Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Why Is the Human Brain So Efficient? Massive Parallelism

nautil.us

Post reply on HN
Print view View original

41–50 of 55 posts

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#41

Post by dlwdlw »

BLKNSLVR wrote: post #39
21 wrote: post #16

Earlier quoted context omitted.

The way other articles describe it, a pro player is not really aware of the incoming ball, lower level non-conscious (reflex) systems track the environment and react before you are aware of it, just like when a regular person dodges a sudden obstacle coming at them and only a bit later they realize what's going on.

Sounds like one of the concepts that inspired the book Blindsight by Peter Watts. The subconscious 'reptile brain' is able to process certain old-school dangers, like an object moving quickly towards us, faster than our conscious brain, and can cause involuntary, potentially life-saving, reactions to the stimulus. The conscious mind processing these things can often get in the way but, in the case of sports, alter th…

One thing that has always bothered me about many AI approaches is that instead of having multiple layers that deal with different levels of urgency everything is delegated to a "rational" mind that thinks vs a combination of thinking predicting and reacting.

The parallel is this obsession with rationality over intuition. Though somewhat ironic because black box AI is just accepted.

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#42

Post by dlwdlw »

fallingfrog wrote: post #25

The article seems to be about how it is that the brain can process so much information at so slow a cycle speed, but it doesn't really address power efficiency at all. Even assuming 100% parallel operation, our current chip designs use something like a factor of a million more energy to do the same amount of computation. I wish I knew why- it's not like we're ignoring power efficiency. Depolarization of an axon must…

I think it's about context. If I say "order pizza" a ton of information is transferred via shared environmental context. I'm not transferring information directly but manipulating contextual levers to map certain types of information.

An example is if I say something to you in say japanese, I also need to teach you japanese in the same phrase unless there is shared context. computers run on a very impotent form of language, logic, that is rigorous and general purpose but has no "escape hatch". (basically the sentiment of Godem, Escher, Bach with regards to rigorous systems. )

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#43

Post by nootropicat »

justifier wrote: post #35
nootropicat wrote: post #26

I recently realized the (obvious in hindsight) fact that general intelligence better that brute force doesn't exist, as intelligence is equivalent to compression. Given the recent discoveries about neurons using mRNA capsids to communicate [0] it's not that farfetched to posit that we are really dna computers [1]. The processing time (for new problems) seems human-like: "The slow processing speed of a DNA-computer (t…

> general intelligence better that brute force doesn't exist What do you mean by this? It sounds like you are assuming nature requires exactness and are really making a philosophical argument about the nature of computation, namely an np solution in p 'doesnt exist' > Memory as dna would explain high-level memory quirks If we are conjecturing than so too could approximate results 'explain high-level memory quirks' we…

>What do you mean by this?

The goal is to reduce n bits of data to xYet it's not possible to compress n bits in the general case. That's because the kolmogorov complexity is a function of your assumed knowledge (assumptions). All you can do is start checking every possible transformation from your assumption starting with the most probable one - the probabilities are based on your knowledge itself.

>we already have many approximate algorithms that are significantly better than brute force

Yes - but that means the algorithm itself, along with its execution, is the shortest (in the used metric, which can include execution time) answer for a particular problem. How do you generate the algorithm in the first place?

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#44

Post by BLKNSLVR »

dlwdlw wrote: post #41
BLKNSLVR wrote: post #39

Earlier quoted context omitted.

Sounds like one of the concepts that inspired the book Blindsight by Peter Watts. The subconscious 'reptile brain' is able to process certain old-school dangers, like an object moving quickly towards us, faster than our conscious brain, and can cause involuntary, potentially life-saving, reactions to the stimulus. The conscious mind processing these things can often get in the way but, in the case of sports, alter th…

One thing that has always bothered me about many AI approaches is that instead of having multiple layers that deal with different levels of urgency everything is delegated to a "rational" mind that thinks vs a combination of thinking predicting and reacting. The parallel is this obsession with rationality over intuition. Though somewhat ironic because black box AI is just accepted.

Building something that mimic how it appears to work, as opposed to how it does work. Like building the symptoms whilst assuming we're building the disease.

Having said that, however, gotta start somewhere. Having said that, however, we might have done a decade of work in the wrong direction by the time we find out what direction it should go...

'tis a conundrum.

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#45

Post by justifier »

nootropicat wrote: post #43
justifier wrote: post #35

Earlier quoted context omitted.

> general intelligence better that brute force doesn't exist What do you mean by this? It sounds like you are assuming nature requires exactness and are really making a philosophical argument about the nature of computation, namely an np solution in p 'doesnt exist' > Memory as dna would explain high-level memory quirks If we are conjecturing than so too could approximate results 'explain high-level memory quirks' we…

>What do you mean by this? The goal is to reduce n bits of data to x Yet it's not possible to compress n bits in the general case. That's because the kolmogorov complexity is a function of your assumed knowledge (assumptions). All you can do is start checking every possible transformation from your assumption starting with the most probable one - the probabilities are based on your knowledge itself. >we already have…

> Yet it's not possible to compress n bits in the general case.

Abstract bits sure, 2^4 objects are unable to represent 2^5 objects, simply due to 16!=32

But what does it have to do with'general intelligence' and brute force?

We were originally talking about rna communication.. where does kolmogrov come in? In the data representation in the rna? But what of it when the mechanism that encodes and decodes is unrestricted in its upper bound complexity? If the disparity between the upper bound of memory being encoded and the mechanism encoding it are great enough then that system could 'compress n bits in the general case'

> All you can do is start checking every possible transformation from your assumption starting with the most probable one - the probabilities are based on your knowledge itself.

This just sounds like you're saying every algorithm is brute force but with different possible states due to assumptions

Would you call Euclid's gcd 'brute force with assumptions'? I would argue algorithm is antonym to brute force

> How do you generate the algorithm in the first place?

Ah, I think I see what you're saying.. are you conjecturing the process of evolving conciousness was itself a brute force process?

Where understanding it's underlying process and being able to implement it ourselves, perhaps even more thermodynamically efficient, is inconsequential due to our efforts being only possible by the original conjectured brute force process that allowed us to abstract to such a degree..? This process being the 'assumption' to be appended to the proof?

But this again seems like a philosophical debate.. one of life and negative entropy

How do you define general intelligence? How do you defend the statement that 'intelligence is equivalent to compression.'?

If you would have to consider all of existence as assumption, then a bitwise representation of our own intelligence would be a significantly small subset of the bitwise representation of all things; expressing this would seem to imply some process substantially more efficient than brute forcing every possible state, or luck is real?, or we underestimate the complexity of 'general intelligence' and in actuality the search is ongoing? Or some undiscussed other?

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#46

Post by antris »

nitwit005 wrote: post #31

We don't know what operations the brain is actually doing, so declaring that it is performing them efficiently is pure speculation. Similarly, the brain surely has limits to parallelism, for the same sorts of reasons a computer does. You only have one mouth, so if two parts of the brain tried to speak, fully parallelized, you'd get nonsense at best. They have to agree on what to say, which is effectively serializatio…

The mouth as an example with regarding to speech doesn't work very well. People can say one thing, but say it in a way that means the complete opposite. So already we have tone incorporated into our speech. Who knows what kind of subconscious thoughts we embed into our speech that are not yet understood well.

Then there's the combination of tone and the words, accent, body language and every other context embed into our language that produces a very complex meaning, even for one word.

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#47

Post by fallingfrog »

dlwdlw wrote: post #42
fallingfrog wrote: post #25

The article seems to be about how it is that the brain can process so much information at so slow a cycle speed, but it doesn't really address power efficiency at all. Even assuming 100% parallel operation, our current chip designs use something like a factor of a million more energy to do the same amount of computation. I wish I knew why- it's not like we're ignoring power efficiency. Depolarization of an axon must…

I think it's about context. If I say "order pizza" a ton of information is transferred via shared environmental context. I'm not transferring information directly but manipulating contextual levers to map certain types of information. An example is if I say something to you in say japanese, I also need to teach you japanese in the same phrase unless there is shared context. computers run on a very impotent form of la…

Oh ok you're talking about data compression. That's also not what efficiency means, in a strict sense.

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#48

Post by Retra »

rmwaite wrote: post #37
coss wrote: post #17

I feel like there must be some way to 'clean' our brains. As we get older it must accumulate 'gunk' that slow it down or weaken connections. I guess sleep is the closest thing we have but to me that's like saying drink water to fix tooth decay.

Drinking water (instead of sugary and/or acidic drinks) is actually a very good way to prevent tooth decay.

Even if you drink sugary/acidic drinks, water helps dilute them and rinse them away.

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#49

Post by rproctor »

iandanforth wrote: post #5

"a professional tennis player can follow the trajectory of a tennis ball after it is served at a speed as high as 160 miles per hour" This is false or at least highly misleading. A reader might imagine that the player's eyes track the ball in flight. This is not what happens. A pro player reads the position of the serve and begins moving before the ball is hit. The predictive power of the brain is much more important…

There is a related anecdote in "The Sports Gene" that suggests that MLB hitters do not track the ball in flight, but instead react to the pitchers' arm motion. As a result they have trouble hitting a motion they haven't visually trained against:

http://www.eugenewei.com/blog/2013/8/23/the-pitchers-who-con...

Re: Why Is the Human Brain So Efficient? Massive Parallelism

#50

Post by jjtheblunt »

iandanforth wrote: post #5

"a professional tennis player can follow the trajectory of a tennis ball after it is served at a speed as high as 160 miles per hour" This is false or at least highly misleading. A reader might imagine that the player's eyes track the ball in flight. This is not what happens. A pro player reads the position of the serve and begins moving before the ball is hit. The predictive power of the brain is much more important…

https://www.imdb.com/title/tt5434972/ is an _excellent_ documentary about just this topic, science and all...and it's funny.
Post reply on HN