Re: List of possible deliverables and position papers for AIKR CG

Another useful paper is:Harnessing the Power of LLMs in Practice: A Survey on ChatGPT and Beyond
https://arxiv.org/abs/2304.13712
The current situation of the proliferation of OpenAi and other original creator derived products is starting to look a lot like a collection of family trees of Linux distros.
Garrido et alii have provided a good starting point, but there are quite a few more family trees.


Milton Ponson
GSM: +297 747 8280
PO Box 1154, Oranjestad
Aruba, Dutch Caribbean
Project Paradigm: Bringing the ICT tools for sustainable development to all stakeholders worldwide through collaborative research on applied mathematics, advanced modeling, software and standards development 

    On Monday, May 29, 2023 at 01:28:29 PM AST, Paola Di Maio <paola.dimaio@gmail.com> wrote:  
 
 Thank you MiltonThe idea of identifying categories of generative AI is valid, imhoFirst step would be to gather the field, ie, make a short summary of existing literature (what about this paper, for example?)A Complete Survey on Generative AI (AIGC): Is ChatGPT from GPT-4 to GPT-5 All You Need? https://arxiv.org/abs/2303.11717https://medium.com/@eduardogarrido90/chatgpt-is-not-all-you-need-a-quick-summary-of-the-generative-ai-taxonomy-e2b8b47a9851is there anything missing in this paper that our list could add?

Rendering in Stratml would be a separate thing, that  can be done at the end (so that the taxonomy ontology can be rendered in any xml format) surely Owen can take care of the stratml part when we get there, but imho the taxonomy should exist as plain text in the first place

On Mon, May 29, 2023 at 5:24 PM ProjectParadigm-ICT-Program <metadataportals@yahoo.com> wrote:

Dear all,
Open AI, Microsoft, Google and Meta have opened Pandora's Box and set in motion an unstoppable development push of generative LLMs and generative audio and visual AI applications.
Countless experts and the "godfather of AI" have made comments on the current state of AI and painted a varied panorama of scenarios ranging from doomsday and human race annihilation to AI as replacing domestic assistants (in embodied form) and other assistive functions (replacing humans in quite a few instances).
There are three issues that will have an impact on further development:(1) the sheer size of the data sets and variable count of the programs and the amount of computing power required for general all purpose LLM chatbot applications;(2) in making online search chatbot powered increases the energy footprints of online search to the extent that chatbot related activities could lead to a 2-3% increase in CO2 emissions and require additional investments and space for data centers;(3) bias, error and unwanted emergent abilities will severe limit open, explainable, accountable, safe and trustworthy utilization in many industries and fields, e.g. in health care, social services and judicial systems.
There are several approaches now being proposed to tackle all three issues.
The focus on few shot, zero shot, limited domains of discourse and reduced equipment and power consumption solutions, and in some scenarios moving the computing from the cloud to the edge.

When we look at how humans learn in school, and other levels of education it makes sense to use learning materials that are (1) curated/peer reviewed, (2) limited in scope by their application or as part of a modular approach to a a specific profession or academic discipline, (3) limited in size of text or audiovisual content.
The learning from practice and experience component in humans is slightly more complicated and is the subject of many overlapping disciplines from biology, biochemistry, cognition and neuroscience to psychology, philosophy and quantum physics.

I would like to propose making an inventory of categories of generative AI applications currently on the market and being developed, where possible mapped to their respective utilization in which fields, and listing for their utilization current issues that are of concern to their openness, explainability, accountability, safety and trustworthiness.
There are several conceptual diagrams for defining artificial intelligence and we could pick a limited number for categorization, one for general purpose like the one from the high level expert group on AI of the EU to some conceptual models used in industry or academe that focus on particular general areas or narrow focus fields of application.
Since the open, explainable, accountable, safe and trustworthy use of generative AI will require academics, industry, civil society and legislators to find common ground in terms of what needs to be addressed, this proposal makes sense.
We can use StratML to help make the strategic planning process automated, and get support of academia in generating literature and systematic reviews of the status quo and state of the art of conceptual models and application of generative AI

And because we as a community group are focused on the knowledge representation, we make this the defining ingredient for the proposed work to be done.

Milton Ponson
GSM: +297 747 8280
PO Box 1154, Oranjestad
Aruba, Dutch Caribbean
Project Paradigm: Bringing the ICT tools for sustainable development to all stakeholders worldwide through collaborative research on applied mathematics, advanced modeling, software and standards development