Re: Knowledge Representation (KR) must be the core

Hi Carl,

A slightly different perspective is where knowledge engineering becomes a collaborative effort between humans and machines.  The human partner is concerned with use cases, curation and scalability. The machine partner deals with the knowledge representation, versioning and ensuring that rulesets are updated to match changes to the ontologies, as a basis for satisfying inevitable demands for ongoing support for both new and old applications.

This changes the focus to the representations used for the human-machine collaboration, freeing the internal machine representation of knowledge to better suit advances in AI.  LLMs have demonstrated that self-guided machine learning is orders of magnitude better than hand crafting knowledge. This is only just the beginning and we can look forward to major improvements in neural network architectures and training techniques.  You can get a feeling for what I am talking about in my recent invited lecture for the University of Bath’s AI Group, see:

 https://www.w3.org/2023/10/10-Raggett-AI.pdf

Collaborative knowledge engineering will be very permissive in respect to representations, e.g. natural language, mathematical expressions, diagrams, pictures, tables, spreadsheets, databases and so forth. AI systems will understand these in a very similar way to how we do.  This suggests that the premise that "Knowledge Representation (KR) must be the core of future AI systems” is flawed and needs unpacking.  What kinds of AI systems are we talking about?  How would this be effected by the emergence of AGI?

Best regards,
     Dave

> On 15 Oct 2023, at 15:22, carl mattocks <carlmattocks@gmail.com> wrote:
> 
> 
> Given that there more types of AIKR I believe our members could help people identify the differences and provide some simple rules on usage e.g. Reasoning supported
> 
> As an explainer this article is focused on " why Knowledge Representation (KR) must be the core of any cost-effective long-term AI strategy  "  and suggests that they need to be "models advise us on what actions to take"   https://dmccreary.medium.com/the-jellyfish-and-the-flatworm-bdad78e6f68b 
> 
> Given this group has already created a document focused on key elements of  AI Strategy .. I would be happy to schedule a series of meeting to expand it towards "
> identify the Knowledge Representation differences and provide some simple rules on usage
> enjoy
> 
> Carl Mattocks
> 
> It was a pleasure to clarify
> 
>> 

Dave Raggett <dsr@w3.org>