chatbox dead giveaway

The Loebner award recognizes that the bots it rewards as most life-like do not pass the Turing test.

But there should be a dead giveway they would not pass the Turing test, without even having to do the test. I am almost certain the prize participants are aware of it because it would be an astounding shortcoming if they are not. On the other hand if they are aware of it, why haven't they rectified it and fixed the problem? And if it is not fixable, then why bother with the Award program at all?

------
four chatbots:

* http://www.elbot.com/
* http://www.icogno.com/joan.html
* http://www.pandorabots.com/pandora/talk?botid=f5d922d97e345aa1
* http://www.jabberwacky.com/
------

The chatboxes or bots listed above lack the notable feature we should from anything that would pass the Turing test: autonomy. The bots mentioned are all only event-driven. Without any input from the user, they do nothing.

At least Elbot recognizes blank inputs, but without any input from the users , blank or not, these four bots are dumb.

As autonomous agents, humans have the potential to talk almost endlessly along coherent lines of reasoning. Being self-motivated, they have the ability to pick subjects, organize ideas and reasoning about them, probe past memory for things they might *want* to remember to mention. All these things contribute to the shape forms and usage of natural language, to the way natural language sounds to it.

Without this dimension, no matter how funny or witty the hardcoded responses thought up by the authors are, there's a very obvious dead endedness or closed quality to the chats.

As if this weren't enough, there is furthermore another difference from human conversation. With these chatbots, the speech responses are alternated between the bot and the user , in a ping pong fashion. In human conversation, it is customary to be interrupted, sometimes ignored, sometimes responses are formed that are spoken over your message, and responses occur in real time as the two parties speak at the same time. In human chat situation, there is no such articificial synchronized input - response sequence. Chatters often type at the same time, sometimes inputs or queries receives other queries as a response due to this lack of synchronization between the speakers.

As a further aside there is also the question of whether and how does XSL and XML deal with temporal reasoning ? If not this could present a possible limitation to the ability of AIML to handle complex reasoning tasks. There are extensions however, and example of where extensions enabled the synthesis of elements like conditional responses, as outlined in t

One of the conclusions is that how can all this be known by the Loebner prize people and yet they proceed with so-called Turing tests. Doesn't the foregoing mean a failure even before the Turing test is started? An a priori failure by default. The conclusion means that the Loebner Prize is really a joke, that the effort is not well-guided, that it simply rewards the wit of funny responses and cute avatar graphics.

That is it takes things like videographic and speech synthesized chatbots as advances, whereas they are only windowdressing with real substance missing and omitted.

The pretentiousness of participants (those that apply for the prie) then is thus clear. They know they are not Turing test worthy, and yet they join just to get cutesy awards of "most human-like". This constitutes a joke, a sham, a waste of time, mock AI. Not AI.

P.S. - 19.xi.08 - Come to think of it, even the name of AIML (which i think was dev'd by Alice' folks) is unfortunate. Maybe W3C should 've copyrighted the use of the partial-acronym ML (just kidding; I'm all for memetic freedom and all). It's certainly an ML, this AIML, but it's *not* AI. At least it is not all the AI there is, if i'm not misunderstanding anything. Even if i'm misunderstanding all the possibilities of AIML, it's still an application within the linguistic or reasoning school of AI (ultimately relying on a von Neumann machine), and not of the connectionist or genetic or distributed paradigms (to say nothing of new AI algorithms for quantumputers - hopefully exploiting other than just massive parallelism).