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..

The final four for 2014 have been announced
 
 
  [ # 31 ]

REF Steve Worswick - Sep 24, 2014:

QUOTE
I see there was a Winograd Schema included for the first time in the qualifying questions:

The car couldn’t fit into the parking space because it was too small.  What was too small?

I wonder if anyone got that correct (apart from a 50/50 guess of course)? I know I didn’t.
Mitsuku would have replied with something like “A pin is small”.
UNQUOTE

This is an excellent example of why I believe my AI system/chatbot(?)
advances the field of AI. Because Simplex does not rely on third party
dictionaries and defines words in a manner suited to use by computers
rather than by people the answer is relatively simple for Simplex to
resolve.

Jim.

 

 
  [ # 32 ]

I had practically the same issue when my program answered “Tall is a giraffe” when asked “Who is tallest?”. It is a matter of context restrictions, not of third party software or knowledge representation.

However, if your chatbot really is as good as you say, I trust it will do well at next year’s Winograd Schema contest. Otherwise you may assume that I don’t believe a word of it.

 

 
  [ # 33 ]

Jim/Don,

“I took the one less travelled by, and that has made all the difference.”

I had intended to ignore your post, partly because I could not be sure
that I fully understood it.

Having brought forward dealing with the question under discussion and
verified my original thoughts I now think I understand what you were
saying.

I also believe that if I was to say that I agree with you ... then we
would both be wrong!

I don’t think it is possible to say there is a practical limit to what a
context consists of. Accepting that view means the question of “context
restrictions” does not arise because the software caters for such. This
I believe Simplex does.

Moreover, the methods employed by Simplex includes new concepts that I
believe do not exist in normal dictionaries designed for use by humans
rather than by computers.

It is this aspect that allows Simplex to clarify the uncertainty assosciated
with the question mentioned in about half a dozen lines or less.

The following dialogue with Simplex demonstrates this fact:
QUOTE
The car could not use the parking space as it was too small.
  Go on.
What was too small?
  That’s the parking space.
What was too large.
  That would be the car.
Which was too big?
  That’s the car.
Which one was too tiny?
  Sounds like the parking space to me.
Which is bigger a car or a parking space?
  a car and a parking space ...
  vary in size.
Which is larger the moon or a strawberry?
  I feel as though I’m humouring you!
  A moon is the larger with quite a differrence.
UNQUOTE
Jim.

 

 
  [ # 34 ]

Firstly, what I meant was that if a program gives an answer like “a pin is small” or “a giraffe is tall”, it is ignoring context, i.e.: it is not restricting its answers to pick only from the subjects that were mentioned in the previous sentence.

Secondly, I find it inappropriate to claim that your program could have done better while you earlier declined to enter the Loebner Prize as it was not in your interest. I challenge you to enter the Winograd Schema contest instead.

Just out of curiosity and against better judgement I have two questions:
1. Did you program anything inbetween your last post and this one to make your program better handle the question?
2. What is Simplex’ current answer to “The car did not fit the parking space as it was too large. What was too large?”

 

 
  [ # 35 ]
Don Patrick - Oct 12, 2014:

...against better judgement…Did you program anything inbetween your last post and this one to make your program better handle the question?...”

‘Simplex’ is an obvious troll, you should really not continue feeding it.

 

 
  [ # 36 ]

Yeah, sorry. Curiosity just keeps getting the better of me.

 

 
  [ # 37 ]

Jim/Don,

“If you can bear to hear the truth you’ve spoken twisted by knaves to make a trap for fools!”

Ref your “firstly” (I’m ignoring the word ‘not’ as surely this is unintentional.)
I disagree with you completely that such answers ignores context. The
“previous sentence” alters the context in such a manner that the response
is limited to “restricting answers from the previous sentence”.

I suspect you have in mind those cases where the pronoun can refer to a
number of entities and so the there is true ambiguity. Such cases I wrote
code for maybe two years ago but I do need to retest that code.

Ref: your “secondly”
I have no recollection of saying “my program could have done better”. Please
do not misquote me. I believe it is something I would not say.

I have repeatedly explained in detail why currently I do not enter contests.
Such context you choose to ignore!

Ref: “Just out of curiousity”
1. Again ... I believe your question is ambiguous.
2. Simplex answered “Sounds like the car to me.” (Do remember Simplex uses
what I term the OTE approach where OTE = Or The Equivelant and hence the
answer can vary but the sense is constant.)

Jim.

 

 
  [ # 38 ]

Jim/Carl,

I’ve been called many things in my time but “troll” is a new one to me.

Incidentally, I did look it up. I am sorry if my posts are upsetting to
anyone. That is certainly not what they are intended to do.

I fully accept and understand that the information they convey may give
cause for concern.

Jim.

 

 
  [ # 39 ]

“This is an excellent example of why I believe my AI system/chatbot(?)
advances the field of AI. Because Simplex does not rely on third party
dictionaries and defines words in a manner suited to use by computers
rather than by people the answer is relatively simple for Simplex to
resolve.”


Jim, like yourself I was convinced that my approach advanced the field of AI.  However, after now entering 2 contests the reality is I am much more behind the curve than being on the leading edge.  I put in a lot of hours preparing for the Loebner, both programming and testing, only to find in the end I did at least 10% worse than I thought possible, which may not sound like much but it was enough to knock me out of the top 4 and rank in at number 7.

I have no clue, nor anyone else except you how well Simplex actually performs.  Simplex has not been tested by an independent source, such as a contest.  Until such time and you have those results you are only guessing at how well it works.  I can state from my own limited experience, it doesn’t work nearly as well in the real world as it does in your own testing.  In my case, I test the program and so do 4 others, and the results are still worse than expected in the real world of contest.

Jim

 

 
  [ # 40 ]
James Curran - Oct 14, 2014:

(I’m ignoring the word ‘not’ as surely this is unintentional.)

Nope, intentional. however, I was talking about my program’s workings in answering these questions, not yours. It was mine who took this test, after all.

Although there are definitely misunderstandings here, I consider my words accurate enough. But just in case there was any doubt in anyone else’s mind: No, I did not literally quote James Curran on what he said.

James Curran - Oct 14, 2014:

Simplex answered “Sounds like the car to me.”

Thank you for giving a straight answer. My curiosity is satisfied, as this suggests your program does examine ambiguity. How you go about your way from here ceases to interest me, I have suggested all I can that might have been of benefit to you, and obviously to no avail.

 

 
  [ # 41 ]
James Curran - Oct 14, 2014:

Jim/Carl,
I’ve been called many things in my time but “troll” is a new one to me.

Talking as if ‘Simplex’ really exists as a program in forum threads devoted to AI programs seems like trolling to me.

 

 
  [ # 42 ]

While I can understand the viewpoint, I don’t think that it’s all that constructive to use “the ‘T’ word” here. If you think that someone here is trolling (e.g. posting with inflammatory comments, or making fantastic claims in an effort to elicit responses, rather than to share information or ideas), then simply ignore them. On the other hand, if you have groundbreaking information or ideas, having a concrete way of backing up your words is not only a wonderful idea, but also highly encouraged, and is a great way to avoid getting summarily dismissed as a crackpot. And lastly, if any of my comments above upset or offended you for any reason, please feel free to get over it, because there was nothing upsetting or offensive that was said.

 

 
  [ # 43 ]

On this issue I don’t understand why the first quote was:

The car could not use the parking space as it was too small.
  Go on.
What was too small?
  That’s the parking space.

And the second one was:

From Don Patrick:
What is Simplex’ current answer to “The car did not fit the parking space as it was too large. What was too large?”

From James Curran:
Simplex answered “Sounds like the car to me.”


With that said, I will move on to what James Adams mentioned, as there are feelings some programmers here might share.

@ James Adams:

How true you are.

What it proves is that entries need more testing, which often means for us, more maturity and more years into it. I was ready for “favorite” and not for “favourite”, for instance; I did not think about the alternate forms in that example. English being a foreign language to me, things that can be obvious to most minds can be unexpected to others. Such errors make a program look foolish (just like one of my other posts here).

We must go over such episodes.
I am sure some of us are still here with that hope and objective to make Loebner contest “complete”.

Preparing for next year will be great, as we’ll bring even more AI in the programs. If our program wasn’t ready for the Silver Medal - and I am sure it was not - then this means we had to continue working on it anyway. I mean, qualifying is one thing, but “getting human” is another.


Increasing the quality and performance of the entries is one way to get closer to that, as it challenges us to more quality. I believe entries like the ones of this year challenge us to do more. So we’ll do more. And we’ll get closer.


Something sad but true is that:
Science doesn’t deal with instant gratification.

But Hugh must know “we are still working on it”. wink

 

 

 
  [ # 44 ]

I am sorry for the lack of edit, as I understand the quotes now, so no need to develop on that point.

Thanks for sharing information about the Winograd Schema contest
Two contests of this scope for 2015… that will be an interesting year smile

 

 
  [ # 45 ]

Christophe—I agree 2015 will be an interesting year.  My program has not performed as well as I hoped so far, which only means I have more work to do.  In comparison to many my program is very young and immature ( 5 years old).  Currently “Nicole” suffers from several problems.

1. Lack of data (which is larger a cat or a kitten)—wasn’t in the database to compare
2. Programming errors - who is taller/shorter - an error in the code prevented this routine from giving the correct answer
3. Parsing—The program knew what is the capital of, but didn’t know what is the capital city.  I also have a problem with “Hot Chocolate”—my program saw it only as “Chocolate”

4.  Then the biggest one—answers from memory.  For example Bob likes to play tennis.  5 questions later who likes to play tennis.  My problem here is that the program will answer the question correctly, however at times when asked a question, the program will answer out of the blue (and after the question is asked)  “Who likes to play tennis”—I know I haven’t explained this well, but it is a problem that I have been unable to resolve.

I am will enter the Winograd Schema contest, however as noted above I have a lot of work to do prior even tackling that one.

If anyone could give me some ideas or clues regarding answers from memory, I would be most appreciative. 

Jim

 

 < 1 2 3 4 > 
3 of 4
 
  login or register to react