Re: First Deliverable, Outline and Draft

Hi Paola,

Are these colleges ignoring well known AI text books, e.g. Russell & Norvig "Artificial Intelligence: A Modern Approach"?  I understand that a new edition is about to be released. Then there are books focusing on knowledge engineering, e.g. Kendal & Green “An introduction to knowledge engineering”, and Liebowitz “Knowledge management: learning from knowledge engineering”.

Many of these books are rather expensive, so an alternative is to look at University courses where these are available online. Search for terms like “knowledge engineering lecture”. It is easy to find lecture notes, slides and videos.

If colleges are ignoring existing resources, why would they take note of a W3C CG Report?

I fully expect that AI will change dramatically over the next few years so that much of the existing courseware will become obsolete. We will see a merging of symbolic and sub-symbolic techniques based upon insights from the cognitive sciences. In addition, deep learning for single tasks will be replaced by modern evolutionary based approaches that cover a broad range of capabilities and which integrate into knowledge based approaches, replicating and improving on human performance.

See the talk I presented to the AIOTI semantic interoperability group this week:

 https://www.w3.org/Data/demos/chunks/chunks-20191205.pdf <https://www.w3.org/Data/demos/chunks/chunks-20191205.pdf>

> On 6 Dec 2019, at 02:02, Paola Di Maio <paola.dimaio@gmail.com> wrote:
> 
> Following on the opportunity to address the AI ALLIANCE, I have requested to the EU some data on t member's  profiles to help us scope this doc, although I envisage - and hope - that this deliverable will
> be a general set of guidelines for others as well.  I ll share this info when it arrives.
> 
> In my work I specifically target the AI research community in colleges, where future generations of AI engineers are formed by instructors who teach mostly statistical methods with nothing in between, however,the deliverable could become a useful educational tool for many others, especially individuals and organisations coming to AI riding the latest wave following a trend, and who have not had exposure to the long history of risks and failures nor to intelligent systems development,  I personally think that leaving AI in the hands of statistical analysis folks is a risk. we need to bring knowledge modelling to the statistical table.
> There are a lot of people and orgs out there talking AI who have never heard of KR 
> 
> I have started a page on the wiki,  with a very indicative outline, feel free to modify it and populate it
> https://www.w3.org/community/aikr/wiki/AI_KR_Guidelines <https://www.w3.org/community/aikr/wiki/AI_KR_Guidelines>  
> 
> Members please claim a section a paragraph or a sentence by entering text, and please do add your initials
> to the text you enter so that we can keep track of who is making the contribution, although the wiki should capture your activity when you are logged in.
> 
> At some point when we have enough tofu (ex beef) we can finalize edits and publish
> could this aim to be a mini spec of sorts?
> Should we try to set some deadlines or calls around this? 
> Co chairs wish to cultivate this draft welcome to, in addition to anything else they may have in mind, if not I ll wrap something up from my own materials and shall submit to the group for approval before publishing
> 
> Obviously suggestions welcome
> 
> Cheers
> PDM
> 
> 
> 

Dave Raggett <dsr@w3.org> http://www.w3.org/People/Raggett
W3C Data Activity Lead & W3C champion for the Web of things