Hacker News new | past | comments | ask | show | jobs | submit login
How the Brain Links Gestures, Perception and Meaning (quantamagazine.org)
81 points by pseudolus 30 days ago | hide | past | web | favorite | 11 comments

I’ve always thought that proponents of “embodied cognition” are onto something. Like the whole field of so-called machine learning is missing something fundamental about how humans learn. This article provides a cogent explanation for what could be missing.

“When children are learning their first language, Macedonia argues, they absorb information with their entire bodies. A word like “onion,” for example, is tightly linked to all five senses: Onions have a bulbous shape, papery skin that rustles, a bitter tang and a tear-inducing odor when sliced. Even abstract concepts like “delight” have multisensory components, such as smiles, laughter and jumping for joy. To some extent, cognition is “embodied” — the brain’s activity can be modified by the body’s actions and experiences, and vice versa. It’s no wonder, then, that foreign words don’t stick if students are only listening, writing, practicing and repeating, because those verbal experiences are stripped of their sensory associations.“

Frankly I find this explanation very convincing. I have absolutely no idea how one would go about trying to design intelligent machines using embodied cognition. Feels like we are light years away from anything close to this, which is why no mainstream AI researchers are working on it or even discussing it as far as I know. Would love to see some counter examples if anyone is more knowledgeable about the space.

You might be interested in the work of Jean Piaget, a constructivist child psychologist. One of his PhD students was a South African scientist you may have heard of: Seymour Papert, longtime collaborator with Marvin Minsky.

There are a couple of companies working on embedded cognition but they are in stealth. Academic work in this area has pretty much faded since the late 1980s.

Among psychologists, Laurence Barsalou is one of the most prominent academics working in the emobodied cognition, or as his lab calls it, grounded cognition, tradition: http://barsaloulab.org/

His work has influenced one of the most distinguished philosophers of mind working in this area: Jesse Prinz: http://subcortex.com/

Embodied cognition is being studied in reinforcement learning with virtual agents in simulated environments. It's not just about the body, but mostly about the environment. Think of the environment as an infinite dataset, always providing new experiences and their consequences to the agent. By comparison, even a dataset of millions of images or gigabytes of text is static.

You might find some of this researcher's work of interest: http://www.cs.otago.ac.nz/staffpriv/alik/research.php

It's not quite cutting edge AI, but it is a lab that takes seriously embodied cognition.

>> (...) in Greece or Turkey forming a ring with your index finger and thumb to indicate everything is A-OK could get you in trouble.

I'm Greek. I'm not sure what that refers to. I think they mean that we use the gesture to indicate the number 0, or, more rudely, the anal orifice (often accompanied by inserting the index finger of the other hand into it to illustrate penetration).

This gesture can indeed be interpreted like that. I think those are rather obsolete meanings though. Mostly I've used the gesture myself to say "everything OK". I suppose this is the equivalent of loan words, in this case, from the more common usage of the gesture in Western Europe.

Edit: Oh. Now that I think of it, there is also a further interpretation of that gesture to indicate a tiny size. Meaning, of the human penis. It's true that its original meanings were always quite rude.

"No other species points, Novack explained, not even chimpanzees or apes"

Except there are many breeds of dogs that point. True it may be partially training, but every dog I've had recognizes the significance of human pointing and there appears to be some kind of innate pointing instinct.

You seemed to have left off the second part of the sentence, which you appear to agree with

"No other species points, Novack explained, not even chimpanzees or apes, according to most reports, unless they are raised by people"

edit: I suppose the clause "unless they are raised by people" could be interpreted as pertaining to either "not even chimpanzees or apes", or "No other species", but I don't think it matters.

Elephants seem to understand pointing without being trained. It's possible that they point too.

> We suggest that success was not due to prior training or extensive learning opportunities. Elephants successfully interpreted pointing when the experimenter’s proximity to the hiding place was varied and when the ostensive pointing gesture was visually subtle, suggesting that they understood the experimenter’s communicative intent.

African Elephants Can Use Human Pointing Cues to Find Hidden Food https://www.cell.com/current-biology/fulltext/S0960-9822(13)...

Do humans point if they're not raised by people? The few cases of such that I've heard about seem to be incapable of most "uniquely human" behaviours.

Possibly not, since we're pretty good at following another person's gaze instead. And different cultures point differently. I knew a guy who lived in Papua New Guinea for a while and he had to untrain himself to point with his finger. People there would just look at his finger. They only pointed with their chins.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact