Re: AI2

> On 21 Feb 2015, at 00:54, Danny Ayers <danny.ayers@gmail.com> wrote:
> 
> what if human intelligence is a local minimum

It is.

I’ve been looking at how to blend cognitive science, AI and statistical learning, and a striking thing about human intelligence is the way our consciousness is highly constrained. We are only able to focus on a few things at any particular time. This is compensated for by the distributed parallel processing in our sensory and motor systems.  It is like using an 8 bit computer for the rule engine run by the central processor and a large number of graphics accelerator chips for the I/O and memory systems.

It is easy to imagine an AI that can focus on many things at the same time in a concerted way, analogous to a closely knit team of people.  Another thing is the variation in creativity, memory and other traits across people.  We could churn out AIs with optimal traits for their intended purpose.**

This isn’t science fiction as I have already started to work on it, although I don’t expect it to bear fruit quickly. The first part is to implement a cognitive architecture based upon work by John R. Anderson and Marvin Minsky.  The second part is to create an evolving taxonomy of common sense and use it to create lesson plans for teaching AIs via natural language. For more details see: 

     http://www.w3.org/2014/10/29-dsr-wot.pdf <http://www.w3.org/2014/10/29-dsr-wot.pdf>

Researchers on AI and the Semantic Web have been dazzled by logic and have largely ignored cognitive science. This will change as the power of the synthesis becomes self evident. In some ways it is surprising how long it has taken us to get to this point!

** Unlike Hawking, I am optimistic about this. If we are to address the challenges of climate change, resource depletion and over population, then AIs and the Web of Thought will help us to maintain a high standard of living.  We will create AIs with empathy, but will need to keep a tight check on those who want to pervert this and weaponize AIs!

—
   Dave Raggett <dsr@w3.org <mailto:dsr@w3.org>>