Revamp Machine learning: Artificial stem-cell to make artificial general intelligence

[ProgrammingGodJordan]

(Part A)

In early machine learning days, researchers coded programs with specific rules to respond to as many scenarios as the researchers themselves could imagine.

However, when unforeseen scenarios came along, their programs quickly broke/failed.

(Part B)

Modern machine learning now learns rules automatically, and fail far less, achieving human level performance on narrow cognitive tasks. However, there are still problems with modern machine learning, and separately, researchers are still manually designing the basis nodes/connections (artificial neurons/synapses) involved in these modern automatic rule learning models.

(Part C)

So, I then investigate whether it is feasible to automate even more, by utilizing stem cell data to "grow" those basis nodes/connections, instead of manually designing them.

(Part D)

What follows is a rough exploration of stem cell proliferation, determination & differentiation as an approach to‘grow’ some degree of software artificial general intelligence.

See my paper here: stem-cell-agi-exploration_research-gate

Or here:stem-cell-agi-exploration_academia

Criticism regarding the topic is welcome. Help me to scrutinize this idea fully. (or build on the idea if possible)

Edited April 30, 2017 by ProgrammingGodJordan

[imatfaal]

!

Moderator Note

 

Moved to CompSci

 

[EdEarl]

Just in case you missed Cultured neuronal networks.

[ProgrammingGodJordan]

Just in case you missed Cultured neuronal networks.

No, I didn't miss it.

In fact, notice I reference neural stem cells on pages 2 to 3 of my paper.

I argued that observing such real, actual cells are perhaps crucial in replicating some degree of proliferation, determination & differentiation of pluripotent neuronal expansion system in silicon.

 

Edit: The source may still be useful anyway. Thanks.

Edited April 30, 2017 by ProgrammingGodJordan

[ProgrammingGodJordan]

ANNOUNCEMENT:

 

After some searching (weeks of searching until now) I discovered today that the idea presented in the original post's paper is an old one.

 

It falls under the topic of Artificial Embryogeny, a sub-topic in Neuroevolution.

Edited May 12, 2017 by ProgrammingGodJordan