(if the above link does not display right, you can try another browser: http://www.pandorabots.com/pandora/talk?botid=c0f4cd29ee374584 )
The best way to understand chatbots and AI is to interact with a chatbot. Above is a chatbot (an AI) named "Alice" and she is "the hostess with the mostest." She is based on the "ALICE" chatbot, winner of several AI awards.
Let me know if you like her. You can type in questions and she will answer. We are always trying out new ideas in robotics and seeing which ones work and which ones don't work.
-------------------------------------------------------------
If you see an empty rectangle, try a different web browser
-------------------------------------------------------------
AI in the form of chatbots started with the psychotherapist program called Eliza.
Chatbots and AI: Go to the ALICE Site
Chatbots and AI. My first few seconds with Chatbots and AI was with Jabberwacky and seemed promising. Its replies gave just enough sarcasm for me to actually question its humanity. At one point it even told me it was human and I was the software, and I almost believed it. But just as quickly, Jabberwacky’s replies completely made no sense, and it soon made up some excuse about needing to go to bed. So much for intelligence. Or was I really that boring?
A chatterbot or chatbot, also called Artificial Conversational Entities, talk bots or chatterboxes, is a computer program designed to simulate an intelligent conversation with human users. It does this with the use of auditory or textual methods that are intended for small talk. Judging from the course of my conversations, very small talk, it seems. Some chatbots use sophisticated natural language processing, a field of computer science and linguistics concerned with the interactions between computers and human natural languages, but many simply scan for key words within the input and pull a reply that match these keywords, or use the most similar wording pattern, from a textual database, often resulting in deflected answers or repeated questions.
Nevertheless, the Chatbots and AI's primary aim is really to fool the user into thinking that the program's output has been produced by a human, which it is able to do at very certain and precise moments. In 1950, Alan Turing published his famous article “Computing Machinery and Intelligence,” which introduced what is now referred to as the Turing test as a criterion of intelligence. This criterion depends on how realistically a computer program can impersonate a human by means of a real-time written conversation. A human judge should be unable to reliably distinguish, based on the conversation alone, whether it was produced by a program or a real human.
The Turing test has influenced the world of AI greatly. The Loebner Prize is an annual competition in artificial intelligence based on the concept. Every year, prizes are awarded to the chatbot considered by the judges to be the most human-like. In each round, a human judge simultaneously holds textual conversations with a computer program and a human being via computer. Based upon the responses, the judge must decide which is which. Hugh Loebner began the contest in 1990 in conjunction with the Cambridge Center for Behavioral Studies, Massachusetts, United States.
The chatbots and AI test also stimulated interest in Joseph Weizenbaum's program ELIZA, published in 1966, which seemed to succeed in fooling users to believe that they were conversing with a real human. Weizenbaum himself, however, did not claim that ELIZA was genuinely intelligent. ELIZA's key method of operation, which has been copied by chatbot designers ever since, involves the recognition of cue words or phrases in the input, and the subsequent output of corresponding pre-prepared or pre-programmed responses. Because this allows the conversation to move forward in what seemed to be a meaningful way, an illusion of understanding is generated, even though the processing involved has been merely superficial.
The program ELIZA has shown that it is actually quite easy for people to buy into this illusion, because human judges are so ready to believe in responses that are capable of being interpreted as “intelligent.” This works because such programs produce responses which are intentionally vague and general so that they can be interpreted as “intelligent" in a wide range of conversational contexts. This is what normally characterizes a program as a chatbot rather than as a serious natural language processing system. Still chatbots and AI have a long way to go.
Aside from participating in contests and fooling people in chatrooms, chatbots and AI are used in integration with dialog systems for various practical purposes such as online help, personalized services, or information acquisition. The fact is that a lot of humans actually exhibit a readiness to interpret computer output as genuinely conversational, even when the conversation is based on superficial and rather simple pattern-matching. Most people prefer to engage with programs that are human-like, hence chatbot-type programs are actually quite useful, mostly in interactive systems that need to elicit straightforward information from users, like online help systems, for example.
One of the more famous chatbots today is A.L.I.C.E., which utilizes a programming language called AIML, which is specific to its function as a conversational agent. It has since been adopted by various other developers, which call themselves Alicebots. One of these, the Pandorabots, is an experimental software robot hosting service where one can create his/ her own virtual personality based on the A.L.I.C.E./AIML free software community. Despite its popularity, A.L.I.C.E. is still only based on simple pattern-matching techniques that do not possess any reasoning capabilities, the same technique ELIZA used back in 1966. This is not strong AI, which would require sapience and logical reasoning abilities.
Some more recent chatbots also combine real-time learning with evolutionary algorithms which optimize their ability to communicate based on each unique conversation. Jabberwacky, I discovered, learns new responses and context based on real-time user interactions, rather than responses derived from a static database. Maybe that explains this excerpt from the conversation we had:
Me: Is it boring talking to strangers all day?
Jabberwacky: Not as boring as talking to a computer.
Maybe that’s why it fell asleep.