A "bottom-up" approach to AI as a more transparent alternative to "top-down" LLMs

post by Paul Jorion · 2023-06-06T18:00:48.282Z · LW · GW · 0 comments

Contents

No comments

At Pribor.io we decided back in June of last year to capitalise on work I've done in AI back in the late eighties as part of British Telecom's Connex Project. At that time I developed ANELLA, an Associative Network with Emergent Logical and Learning Abilities. As I state in the post's title here, although LLMs were unknown at the time, seen in retrospect, ANELLA was providing a "bottom-up" approach to AI as a more transparent alternative to "top-down" LLMs, also more economical by several orders of magnitude than LLMs. ANELLA was however very much in the same spirit as LLMs of letting the self-organisation that lies in a language's lexicon to bubble up to the surface.

A brief introduction to ANELLA can be found in Luca Possati's The Algorithmic Unconscious (2021). An extensive account can be found in the French book I published at the time: "Principes des systèmes intelligents" (1989), still available as it was reprinted in 2012.

A couple of days ago I started on my YouTube channel a course of lectures in English entitled "Thought as Word Dynamics" where I plan to explain in detail my alternative approach to AI (5 videos have been recorded by now, 3 are already available on YouTube).

Here the first lecture, where I give a historical account of my work in AI:

 

Now to offer a glimpse at where we currently stand at Pribor.io, here is the first video we've released in February entitled "Self-Aware Machines" (the second video is forthcoming later this month):

0 comments

Comments sorted by top scores.