Thanks for the correction! I definitely need to be slapped in the face
like this from time to time. Very recent paper too. =P
Yeah, the psychology shows that humans definitely operate on a stack WRT
problem solving, it has also been demonstrated that humans can deal with
a roughly 7-item short term memory... And then there are long term
episodic and procedural memories...
An AI billed as human equivalent should closely match those
capabilities. Related questions are what types of memories are required
to produce a general problem solver/general intelligence even at a
sub-human level performance wise, and what kinds of memories would be of
benefit to a super-human intelligence.
James Bowery wrote:
On Fri, May 19, 2023 at 12:28 PM Matt Mahoney <[email protected]
<mailto:[email protected]>> wrote:
...
Actually *transformers are not just feed forward networks*. They
implement an attention mechanism as a winner take all network
using lateral inhibition. If you make these connections
programmable then you have a fully connected network and you can
have an arbitrarily complex hierarchy of features and short term
memory using loops and delay lines just like real brains.
As usual there is a terminology problem
<https://arxiv.org/pdf/2207.02098.pdf>:
image.png
*Artificial General Intelligence List
<https://agi.topicbox.com/latest>* / AGI / see discussions
<https://agi.topicbox.com/groups/agi> + participants
<https://agi.topicbox.com/groups/agi/members> + delivery options
<https://agi.topicbox.com/groups/agi/subscription> Permalink
<https://agi.topicbox.com/groups/agi/Te1610d7fc26c4586-M12ac53aa79c294ca3e13beda>
--
Beware of Zombies. =O
#EggCrisis #BlackWinter
White is the new Kulak.
Powers are not rights.
------------------------------------------
Artificial General Intelligence List: AGI
Permalink:
https://agi.topicbox.com/groups/agi/Te1610d7fc26c4586-Mc846f33a9b530518a89cdc1f
Delivery options: https://agi.topicbox.com/groups/agi/subscription
