Denis Robert - May 10, 2017:

Just at this time, the judge decides to send a second message. The judge program will open the file, and append the second message.

The judge cannot send another message at this time. He must wait for the bot’s response (step 2). He can only send a message once the contents of bot.txt have been displayed and the files deleted.

The bot should only delete judge.txt once it has written its reply into bot.txt (step 7). It should not keep trying at intervals to remove this file.

The Process should not rely on any specific manual action of the judges or confederates. It introduces too much chance of error.

Steve,
what programing language is your/Pandorabots Loebner bot programmed in?

It (along with the other AIML entries) is written in Java.

If I do manage to enter this year, I will still insist on manually restarting my entry in order to flush out the previous conversation and lines that have already been used. It will look rather silly calling judge 2 by judge 1’s name.

Denis, I think that within the one second that it takes a judge to type a second message and move his hand to the enter key, a bot can have processed the first message ten times over. If you are really worried about the microseconds inbetween a consecutive fclose() and remove() call in a single program loop, you could also number the files, or have the bot perform a rename() on the text file before processing it. Again, basic functions that pretty much every programming language has.

Don Patrick - May 10, 2017:

Denis, I think that within the one second that it takes a judge to type a second message and move his hand to the enter key, a bot can have processed the first message ten times over. If you are really worried about the microseconds inbetween a consecutive fclose() and remove() call in a single program loop, you could also number the files, or have the bot perform a rename() on the text file before processing it. Again, basic functions that pretty much every programming language has.

Indeed. However, as the judge can’t send a second message until the bot has replied to his first message, it’s a moot point. The judge sends a message then the bot replies. Nobody sends 2 or more messages in a row. I am unaware of any (decent) bot that works that way.

To be honest, I’m going to wait for the official announcement on the AISB page before concerning myself any more with this. The page currently has no mention of the​ new protocol and so my entry will work under their current entry conditions.

Steve Worswick - May 10, 2017:

Nobody sends 2 or more messages in a row. I am unaware of any (decent) bot that works that way.

The bot “Shakespeare” ?

Juliet.
Ay me!

Romeo.
She speaks.
O, speak again, bright angel, for thou art
As glorious to this night, being o’er my head,
As is a winged messenger of heaven
Unto the white-upturned wondering eyes
Of mortals that fall back to gaze on him
When he bestrides the lazy-puffing clouds
And sails upon the bosom of the air.

Juliet.
O Romeo, Romeo! wherefore art thou Romeo? 
Deny thy father and refuse thy name;
Or, if thou wilt not, be but sworn my love,
And I’ll no longer be a Capulet.

Romeo.
[Aside.] Shall I hear more, or shall I speak at this?

Juliet.
‘Tis but thy name that is my enemy:
Thou art thyself, though not a Montague.
What’s Montague? It is nor hand, nor foot,
Nor arm, nor face, nor any other part
Belonging to a man. O, be some other name.
What’s in a name? That which we call a rose
By any other name would smell as sweet;
So Romeo would, were he not Romeo call’d,
Retain that dear perfection which he owes
Without that title. Romeo, doff thy name,
And for that name, which is no part of thee,
Take all myself.

Seriously, if a judge asks “Tell me a poem”, it is more easy and “human like” to send four messages (for example) than one with the entire poem.

Maybe so but I know of no bots that behave that way. If we are saying that’s not humanlike then we might as well stick with the old letter by letter protocol, as surely it’s more humanlike to type each character one at a time…

As I say though, until something official is announced on the AISB page, I prefer to spend my time improving my chatbot than worrying about this.

AIML bots 2.0 :

< br / >

Creates a line break in the response string. Identical to the HTML < br > element.

Two lines, therefore two messages, no ?

No. The br tag is used as a line break for formatting purposes. The contents of the text file will still only contain one message from the bot. Granted, this message may be made up of several responses but there’s still only one message to pass back to the judge.

Example:

Judge: What is your name and how old are you?
Bot: My name is Chattybot.
      I am 5 years old.

A text file can contain many lines. My reply here is 11 lines long but I have only posted 1message.

Then, if a judge see a multi-line response, he will know that it is the bot, because a confederate cannot send a < br > in his message… 

Indeed, so it’s up to the developer to remove line breaks to keep the illusion going. Either that or the confederate composes his message with line breaks and hits a submit button on his form, in a similar fashion to how we post messages to each other on here.

I am following this thread with increasing astonishment and disbelief. It’s less than 2 months to the closing date and we are replacing the ‘old’ LPP with a new one that is undocumented and does not seem to work reliably. In desperation we are now talking of an interim ‘LPP’ that no-one can even agree on.

Hugh Loebner was wiser than we knew. He fully realized the LPP was horrible, no-one liked it and everyone thought it should be replaced. The only problem was no-one could agree on something to replace it with, so in the meantime everyone successfully used the LPP while complaining about it.

Let me make something clear - the old LPP works properly, and when used on a network a Windows registry setting is needed to ensure the LPP named folders are not ‘cached’. If anyone doubts this, I was at Bletchley in 2014 and Uberbot worked perfectly with the LPP (and came 3rd)  as I dealt with the cacheing issue. You can check the transcripts and see Uberbot’s input and output was not garbled, even though the other AI’s were garbled. I have even set up a ‘testing’ network using the LPP and the judge program, and it works fine.

So I would repeat my plea, don’t ditch the old LPP as it will work fine with the registry setting (which takes 1 minute to apply)

I think we’ve heard everyone’s opinion now repeatedly. The next word of insight should come from the AISB.

Will Rayer - May 11, 2017:

Let me make something clear - the old LPP works properly, and when used on a network a Windows registry setting is needed to ensure the LPP named folders are not ‘cached’. If anyone doubts this, I was at Bletchley in 2014 and Uberbot worked perfectly with the LPP (and came 3rd)  as I dealt with the cacheing issue. You can check the transcripts and see Uberbot’s input and output was not garbled, even though the other AI’s were garbled. I have even set up a ‘testing’ network using the LPP and the judge program, and it works fine.

So I would repeat my plea, don’t ditch the old LPP as it will work fine with the registry setting (which takes 1 minute to apply)

I would say the old LPP/configuration does NOT work properly.
If the fix was as easy as you say, we should have seen no issues in 2015 or 2016.

Although my bot came in third in 2015, 2 rounds were lost completely because of the LPP/configuration. My bot generated a secondary log which allowed me to verify that it was running correctly, but the judges were not getting the results.

In the qualifying rounds:
2014 - 20 bots tested - 1 failed
2015 - 16 bots tested - 1 failed
2016 - 16 bots tested - 3 failed (for unknown reasons)

Errors in finals in 2014, 2015 and 2016.

I cannot point to any year that finals have been run without a communication glitch. That type of failure prevents a good bot from competing effectively.

It is up to the AISB to provide direction, but we should not assume that the status quo works, because that has never been shown to be the case.

“If the fix was as easy as you say, we should have seen no issues in 2015 or 2016.I would say the old LPP/configuration does NOT work properly.”

Maybe the fix wasn’t applied, I wasn’t there in 2015 so I don’t know. But I do know if the fix is appied, the LPP works fine. In essence the LPP is a simple protocol, and the cacheing is a known Windows issue with a well known solution.