What is Uberbot?
In a nutshell, the Uberbot is an AGI (Artificial General Intelligence) program that lets
you talk to your computer. Uberbot stores concepts (or ideas) in its computer brain. You talk to
Uberbot by typing in text, then it considers its best response and replies by text. There is also a
text to speech convertor allowing spoken output.
What questions does Uberbot understand?
Uberbot understands simple questions about maths, dates and times, properties of common objects,
geography, spelling, and many other types of things. Uberbot remembers things you tell it, such as
your name, and has its own preferences. For example Uberbot can answer questions like these:
- Two hundred plus two hundred = what?
- What is the derivative of x squared?
- What is tomorrow's date?
- Is an apple bigger than a grape?
- What color is a frog?
- Name something that is red
- What is the capital of Norway?
- My name is Fred. Do you know my name?
- What is your favourite book?
Can Uberbot (or any computer) think?
This is a very complex question that touches many areas, such as the nature of thought,
consciousness, awareness of self, and whether there can be different levels of consciousness. A lot
of discussion has taken place on these questions, including ideas from learned professors and
academics. At the present time no-one knows whether thinking machines will be possible, nor do we
understand the nature of consciousness or thought.
The question 'Can computers think' is not new. Around 1950 Alan Turing, a British mathematician
and code-breaker who was involved in the design of the first computers, asked the same question.
Turing said it was impossible to define 'thinking' in a clear manner, so he proposed a different
way of answering the question. Instead of asking 'Can computers think', he proposed a test now
known as the 'Turing Test'.
What is the Turing Test?
The Turing Test was invented by Alan
Turing around 1950. It measures the ability of a computer to think, by comparing its conversational
intelligence to that of a person. The way it works is a Judge uses a text interface to 'talk' with
a computer and with a person, without knowing which is which. If the Judge cannot tell which is the
computer and which is the person, the computer's conversational intelligence is considered similar
to a person's. The beauty of the Turing Test is, it's easy to understand and it's easy to run a
test. Also it avoids deeply complicated questions such as the nature of thought.
Some computers have passed simple Turing Tests where the conversation is limited to five
minutes, or is limited to a particular topic. But no computers have yet passed a thorough version
of the Turing Test.
What is the Loebner Prize?
Every year, the Loebner Prize runs a
live Turing Test. This works by setting up a Turing Test between four computers and four people,
with four judges taking turns in the conversations. During the test each judge talks to one
computer and one person at a time, and has to tell the two apart. At the end the scores are
assessed and a prize is awarded for the best computer. The Loebner Prize has been held annually
since 1991, but so far none of the computers have fooled the judges.
Has Uberbot been tested?
Yes. Uberbot has been submitted to the 2013 Loebner Prize where it came 5th, and to the 2014
Loebner Prize where it came 3rd.
How does Uberbot work?
Uberbot uses a technique called "computational behaviourism". The principle is user input
(stimulus) triggers a search for the best response across a large number of stimulus / response
pairs. Responses can trigger other searches while answering the user input. A stimulus / response
pair is treated as a single 'concept' by the Uberbot, and concepts have the characteristic that the
same response can be triggered by multiple stimuli. The full set of stimulus / response pairs
constitutes Uberbot's computer brain.
In terms of the underlying technology, Uberbot is an AGI (Artificial General Intelligence)
application that runs on a Windows PC. The core elements are: (1) definitions of concepts provided
as XML and Excel spreadsheets (2) an AGI engine using SQL and XQuery that stores concepts and
searches for the best response to input stimulus (3) a client program written in Ubercode to handle
interaction with the AGI engine and the asking of questions and display of responses and (4) a text
to speech engine for spoken output. The client program is able to handle interactive conversations
with its users, also to handle the protocol needed for the Loebner Prize.
Interestingly there is a basic similarity to a biological brain. In the Uberbot, concepts can be
activated by multiple stimuli, and concepts have a single response which can potentially trigger
other concepts. When the complete 'tree' of concepts and sub-concepts is played out, the composite
response is returned as the 'answer' to the original input. In a biological brain, neurons can be
activated by multiple dendrites, and when activated generate an 'output' signal on their single
axon. The branches of the axon potentially connect to other neurons via their dendrites, producing
a cascade of neuron activation.
How are Uberbot and Ubercode related?
The core Uberbot is an AGI engine for answering questions, and it uses an Ubercode client
program to provide an interface to handle questions and show the response. The Ubercode client
program also handles communication using the Loebner Prize Protocol (LPP) as required for entry to
the Loebner Prize, and handles text to speech conversion when required.
How do I get hold of the Uberbot?
Evaluation copies are available on request, by email to info@ubercode.com
What does AI and AGI mean?
AI stands for "Artificial Intelligence" and AGI stands for "Artificial General Intelligence". In
the earlier years of computing, AI refers to attempts to make "intelligent computers". Eventually
it was realized this attempt was not well defined and was very difficult. AI now refers to a very
wide field covering many areas of potential computer intelligence, such as computer vision,
character recognition, representation and manipulation of ideas, robotics, text to speech, speech
to text, self navigating vehicles, self organizing systems, and domain optimized applications such
as chess playing programs and knowledge base systems.
As AI now encompasses a wide range of areas, the quest for 'thinking computers' has receeded
into the background. This is partly because the field of AI has expanded, also the 'thinking
computer' is simple in concept but has proven impossible to realize. Thus 'AGI' now refers to any
kind of system that can cover multiple areas, such as conversational computers as entered to the
Loebner Prize.
|