Hacker News Reader

Comment by gwern | Hacker News Reader

I don't think kids have a bullshit detector (but maybe you agree and that's why you are quoting that article - if you need to develop a bullshit detector, that implies you don't already have one). Kids are completely gullible, and trust anything you tell them about Santa Claus, ancient Assyria, the continent of Lemuria, chemistry, God, traffic laws, the Easter Bunny, poisons, Jesus etc.

That makes perfect sense, since they spend so much time being taught true things - they have a lot to learn (think about the infamous 'why' question, or the general progress of AI), they can't afford to disbelieve and question unless something seems really out of whack.

The listed examples seem more like exceptions proving the rule, possibly combined with a bit of naivete and 'beginner's mind'.

View on HN

"the general progress of AI"

Out of curiosity, what are you referring to here?

A lot of what we perceive as obvious common sense relies on a huge amount of background knowledge that we don't really think about when performing them. Take for example the seemingly simple action of pouring a glass of water. This relies on knowing the glass has to be oriented the right way up, that liquid is subject to gravity, the general behaviour of liquids, viscosity and lots of little things that we take for granted.

For some good writing on the subject have look at Douglas B. Lenat's essays about knowledge pumps and Cyc.

Knowledge and hardwiring. Consider the feather-and-cannon-ball experiment in counter-intuitive gravity. Our unphysical hardwired heuristic assumes that heavy things go faster (and with air drag, they do). But also experience. We know water stays in upturned but not overturned glasses because we've seen it do so.

A lot of knowledge seems hardwired. Infants who can barely see already knows things like inertia and persistence of invisible objects, and then there's the whole Chomskyian linguistics argument that kids learn language too fast and well.

> Infants who can barely see already knows things like inertia and persistence of invisible objects

Do you have evidence / references for that? (Genuine question; I'd like to see it if you do.)

As far as language learning goes... there are plenty of linguists who disagree with Chomsky on that point. Among others. In particular, the cognitive linguistic school is fairly at odds with the generative grammar school.

There's a lot of research on infants these days. There's a great deal of discussion on what infants do and don't know in _The Philosophical Baby_ (because the issue has obvious philosophical relevance to Kant's categories and Locke's tabula rasa).

Just yesterday I read of new research covering probability: http://www.sciencedaily.com/releases/2011/05/110526141357.ht...

> Infants who can barely see already knows things like inertia and persistence of invisible objects

No, they do not. Object permanence is a good example of how even fundamental principles about reality are learned. In this case, it happens when a child is about a year old:

http://en.wikipedia.org/wiki/Object_permanence

It's quite possible that if you dropped a newborn into a simulated reality that has completely different rules from our world, this child would learn the "fake" rules without knowing the difference.

A year old? Try 3 months old: http://en.wikipedia.org/wiki/Object_permanence#Contradicting...

And in the vein of hardwired preferences, notice the mention in the later section of that article of 7 minute old babies preferring faces - unless you want to suggest that unbiased algorithms would learn the importance of faces in that little time?

Yes, Cyc was what I was thinking of. A 3 year old knows an awful lot.

Reply on news.ycombinator.com