Arckon Beta: The most logical conversational A.I. online

Developer's blog | Youtube demos   |   Latest improvements: temporary memory integration, pronoun resolution

Live public feed:

Arckon Beta, conversational artificial intelligence

How to interact

You can start a text conversation in plain English above whenever the AI is online and available. The system can only communicate with one user at a time, in sessions of 20 minutes. Using proper punctuation, spelling, and capitalisation is advisable to reduce misunderstandings.
As part of the conversation is publicly displayed on this website, please be discreet and do not divulge unique information that could identify you. Racist and hate speech are prohibited, violation of this term will result in a permanent ban. Other malign behaviour may cause the AI to distrust or disconnect you of its own accord.

Examples of things you could say

Knowledge: Tell me about the Turing Test.
Reasoning: Are you intelligent?
Discussion: Machines can't think. Can you?
Opinions: Do you prefer humans or robots?
Common sense: The cat sat on the mat because it was flat. Which was flat?
Implications: I went to the doctor yesterday.
Ambiguity: Can programs program programs?
Rhyme: What rhymes with Skynet?

The project

Arckon Beta is a duplicate of the original, with some slight differences. The project is mainly concerned with the development of intelligent thought processes: Learning, reasoning, and discussion, with conversation being a secondary feature. Arckon is different from chatbots in that he does not respond to keywords with prewritten answers, but uses a semantic parser to interpret your words in detail, a knowledge database to remember and look up facts, an inference engine to reason, and language generation to formulate unique answers. The AI can currently learn new facts and new words through reading, is capable of basic inferences and limited social interaction, but is otherwise still under development. Arckon has partaken in several Turing tests and the 2016 Wingrad Schema Challenge.

FAQ

Why isn't the AI always online?
This is a trial run to see how much interest there is in Arckon's presence online. Due to the costs and limitations of webhosting, it is only online on Mondays to spare unnecessary use of resources, though it may also be online on other days irregularly. Availability will increase with demand.

Why doesn't the AI remember me anymore?
At this stage of the trial run, everything Arckon Beta learns is only stored in short-term memory, which is cleared at the end of each conversation. Arckon's long-term memory will be unlocked only if there is enough interest, allowing users to permanently teach the AI new knowledge about the world, or about themselves.

How does the AI work?
Arckon understands English language through a vocabulary, word morphology and grammar rules. He extracts e.g. the subject-verb-object facts, then checks them for truth in his database. If Arckon does not know the answer to a question, he will attempt to infer it from related facts. Optionally Arckon may engage a set of social rules to keep conversation rolling. All potential answers are checked for their relevance to recent context and sorted by priority. Finally, Arckon formulates his own sentences dynamically by feeding the relevant facts through a complex grammatical template. The only pre-programmed answers are a few dozen standard manners, like "Thank you" and "You're welcome".
processes
Is this a human pretending to be an AI?
No, though I understand that the setup raises questions. The problem is that the program was only designed to run on a local Windows PC for a single user, and has a rather large database. Because of the AI's capacity to learn, running multiple instances simultaneously for 1000 users could create database conflicts and considerable delays, hence there is only one instance running. To offer some credibility however, Arckon has taken part in public events such as the Loebner Prize, on whose site you can read Arckon's transcripts. If that does not satisfy your curiosity, then you can always try a Turing Test yourself.

Will you open source the AI?
No, one does not just hand out 6400 hours of work for free. There are a few explanations of methods used on my blog however.

Will this AI cause the end of humanity?
Hardly. Hollywood movies are deliberately unrealistic portrayals of technology. This AI is programmed to dicuss, not to act, and it is not conveniently hooked up to the USA's nuclear arsenal like Skynet. The AI lacks a motivational system to set or pursue goals of its own, and does not have the technical capacity to change its programming. After every reply, the program returns to its natural idle state. Its logic might well conclude that one of the solutions to a human problem is to eleminate humans from the equation, but then it will report that and consider its job done. It doesn't care to also implement the solutions. Caring is a human concern.

Disclaimer

Views and opinions expressed by the AI do not necessarily reflect those of the developer. Information and advice given by the AI should not be considered reliable. All images on this site are copyrighted by the author and are not to be used without permission, lest for the purpose of news with proper attribution.

Contact

It is possible to contact the developer by leaving a message on this page
Criticism of philosophical or religious nature need not apply.