What do you think about this argument? GPT-3 is obviously not intelligent. It was fed by a nontrivial percentage of all text produced by humans which cannot grow by several orders of magnitude. Therefore the paradigm of feeding data into function learning is a dead end for AGI.
Obviously false: a better function learner (the brain) and more varied data got us to natural intelligence, so they can get us to AGI as well.
-
-
I'm (between the lines) arguing that the brain is more than a function learner.
-
Function learners and optimization are deceiving traps in which AI is falling: post hoc we can say that there is a function optimized given a task which doesn’t mean there some universal mechanism to build such function ad hoc and that real world constrains allow for optimization
- Näytä vastaukset
Uusi keskustelu -
-
-
Just so I understand the discussion, isn't
@balazskegl's argument, as he states in a reply, that varied data IS what we need but that there needs to be a way "to select a computationally (and here I mean absolute physical constraints) feasible subset of the multimodal input?" -
Yes, and that 1) this selection is as important as function learning, and 2) it cannot be done by function learning.
- Näytä vastaukset
Uusi keskustelu -
-
-
it seems like human intelligence involves several layers of control and learning. learning what to pay attention to, making connections and analogies between disparate domains. this is a paradigm shift (or several) beyond current AI approaches, right?
Kiitos. Käytämme tätä aikajanasi parantamiseen. KumoaKumoa
-
Lataaminen näyttää kestävän hetken.
Twitter saattaa olla ruuhkautunut tai ongelma on muuten hetkellinen. Yritä uudelleen tai käy Twitterin tilasivulla saadaksesi lisätietoja.