DAYDREAMER: Re-introducing an old goal-based agent
DAYDREAMER: Re-introducing an old goal-based agent

By KermitDZ | ColorlessCoder | 25 Oct 2019


Recently and while I was exploring GitHub by searching for random words, I stumbled upon this interesting project...

DAYDREAMER is a rule-based and goal-oriented AI. As the name suggests, it mimics the phenomenon of daydreaming in humans. What's cool about this, is its employment of rules and algorithms that are adapted from theories of cognition, psychology, and emotion, among other things.

Though it is built on dated and maybe inaccurate understandings of the human mind, and might not be the best model/simulation out there, it nevertheless remains an interesting project. You should find it interesting if you are a fan of rule-based approaches to AI, and don't particularly 'love' recent trends in AI -- where everything nowadays is a mysterious black box that consumes absurd amounts of data and magically learns to produce "correct" results.

(I know ML is huge and important and a serious research area; I'm just not really into it.)

The project is written in Common Lisp and is open source, check it out on GitHub.

And make sure to read its non-technical conference paper (which is available on arXiv) to have a general idea of how it works.

Testing
I have tested it with Allegro CL v10 on Ubuntu 18.04 x64.

After installing the Common Lisp environment, run:

# clone or download the project's source code
git clone https://github.com/eriktmueller/daydreamer

cd daydreamer

# enter the Allegro CL REPL
alisp

# load and execute DAYDREAMER
(load "dd.cl")
 

And that's it... It will now start daydreaming:

351665157-9a6dc24bad32e639ee3b3597349f804872279f098ee2649996ce48f502518339.png

 

Thanks for reading. This was actually my first post, I'd appreciate it if you could comment and tell me what would improve this post.


KermitDZ
KermitDZ

Algerian nerd


ColorlessCoder
ColorlessCoder

about coding and stuff

Send a $0.01 microtip in crypto to the author, and earn yourself as you read!

20% to author / 80% to me.
We pay the tips from our rewards pool.