h a l f b a k e r y0.5 and holding.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Ravens & Chimps both perform broadly as well as a two year
old child in intelligence tests, of course in specific test areas
both outperform a two year old & one another but averaged
across all
tests
that seems to be the consensus for roughly where
they fall.
Not very surprising with
the Chimp which has a brain only a
little smaller than a two year olds (around 384g) but the
raven
does it with a 15g brain.
That's 25.6 Raven brains (by weight) per Chimp brain,
they're
allegedly doing as well with near enough 4% the brain mass.
When you fly there's very strong evolutionary
pressure
to cut weight so it's not too surprising that they've found a
few
ways to achieve better results with less.
I have read articles that suggest some of the tricks birds
have
found are to do with a greater neuron density .. but how
they
do it isn't really important
to
this idea .. only that they do.
What I'd like to know is how well does it scale?
An adult human brain is around 1352g, could a 54g bird brain
be as intelligent as a normal human?
So I propose we selectively breed birds for bigger brains just
to
see.
Ravens aren't particularly well known for their fecundity so
they're probably not the best test subjects for this.
Chickens are something we've already selectively bred
to
reproduce at excessive speed & don't tend to fly away
as
much making it a lot easier to keep tabs on your
experimental
subjects.
So
chickens it is then.
Essentially this will be The Intelligent Mouse Project.
But
with chickens.
Accept that the Intelligent Mouse Project by design only
selects for
intelligence rather than any particular features or structures
because it's
specifically interested in seeing how evolution will go about
doing it
without
deliberate direction from us.
While what we're interested in is the scalability of the tricks
birds
use to get more out of smaller brains (their greater neuron
density etc).
So rather than selecting
only for intelligence in our experiment we'll also be selecting
for
physical
features such as brain topography (all those little pleats &
folds) and
a 2% body mass ratio to match the human brains with a
target
brain
weight of 54g.
The Intelligent Mouse Project
The_20Intelligent_20Mouse_20Project
[Skewed, Sep 08 2021]
Joscha Bach talks to Lex Friedman
https://www.youtube...P-2P3MSZrBM&t=5725s
Joscha Bach talks about a wide range of things with a unique perspective on how our brains work. [zen_tom, Sep 09 2021]
Hens Brain Size
https://www.science...i/S0032579120302133
[Skewed, Sep 09 2021]
Birds have primate-like numbers of neurons
https://www.pnas.org/content/113/26/7255
[Skewed, Sep 09 2021]
A very confused parrot.
https://www.youtube...watch?v=9T1vfsHYiKY
[Skewed, Sep 09 2021]
Stupider Chickens
stupider_20chickens
[Voice, Sep 09 2021]
How to curcumscribe a cock.
https://en.wikipedi...he_Headless_Chicken
[2 fries shy of a happy meal, Sep 09 2021]
Flight Muscle Contractility
https://www.frontie...hys.2020.01038/full
[Skewed, Sep 10 2021]
Norm Macdonald: Professor of Logic
https://www.youtube...watch?v=hh3TI3iMb1E
[zen_tom, Sep 15 2021]
Please log in.
If you're not logged in,
you can see what this page
looks like, but you will
not be able to add anything.
Annotation:
|
| |
If a brain's function is to model the world such that the brain-host gets to predict and anticipate a complex environment, gaining
advantage as it does so - then I think you have to tune your selection/reward process towards this metric, rather than just brain
size/topology. I'm not convinced that breeding more "cauliflower" shaped brains is going to provide the benefits we're looking for. |
|
| |
But doing that is complicated, and depends on your definition of intelligence. |
|
| |
For some, consciousness is something that emerges from systems with enough intelligence - and is a symptom of a system trying to
model other systems similar to itself - the system builds a model of what it is like to be a system like itself, and uses the data it
gathers from the experience of that to help inform and predict what other systems might do under similar circumstances. |
|
| |
(This is a super interesting idea in-of itself - our perception of what it is like to be ourselves is merely a simulation our brain
constructs so that it is better equipped to predict what other brains might be up to in the same environment) |
|
| |
So to some extent, the problem is (to pardon the pun) a chicken and egg problem. If having consciousness is a result of modelling
other conscious entities, then developing such models requires other entities that the system is going to interact with (and therefore attempt to model), to already
have consciousness themselves. And chickens (or mice) currently don't fit the bill. |
|
| |
If intelligence is about moving levers and solving ever-more-complicated puzzles to receive a prize, then I think the Intelligent
Mouse Project just might work and you'd end up with a collection of super smart problem solvers - but until you encourage them to
interact, form tribes and communities, lie, cheat and play games as a means to solve their problems, my intuition here is that
they're going to remain uninteresting from a consciousness perspective, even if they have super convoluted, densely connected
brain topologies. |
|
| |
In other words, I think you get the brain suited to the evolutionary problem - if that's (relatively) simple case of gathering nuts,
then only a degree of intelligence is necessary, wheras if you're on the crest of an ever-evolving wave of 10,000 years of cultural
evolution and need to process that in order to survive and reproduce, you're more likely to evolve the necessary. Of course, you're
probably going to need 20+ years to become properly effective in such an environment, and so conducting such an experiment
becomes impractical. |
|
| |
What you need is a way to encode some ongoing knowledge or culture into your experiment so that later generations have to
onboard and transmit that additional data (language, rules, technology etc) but in a controlled environment with a generational-
timescale much lower than 20 years. If you can kick-start that system, I think you'd be onto something. |
|
| |
Whether we'd be able to communicate with consciousnesses that inhabited such a system is another question entirely - it would
certainly be a weird paper to peer-review. |
|
| |
Just place them on one side of a road and wait on
the other side with your clipboard. |
|
| |
I think your talking about 'what is intelligence' & 'how do
we test for it' [zen]? |
|
| |
We don't really care what it is. |
|
| |
It's not the subject of this experiment so it doesn't matter. |
|
| |
As for how do we test for it we already have that covered. |
|
| |
We'll just subject them to the same developmental
benchmark tests we do children & breed from those that do
best. |
|
| |
All we're doing is selectively breeding brighter
chickens to see if their greater neuron density will achieve
human level intelligence with a smaller brain than ours, it's
a very tight project objective. |
|
| |
We're just not interested in philosophy & all that other
fluff :) |
|
| |
I agree, I have zero interest in philosophy either, but
knowing what a thing is isn't "fluff"! Especially if you're
proposing a method for generating it. If your target
objective is to answer that specific question, then I predict
that the answer will be "no" for the reasons mentioned. You
will likely have some very smart chickens capable of all
manner of nut-retrieval tasks, but none will show "human
level intelligence" except on some very narrow metric. |
|
| |
//Especially if you're proposing a method for generating
it// |
|
| |
Very much my point, we're not, chickens already have
intelligence, really small amounts but they have it. |
|
| |
This isn't
an AI project to create intelligence from nothing (which
your dissertation seems more suited to, to me) & there's
little profit in comparing the two, when it comes to the
question I want answered (can birds greater neural density
achieve human-like intelligence with smaller brains than
ours) none. |
|
| |
We're just breeding for
more
of something they
already have. |
|
| |
Something already well defined (if not well understood) in
human terms. |
|
| |
It's in
human terms rather than any other that we want it & for
the purpose of this experiment we
don't need to understand it beyond their ability to succeed
at the tasks we set
them. |
|
| |
So most (or all?) of your anno is superfluous,
interesting, but superfluous. |
|
| |
What's your selection criteria? Looking at the work on
crows/ravens etc. they pass tests that things like chickens
don't. What we need is a test that the smarter chickens
pass. What do you suggest? It needs to be high-throughput,
fast and cheap. Later on things get trickier. As intelligence
grows, the tests become longer and more granulated. |
|
| |
//As intelligence grows, the tests become longer and more
granulated// |
|
| |
Yep, probably starting with clicker training & similar
that they are capable of, those who perform best
& learn fastest would be the obvious criteria but we're
open
to suggestions. |
|
| |
When the first one gets a degree from Yale we can stick a
fork in it & say we're done. |
|
| |
We'll have a chicken dinner to celebrate. |
|
| |
Anyone who thinks they are smarter than a chicken
is welcome to set up a camera and record
themselves laying an egg. |
|
| |
//And as it is in human terms rather than any other// if that's your perception, it was certainly not my intention - which was to be
general, if specific. Much of the practical, non-philosophical thought about intelligence is based on empirical efforts in computer
science, so maybe some of the language around that sounds "computery". But intelligence is still intelligence, even if it's tricky to
define. |
|
| |
But testing for a thing, as other annos point out is tricky if you don't define that thing. |
|
| |
My dog is one year old, but she out-performs a one-year old baby on most intelligence tests I can think of. (Following commands, finding
hidden treats, anticipating the movements of other creatures, making simple plans to carry out her intentions). |
|
| |
I don't think "Human Level Intelligence" is just a point on a linear scale that you can turn up to eleven. It's a very specific series of
evolutionary adaptations that allow individual humans to process thousands of years of information (via stories, teachings and culture)
into their survival strategies that would otherwise not be available via purely biological pathways (DNA etc) So to read symbols, parse
them into concepts and use those to build internal mental models that can predict complex behaviours that have never been directly
experienced by an individual are all very specifically human. Unless you include this kind of historical pathway for success in your
experiment, you're not creating the evolutionary pathway for such functions to evolve. |
|
| |
Equally, given a nut-finding task, how would you design an incremental test that both a smart chicken and a human would pass, but a
stupid chicken would fail? |
|
| |
But, if you believe that intelligence is just a simple sliding scale based on brain-mass/neural density (though I'd agree that these might
be necessary, if not sufficient) then this might be an opportunity to demonstrate just that. |
|
| |
There are examples of "wet-ware" computers that use biological neural material to make decisions after being trained on data, but I
think the sense there is that these are just machine learning techniques implemented in a non-standard infrastructure. This though is
interesting because human intelligence allows people to make decisions not on statistical data, but via the abstract models they've
learned about second-hand through their language and culture. (Demonstrating from the opposite direction what "human intelligence" is, and what it
probably isn't.) |
|
| |
On an admittedly philosophical end-note, how intelligent would a human be if they had never been exposed to language or culture? How
would you compose a test that this human would pass, that another creature would not? I think one reason why intelligence is so tricky
to pin down is this symbiotic relationship between biological and cultural environments and adaptations. For humans, and by extension, human
intelligence, the two have gone
hand in hand for so long, they may well be impossible to decouple. |
|
| |
Not tricky at all. We'll be defining it by (comparative to
humans) results. As I've been trying to tell you, for our
purposes here we simply don't need to
define it any further than that. |
|
| |
When they can get a basic high school
qualification, a score
of 100 or better on a normal adult IQ test & hold down an
entry level desk job
we'll say they're as intelligent as a
person. |
|
| |
See what I mean by superfluous? |
|
| |
//When the first one gets a degree from Yale// |
|
| |
//We'll have
a
chicken dinner to celebrate// |
|
| |
//See what I mean by superfluous?// |
|
| |
So functionally, you need it to eventually process language, make abstractions, and
build mental models both of things it experiences out in the world, but also of more
conceptual things that it will never have direct experience of. |
|
| |
For these things to happen, it needs to be able to live through some kind of cultural
developmental stage, go to school, have friends with other like-minded chickens so as
to learn about inter-creature relations (nobody is going to employ an ill-mannered
chicken who doesn't gel with the team) and develop the kinds of motivations we
humans satisfy through the attainment of wealth through gainful employment (rather
than, for example, simply scratching about for another juicy worm). |
|
| |
You can either provide these contextual/environmental things as and when they're
necessary, like a great omniscient benefactor, or you can allow them to to evolve and
emerge spontaneously. There is the problem of you ruthlessly murdering a percentage
of the population each generation which they might, if all goes according to plan, start
getting suss to. |
|
| |
Either way, in practice, I think some of that might end up being a tall-order. |
|
| |
/How do you want yours?//
Chicken katsu curry, over rice please. |
|
| |
//ruthlessly murdering a percentage of the population// |
|
| |
[Slightly worried guilty expression] |
|
| |
Well we thought we might just dump all the rejects on the
Isle of Man & let them get on with it themselves. |
|
| |
Just eating them would be unconscionable, of course. |
|
| |
[Carefully doesn't look at plate of drumsticks on table] |
|
| |
[Pushes plate slowly out of shot, still not looking at it] |
|
| |
//might end up being a tall-order// |
|
| |
We'll probably have to knock it all on the head if we ever
reach the three year old infant benchmark to be fair. |
|
| |
You know what all the bleeding heart types are like, they'll
make a fuss if they ever get wind of it at that point. |
|
| |
Without defining anything, the process results in the monkey typing exercise and infinite time. By trying to defining 'the fluff', insights to short cuts to training/tests become apparent. |
|
| |
Might I suggest starting with something other than
chickens? We have some, and they frequently astound me
with their stupidity. I assume it's because we've bred any
intelligence out of them. The "smart" ones escape and get
eaten. I wouldn't count on chickens having whatever brain
density advantage that most birds have. |
|
| |
On the other hand, creating an environment for any of the
intelligent birds will be quite challenging, whereas with
chickens, if you could create a peck screen interface and
partner with a traditional caged egg laying operation, you
could roll it out on a massive scale for a relatively low cost. |
|
| |
//For some, consciousness [...] is a symptom of a system trying
to model other systems similar to itself // |
|
| |
No offense to you, [zen_tom], but that is a terrible, terrible idea,
and I'm disturbed to see it still in circulation. Can you provide
any links to recent sources for it? |
|
| |
No offence taken [pertinax] it's an idea I heard from
listening to cognitive scientist Joscha Bach, it resonates
quite well with me anyway, evidently you're not of the same
opinion. I am sure it would be interesting to learn why
you're so vehemently opposed to the idea. |
|
| |
It's also an idea not a million miles from those explored at
length by Douglas Hofstadter in various of his books. It fits
very well much within the "Strange Loop" context that he
sets out. |
|
| |
//wouldn't count on chickens having whatever brain density
advantage that most birds have// |
|
| |
They do I'm sure, we should maybe check b4 starting but as
it's likely an adaption of birds for flight (to cut
weight) it
would be surprising if they didn't. |
|
| |
They have pretty small
brains <link> only
2.4-
3.9g compared to a Ravens 15g brain but we'll bulk that up
by
selective breeding as part of the project, the ease of bulk
breeding should more than make up for the poor
starting material & ultimately result in a much shorter
project lifespan beginning to end, we can of course select
for neuron density as well, they'll have the same
'tools' as
other birds in their genome for that so given time we
should be able to get
it up to Raven levels in their forebrain if it isn't already (I'm
guessing it's not). |
|
| |
I expect they have better neuron density than
mammals (but in a much much smaller teeny tiny brain of
course), but I'd guess their neuron density might not be as
good as a Ravens. |
|
| |
Here <link> ("Birds have..") this study included chicken
brains &
they make
no mention I can see of chickens being surprisingly unusual
in not
having this adaption that other birds do. |
|
| |
From that study: "We found that the bird brains have more
neurons than mammalian brains and even primate brains of
similar mass" & also, in specific reference to results
from
"red
junglefowl" (aka a chicken), they say "Thus, high neuronal
density
in the telencephalon appears characteristic of all birds" |
|
| |
Thank you for telling me where you heard it, [zen_tom]. |
|
| |
To understand the origins of the idea, (and hence the reason
for my hostility to it), we have to go back to the state of
psychology in the mid-twentieth century, when the field was
dominated by psychoanalysts. |
|
| |
Hang on; I have to go and do some other things now: I may be
able to expand on that later. |
|
| |
We did try other species. |
|
| |
The Kakapo seemed like a good
idea because it can't fly & is really quite friendly <link> |
|
| |
Success was limited, no eggs you see & there's not a lot of
them left anyway, was getting difficult to smuggle them
out too. |
|
| |
So we dropped the
hybridisation efforts & switched to
chickens. |
|
| |
//I'm not convinced that breeding more "cauliflower" shaped brains is going to provide the benefits we're looking for. // |
|
| |
Even if it doesn't it's sure to give us something scientific and culturally interesting. |
|
| |
We should also try breeding for purely larger brains regardless of intelligence. I'm sure we can get chickens with relatively titanic brains in a relatively short time and learn something from that. Actually I honestly think that would be a worthy study. |
|
| |
...and yet living headless chickens might be the way to go.
Meet Mike the headless chicken. |
|
| |
//breeding for purely larger brains regardless of
intelligence// //would be a worthy study// |
|
| |
Of course, though it's largely built into this one anyway,
we'll
be measuring for neuron density, brain size (regardless of
neuron density) & brain to body weight ratios to select for
them in the breeding program as well as
testing for intelligence thus providing (with others) the
data
sets you'd be
looking for with a pure brain size study, I'd expect it to
prove neuron count (particularly in the forebrain) rather
than overall brain size or brain to body weight ratio as the
pertinent factor for intelligence & help determine a
maximum neuron density. |
|
| |
//My dog is one year old, but she out-performs a one-year
old baby on most intelligence tests I can think of.// |
|
| |
That's because the dog is essentially complete, it's still
got far too much juvenile energy but it's more or less an
adult. The human is 10-19 years away from that point,
the brain is just about getting the hang of moving about
and handling the firehose of information that the eyes
provide. |
|
| |
I wonder if longevity is a necessity in the biological
intelligence game? All the intelligent creatures are
relatively long lived: Ravens, dolphins, us, chimps,
elephants etc. Are there exceptions to this? Is there a
mayfly that gets to the intelligence of a cat in a week? Is
there a maximum rate at which neurons can grow a
functional network? |
|
| |
Actually, separate those, is there a maximum neuron
growth rate? Yes, growing billions of them takes time,
although the largest cells in the world are neurons in
things like the sciatic nerves in blue whales that clearly
grow at a tremendous rate in absolute terms. |
|
| |
Is there a maximum rate at which a functional network
can be assembled? This looks like it might be the main
problem. It seems to take many months to years to
develop the sort of intelligence we're talking about. That
process is essentially expose to stimulus>wait for the
brain network to respond looks like it's reasonably fixed in
terms of speed - at least within a factor of 10 or so. |
|
| |
So while chickens might be long-lived enough, and you
could breed them to be more intelligent, I expect that
they would just have longer and longer juvenile stages. |
|
| |
//bird brains have more neurons than mammalian brains
and even primate brains of similar mass// |
|
| |
There will be a reason. It's not like mammals are just
wastefully filling the brain with non-neuron cells. There's
tremendous selective pressure on brain volume in
humans, we have huge heads that are only just able to fit
during birth, then they keep expanding afterwards. Those
non-neuron cells are doing stuff, stuff we're still working
out, like physical and metabolic support for the prima-
donna neurons. As brain size increases, it's probably that
the support network increases disproportionately, and
unavoidably. I wonder if the 40C body temperature of
birds helps in this at all? |
|
| |
//There will be a reason// |
|
| |
Of course there will, & it will be flight .. lighter is easier &
that
was pretty obviously the driving factor for natural selection
of lighter brains in
birds the same as it was for their hollow bones, I would be
unsurprised to find that there are reasons
it
won't scale indefinitely, for a start there's going to be a
absolute limit for how small you can make a neuron &
how tightly you can pack them, there may perhaps be
cooling
issues with all those neurons sparking away that cause
problems in larger brains than any birds have. |
|
| |
I'd like the
answers to those questions & this project should help get
them. |
|
| |
//I wonder if longevity is a necessity in the biological
intelligence game?// |
|
| |
[Makes a note to include the chickens ages in all the
collected data sets] |
|
| |
We'll let you know when we're done. |
|
| |
// there may perhaps be cooling issues with all those
neurons sparking away that cause problems in larger
brains.// |
|
| |
That's been solved. But you're astute to identify that it
should be a problem. The brain is water (blood) cooled,
and the relative amount of blood vessels has to increase
with brain size. Then there's the metabolic tricks: by
running on largely glucose>lactate neurons offload the
heat generating part of metabolism to the mitochondria
in the liver (and glia etc.). By minimizing the amount and
utilization of mitochondria, neurons likely minimize the
damage associated with high-energy O2 reactions. But,
neurodegenerative disorders almost always feature
malfunctioning mitochondria, so there're very much
needed. |
|
| |
I wonder if stepping up temperature might be a way to
go, just going from mammalian 37C to bird 40C gets you a
~5% increase in diffusion of most of the important ions &
molecules. |
|
| |
//flight// //the driving factor for natural selection of lighter
brains in birds// //as it was for their hollow bones// |
|
| |
I'd be unsurprised to find they've developed similar
advantages for muscles as well .. another study perhaps,
if there isn't already work out there on that. |
|
| |
//if there isn't already work out there on that// |
|
| |
There is, some at least <link> (Evolution of Flight Muscle
Contractility and Energetic Efficiency) |
|
| |
//they've developed similar advantages for muscles as
well// |
|
| |
Not really. Again, there's a lot of selective pressure on
muscle in, I assume, all species. They recycle a lot of the
same tricks, but there's no getting away from the
fundamentals. Muscle contraction requires actin/myosin
which requires Ca2+ and ATP to cycle. Those require
mitochondria burning fuel in oxygen. You can move things
around in time and different parts of the anatomy, but
you need what you need. Bats fly, and their muscle is
indistinguishable from rat muscle, they don't have most of
the bird tricks either, bones are standard. |
|
| |
The extremes get interesting. Really fast muscles have to
pull some strange tricks. Rattlesnake rattle muscles
(~70Hz) & Cicadas (150+Hz) have to swap out most of the
actin/myosin force producing stuff and cram in Ca2+ &
ATP machinery (SR & mitochondria). |
|
| |
Bees are nuts, they flap their wings over 200Hz, which is
faster than neurons can operate, meaning they have no
control over the frequency, they just turn the wings on &
off. The muscles seem dependent on stretch activation,
the upper muscles stretch the lower, which contract,
stretching the upper etc. which sets up an oscillation of a
relatively fixed frequency. There's a whole lot we don't
know about how that works. |
|
| |
//cooling issues// //That's been solved// |
|
| |
Solved may not be quite the right word? it's been
Identified as a (very) probable issue certainly & I'm sure
there are papers that offer theories on what & where the
limits
it imposes are. |
|
| |
But I don't
know as there
have been any actual practical studies to positively locate
(define?)
& prove
those
limits? |
|
| |
What I was getting at, was that nature has solved the
problem, or worked around it at least. |
|
| |
Those poultry farms are on a large scale. At least we can eat the genetic losers. |
|
| |
Any pâté-like recipes for the excess in chicken brains? |
|
| |
If nuts are also bees, we could have a new recipe for nougat. |
|
| |
But peanuts are not nuts. Therefore, by the operation of logic, bees are not peanuts. |
|
| |
// by the operation of logic, bees are not peanuts//
That reminds me of the Norm Macdonald (R.I.P.) joke
about the professor of Logic at the University of
Science (the telling of which [link] coincidentally
features chicken) |
|
| |
[Pertinax] Can you follow up on our comment about why the
idea in [zem_tom]'s annotation is a // terrible, terrible idea
//. I'm not saying I agree with the idea, but it sounds like an
interesting hypothesis about consciousness, and I don't see
right off why that idea could be considered terrible. Based
on the brief start of your comment, I wonder if this idea was
used as part of an excuse for some inhumane treatment of
mental patients? |
|
| |
[scad_mientist], I took that discussion with [zen] to email,
because it is off-topic, complex and likely controversial. But,
yes, mistreatment of mental patients is part of it, but also only
the tip of an iceberg. You can find my email thinly disguised on
my profile page, if you don't mind listening to me whinge and
rant about it, but I thought best to keep it off this comment
thread, and maybe of the HB altogether. |
|
| |
I called a man a bird brain once. Strange that he took
offense to it given their neuron density. |
|
| |
Bees must somehow interject some level of control
by queueing up a neuron or 3 that get on the CAN
bus and interrupt or slow the process a bit. |
|
| |
^Auxillary muscles might just move the oscillating wing into a strained position that slows up the back and forth allowing nerves to get some messages in. |
|
| |
Though, it's the right way around, right?. It's ok for the effector to be superspeed rather than have the controller superspeed and a slow effector missing calls. |
|
| |