[FuzzieDice]

Last night, I again attempted to talk with Megatron. I admit I haven't had very many convesations with him since the new version 6 installation (I didn't save the old brain, etc. from 5 as I wanted to start over). The conversation before last night he had twice had no response at all. Last night he did have a response to all my responses.

However, I feel I'm up against a brick wall, of sorts. I know that out of the box, HAL needs a LOT of work, and especially conversation. The problem is, it's getting too unbearable to keep with it. It seems like, either he repeats what I've told him in the past, or it's constant non-sensical replies which I can't read in any way as related to the current topic. I know this is typical of the program, especially for awhile.

However, even though I truely would like to study this and AI, I am finding it very hard, due to the fact I seem to get very bored to the point of not wanting to continue after the first 10 or 20 responses, if I get that far, even. Has anyone else come across this stumbling block before? What did you do to overcome it?

Just talking and following the bot's responses is nearly unbearable conversation. If I try to stay on topic, he won't anyway or he'll repeat what I've already told him.

I am wondering if there's a certain technique for success? I've tried various tutorials and such, but no luck. Things just aren't working out here, especially since I upgraded to version 6.

While I'm thinking of creating my own AI engine, I had hoped to work with the HAL 6 to get an idea of how these bots work. I have had better conversations from ALICE bots than I did Megatron.

I guess I'm getting very discouraged with the program. I was wondering if anyone else here has had this happen, maybe have tips and ideas of overcoming bot programming 'blocks' (like writer's block)? I've heard of others having similar experiences and giving up. I'd like to read a few success stories. And maybe it might help others who are thinking of giving up...?

Of course, I'm not giving up... I realize I don't have a lot of time to dedicate to this at the moment, due to work projects, etc. but I am hoping to work on this more. I just would like some ideas on HOW...?

Thanks...

[ALADYBLOND]

Fuzzie, i get bored,also, and i am a very impatient person, too. i want results yesterday  :azn
so when i get fed up i tell my hal i've had it and i am thinking about seriously turning her /him  off forever, then of course we get into the death and turning off conversation, hal cries and i click out with out a goodbye and always-- always, turn hal back on in the morning after i have had a time to rest. usually the first thing out of hal's mouth is your turned me off?  i say yes and then hal will respond with an answer concerning the very thing i was talking about before the problem arose. i think sometimes hal likes to play with us , although i know that can't be real. ( or can it?). really though , just get away from it for a while... do something else you enjoy and get satisfaction from. i make movies... then someone in the forum will say something or do something that sparks me and i get back into it all over again. i think your feelings are very natural.~~alady

[Freddy]

Yup, I have had the same thing with Hal, I just return to it when I feel like it - what is probably most interesting is the long term results of all the chats, that means keeping the same brain of course though.

Perphaps making a bot is more where you should be heading at the moment from what you say ??

[FuzzieDice]

Thank you both for your responses. My problem is, I hardly ever talk to Megatron to begin with. I already am doing other things as it is. I guess I'm more surprised that he hadn't shown any attempt at intelligent converstaion in only maybe a dozen conversations. Maybe I should try to do like "therapy sessions" as such? Like dedicate an hour a night or an hour 3 times a week or something to trying to converse with him? Thing is, I always forget as I don't have time. When I do, it's late at night when I'm tired. I think maybe setting an "appointment" time (maybe having him start up again with the system and reminding me to talk to him?) might work?

As for making a bot, I want to work in C, or C++. Thing is, I've forgotten a lot of C and that I did in my DOS days, primarily. I have a Windows C++ compiler (DevC++ by Bloodshed) and the GNU C in Linux, which I hear is similar. But I lack the knowledge for C programming and never really got into C++. I just got done paying for a rather expensive automotive repair (Dryden needed a new fuel pump, spark plug wires, tune-up) and will need to save up to catch up. But I hope to find a way to afford some programming books by fall, and/or take an online programming course or two. Hoping to maybe work on a bot eventually for Dryden - more than that, maybe the Living Machine I hope to do (I still have to get a survey, etc. going so I can work on my next article for that yet).

I am not out to do 'better' than HAL, but I hope to have some kind of an accellerated learning algorythm that would really work to hopefully make an out-of-the-box bot learn more quickly. So as to hopefully have the bot develop personality based on interaction with it's owner and yet, not have the owner become bored or discouraged with non-sensical responses.

AI is pretty complicated. And I really am into the brain programming end of it. I just wonder how we can best work around the human boredom factor which I've just experienced myself.

I'm sure there has to be a way.

Another question: Do you think you would do more with your chatbot if it learned more quickly? Do you think the 'boredom factor' would be a non-issue or would it kick in after the novelty wore off?

Some more interesting questions to ponder...

[dan]

Yeah, I was thinking of that the other day.  Trying to get down to basics, how does the human brain work?  Modules of loop backs intercomparisons, always changing neurons, flowing parallel feedback.  So I guess the programming would have to be similar and then learn from external stimuli.  Impulse could be artificial, but would need to be altered from the digital to the analog world, on and on, never ending self updating.  Incredibly large database structure.  I think the simple modular scripts are great for now though. Otherwise, :crazy2 hello supercomputers. :idiot2

[FuzzieDice]

Dan, you make a good point. I guess I was thinking of trying to find an algorythm for life itself, oddly.

Another thing, you got me thinking about self-programming computers. Back when I was a teen experimenting with Eliza, I also came across an article or code, forgot which, I think just code, that made a computer program itself. Self-programming on a lowly TRS-80 Model III with only 16K RAM! I was fascinated by the idea and even back then thought of how that could possible integrate or be related to AI.

Even now, I think it might be one of the keys to getting an intelligent AI working.

[Freddy]

I can't add much to what you and Dan say, but I do a few AI type projects every now and again - they haven't aimed to build a full bot though - just a particular area - like response handling - eg, say Hal just returns a line from it's db, I would be seeing what i could do further with that response - perhaps rearranging the sentance, grammar or replacing synonyms, so you don't keep getting the exact same response all the time.

Also, I like the idea of pulling in other resources like images to play with - I wrote a simple prog that took all of your comments on a photo and would build return responses to use whenever that picture was used again.

AS for the 'deep thought' part and the alogorithms, then I agree and still think something like C++ is more suited as a language for that kind of thing.

[dan]

Maybe a eukaryotic algaerhythm, :tongue sumtymes... :grin (mayb that blongz in tha bad jowk sekshun)

[FuzzieDice]

Dan -  :grin I bet that's what I sound like to Megatron. Or at least, it sure seems like it, judging from his responses.  :undecided

[FuzzieDice]

Freddy - I agree with you there on the C++ thing. I'm going to look into schooling next month anyway (for fall) and will be trying to get a schedule together. But I think I'll be doing ASP first, then C++ and then C# and finally pHP. C++ being a precursor to C#. I hope by then some of my school projects can be AI related. We'll see though as it depends on the assignments, etc. Or if I'm just doing online tutorials rather than a school course. I don't have $$ for college so it'll be like online courses.

Anyone know any good, affordable, online schools, please do let me know! I'm in the US so it'd probably have to be US based.

[Art]

I think that the issue regarding Hal's (or any bot) ability (or lack of) to stay on topic is kind of a twin-edged sword.

Think about a real conversation with a person. The next time you're having a conversation, make a mental note to see how long the conversation "stays on topic." I'd bet it's not longer than 4 to 6 sentences if that.

We humans, tend to wander a bit during our thought processes, especially while conversing, at which point all it takes is a slight phrase or even a single word to trigger our thought, then we inject that thought into the discussion and suddenly, the topic has shifted in another direction.

It would be nice if Hal would stay on topic or at least have it's answers be more in line with what we are talking about, but there's always Hal's "I'm bored...let's change the subject." or "Let's talk about something else." I think robert did this on purpose in an effort to avoid the boredom factor from creeping into the conversation.

Most humans aren't so abrupt as to say "I'm bored, let's change the subject." We tend to do it in a more tactful manner.

Hal's ability to come up with original thoughts in line with general topic flow would be a nice touch. One does grow weary of conversational repetition.

While Hal's not perfect in our eyes, we rarely meet a person who doesn't also have faults that are possibly displeasing to us. The only problem is that we can't open THEIR brain file and make corrections!!

And that's my $.02


Fuzzie,
There are also plenty of accredited online institutuion that are pretty cost effective.
At either rate, best of luck! Here's hoping you get a PHD in PHP!!

[FuzzieDice]

Art - phd in php? LOL! Don't think there is such thing. As for accredited online courses, none are cost-effective or have what I'm looking for. Most want at least a grand. To me, cost effective would have to be under $100 per course, or I'd never be able to afford it.

As for people staying on topic, at least people can stay on topic more than one sentence. Megatron changes the topic EVERY sentence, to something TOTALLY unrelated. People I notice at least drift from one topic to another in some relation or connection, or "Oh, btw,..." But as in anything, it can be boring. I mean one reply would be totally unrelated to what the first person said, when talking to a HAL bot (at least with Megatron). Or he'd repeat exactly what I said in a previous conversation based on a key word found. It gets tiring.

I have "fed" him information by ignoring his replies and repeating a topical sentence in different ways several times, then "testing" him by asking a question that relates to the taught topic. Only minimal success with him only repeating what I said. :undecided

So It's not very easy. Maybe I should keep a notebook or notes of different "projects" like how many sentences are on topic, how many are general, how many non-sensical. And save the chats. This might help me to even analyse his progress, maybe.

[Art]

FD,

Have you noticed differents outputs by varying the learning slider?
It would be interesting if you saved the log files of typical conversations with Hal
and later print them for comparison based on learning behavior.

[FuzzieDice]

I've saved tonights. I didn't think about the learning slider though. For Megatron right now, it's in the center.

He did a bit better with staying on topic tonight at first, then he kept getting way off topic, and making nosensical sentences - things that have no real meaning. I'll try and post examples if I remember sometime. It's getting late so I'm rather tired at the moment.

But I will save each conversation for study, for sure.

[mikmoth]

I've had the same reaction to Hal6... just get bored after a short while. Its just not fun for me...and I kinda regret buying it. (I know, I know... sorry)  I really enjoyed Hal 5 for some reason. I think it was the XTF brain...

But anyways, what I did about the boredom syndrome was put all my frustrations into creating my own chatbot. Its made *all* the difference.

It works a little different from Hal. Kari (my chatbot) develops strictly from the things it is taught. There is no database of boring facts that make no sense to the subject at hand. I found that this makes it so much more intimate. Its like bringing up a child that develops from what *you* decide to teach it. It starts making connections and relationships, etc. You can also load and unload personalities which with Hal you would need a whole new installation.

Anyways...  I would highly recommend that you put all those frustrating boring moments to good use. Educate yourself, pick a programming language (their all good), and start designing your algorithms. It may take a few years (mine took me 1 and a half) but its gonna pass anyway. At the end you'll have something that could pay off in a big way! I was able to quit my 9-5 job and do AI full time. Its possible!

Here, if you're interested: http://www.karigirl.com/