The Real AI Institute is a
nonprofit organization, dedicated to the creation of the real AI and its
dissemination and development for the benefit of all living beings.
What do we mean by " real" AI?
Not just software that displays intelligent behavior in one particular domain like
chess or financial prediction. This kind of software is
fascinating, but it doesn't meet our criteria of intelligence.
Our goal is a generally, flexibly
intelligent software system. One with a sense of its own self and
its own digital embodiment in the world, with its own goals, feelings
and motivations. One that learns experientially based on its
interactions with humans and other computer programs.
No such software program
exists. Our goal at the Real AI Institute is to create one, and
allow it to interact freely with Internet-using humans like
yourself.
One might think there
currently existed similar R&D projects within major corporate
research labs, or big-name academic institutions. One would be
wrong. By all appearances, we are the only group in the world
currently pushing hard toward a fully-functional, autonomous digital
intelligence, based on a detailed design describing how all the aspects
of intelligence can be realized in software in an integrated
way.
At first our digital mind
will be a baby, but it will learn through interacting with us, and even
in the beginning it will have some valuable insights, due to its
different way of thinking. We will feed it many kinds of
information, and see which types are more agreeable to its evolving
mind. So far we've been playing with financial and biological
data, and text describing this data. Already our system's various
components demonstrate formidable abilities.
The Real AI Institute is a
new organization, but the project it embodies is not new -- the Real AI
Institute team has been together for 3 years, within the AI Development
and Research Divisions of Webmind Inc. (originally named Intelligenesis
Corp.). When Webmind Inc. dissolved on April 1, 2001, the
consensus among the firm's core AI scientists and engineers was clear:
we had come too close to the end goal to give up the quest, company or
no company. A group of us resolved to keep the real AI aspect of
Webmind Inc.'s work alive -- and thus the Real AI Institute was
born. Webmind Inc.'s AI staff numbered 45 at maximum; the Real AI
Institute diehards number less than half that. But what we lack in
numbers, we make up in confidence, dedication, insight and sweat.
Within Webmind Inc. we
produced 750,000 lines of Java code embodying a first attempt at a real
AI system, the Webmind AI Engine. We got perhaps 3/4 of the way
toward the end goal of a real AI system, but we found ourselves bogged
down in the kind of technical problems that confront many large software
projects. Now, taking everything we learned from this experience,
we are starting over, and building a leaner, meaner real AI, in a
combination of C and Java. After 2 months of effort, we have
replicated about 1/3 of the core AI portions of the Webmind AI Engine,
in many fewer lines of code and using a vastly more efficient software
architecture. Since Webmind Inc. is no more, the name
"Webmind" is free, and we now call our real AI system simply
Webmind, as was done with prototype versions of the system in 1996
before Intelligenesis Corp. was formed. Webmind Inc. is gone, but
Webmind is cruising along nicely.
By the end of March 2002,
we believe, we will be able to launch on this website a baby digital
mind, our first baby Webmind. Only one thing can stop us: we all
need to eat, and some of us have families to feed as well. This is
why, as mentioned above (we never claimed to be subtle!), the Real AI
Institute is seeking donations. We
need these badly. We hope to raise $500,000 by the end of August,
2001.
This initial $500,000 will
cover living expenses for a year, for the minimum core team needed to
get the job done: 2 scientists based in the US (Ben Goertzel and Pei
Wang), and 12 software engineers and AI scientists based in Belo
Horizonte, Brazil (where the largest Webmind Inc. office was
located). Of course, a bigger team than this would be wonderful,
if donations exceed our expectations. And if donations fall short,
well, a team of 3 would be better than a team of 0. But we believe
that a team of 15 or so is about the minimum needed for smooth progress
toward our goal. We will initiate another fundraising round toward
the beginning of 2002, aimed at raising funds to support us through
early 2003.
We don't believe that real AI should be
owned by any commercial organization. But this is not to say it shouldn't be
commercialized. To fund our work, in addition to seeking
donations,
we are interested in developing co-development relationships with
commercial firms operating in particular market niches. However,
we caution that the
commercial applications of a real AI should not be confused with the
essence of the thing. An AI, just like a person, can work for a
corporation and make money for itself and for the corporation. But
a person is not his job and a real AI is not its commercial
application. The Real AI Institute exists to do what no
profit-focused organization can do: focus on bringing real AI to the
world at large, in the service of everyone, rather than in the service
of the shareholders of some particular corporation.
Although we believe we have
our project well under control, we are always eager for engineering,
design or conceptual assistance from relevantly skilled and
knowledgeable individuals. If you're a computer science,
mathematics or cognitive science expert and are interested in
devoting some of your spare time to helping us create real AI, please
click here.
Finally, as our schedules
permit, we will publish more and more of the concepts underlying our
work. For now, a brief summary of some of our
work, focusing on our long-term goals but also touching on some
nitty-gritty software architecture and AI theory issues, can be found in
the article Webmind: A Digital Intelligence in the Making.