The study was done in Taiwan. Over the last year I have had a surprising number of international visitors. A high school class in Korea interacted with Skynet for a class project and recently a user has had a number of lengthy conversations in German. It is a bit surprising since Skynet-AI only speaks English (but many of the search functions can handle other languages). I have had a number of University and highschool classes in the US use Skynet-AI as part of a language, ethics, or computer science class. It helps though if they tell me about what they are doing so that I don’t change/break something in the middle of their project.

One of the advantages for people who speak English as a second language is that Skynet-AI is fairly robust to errors in grammar and simple spelling mistakes.

Skynet-AI has joined the the International Chatbot Contest.

It has been updated to version .0053.

Among the latest features is it’s attempt to display context aware Wikipedia pages and its ability to handle “relationship” questions like:

Tom is taller than Tim. Who is the tallest?

Some of the operating systems/browsers Skynet-AI has run on:

Firefox
Internet Explorer
Safari
Opera
Android Browser
Mozilla Compatible Agent
Opera Mini
Mozilla
IE with Chrome Frame
SeaMonkey
NetFront
Safari (in-app)
Playstation 3
RockMelt
BlackBerry8520
Camino
j2me
Lunascape
SonyEricssonJ108i

Some of the mobile devices:
Apple iPhone
(not set)
LG P999 T-Mobile G2x
Apple iPod
Apple iPod Touch
Acer A200 Picasso_E
Acer A500 Picasso
SonyEricsson LT15i Xperia Arc
HTC APA7373KT EVO Shift 4G
Google Nexus S Samsung Nexus S
HTC ADR6350 Droid Incredible 2
Samsung Galaxy Nexus
Samsung GT-I9100 Galaxy S II
Samsung SGH-I747 Galaxy S3
Archos
Asus TF101 Eee Pad Transformer TF101
BlackBerry 8520 Curve
Creative ZiiO 10
HTC 9292 EVO 4G
HTC ADR6400L Thunderbolt 4G
HTC Desire
HTC Desire Z
HTC G2 HTC Sappire
HTC PJ83100 One X
HTC SensationXE Beats Z715e Sensation
HTC Z710 Sensation
LG LS670 Optimus S
LG LS840 Viper 4G LTE
LG P920
Motorola DroidX
Motorola MOTWX435KT Triumph
Nokia 5230 Nuron
Nokia Lumia 920
Pandigital Planet
RIM BlackBerry 9780
Samsung GT-I9105P Galaxy SII Plus
Samsung GT-I9300 Galaxy S3
Samsung GT-I9300 Galaxy SIII
Samsung GT-N7000 Galaxy Note
Samsung GT-P7300 Galaxy Tab 8.9
Samsung GT-S5830 Galaxy Ace
Samsung SCH-S720C Galaxy Proclaim
Samsung SGH-i717 Galaxy Note
Samsung SGH-I897 Galaxy S Captivate
Samsung SGH-I997
Samsung SGH-I997 Infuse
Samsung SGH-T839 T-Mobile Sidekick 4G
Samsung SPH-D710 Galaxy SII Epic 4G Touch
Samsung SPH-M830 Galaxy Rush
Sony Tablet S
T-Mobile myTouch4G

You can add Samsung Galaxy Y to that list Ken. I just tried it on my phone.

Thanks.
Also left off the list are some of the other game consoles and web TVs, it has run on.

Skynet-AI continues to prove it is the fastest, most flexible AI.

The latest devices it has run on:
RIM Z10
Samsung SGH-T889 Galaxy Note II
Samsung SGH-T989 Galaxy SII
SonyEricsson LT26i Xperia Arc HD

Skynet-AI has entered the Chatterbox Challenge.
It has been updated to version .0054.
Included in this version is access to the “Colossus” update, a 75MB AI data set.

The Chatterbox Challenge (CBC) which began in 2001 is an annual contest for chatbots.
http://www.chatterboxchallenge.com/

OK, I’ll bite… what is ‘the “Colossus” update, a 75MB AI data set.’?

Skynet-AI has focused on speed and flexibility. The Ai interpreter and personality are only about 600k.
To keep the size small, I intentionally left out data about common nouns. This keeps the download times reasonable, but prevented answering questions like, “What is a hammer used for?”

The “Colossus” update adds a mechanism to handle the most common of these without needing to download the entire data set. This is in preparation for a 1 GB update I hope to have later this year.

I found that things like word net or some of the on-line services did not meet my needs, and of course Wikipedia is too big to integrate into a bot directly. The goal is to add the most commonly used human knowledge without the burden of all the overhead of things never used in normal conversation.

Merlin - Jun 9, 2013:

The goal is to add the most commonly used human knowledge without the burden of all the overhead of things never used in normal conversation.

Why not just use Wikipedia then?  Any good AI is going to have to be internet enabled to be truely useful and “current”.  What is the point in hard coding common knowledge when it can be queried from a Wiki, where is it constantly updated/edited? Or do you have a different meaning of “common knowledge” than what I am thinking of (which is, as a just descibed, basically Wikipedia). It seems an acurate and precise NLP system is, at its core, what any really good info bot needs.

Merlin, besides for offline purposes why would you want to have the client download the entire data set when only a small percentage of it would ever be used anyways in a single session. I understand wanting your ai to work on as many devices as possible which every developer should take into consideration, however almost all newer devices have access to the web which to me when working with a large database seems like it would be more practical to make use of calls to a remote database. Having access to the most common things is a good start, but to me, those things don’t make a conversation interesting. In actual conversations, what is being discussed can change numerous times traversing through a wide range of topics. Having information readily on all of those things isn’t always going to be possible so I have to agree with Carl that it is a must for a chatbot to have access to the web for information and recent events.

Merlin - Jun 9, 2013:

I found that things like word net or some of the on-line services did not meet my needs.

Without divulging any Skynet secrets… That’s certainly not what this question is about. May you kindly elaborate in what ways wordnet may not have met your needs in general? Feel free to discuss this topic in terms of a rudimentary chatbot. Can you make any comments about the Synonyms, Hypernyms and Hyponyms, Parts of Speech, and other wordnet features?

Thank you.

Princeton University “About WordNet.” WordNet. Princeton University. 2010. <http://wordnet.princeton.edu>

Carl B - Jun 9, 2013:

Merlin - Jun 9, 2013:

The goal is to add the most commonly used human knowledge without the burden of all the overhead of things never used in normal conversation.

Why not just use Wikipedia then?  Any good AI is going to have to be internet enabled to be truely useful and “current”.

Carl, in many ways I agree. Skynet does use wikipedia (I think I added that functionality in version .005) and if I may say so I think it is doing a good job displaying results that are relevant. Sometimes the serendipity factor (displaying an unexpected wiki page) can be quite amusing.

Carl B - Jun 9, 2013:

What is the point in hard coding common knowledge when it can be queried from a Wiki, where is it constantly updated/edited? Or do you have a different meaning of “common knowledge” than what I am thinking of (which is, as a just descibed, basically Wikipedia). It seems an acurate and precise NLP system is, at its core, what any really good info bot needs.

I think I have proven over about a decade of integrating web search engines/services/pages that this is the preferred way to go to answer “factoid questions”.  I have employed “cascading search engines”, if one search fails, a fallback search takes place. Different search engines provide different quality of results based on the query. I now direct different queries to the best available services. Most of the advanced chatbots are now starting to adopt this practice.

One practical reason to hard code common knowledge is to score better in chatbot competitions. In the latest competitions, organizers are giving higher scores to “embedded knowledge”. Although I disagree with this, I am not the one who writes the contest rules.

More importantly, I believe that by taking knowledge and formatting/modifying the data I can make it much more “bot friendly”. At minimum, the bot can access the data faster. Also, it may allow me to dynamically download topic data and store it on the user device on an as needed basis.

After 6 months of displaying wiki pages integrated into the conversations, users are now asking the bot to parse the wiki page and integrate the information into the conversation. I have started looking into this, but I have yet to convince myself that I can do this on the fly during the wiki page download. Currently I am more optimistic that I can preprocess wiki data for later use. Taking that step off-line allows me to retain the speed of the bot.

Chad J - Jun 9, 2013:

Merlin, besides for offline purposes why would you want to have the client download the entire data set when only a small percentage of it would ever be used anyways in a single session. I understand wanting your ai to work on as many devices as possible which every developer should take into consideration, however almost all newer devices have access to the web which to me when working with a large database seems like it would be more practical to make use of calls to a remote database. Having access to the most common things is a good start, but to me, those things don’t make a conversation interesting. In actual conversations, what is being discussed can change numerous times traversing through a wide range of topics. Having information readily on all of those things isn’t always going to be possible so I have to agree with Carl that it is a must for a chatbot to have access to the web for information and recent events.

Chad,
Unlike virtually all other bots, Skynet-Ai’s AI interpreter is always downloaded an runs in the target device. This allows the AI to be disconnected from the net after start-up. It also allows speed unattainable in client/server systems. Average response time for Skynet-AI is <10ms. This is less than it would take to round trip to a normal web host (without processing). It also eliminates virtually all overhead on the server, allowing Skynet-AI to support an almost unlimited number of users with no drain on the host server.

I don’t anticipate downloading the entire data set (although there is a case to be made for a 1 time set-up if I made the bot into something like an APP). More likely (and what is currently implemented) are micro downloads (<1k to less than 200k) where the results can be cached on the user machine. The speed of these seem reasonable so far.

In the future, I plan on using this method to enable on-the-fly topic/conversation sets.

8PLA • NET - Jun 10, 2013:

Merlin - Jun 9, 2013:

I found that things like word net or some of the on-line services did not meet my needs.

Without divulging any Skynet secrets… That’s certainly not what this question is about. May you kindly elaborate in what ways wordnet may not have met your needs in general? Feel free to discuss this topic in terms of a rudimentary chatbot. Can you make any comments about the Synonyms, Hypernyms and Hyponyms, Parts of Speech, and other wordnet features?

Thank you.

Princeton University “About WordNet.” WordNet. Princeton University. 2010. <http://wordnet.princeton.edu>

In many ways, wordnet is overkill for what I am doing. WordNet’s 117 000 synsets is larger than the entire Skynet-AI brain. I don’t try to identify parts of speech or specific word meanings. What JAIL (Javascript Artificial Intelligence Language) is built for is to be fast, small, and to be able to extract the “meaning” of a sentence.

If I look at the word net listing for “hurt”, most of the data is superfluous to creating a response. The interesting pieces are the synonyms, but they can be put in a simple list.
http://wordnetweb.princeton.edu/perl/webwn?s=hurt&sub=Search+WordNet&o2;=&o0=1&o8=1&o1=1&o7;=&o5;=&o9;=&o6;=&o3;=&o4;=&h=0

Although Wordnet has on-line interfaces, they are to slow for my needs and I can do better with search engine services. I have a different conceptual organization structure than word net (noun hierarchies), and as such tend to “roll my own” data structures to meet my needs.