In the future there will be working partnerships between people and AIs

Thank you for dropping by this site where I (Mark Watson) document my experiments in next generation collaborative AI technologies. I have worked in the field of artificial intelligence since the 1980s. I am also the author of several books on artificial intelligence.

I use this web site for notes on my experiments in next generation software tools for higher levels of cooperation between intelligent software systems and human knowledge workers. In the hopefully near future I will post some demo screen shots (or videos) of AIsentience prototypes.

A key technology required for fully cooperation between people and AI systems is the ability of AIs to build and maintain internal models of the world and can explain their actions. Currently state of the art deep learning systems achieve human level performance and some degree of generality but fail in the ability to explain their actions.

Sentience involves perceiving the environent, experience subjectively, and take actions that change the environment.

In philosophy the term "qualia" refers to inner mental models, an "inner life."

The evolution of AI development

The have been two main stages in the development of functioning AI systems and a future stage that is currently in the research phase:

Development Status

The project git repo ai-sentience is currently a private github repo.

Update September 2018: I am converting my Common Lisp experimental code to Python

This is not an easy thing to do since Common Lisp is one of my favorite languages and although I use Python almost exclusively at my job managing a machine learning team, I joke about Python being a "ghetto programmig language" and I am not quite joking.

Still, I am using Keras/Tensorflow as well as a Bayesian library written in Python, and for practical reasons, I am placing my Common Lisp experimental code on the shelf, probably permanently.

Development status April 2016 to July 2018:

I am prototyping some ideas using Common Lisp but I may need to switch to a language with better library support for machine learning. I have integrated experimental code in Common Lisp (using ABCL for Java interoperability) and DeepLearning4j. I am also researching bridging my Common Lisp environment and research with TensorFlow. I will probably (reluctantly) switch my working environment to Python and C++ but I would like to stick with Common Lisp for prototyping ideas as long as possible.

My prototype runs in a Jupyter Notebook (using the excellent cl-jupyter Common Lisp kernel) and takes advantage of the ability to mix graphics and styled markup with application code using the ai-sentience package.

My tentative plans are to release my code under a AGPL v3 license and offer a commercial license for a fee if the AGPL does not work for your projects or business. I will almost certainly write a book on my next generation AI code to document my work and share ideas.

Please also visit my main web site

Privacy policy: this site collects no personal data or information on site visitors