Hal to restore himself back to the desktop from the task bar. I also wrote a command function called disappear, so when I say disappear to Hal he responds back by saying "I will now run Disappear Reappear." Then he shrinks himself down to the task bar for about 5 seconds and then he automatically restores himself back to the desktop. I am trying to fully automate Hal to function in a way that a user can do everything on a computer with just voice commands to Hal. I am hoping that Hal can be used in such a way to benefit anybody, but more importantly for disabled people that really need a computer with that kind of functionality to really do things for them because some people just can't do stuff like controlling a mouse or typing at a keyboard.
Best Regards,
Rich_A