techcrisp

Alison Gopnik developmental psychologist at UC Berkeley

“Until we solve the basic paradox of learning, the best artificial intelligence’s will be unable to compete with the average human four-year-old.”

*|MC:SUBJECT|*

Present state of Artificial General Intelligence – A thought experiment !

10 Minutes: Imagine, taking a 3 year old for baby sitting from grateful neighbor. Place in comfortable room. Borrow state of art AGI Super Computer with robotic arm from Silicon Valley, place in separate room.


15 Minutes: Place a set of staplers, 3 inch scissors, 20 sheets of art paper, 10 color markers in both rooms. In addition place large box of candies and one First Aid Box in baby’s room.


20 Minutes: Innate curiosity, recognition of an external environment and the need to engage and explore with it is self driven in child. He shall start picking up various objects and play and explore with them at random. Child shall for example explore stapler, push it, open it, poke in his ear and may insert even one or more staples in his palm. Baby may also play with scissors and injure himself. Immediately give First Aid and placate baby with candies. This process will create a 3 D Kinematic model in his brain and in future the same model shall be used for any size, shape or color for operating a Stapler.

AGI Super Computer shall remain in comatose state. It has no curiosity, recognition of the out side world and no ability to play with the Stapler and make a 3 D model  by experimentation. Even the most powerful computer and AI software does not have that capability.


40 Minutes: Baby shall start stapling pages, cutting pages and use color markers on paper. May create different designs with stapled pages, scissors and color marker pens .

AGI Super Computer shall remain in comatose state


60 Minutes: Baby shall create more intricate collages.

AGI Super Computer shall remain in comatose state


70 Minutes: Return baby to grateful neighbor after putting make up on bruises along with box of candies and soft toy.

Wait 2-10 years for Artificial General Intelligence Software and Hardware to come up to level of 3 year old.


This is the present state of Artificial General Intelligence, we shall need very different hardware and software tools unlike those used in Deep Learning and LLM’s for AGI.


Further reading:

SuperIntelligence by Nick Bostrom, Professor, University of Oxford

Possible Minds, Twenty Five Ways of Looking at AI Edited by John Brockman


Alison Gopnik developmental psychologist at UC Berkeley

“Until we solve the basic paradox of learning, the best artificial intelligence’s will be unable to compete with the average human four-year-old.”


Comments

Leave a Reply

Your email address will not be published. Required fields are marked *