Research: A Plan to Discover the Fundamental Principles of Intelligence
"The intelligence of a system is a measure of its skill-acquisition
efficiency over a scope of tasks, with respect to priors, experience, and generalization difficulty". - Francois Chollet.
(On the Measure of Intelligence).
The study of intelligence should be concerned with concepts, such as the need for embodiment, general task (prediction), misalignment between input & output dimensions, need for self generated open ended abstractions,
dynamics v/s statistics and much more.
Dynamic Connectionism is a new paradigm proposed by Filip Piękniewski in the the research paper Predictive Vision Model(PVM)
In PVM, Filip proposes a new paradigm of AI based on dynamics not statistics.
To know more, refer to Filip's blogs . Concisely, the wisdom and
exciting set of ideas in those blogs is the best in the field of AI I have found till now.
I am taking a deep dive into the field of contemporary neuroscience, which would take 4-5 years, and use it for
inspiration to create new AI systems. Also, how to encode Human Priors as specified in
"On the Measure of Intelligence" is an unsolved problem for now.
Intelligence's first necessity is embodiment. "Embodiment is constant access to the real world through a
high bandwidth sensory input." - Me ; )
To solve intelligence, I am starting with the frontier of perception inspired by the fundamental problem in AI
i.e. Moravec's Paradox. Dexterity and Mobility will come in soon due to
affordances being necessary for perception. Language is a tool for communication and thought, but not necessary for
intelligence.
If this resonates with you, let's connect.
Extras: Cajal's Butterflies of the Soul,
What is reasoning?,
Ideas on how to be a better scientist
Intelligence, The Bitter Lesson and the Better Lesson
|