My last post tangentially stimulated another important redefinition of AGI –
and the nature of a real world agent’s activities/skills.
At the moment, we think of AGI mainly as about being able to master one
activity and acquire one skill after another. That is obviously what real world
creatures do.
But implicit in this is the assumption that each activity/skill is largely an
island unto itself.
Yes, we talk about AGI involving the “transfer of skills” - an AGI will be
able to transfer some of what it has learned in one activity to another – but
we don’t really develop or explore this idea.
The reality is that all real world activities involve applying complex
patchworks of skills and bodies of knowledge to any activity – patchworks which
develop and become more complex over time.
For example, playing tennis is not simply about hitting balls around a court.
For a human, it will also involve the application of some
*psychology - understanding an opponent’s motivation, thinking and behaviour
- understanding and developing one’s own motivation,
concentration
(the player may study courses, read books etc)
*cultural knowledge of the game – of past and present tennis players and games
and competitions
*nutrition - what to eat/consume generally and before and during a game
*exercise/ exercise physiology – how to train for tennis
* commerce - what equipment to buy and where
* fashion/clothing – what to wear
*healthcare – how to treat injuries
... and still more skills and knowledge. These of course are not applied all at
once, but gradually developed.
There is never a definitive *set* of skills to be applied to any activity.
Ultimately there is a potential infinity – a worldwide web - of skills that can
be applied. We can not only “always learn more” about an activity – we can
also apply more skills.
The real world agent has to decide arbitrarily – to make “executive decisions”
– about what constellation of skills he will acquire/apply .
He must choose to patch together a “patchwork* of skills for an activity – a
reasonable but arbitrary selection. How much attention, if any, should a tennis
player pay to any of the skills listed above, for example? There are no rules.
Note that as he develops - the patchworks of skills that any real world agent
can apply to any activity – grows ever larger. (There is certainly no neat
“pattern” of skills).
Think of those currently engaged in “AGI computing” – everyone here brings very
complex patchworks of skills to this activity – and each individual also brings
different patchworks. Some are more visual/robotic/logical etc than others;
Some altogether exclude certain skills that others consider fundamental. Some,
like Ben, consistently acquire and apply new skills here. Others are relatively
stuck in their ways.
So AGI is about applying patchworks of skills (and actions) to produce
patchwork courses of actions in patchwork activities.
-------------------------------------------
AGI
Archives: https://www.listbox.com/member/archive/303/=now
RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424
Modify Your Subscription:
https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657
Powered by Listbox: http://www.listbox.com
