AI Zone Admin Forum Add your forum

NEWS: Chatbots.org survey on 3000 US and UK consumers shows it is time for chatbot integration in customer service!read more..

New Loebner Prize Protocol announced
 
 
  [ # 121 ]
Steve Worswick - Jun 14, 2017:

Can I ask if anyone has managed to make an entry which just uses the new LPP rather than hooking into the old LPP?

I believe Merlin, 8pla and Dave Morton may have done.

Steve, I’ve managed to make a preliminary entry which uses a basic chatbot engine.

This is not the final design, but take a look… The chatbots input textbox is gone.  Instead

it listens for a new message directly from the judge.  That’s a frameset with four webpages

to help make alpha testing faster and easier.

 

 

 

Image Attachments
for_discussion_purposes_only.png
 

 
  [ # 122 ]

If you can get it to work successfully 8man, I would definitely enter it. Anyone who manages to enter this year should have a great chance at getting through to the final 4.

 

 
  [ # 123 ]

Steve,

The new LPP is has a lot of potential.  Chatbots run fine in DOS on a webpage.

Reference:  http://chatbot.tk/newLPP  ( Eliza running in a DOS box )

They should be able to connect to the competition with custom programming.

 

 
  [ # 124 ]

I set up Denis’ interface (Windows 10, IE 11). The following settings appear to be the minimum required:

“internet options” -> security -> custom level -> “Initialize and script ActiveX controls not marked as safe for scripting” must be set to “Prompt”
“internet options” -> security -> custom level -> “Run ActiveX controls and plug-ins” must be set to “enable” (is enabled by default)
“internet options” -> security -> “Enable protected mode” must be turned off, or you’ll get a “Permission denied” error upon connecting.
“Safety” -> “ActiveX Filtering” must not be checked (is off by default)
After applying these changes, close Internet Explorer and launch everything, and click “yes” when prompted to allow ActiveX. From there on everything works just fine.
Dave, I’d love to hear if you get this working.

I have to say though, this is much to ask of the tester. How many entries beside Denis intend to use this interface? There is something to be said for safety in numbers, but otherwise I think I’d better stick with my own hack.

Image Attachments
LPPactivex1.pngLPPactivex2.png
 

 
  [ # 125 ]

After much toil, and losing enough hair from the top of my head to fully clothe a Poodle, I’ve come to the decision (with Bruce’s suggestion/support) to call off my attempts this year to get the current LPP to work universally in all browsers and platforms. However, I’ll continue to hack away at this over the next few months in order to have a more robust, user friendly version ready for next year. It saddens me that some of the top competitors won’t be entering this year, and I worry that this could possibly mean the end of the Loebner Prize, but there just isn’t enough time remaining to get a competition-ready “out of the box” solution ready that everyone can use. downer

 

 
  [ # 126 ]

Assuming the new LPP chatbots
are entering the competition:

SkyNet-AI
__________________________
8pla.net
__________________________
???
__________________________
???
__________________________

We need four or we may as well
skip the qualifying round.

Who else is planning to enter ?

__________________________
This is all for the sake of discussion only

 

 
  [ # 127 ]

Denis Robert and I are entering, surely. José Ignacio sounded like he got something to work, and any javascript-based bots shouldn’t have any difficulty with the current protocol.

Dave, you may want to hold off working on it for next year. Should I get into the finals, I intend to offer the AISB a simpler interface of my own making next year, using plain text files and shared folders along the lines of what Steve discussed earlier in this thread. I reckon I can make something both simple and reliable for everyone.
I lament the absence of Steve and Bruce’s entries this year. Although it frees two of the top spots for others, it’s hard to gauge progress when the bar’s missing.

 

 
  [ # 128 ]

The live contest round experience gained with the new LPP is worth it for Steve, Bruce and others to build a new entry for this year. It may be seen as good sportsmanship, but I am certainly not criticizing anyone.  No offense intended.  I understand the difficulty of the new LPP.  With that said, the new LPP has tremendous potential.

 

 
  [ # 129 ]

Hi Guys,
My name is Steve and this is my first post to this forum or any other for that matter.
My chatbot is Talk2Me and is written in java.
I hope this might help anyone considering entering this years LP.

I had never heard of Socket.io before reading this thread, I felt I had to give this new LPP a try.
See Socket:IO website for installation instructions.

I ran node server.js
Then opened IE on Windows 8.1, opened controlpanel, judge & confererate on seperate tabs.
Then opened ai_old_protocol(thanks Denis), entered name, secret & commuications dir, & hit connect.
All 3 ‘Round 1’ labels on controlpanel are now green.
I opened Talk2Me, set the communications dir(as old LPP),
On controlpanel hit ‘New Round’, ‘Start Round’, then started typing in the judges input box’s,
to get responses from Talk2Me.
Everything seemed to work very well, and the beauty is I made no changes to my chatbot(thanks again Denis).
I think that with Denis’s script, any chatbot that worked with the old LPP, should work OK with the new LPP.

Ideally I would prefer to connect my chatbot directly to Socket:IO, so I have tried several java-Socket.io implementations mostly from Github, but I can’t get any of them to work, they either won’t compile, or I get handshake/I/O errors, or server.js doesn’t detect my connection. I’ll keep trying, there must be a way to do this.

Personally I would go along with the idea of a plain text file communications. The preliminary round is essentially an exam, so place exam.txt(20 questions) in each chatbots communications folder, point chatbot to that folder, the chatbot sees exam.txt and goes into EXAM_MODE.(if it doesn’t see exam.txt, then it goes into LPP_MODE). It then appends it’s responses to ‘exam.txt’, saves the file, then renames it to chatbot-name.txt and shuts down. The examiner can then grade each paper, and the best four go into the final.
These papers could easily be posted on the net, as is, complete with scores.
It must be easier for everyone concerned running it this way & everyone gets exactly the same set of questions(no typo’s).
For the final I would also go along with the idea of a plain text file sent to the communications directory.

Anyway for this year I will enter my chatbot and request Andrew to use Denis’s script, and hope for the best.
It won’t be the same unless all the top names enter, but good luck to all who do.

Regards.

 

 

 
  [ # 130 ]

Hi, Steve, and welcome to chatbots.org! smile

I’m glad that you were able to get something going, and I wish you luck in the upcoming competition.

@Don: I will, of course, consider your suggestion, but I agree with 8pla that the LPP has a lot of potential if we can get it to be a bit more robust and universally usable. I may never have entered the Loebner, but I’m as invested in it’s success as anyone else here, so I’m willing to work with anyone who wishes to make whatever will be used in the future. wink

 

 
  [ # 131 ]

Ah yes, I forgot you had bigger plans for the javascript server interface.

Thank you for your input, Steve. I’m glad to hear that not everyone had a hard time with the new protocol.

 

 
  [ # 132 ]

JavaScript has the most crippled regular expression engine, according an article, “I know you don’t want to hear bad news” it continues, “among all the major regex engines, JavaScript is the worst” and it compares, “Python’s default [Regular Expressions] engine—which occupies the second-worst position—is far less crippled than JavaScript

Due to its crippled Regular Expression engine, JavaScript lends itself to light duty Artificial Intelligence design, I think.  What do you think?

Reference: http://www.rexegg.com/regex-javascript.html

NOTE: Sharing this article is meant to be on topic, as it may relate to building JavaScript contest entries.

 

 
  [ # 133 ]

I once tried coding a grammar parser in Javascript. I quickly ran into the issue that Javascript can’t save anything to files, there was a low limit to array sizes, low speed, different limits across browsers and so on, meaning Javascript is not fit to process extensive world knowledge. On topic, it says here that one would need Ajax and a server to interact with files and databases, but that one can only write cookies on client side. I experimented with that way back in 2003 and iirc cookie files do lend themselves to relaying messages such as in the LPP, but I do not recommend it.

 

 
  [ # 134 ]

JavaScript has progressed a lot in the last 10 years.

∞Pla•Net - Jun 23, 2017:

Due to its crippled Regular Expression engine, JavaScript lends itself to light duty Artificial Intelligence design, I think.  What do you think?

Reference: http://www.rexegg.com/regex-javascript.html

I have not had any big problems using the JavaScipt RegEx engine. One of its benefits is the extreme speed, which rivals compiled languages. I like it better than the Python engine. There are features that I would like to have, but nothing that has prevented me form developing my systems.

JavaScript can do heavyweight AI and I created JAIL (JavaScript Artificial Intelligence Language) to specifically speed the process of creating chatbots and doing AI.

 

 

 
  [ # 135 ]

Merlin,

You certainly have a point. Skynet A.I. (JavaScript Artificial Intelligence Language) is a great achievement, just like RiveScript, ChatScript, Arckon, and AIML.  All of which may use JavaScript.

In discussing applied techniques, my intention is not to try to trivialize JavaScript. May I ask your technical opinion as a JavaScript expert?

JavaScript lacks regular expressions support for negative lookaheads. So it is extremely impractical to match a letter _NOT_ followed by another letter. How then would a JavaScript A.I. (blind to digraphs and trigraphs) compromise a cipher, for example?

EXAMPLES:

digraph: regex

ch: c(?!h)

trigraph: regex

scr: s(?!cr)

There may be better regular expressions examples.  My examples are inspired by ChatScript

 

‹ First  < 7 8 9 10 11 > 
9 of 11
 
  login or register to react