Sujet : Re: AI superhuman cognition
De : ram (at) *nospam* zedat.fu-berlin.de (Stefan Ram)
Groupes : comp.miscDate : 11. Apr 2024, 20:33:59
Autres entêtes
Organisation : Stefan Ram
Message-ID : <reason-20240411203339@ram.dialup.fu-berlin.de>
References : 1
Richmond <
dnomhcir@gmx.com> wrote or quoted:
"OpenAI and Meta are on the brink of releasing new artificial
intelligence models that they say will be capable of reasoning and
planning, critical steps towards achieving superhuman cognition in
machines."
It seems an ordinary chatbot from the Web already is capable
of this. Me,
|There are two cards lying on a table. Please make a plan how
|to pick up the lower card when you can only ever lift one
|card at one time.
, chatbot,
|To pick up the lower card when you can only ever lift one
|card at a time, you can follow this plan:
|
|1. Lift the upper card and place it aside, exposing the lower
|card.
|
|2. Now that the lower card is accessible, lift and pick up the
|lower card.
|
|The key steps are:
|
|1. Remove the obstacle (the upper card) to access the target
|card (the lower card).
|
|2. Pick up the target card (the lower card).
|
|This plan ensures that you can pick up the lower card by first
|removing the upper card, which is blocking access to the lower
|card. By lifting the cards one at a time, you can successfully
|pick up the lower card without violating the constraint of
|only being able to lift one card at a time.
.