Hi all,

I recently posted this to the BOINC mailing list:-

>>>

There are reports that a worm exists on 'AIM' which 'talks' to the user to persuade them to download a file of itself and then communicates to the users 'AIM' buddies in order to replicate itself.

http://news.bbc.co.uk/1/hi/technology/4520766.stm


In this article, its author suggests that this worm is close to fulfilling the criteria for the Loebner Prize.

http://www.loebner.net/Prizef/loebner-prize.html


"The Loebner Prize is the first formal instantiation of a Turing Test
<http://cogsci.ucsd.edu/%7easaygin/tt/ttest.html>.
The test is named after Alan Turing <http://www.turing.org.uk/turing/>
the brilliant British mathematician. Among his many accomplishments was basic research in computing science. In 1950, in the article /Computing Machinery and Intelligence / <http://loebner.net/Prizef/TuringArticle.html>
which appeared in the philosophical journal Mind, Alan Turing asked the question "Can a Machine Think?"He answered in the affirmative, but a central question was: "If a computer could think, how could we tell?"
Turing's suggestion was, that if the responses from the computer were indistinguishable from that of a human,the computer could be said to be
thinking."

It occurs to me that BOINC may be a suitable candidate for development to try and win this prize.


For example; when the user starts to attaché to a project, it could ask the user, using the computers name, if they would like it to find the best project to attach to. It would then search all projects and, say, pick the one with the lowest number of users, and offer the user that project. An interactive speech option could also be used.

This is just a quick example, but a properly thought out dialogue could be developed for each function in BOINC. It could also include reports, alerts, and suggestions as to best settings of preferences for that computer, when, and when not, to run benchmarks, running them on request etc etc.

If this is technically possible, and people are interested in developing it, I would be happy to help with the dialogue part, but have no skills whatsoever on the tech side.

Also, perhaps BOINC Alpha would donate an area on the alpha message boards which can be used as a forum for such a development.

What say you?

<<<

David replied as below:-

>>>

From: David Anderson
Subject: Re: [boinc_alpha] [boinc_dev]BOINC and The Loebner Prize


Grid Republic, which will be launching soon,is a step in this direction; it provides a unified,super-simple interface for running multiple BOINC projects.

Somewhere down the line, when the credit system is extended to reward RAM, disk, and network usage,
I've been imaging a "wizard" that tella you which projects to run on which machines,in order to maximize credit per day.

The BOINC message boards: http://boinc.berkeley.edu/dev/
would be a good place to discuss this stuff.

-- David

<<<


What we could discuss here is

1/ is it technically possible to do, given the current levels of state of the art programing etc.?

2/ is this a direction BOINC should/could be going in?

3/ if 2/ above is 'Yes', then how would you visualize it, in terms of things it should do.

4/ any other thoughts, ideas or suggestions.

Knightrider.
TeamMacNN

...

---

[quoteWhat we could discuss here is

1/ is it technically possible to do, given the current levels of state of the art programing etc.?

2/ is this a direction BOINC should/could be going in?

3/ if 2/ above is 'Yes', then how would you visualize it, in terms of things it should do.

4/ any other thoughts, ideas or suggestions.[/quote]
1) It is possible. Many ways to do it. However, that is not the real problem at this time.

2) Probably. We are trying to attract more participants and to do that we need to make things easier.

3) You need to work out which problem you want to address, then work out the logic and the dialogs. For example, if it is the connection to projects ... then you need to work through the process.

For example:

a) Test CPU speed, memory, disk space
b) suggest 3 candidate projects

oops ...

a) Test CPU speed, memory, disk space
a1) Determine list of suitable projects based on: participant base, sign up allowed, work size, work time requirements and the client computers capabilities
b) suggest 3 candidate projects and explain the characteristics and the reasons they are suitable
c) etc ...

Believe it or not, this is a hard problem (one of the reasons it is not being addressed). It is difficult to work out a logic tree that makes sense ...

---

Hi Paul,

Yes, your right in your comments.

The key, or hook, to understanding this is that a computer doesn't 'Think' only that it has to appear to 'Think'. This could be said of human beings of course, in that our brains are similar to a computer, sending electrical impulses down nerves etc results in some action. But we don't want to go to far down that road.

The objective here would be to create a software that would be capable of verbal and visual interactivity with the user and then to perform some function(s} as a result of that user interchange. Therefore, rather concentate on a specific task, such as, say, asking the user if they would like to set preferences, and if so, how, and here are the options, user selects options. The options could then be set in the background on both the BOIN manager and on the server account settings without any further action by the user. All of the tasks needed to assess optimum settings would then be carried out to perform this action on the bases of the users preferences. Such a thing would no doubt have to be done with dialogue boxes, but with speech it would be, as they say, way cool.

In other words, everything that needs to be done to successfully run BOINC should be mapped out as a - 'user/boinc/' to 'function/action'- map. This could then be mapped out vertically and hoizontally and integrated where required.

This is probably to simplistic, but all I am seeking do, is to first get an overview of the shape something like this would have to take so as to work in such a way that BOINC appeared to be thinking about what it was doing.

K.

---

Good luck ... :)

If you post your work and let me know I will be glad to critique it ... can't promise more than that ...

---

HI guys! need help! can you please explain to me how turing test is being conducted and how an AI is being scored. also, are there any alternatives on how to rate an AI based on its intelligence. we currently conducting our thesis related to AI.we dont know how turing test is being scored and what criterias are being considered. thanx for the help.