Slashdot Mirror

Daydreamer is a computer model of the stream of thought I developed at UCLA from 1983 to 1988. It implements the following:

* daydreaming goals: strategies for what to think about
* emotional control of thought: triggering and direction of processing by emotions
* hierarchical planning: achieving a goal by breaking it down into subgoals
* analogical planning (chunking): storing successful plans and adapting them to future problems
* episode indexing and retrieval: mechanisms for indexing and retrieval of cases
* serendipity detection and application: a mechanism for recognizing and exploiting accidental relationships among problems
* action mutation: a strategy for generating new possibilities when the system is stuck

... would perhaps be to try to redirect the money to The Internet Movie Project ...

Begin Quote
Our dream is to create a movie with the POV-Ray raytracer,
as a collaborative effort of many people from all over the world,
just for the fun of it, "because it can be done" -
very much like POV-Ray itself is developed.
End Quote

or a similar instance ...

Yes, I am daydreaming.

CC.

Obviously this is a developing field. The best models seem to use phrases from the original text, anyway the Mac OSX example above shows that it is useful to users willing to take it with a massive grain of salt, even if we are not into full computational sentience yet.

When it works even a little better it will replace all those awful grade school teachers who assign paraphrasing as a homework assignment. The reporters who might have been replaced by it will have already lost their jobs, except for the ones in AhaIndia of course who will paraphrase for the rest of us, usually at a marginally better level than the machine.

The research is interesting - and I'd like to understand Barzilay's notation is that APL or calculus of statement? - in the paper (pdf) I found on google. Also see the papers on her site.

Of course structured text is easier, and news stories are known to have most of the meat in the beginning, but this is great stuff.

One interesting older system is ThoughtTreasure which was built to understand a story and answer questions about it. The author also did work on news analysis ("NewsForms") too. There are tools out there, I've been making a survey myself too. If anyone has information about practical NLP tools for real world tasks please post.

The AI and cognitive science fields already have such a large body of published theories and experimental work that I think this guy has basically wasted his money getting himself a vanity patent, and demonstrated his own deep level of ignorance about the whole field in the process. The first time he tries to collect his millions of dollars he's going to discover what's lurking in a field of study with hordes of earnest researchers and a 50 year history.

So I'm not worried about him and his patent, it will blow away with the first little breeze of reality, but I am profoundly disturbed about a U.S. Patent Office which hands out BS like this to anyone with a filing fee and the right format for the paperwork. Now, that's the real travesty here.

Russian puppets - forgot the name
Babushkas. If you want some, there's always Google.

Consider this: you save your spreadsheet today as "Yearly Report 2002", and two days later you want to call it back your mind just doesn't say "Yearly Report 2002", but more like "Financial Data last year". Then your nice database-filesystem won't find it either. Unless there is some serious AI backing it.
Now that would be an interesting file storage abstraction. I've played with the idea of a relational file structure, that would enable one to save meta-information on a file and later find it by information that relates to it. Implemented correctly, you could save your "Yearly Report 2001" and later find it by asking for "financial data two years ago". Something that combines newdocms and ThoughtTreasure.

ThoughtTreasureTM is a relational information storage handler combined with a (semi-)intelligent AI. You can supply information like "Peter loves Paul" and "Paul hates Cahtrine." You can then ask questions like "Who does Peter like?" and "What relationship are there between Paul and Cahtrine?" If you say stuff like "Peter dislikes Paul" it complains like "But I thought Peter loved Paul." But it goes far further than that. You can have it parse a movie review, and ask about information about the movie "Who directed Pulp Fiction? Who starred it?"

Combined with a file storage solution, this would open quite interesting, new forms of computer file storage.

There's a huge amount of open-source NLP resources and software for many languages on the web.

• Thought Treasure - a bilingual database of 25000 concepts including 55000 English and French words and phrases. Compiles under Linux and Windows

• Leeds University NLP Research Group (mail webmaster re broken links to software)

• Japan's ICOT/5th Generation Computer Project archive of free software

• Linguistics Toolset at Vaasa University

Last but not least:

• A well-annotated collection of NLP links including NLP, NLU, Speech*, MLT, Fuzzy*, MLPs, SVMs, etc

Will.

ThoughtTreasure understands questions and gives answers. Uses assortment of methods to analyze text, understand problem, apply common sense, and find answer to problem.

CLIPS rule and object expert system tool.

Knowledge Server Toolkit is a Perl-based system for monitoring and acting on continuous information flows, such as alerting when telemetry indicates unusual conditions.

An AI would pass the Turing Test, and I'm quite certain this won't. This is a project specific system that has been designed to operate a spacecraft.

Oh, you prefer a Linux program that can answer common sense questions?