Status Updates From Jak se děti učí významu slov

Brown (1957): children use syntax as cue about what ontological category to pay attention to (if a verb, then an action) in word learning. Some corroborating results from McPherson (1991). Some controversy: maybe they use syntax to access, not learn, word meanings (Pinker 1989)

Nouns don’t refer to objects and verbs don’t refer to actions; NPs and VPs (respectively) do

Spelke-object bias in word learning overcoming Quines problem. Other possible solutions (ToM, syntax) not as precocious. US children generalize simple objects and substances on the basis of shape, but Japanese children don’t: possible explanation is Japanese doesn’t have the count/mass “many/much” distinction. Pblm for this solution: how do children learn words for things that are NOT Spelke-objects?

Pragmatic explanation of the mutual exclusivity effect (PEMEE) (Clark, Markman): problem: when you make your intention to refer to the more familiar object using the less familiar word (out of a set containing a shoe and a whisk, point to a shoe, say “give me the item”), children still give you the less familiar thing (the whisk) (Woodward & Markman 1997). But this is compatible with PEMEE: child is just confused

Arguments against associationism about word learning: errorless fast mapping, cultural variation, blind children do fine, chimps suck

Fast mapping: applies to facts and words, not specific to word learning alone. It’s specific to unobservable properties of the datum

About half way through this one, and I find Blooms’ contention that very young children are able to infer the intentions of others to be a stunning argument. Evolution has shaped us to give off physical cues about what we are directing our attention toward, and young children appear to have the ability to infer these intents. This will be a crucial piece of the AGI puzzle.