K base and MS Agent

[johnphoenix]

I'm checking out verbot 5 and hoped she came with an updated default knowledge base. I had also hoped that she would have lots more added functionality. It's been 5 long years between versions 4 and 5.

Since 2004 hasn't there been any one person who has 'taught' the verbot to learn in all this time so we could have a really extensive knowledge base and better simulated AI in verbot?

Why hasn't this been done already?

That also goes for all the other 'learning' programs, Ultra Hal etc.

It seems to me the folks at Conversive could have done this from the start so Verbot might really be something today.

About MS Agent. Since MS is dropping support for Agent will the next version of Verbot be replaced with something else?

When I first heard about Sylvie3.0 back in 2003 I called Virtual Personalities to find out when I could buy her for my home PC. I was told at the time Sylvie was only being developed for Industry related Applications and the license would cost me $20.000 (twenty thousand) US dollars. (hey you can tell people you have 20.000 dollar software on your pc <grin>) I have waiting all that time to see if the girls would make my dream of a talking desktop come true.

[Matt]

I'm checking out verbot 5 and hoped she came with an updated default knowledge base. I had also hoped that she would have lots more added functionality. It's been 5 long years between versions 4 and 5.

Since 2004 hasn't there been any one person who has 'taught' the verbot to learn in all this time so we could have a really extensive knowledge base and better simulated AI in verbot?

Why hasn't this been done already?

That also goes for all the other 'learning' programs, Ultra Hal etc.

It seems to me the folks at Conversive could have done this from the start so Verbot might really be something today.

I understand your frustration John, I would have liked to devote more time to the verbots because it's fun, unfortunately since our company is really focused on large enterprise solutions, the time we can spend on verbots has to be largely on our own. We've wanted to make the verbot free for quite some time because we've had so little time to work on improvements, and even though Verbot 5 doesn't have major feature enhancements we've updated it to a more modern .NET runtime, future proofed it for Windows 7, and best of all, now it's completely free, which may not seem like a huge leap forward, we thought it was appropriate to up the major version number.

Perhaps if we have time in the future we can look at getting a better set of knowledgebases to include in the installer, but since we don't really have that set of kbs developed it would take some time and we wanted to release the Verbot 5 version while we had time.

About MS Agent. Since MS is dropping support for Agent will the next version of Verbot be replaced with something else?
...

In windows 7 you will not be able to use the MS Agents, but you can still use the Conversive Characters (and I notice on windows 7 there is a nice new, better sounding SAPI 5 TTS that you can use with the Conversive Characters.

-Matt

[johnphoenix]

Don't get me wrong Matt I'm glad Verbots are free now.. have been since 04 right..I suppose Im as guilty as anyone as part of the community because If I had the time over the years I could have built up my own KB. I know its hard to find the time and someone willing to just sit there and write scripts for verbots all day based on encyclopedias or other tomes of learning. Hey what about that janitor guy who always offers you one of his wife's homemade muffins in the morning? Kidding of course.

I know at one time you guys wanted to use verbot tech on a broader scale in business. I got the impression from speaking to whoever it was I spoke to back in 03 that you wanted to see verbots play more of a role in office work.. Is this correct?

What made you guys stop developing verbots for business and switch to web applications instead?

Or for that matter what made you decide to stop serious development of verbots when at one time verbots seemed like your flagship product?

And now that Banner Bots are gone.. what exactly do you guys do? What projects are you working on and how is it going?

[Edit] Trying to research the history of Verbots I just came across a link for Dr. Michael Mauldin. I didn't know the Verbots were based on work from the founder of Lycos. Is Conversive still owned by Michael Mauldin ? If so, I think I can see where the time is spent.


Oh and I'm going to try Verbots with my copy of Windows 7. Currently I use Julia with Cepstral Callie's voice. It works really nice.

[Matt]

What made you guys stop developing verbots for business and switch to web applications instead?

Or for that matter what made you decide to stop serious development of verbots when at one time verbots seemed like your flagship product?


And now that Banner Bots are gone.. what exactly do you guys do? What projects are you working on and how is it going?

Conversive has been moving away from the consumer market and developing Enterprise applications since around 2001 (which is why we changed our name from Virtual Personalities), we use similar pattern matching technology that the verbot uses, but much more powerful and sophisticated and target the enterprise customer service and sales space. Our main product is a Live Automated Web chat product; When a web user asks a question that is handled by the knowledgebase it is able to answer automatically, if there is a question that the automation is unable to answer, our software seamlessly engages a live Customer Service Representative to answer the question, and even once the conversation is escalated to the live CSR, the automation continues to help the CSR by autoanswering highly-confident answers and providing suggestions for less confident answers.

Trying to research the history of Verbots I just came across a link for Dr. Michael Mauldin. I didn't know the Verbots were based on work from the founder of Lycos. Is Conversive still owned by Michael Mauldin ?

Dr. Michael Mauldin is still involved in the company at a high-level, he is a director on our board of directors.

If you are interested in finding out more about our main enterprise products just visit our website: http://www.conversive.com/

-Matt

[POWERPLAY07]

Well John,

I uploaded some zips 2009 for the conversation base, to make it work better. Variety 1---5

That should help a little

Catcha Later
Tom

[MikeA]

I realise this is an old thread....... but if any new users see this thread they may be put off, and I don't think they should be.

About johnphoenix's original point about learning, this is not as good as it may seem. Just check out the learning bots. What they do is very good as far as learning goes, but the quality of output is not better.

I used Ultra Hal for a while before trying Verbot. What you have to realise is how much UH depends on a massive database full of random (and sometimes bizarre) sentences that form what some call it's knowledge, used for pattern-matching. The learning it does as far as how that affects output is quite minimal, and it takes a long time to see results (on their forum it has been stated as years). The interesting thing is, when using UH, what I really wanted to do was add direct input / output rules..... because the learning was going to take too long. I also experimented with deleting the random sentences, because many of the replies were based on these and to me were unnatural. After doing this, I realised how much time I would need to spend teaching the bot. So I turned to Verbot.

The thing to understand about self-learning bots is that, all you are really doing is building up a knowledge base to do pattern matching from. UH is a bit smarter than that, and can use learned data in a number of ways. But in the end, it is still matching sections of input with prescribed output. The UH learning system is clever, but overall not what I would call an advancement in AI.

The same applies, in my opinion, to the uploading of conversation bases. Verbot can do this, and is then working in a very similar way to most AIML bots. This is fine, but is far below the actual power Verbot has.

Where Verbot really stands out from the crowd is 1) the ability to use c# within the process 2) the interface which enables easier setup of more complex input / output pattern-matching 3) the new before and after rule firing commands that can capture information.

Put together, I believe Verbot is actually the most advanced bot as far as pure AI is concerned, because we can set Verbot to analyse inputs based on programmed calculations, and only then forming an output.

Just my 2 cents.

[Matt]

Thanks for writing that Mike! It's nice to hear that the features of the Verbot stand up against the competition in some regards, especially if you like the power of C# and the not so random as UH rules structure.

-Matt