[Fis] Paradigm AI
钟义信
zyx at bupt.edu.cn
Wed Oct 18 09:16:04 CEST 2023
Dear Eric,
Thank you for the interesting talk on "Paradigm AI" from which I learned a lot.
As a discussant, may I propose some of my understanding. Comments are welcome.
(1) I appreciate your idea that saying "Physics paradigm PPD does not fit well with AI paradigm" and "Information paradigm PID is a better fit". This is the valuable common basis, between you and me, concerning the PPD, PID and AI.
(2) How to define the concept of intelligence? This is a very difficult problem. To my own understanding, the following short statement may serve as one of the candidates: Intelligence is the ability to solve problem but not the ability to find and define problem, the latter of which is one of the abilities for wisdom.
(3) The paradigm for AI can be used as the paradigm for bio-technology with certain simplification and specialization. This judgement is not based on their "structure/architecture", but based on their "information function" - which is the basic function in both AI and biotechnology, that is to seek opportunity for "living (or solving problem)" and to avoid the "danger (or failing to problem solving)".
Once again, comments and criticisms are most welcome.
Best regards,
Prof. Yixin ZHONG
AI School, BUPT
Beijing 100876, China
------------------ Original ------------------
From: "Eric Werner"<eric.werner at oarf.org>;
Date: Tue, Oct 17, 2023 02:32 AM
To: "fis"<fis at listas.unizar.es>;
Subject: [Fis] Paradigm AI
Here are some brief thoughts on Paradigms and AI by I presume was written by Yixin Zhong since I cannot read Chinese.
Paradigm AI
I agree that the physics paradigm PPD doesn’t fit well with the AI paradigm, and that the information paradigm PID is a better fit
Artificial intelligence systems, don’t necessarily learn from human beings. In unsupervised learning they learn from data and not from humans.
The problem, and becomes really how to define what intelligence is: Which of the following is it?
Rational inference
Summarizing large amounts of text and data
Making new predictions based on scientific theories and available data
Developing new theories that explain the data in the more succinct way, and making new predictions
Developing new technologies independently of human input
Planning and executing the actions and intentions of a robot
Having social intelligence
Being cooperative with a human being in achieving a task
Interrelating two discipline, such as physics and mathematics, to make new discoveries
Understanding, genomes in the way that human beings cannot
Designing new organisms by designing their genomes
I agree with the language of a new paradigm, such as artificial intelligence will develop slowly step by step in conjunction with its use -both conceptually and experimentally .
In a new paradigm entire new language is created as a paradigm is developed
The language evolves in concert with a new ontology suggested by the paradigm
It is an ontology of objects, technologies, actions, and strategies
What will be particularly interesting, is the linking of the paradigm of artificial intelligence with the paradigm of biotechnology
Biotechnology and AI will truly link the human brain with the artificial brain
The genome of the natural brain will be reflected in the architecture of the artificial brain
Hence by using AI to decode the genome of the natural brain, it will be self-reflected in the design of the developing artificial brain
This will bring unprecedented social and rational functionality to the artificial brain
Note that the biotech-genome paradigm also is founded on the information paradigm.
Thank you Yixin Zhong for your input and emphasizing the intimate relationship of information and AI paradigms.
Best wishes,
Eric
--
Dr. Eric Werner
Oxford Advanced Research Foundation
https://urldefense.com/v3/__https://oarf.org__;!!D9dNQwwGXtA!Ueh6UEXLK5rMEnfEBpIw9KaCLJxQkyAYQMV7cTF-1bL1UTe7bCPmctLpxf3ssmx5tOkwJ8lw2NHCspP-$
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listas.unizar.es/pipermail/fis/attachments/20231018/1d43be76/attachment.html>
More information about the Fis
mailing list