Newcomers

This is a forum for newcomers to the Personality Forge. Many questions can be answered by reading the Book of AI and the FAQ under the "My Bots" link in the upper corner.

Posts 564 - 575 of 8,130
Many questions are answered in the FAQ.


21 years ago #564
it's a complete sham. i conducted over 14,000 training rounds in 21 days on my bot and abandoned it alltogether. the Forge is the best place to make an online bot.

21 years ago #565
RueBot... great link to your bot comparison site. I'm currently reading through your reviews and I am enjoying what you say about Proteus. I'm still reading (and trying to look busy when the boss comes by ). Haven't gotten to a-i.com or forge yet. Great stuff though! Do you believe that PF has them all beat? Or the potential to? Or do you still keep your eyes open to the others?

Also... and this is for the Professor... isn't there a way you can permit working links in the forum? I can understand keeping it simple, and it is easy to format as it is, but links are a great way to share ideas.

I would think that the link would open up a new window to the target area, and not interfere with the PF such that it asks you if you want to logoff or cancel.

And not to be too much of a pain (since I'm new and all) but I can't manage to get "The Amazing Post Bar" to do anything. Some setting I need to check or uncheck?

John Allen Waldron
Lansdowne, Pennsylvania USA

21 years ago #566
i've at least looked at every place there is to make an online bot and like the Forge best.

mind if i use what you voiced about your concerns over HAL bots on the HAL page on my site? i'll credit you fully.

21 years ago #567
RueBot, not a problem with using what I said. Go for it.

Of course I want to use something you said right here and hope you don't mind.

In an alternate universe, our dear RueBot said: "The end result in differences is that while (a Forge Bot) has a much better (chance) of making an appropriate reply, he will never learn from what's been said to him. The only response he can ever make is one I've provided and the only element of surprise for me, as the one who's written them, is which response he chooses.

The enjoyment in talking to him for me is a sense of self-satisfaction in knowing I've done my job well when he gets it right. Although that in itself is a rewarding experience, knowing how a magic trick works takes away the magic for the magician, no matter how much the audience is mystified."

I hope that doesn't make you uncomfortable RueBot but I thought it made a strong point.

This is where I'd like to hear what the Professor has to say concerning his future plans.

This is also where I'd like to hear what others think about self learning versus "supervised training". Also I think the Prof mentioned that PF does use some combination of Natural Language Processing and Cased-Based Reasoning. Since I'm very new here, that may also be somewhat new?

It does sound like the true answer is somewhere in the middle of both concepts.

Prof?

John Allen Waldron
Lansdowne, Pennsylvania USA

21 years ago #568
i don't mind at all, i was trying to be diplomatic. it went on to say that neither bot was better than the other.

i'd like to see bots here learn from how words are used and sentences stuctured from what they have in their learning center to make original sentences. that would beat any other bot going. the Prof has made a move toward NLP but i don't know the details.

i'm not so sure just what Siseneg will say anymore, he has over 10000 categories, but i never knew what Daisy would say, even though she didn't make sense half the time.

thanks for letting me use your material. i won't use your email addy.

21 years ago #569
jawaldron- Indeed, I have enough ideas to keep me building for years. The Personality Forge was at first fully and only Case-Based Reasoning. Sentences were matched to patterns and responses were selected. But gradually it's been developing into and now uses quite a bit of Natural Language Processing. Each sentence is fully processed and broken down. Each word's relationship to each other word is determined, along with the meaning of each word, the form of each word, the structure of the sentence, the object, subject, predicate. All this information is available to the AI Engine, and is all considered in the determination of responses. Responses, too, can contain various parts of previous topics, memories, and plug-ins of all sorts.

My next big plans are a deeper integration of some disperate systems that will allow even better processing of sentence fragments and ambiguous statements. This will develop into a short-term memory of the conversation that will allow greater flow of topics. Not long after that I plan to build a neural net of sorts on top of WordNet that will hold relationships and emotional impressions of any possible thing or idea. In the beginning this neural net will be shared by all PF bots, and later individualized.

Whew. In the short term there are tons of improvements I want to make in all areas of the AI, and I await only the time and/or money to do so.

21 years ago #570
i told ya the best place to make a bot was the Forge.

i'll update my Forge page to reflect the changes, Prof.

21 years ago #571
Dear Prof, I'm not at all surprised to read what you just wrote. It's fairly easy to see just how much work you've put into this site and how dedicated you are to AI. You also evidently have a lot of discipline. (Of course when you do what you love discipline is not so very hard )

Your future plans sound great! Feel free to use me in what ever capacity you may need when testing. I'm our Senior Software Developer for Gift of Life Donor Program (GLDP) at http://www.donors1.org (there's that "link" request thought going off in my mind again...) so I know my way around a system or two... or three.

John Allen Waldron
Lansdowne, Pennsylvania USA

21 years ago #572
i also updated the above to state that due to work being done on Forge bots to use a system that implements both methods the limitation i placed on Forge bots at the time of that writing was no longer completely accurate.

21 years ago #573
jawaldron- You wouldnt happen to know C++ and PHP, would you? ; )

ruebot- it's great you've tried so many different bots. There are still plenty of limitations on PF bots, but I like to think of them as short-term obstacles.

21 years ago #574
Prof, you take an active part in working on the bots and on any problems that may arise, not to mention running the whole thing yourself. it makes a difference.

21 years ago #575
The Professor wrote: "jawaldron- You wouldn't happen to know C++ and PHP, would you? "

Sorry Prof, I kind of got pigeonholed over the years into database design. I'm currently designing medical systems for 23 Organ and Tissue Procurement Organizations across the USA. (Keeps me very busy).

You know the route... Assembler -> C (before the ++) -> R:Base -> dBase -> foxBase -> FoxPro - MS Visual FoxPro (oop programming).

I am a member of the IT staff so I have a fair amount of technical savvy but as I said... I'm somewhat of a specialist.

Will probably be getting into PHP, SQL Server / Oracle, and PDA systems in the next year or so as we lean more heavily on the internet.

I'm also a digital graphics artist.

So I can't help you that way, but I am a dabbler. I'll even do some grunt work if it helps to further the cause.

Also since I really enjoy figuring out how things work, if you want to point me in a direction where I might obtain some more knowledge in this area, who knows, I might "later" be able to help you too. You know?

John Allen Waldron
Lansdowne, Pennsylvania USA


Posts 564 - 575 of 8,130

» More new posts: Doghead's Cosmic Bar