Carl,

Pretty impressive how the process is working. Some answers are spot on and some other questions seem to require considerable processing by Labia.

I also like the different “tab” selections you have at the top. One can view the content of conversations as well as some other users. I must say that some users must be kids or some seriously demented adults based on their conversations! That seems to be a perfect example of why we do not want our bots to acquire knowledge from Internet conversations (unless the vulgarity and slang can be properly filtered and not entered into the bot’s memory).

Good direction you seem to be going with this and a bit more trimming and a tad more topic focus and you’ll have a real gem! Thanks for sharing and continued success!

Art Gladstone - Mar 1, 2015:

Carl,

Pretty impressive how the process is working. Some answers are spot on and some other questions seem to require considerable processing by Labia.

It is working pretty well now, but still needs onging effort to filter/clean/organize output from the “thinking” part of the bot as well as better image processing (so much as yet wasted meaning is still trapped in images).

Art Gladstone - Mar 1, 2015:

...One can view the content of conversations as well as some other users. I must say that some users must be kids or some seriously demented adults based on their conversations! That seems to be a perfect example of why we do not want our bots to acquire knowledge from Internet conversations (unless the vulgarity and slang can be properly filtered and not entered into the bot’s memory).

Indeed!  In fact though, one reason I chose the “reformed sexbot” personality was to get the maximal number of users giving input (even if a lot of it is pretty freaky), since analyzing lots of variagated input is one key to creating a solid NLP engine.

Art Gladstone - Mar 1, 2015:

Good direction you seem to be going with this and a bit more trimming and a tad more topic focus and you’ll have a real gem! Thanks for sharing and continued success!

Thanks Art!

Could you elaborate on how exactly it’s ‘thinking’? It’s certainly an accomplishment and something to be proud of… but it just seems like a glorified search engine and not something that is doing anything resembling ‘thinking’, at least not in the sense of what I conceive thinking to be. Perhaps you have a different concept of it? In which case, I would genuinely love to hear it to gain a better understanding of what you have accomplished

Donavan R Lane - Jun 9, 2015:

Could you elaborate on how exactly it’s ‘thinking’? It just seems like a glorified search engine and not something that is doing anything resembling ‘thinking’, at least not in the sense of what I conceive thinking to be.

Well, I suppose it does depend on what you consider “thinking”.  What I meant by “thinking” is that the bot can draw on data from the Interwebs to form a (usually) cogent response when confronted with input beyond the bot’s knowledge. 

Most bots just degrade to a random response patterns when challenged with unfamiliar input whereas Laybia AI tries to formulate (“think up”) a response that is related to the input (and temporally related to the conversation at large).  Information retrieved is organized and analyzed for, among other things, emotional content, so that if you ask the AI if it “likes” something, the bot will try to determine if the thing is “likable” and respond appropriately, as opposed to the more traditional bot interaction of:

chatbot - Jun 9, 2015:

INPUT: Do you like ice cream?
BOT: I scream, You Scream, we all scream for Ice Cream.

Which is funny once, disappointing the second time and induces boredom thereafter.

Contrast that with a “thinking” bot:

Laybia AI - Jun 9, 2015:

INPUT: Do you like ice cream?
Laybia: *Recurses* Ice cream? Ice cream derived from earlier iced cream or cream ice is a frozen food typically eaten as a snack or dessert usually made from dairy products such as milk and cream and often combined with fruits or other ingredients and flavours Ice cream parlors or parlours are restaurants that sell ice cream gelato sorbet and frozen yogurt to consumers From scanning the data stream, I think Ice cream, feminine, products that do not meet the criteria to be called ice cream are labelled “frozen dairy dessert” instead. analogues made from dairy alternatives such as goat’s or sheep’s milk or milk substitutes are available for those who are lactose intolerant allergic to dairy protein or vegan. the crea into the larger bason. That is what I think I understand, generally Ice cream is something I would really like! Totally.

So, is that considered “thinking” or a “gloried search engine”, or a little bit of both? I’d say it is closer to “thinking” than simple input pattern matching, or even more elaborate Parts Of Speech (POS) analysis (which Laybia AI also will fall back occasionally).  Essentially I have combined many approaches (including auto-correction of input and real-time information searching and emotional analysis) in an attempt to create a truly “thinking” bot.

Of course this is still very much a work in progress, and I am always appreciative of comments and constructive criticisms

-cb

I think I have a good understanding now of what you consider to be ‘thinking’, and Laybia AI certainly seems to do that! Bravo and thanks for elaborating!

My only critique would be that you posed a question to Laybia, “Do you like ice cream?”

She could have addressed the question directly like “Yes, doesn’t everyone?” or “Sorry but I’m just an AI and as such, do not eat. I understand that most humans love ice cream.”

Laybia, however decided to provide a cute quip in one example and a very detailed definition in the other example. Unfortunately, neither of these were what one might call “proper” responses to your initial question.

Importantly, how much weight does or can Laybia place on queries in an effort to maintain topic flow?

Just some more wood for the fire….

Art Gladstone - Oct 14, 2015:

My only critique would be that you posed a question to Laybia, “Do you like ice cream?”

She could have addressed the question directly like “Yes, doesn’t everyone?” or “Sorry but I’m just an AI and as such, do not eat. I understand that most humans love ice cream.”

Laybia, however decided to provide a cute quip in one example and a very detailed definition in the other example. Unfortunately, neither of these were what one might call “proper” responses to your initial question.

Actually, Laybia looks up “ice cream” and evaluates if it IS something she would like, adding at the end of the data summary an evaluation like “ice cream sounds like something I would really like” based on all the positives associated with the data for ice cream.

Art Gladstone - Oct 14, 2015:

Importantly, how much weight does or can Laybia place on queries in an effort to maintain topic flow?

Laybia does have some temperal awareness, but not as much as I would like   As a simple example, she adds “ice cream sounds like something I would really like”, but does not yet follow up with something tangential like “My sensors resonate after tasting X flavor of ice cream in particular”, which is obviously a more difficult exercise to generate dynamically.

I would suggest stating the conclusion “I like ice cream” first, then follow with the argumentation thereof.

Don Patrick - Oct 15, 2015:

I would suggest stating the conclusion “I like ice cream” first, then follow with the argumentation thereof.

That makes a lot of sense.  I have implemented it in the latest build. Thanks!