good Monday morning this is again like
Fredrik is back from last week we’re
actually recording these back-to-back
but and we are last week we looked at
pics link which is a project where kind
of like a neural network sandbox so
let’s be yeah it’s like an evolution
simulator where the fixings agents or
the creatures in it has been your
networks yeah it’s a game I might like
it’s a little game sort of we can it’s
kind of like a Minecraft for kind of
programmers that want to learn about
about neural networks and evolution so
no well if you like you don’t have to
watch that episode to to watch this one
body my cap so you can find it there or
there I will link it it’s also in the
episode description so what are we gonna
do today PJ from the future here to tell
you that today’s sponsor is brilliant a
problem solving website that teaches you
how to think like a computer scientist
if you are interested in learning
computer science fundamentals binary
trees quicksort big ole with that kind
of stuff instead of just passively
listening to video courses they have
interactive challenges with increasing
difficulty so it’s kind of like a game
with a sense of progression it’s very
satisfying use brilliant dot org slash
FFF when you sign up to let them know
that you came from here also if you are
one of the first 200 to use that link
you get 20% off the annual subscription
if that sounds interesting to you make
sure you check it out after the show
that link is also in the episode
description so what are we gonna do
today yes the last time we started
building a simple world and we started
giving the pics things some abilities
yeah and maybe for the people that
didn’t watch the last episode quickly
like what you do in this game is that
you build worlds and you put these
creatures into them so we started
building one of these worlds and the the
worlds are online this is where
editor so we’re gonna link this world
that we’re playing with right now in the
episode description so that you can
check it out and play around with it as
you as to watch this video basically
yeah so I think today since what we did
last time we started giving them the
ability to kill each other and also give
them this kind of weird ability to sleep
and do nothing
yeah and I think also humans have yes
it’s such a weird ability so we were
thinking maybe let’s explore sleeping a
little bit and see what we can do with
it I think we don’t have like really a
plan of no this this is like last
episode we have planned out a little bit
like figuring I would have a couple of
like we’re gonna do this and this and
this this is gonna be more like how you
would actually use pick sling yeah like
yes hey I have this like why do we sleep
how did that came to be an evolution
like it’s really strange really because
I’m reading this book now called why we
sleep which summarizes like the last 20
years of sleep research and it’s only
recently like we have figured out why we
sleep earlier we were just basically
dumbfounded like we are stick okay
animals sit down for eight hours and do
nothing like and they shut down
everything so they were super easy to
eat and we don’t know why and all
animals do it like there’s no like it
all animals that didn’t sleep they got
like terminated by evolution right so it
seems very beneficial but we don’t know
what so let’s see if we can make the
pics things sleep them and see if they
can somehow get any benefit from this
yeah or maybe they won’t know we might
not be able to do this like who might
fail so let’s try to do this let’s get
into it I think the first thing we’re
gonna do is that we’re just going to
open up the editor and take a look at
what we already have
alright let me pause this thing ah
they’re cool
I’m going to check this out like this I
wanna look inside the pics link I love
it yeah because we didn’t do that show
advanced yeah go for it
so here you can see okay so it has
abilities that we defined before like
the killing properties yeah why does it
have to
so actually what happens that is that
sometimes they randomize the abilities
that I have so in this case this guy
actually ended up having the same
ability twice no but you can actually
changed the second one to a sleep
ability oh and now he’s got two
different abilities cool and then it has
a bunch of properties like State and hat
and where it is and in the world yep and
it has position and it has brain inputs
right so this is where we get into the
whole neural network world yeah because
if you didn’t watch the previous episode
every single pixel in has its own little
own little brain has a neural network
inside of it exactly and they’re a
little bit like you know living in the
matrix that they only the way they
perceive the world is used through
numbers so their inputs like their eyes
or their senses it’s just a long list of
numbers yeah and the senses are
basically these not quite okay that’s
the properties and then down here or
actually the is the brain inputs and
some of them are some of the properties
some of them are things around them in
the world yeah we’re gonna see how we
define the environment of the pixels
right so all right so it takes like a
current state of the world and what he
sees and that’s the brain inputs like
eyes ears there and then you go like get
these brain outputs which is move and
focus and writing to memory
exactly so exactly he takes his inputs
processes that through his neural
network and improves produces basically
his choices or is what he wants to do in
this world the movie decides which
direction he wants to move and then at
the bottom you see their ability slots
that’s deciding which ability he wants
to use in the certain situation that
he’s in and if we keep going down we can
actually actually see his brain yeah
alright so this is yeah this is a neural
network basically this is a you know all
the weights for a neural network exactly
so the way neural networks works is that
you take the inputs and then you run
these mathematical operation
together with the weights yeah and then
that is what produces the output and the
cool thing is that these each little
weight here matters for the output yeah
so if you change one of these numbers
the behavior of the pics thing might
change drastically yeah and that’s the
way the the inheritance and the
evolution of this works because the
children each of like their weights
which of them might have a small chance
to actually change a little bit so like
when we say like all this brain like
these weights and we take this inputs
and get these outputs and we somehow use
this way is this very complicated math I
would say not super complicated it’s
there’s a few operations it’s matrix
multiplication and then you have basic
summation with the biases yeah and then
you have what’s called an activation
yeah and matrix multiplication is for
like I don’t have a math background like
you like it’s basically our matrix is a
multi-dimensional array exactly it’s
worth two-dimensional it’s kind of like
a table of numbers yeah and then you
just matrix multiplication it’s
basically just taking a matrix and then
another matrix and then multiplying them
which means like multiplying that value
with that value in a certain order
exactly exactly yeah you use regular
multiplication and addition and you
combine these numbers yeah so there’s
it’s basically just times it’s like it’s
not it’s like some kind of weird simple
no no it’s very simple and then the the
final thing uses what’s called a real
ooh is the fancy name exactly as the max
and at the end of these so it’s kind of
this thing where you take where you kind
of cut the number off and if it’s
negative then it becomes zero and if
it’s positive it stays the same okay so
basically is the way of normalizing the
numbers to a certain format sort of yeah
yeah and that’s basically the neural
network of these pics things all right
cool yeah there was a digression I just
wanted to like show that are they the
interior of this thing is it’s a bit
complicated but not at all that’s
complicated that you might think yeah
exactly
neural networks is one of those things
that sounds insane
when you
you hear about it like oh we’re building
a brain and moving on the brain and like
this axis norms and dendrite it’s really
a pretty simple math it’s very simple
actually and I would add one very
interesting things about neural networks
so neural network is it’s it’s like a
function right it takes inputs produces
outputs and it’s been proven that if you
have a large enough neural network it
can actually model any other function in
the world oh so the code was there
what’s called a universal approximator
and the reason this is interesting for
fixing world is because that means that
there is no behavior that these pixels
can have with their inputs but that they
can’t have like there’s there’s there’s
no limit to what they can do as long as
the neural network is big enough of
course they know that they’re quite
small but if it was big enough then it
could essentially do anything that is so
cool yeah it’s just an airline your
networks are insane it’s very
interesting such a little simple thing
that isn’t cool all right okay yeah
let’s get into it let’s give them let’s
give this a bit more detail yeah so I
think the first thing we’re going to do
is that we’re going to give them an
environment to interact with alright so
if you go ahead and hit add environment
on the Left yep
so now we’re getting to the editor and I
think then right we’re going to create
this something for them to eat so I
don’t know we can change the colors and
we can change you can see the masks up
here yeah and there’s a couple of things
that we can choose from yeah we’re going
from from function theme we should have
like Buddha but we don’t have that like
it’s not even okay yeah I suppose like
up yeah exactly
upload graphics area would be like well
but I do i gold nugget seems like it’s
closest thing sure let’s go with it
and then you guys should change the
color so you can zoom in on the right
here to see a preview of this
environment all right
oh this it’s not too far off actually
yeah it’s pretty good you can change the
min and Max colors to whatever you like
and then what we’re going to do is a
very simple operation with this
environment so each environment way it
works is that it’s it’s just a single
value for each cell in the world okay so
what we can do here is that we can
increase that value for each cell in the
world at each time step see so what do
you mean like there’s more bootlegging
one second it’s certain exactly yeah
we’re gonna grow pull out in in the
world of to fix things so if you go
ahead and hit Plus on each step then you
can change that set value to an
increased value now just increase I
guess yeah and I think we’re pretty
happy with that and then finally we want
to add one more thing here which is
gonna be a clamp yeah change this one so
the guys who never rises above a certain
value exactly so it can’t go above 100
or below and if you hit update now we’re
gonna get this in our world and we can
see it at now is standing on Clara its
bola and if you play here
we can actually see that right yeah okay
so alright cool
I wanna yeah all right I’m with you all
right so this is I know I need to change
yeah like are these colors good yeah I’m
good with the colors it’s gonna be fine
so like this goes from brown when it’s
like almost no bola and then like really
gold when it’s a lot of bullet okay so
the second thing I’m going to do now is
that we want to give the fixing some
kind of property that they can so they
can eat from the ground and transfer it
to their property you know like in the
game you might have cows or mana we’re
gonna give them something like this so
if you open up the editor again here
yeah they all actually lead to the same
kind of place but okay so you go into
properties oh so pics ling-su is
basically all of the pics links and then
like this years happens to be selected
no it’s more that this is kind of like a
prototype picks thing it’s like the
Sheep the Dolly the sheep of pegasi and
he only exists in the editor but then
you can take him and create many
instances of him in the world I see this
is actually where prototypal inheritance
sort of start to make sense in this
context exactly yeah all right cool yeah
let’s go to properties and then we
create a new property and you can give
it whatever name you want all right
we’re gonna use this to clicked Bullard
yeah okay so like it smarts because
we’ve become smart if you perfect we can
change the arm reproduce too let’s make
that a split what does that mean so if
one pic sync clicked it a lot of Buller
and then he reproduce then the child is
gonna get half of them and he’s gonna
keep my first time it’s a sharing
economy all right so yeah but that’s how
I don’t know that’s how inheritance
works in Sweden yeah cool and then
unless there’s a spouse like all right
and then let’s actually give him an
ability to eat eat these things then so
if you go to everybody’s yeah we let’s
actually change this whole thing that
would be last time so in certain if up
here yeah let’s change that into a
transfer up to yeah perfect
transfer to yeah right they’re similar
but and you can do it from so you can
change this focus fixing property if you
click that and then go environmental
pixels position environment InFocus I
know the top one air oh okay
and then you change the environment from
energy to the environment that you
create and then the Pixum property you
want to keep that now when you want to
change generally to smart and then you
can adopt it and then you can create a
few more of these guys all right cool
oh my god and you’re gonna hit play up
here and actually you can change this
alright we need to change this more oh
no this is the I picked the worst color
like it’s just hiding all the pics links
oh it burns there and what would be a
good color that is less I think actually
if you change the mask maybe to like a
neutral mask maybe grass for now and
then it’s gonna be a bit easier to see
no that’s cool that’s better so if you
go ahead and hit play and if you change
the extreme to maybe just hyper so we
can see now that in the background of
these pics links to grab the bar that we
defined is lower for some of them right
yeah they kind of like eating this and
then they are coming back up again yes
that’s how we built the environment but
and you can actually sue now again yeah
we can see it on these guys here yeah
and you can actually see these like
health bar something right but I have
filled up their health bars because
they’ve been eating from the ground all
right so it’s like and the health bar is
it like I guess at that points to some
default property so it’s if it’s just
one property that’s gonna be the health
bars or yeah actually you can you can
have multiple health bars okay if you
had more have properties than each of
them can actually be showing off how far
alright but they’re actually not using
this this but here we see like this
these just flying around and eating yeah
exactly but there’s no there’s actually
no point to the meeting right now they
are not they’re not using this for
anything
so it’s jumping to the editor again and
then we’re gonna actually make this
requirement that they have to eat to be
able to reproduce oh yes let’s jump into
abilities and then we’re gonna create a
new ability we can actually change the
name of this one that’s right alright
yeah which one so we can change the name
of the one we just created oh yeah
no no no no no yeah perfect yeah well if
you don’t know what that is
you’ll ask in the comments cool and
let’s give them one more ability to use
this bowler and reproduce yeah all right
yeah I guess like birth yeah and what we
want to do now is that you won’t even
use an if statement so yeah yeah and the
condition is gonna be that they have to
have eaten enough blood
otherwise they can’t reproduce oh yeah
if so if hang on maybe I can do this
if see yeah if any of the cup following
conditions true maybe okay no I can’t do
it this so you can see because I was
looking for like if a property is yeah
you were actually close to it so if you
go back here you can see that this first
one so I comparison oh okay
so like if pics link property smart is a
bomb like I don’t know 10 perfect then
it can reproduce no yeah okay I need to
delete this that because the den run
says it here this is perfect
let’s try this and you might have to at
this point we might want to restart to a
simulation because this is lot of yeah
yeah these don’t have these abilities
right exactly they want to have the
reproducibility as mouse to animate yeah
so it doesn’t change because it’s it’s
not a class based model it’s a prototype
based model so like they it will like
only new ones will be spawned from a
prototype it won’t change like the exist
exactly so if you change the code inside
of an ability then that will update if
yes the abilities a global but like as
if it’s assigned exactly exactly alright
brilliant
I’ll restart it and we’ll paint some new
page links whoops
click this paint paint pics link button
here now get some cool and then we can
see to the right here that they’re
spreading over here yeah so yeah these
are the they learn to they have learned
to eat or they were pics links that that
ate because of randomness and that means
that they they spread a lot more yeah Oh
safe home yeah I just I want to say
feature I don’t know what to say that’s
nice but there’s a bunch of problems
with this world like it’s such a simple
world for instance there is no way for
them to die right now know what just
well yeah exactly and that I guess that
creates a problem with a world where
things don’t die is the this is actually
an interesting has become like the
purpose of death like it becomes like it
becomes very stagnant after a while
exactly because there’s no developments
if you if you don’t have death they were
going really deep really but yeah if you
don’t have death in a world like this
then I don’t think that death is very
deep in the context of an evolutionary
system that like super basic yeah that’s
true that’s true but yeah like really
for them to evolve and become more
interesting you need to leave space for
like the next generation yeah and right
now they’re not really doing that so I
think that’s actually a probably good
step for us to to add some way for them
to die as well yeah perhaps if they eat
too much blue that’s all right so so
let’s see environment rule no no okay
abilities see how if I can do this
actually you were it’s an ability to die
well actually you were on the right tab
that rules might actually be a good way
to use because rules are kind of like
these global things that you can run
yeah and they always run they don’t the
pieces don’t choose to
or not yeah but I figured I could use
this as a no I could model this as an as
an ability I think all right yeah like
because now I want to use what I’ve
learned I’m sure there’s a better
version way to do it but I’m gonna like
do the ability overeater and so if this
like this also no it’s it’s bigger than
a fixing property smart okay we need to
change that to like fat instead it’s
above no I got it it’s about oh yeah
this we want a constant here yeah
where’s the constant oh it’s not a
constant it’s this oh this is not in
alphabetical order
already some special yeah there’s no
real good order at all right now it’s a
little bit of miss yeah but this is beta
very beta software yeah but it works
really well anyway it hasn’t crashed yet
well fingers crossed but yeah so if I
could say that it’s like 25 or something
yeah and then wrong die yep so it might
not necessarily like some fix links
because these are chosen by the pics
links like they do have free will
picking between the abilities so I won’t
always die when it hits 25 but it’s like
it’s likely to die after eating one yeah
yeah and I want to change this property
now to call it fat instead cold we’re
gonna have to just update them in all
right all the places where you use them
because it’s yeah it’s me it’s weakly
typed it does not matter it fat but it’s
generalised that it gives me like the
errors and because everything is running
all the time I suppose yeah
well actually the game get becomes post
when you open the editor but not here
okay inability overly snow yeah but
there we go debugging cool so I’ll let’s
restart the world and see yeah so yeah
so now we have like some I don’t wonder
if we get some pics links at overeat
here yeah I see that
so I can I can tell you right now that I
think there’s gonna be one problem with
this yeah
because as you create this and as an
ability they choose the wrong ability
yeah so if there’s there’s very likely
to they are very likely to reproduce
during the period between when having
enough fat I think you reproduce and I
think both that but it’s also that they
would have to choose to kill themselves
oh that that strain yeah so this strain
the breed of pick slings that or
choosing to kill themselves when they
overeat they will very quickly die out
it’s like he gave them a gun and you
made it so they can only shoot
themselves and of course that’s amazing
because I didn’t think of that because I
don’t think in in AI I don’t think it
your record I think in programming so I
just think oh the choice of course
that’s just a random random function but
no they will they will they learn
actually learn or they they won’t learn
it’s just that they genetically they
will learn exactly I mean the ones that
choose to kill themselves they won’t
spread so it’s just yeah that’s gonna be
enough using that one that’s oh that’s
actually such a good example I actually
died I learned something okay so yeah
okay so let’s that didn’t work we have
to hold on you can actually what you can
do is that you can actually copy this Oh
see if you right click little thing here
and only copy all right
and then now we can delete it but then
we have this in our clipboard right and
now I can create a rule yes all right so
if this cap paste here no actually but
like so it’s a little bit buggy still
but if you click the plus on the second
one on the one below oh and then click
the just click it normally like yeah
left click and then right click this one
and then click oh there room so a little
bit cumbersome guys it’s gonna become
better in the future
alright but run the following every step
ok
does this look right if the over pics a
property fat then wrong die exactly and
then I made it like this so now they
don’t have a choice to run this like
it’s always gonna run yeah so now they
better watch out for overeating alright
so like the old overeating rule yeah so
I love that little it’s like I love that
so let’s maybe restart it and see if we
can see what we can have happen alright
let’s see what happens and I I don’t
know what’s gonna happen because I’ve
never tried this either now that’s a
very quickly new experiment it seems
like they are not actually eating from
the NIE
anymore or not I won’t expect the shorts
happening yeah and they’re not
reproducing either they are not perhaps
if we catch some more because this is
something that we learned I learned from
the from the other episode that
sometimes you just need more six links
yes it’s all like how our life will like
life in the universe is very unlikely so
we needed basically an infinite universe
for life to happen
exactly so let’s paint a bunch more
fluffykins all right they seem very
reluctant to tewi reproducing so no they
say yeah exactly the scene no acts like
they seem reluctant to eat in the first
place right Yeah right right so why
might they not be eating all right let’s
and let’s have rules
so if pixie Papaji fat is above 25 then
run die
should we just increase the number or
maybe that’s okay so let’s say that they
have to beat 250 but it’s still weird
it’s still like it they they can
reproduce after 10 yeah right it’s kind
of odd all right it’s pretty science oh
now it’s all right and then a bunch of
them use died there as well yeah so they
like their ro oh yeah exactly they are
overeating so actually what you can do
is that you can you can actually lower
this number now yeah and it will affect
see if you run open up the editor and
make it 25 now it might be that they
have evolved enough that they will
survive oh that’s pretty cool so they
can now yeah so they’re now clever
enough to survive the harsh credit yeah
now a lot of them died out but some of
them are still surviving so presumably
babe yeah but I mean over time there
will I think that they will if not
everyone dies that means that some
strain well figure out how to not
overeat right yeah maybe all right yeah
because this is a kind of like not it
turns out that like maybe not eating too
much is a very tricky thing to do well I
think actually what they will never
learn I think I mean I mean it reflects
like the problem that we have in where
food is abundant like they have so much
food it just grows in men sleep fast so
they they their food is never scarce
here right all right so I mean it’s it’s
it’s kind of one of these things that
it’s actually not too hard for them to
learn because the rule is so simple it’s
like do I have too much food yeah don’t
eat any more so it’s usually yeah work
slots as you see like it also it seems
like they have a trouble finding the
balance though because no no breed here
is is dominant
that could be a breed that was very good
eventually at eating just the right
amount then they would completely
dominate exactly yeah but it’s not like
they’re there it seems to be like they
seem to be like not yeah you’re right
stop eating basically yeah it’s
interesting and stop reproducing
maybe we should punish them if they stop
doing like because it seems like when
they’re likely what you’re saying that
they’re just not doing anything right
now no exactly they’re not starving yeah
so we’ve perhaps you should add
starvation why do I notice that we are
completely like gone away from the sleep
thing here that will happen well yeah
all right actually one thing I can teach
you here is that if you click on one of
them if you use this tool and the other
one though
yeah the selector will pick this one so
we can see that he’s actually he’s got
only two abilities and he’s alternating
between them yeah and it seems that it’s
actually it doesn’t even have the eat
ability has only got the sleep and a
birth ability oh so okay I see so that
is that is tragic reminds me of my old
high school friends anyway it’s also
it’s very slow right now like in
summation there’s our that’s about like
when the Select window is open like it
becomes the simulation becomes much
exactly it slows down a lot because it
needs to read back information all the
time all right cool but what we can do
is that we can give them at least the
three abilities because now these guys
are just surviving by doing nothing and
kind of boring so you can you can do
that by opening up the auditorium all
right and then going into picks thing so
and then changing the ability slot count
here to three abilities like that okay
it’s here and then we can actually
configure exactly which ones they are
gonna come born with one we Oh games so
you can set set them up to have all
these three and then a final thing you
do is that we can change the mutations
abilities chance to zero oh okay so what
that means is that they’re gonna always
keep this one oh yeah otherwise there’s
a chance for them to like replace one of
them oh so they basically evolved out
like fika in the first place or
something
exactly all right cool
so I’m just gonna restore it because I
don’t remember anything
and it’s paint so many pics links and
see what happens hmm okay they seem like
oh there we go there’s something on the
address Jesus yeah somebody is like they
I figured out how to you yeah this is
what I would have expected to happen
actual right yeah yeah like they seem to
oh they’re clustering in a interesting
way as well you all right it’s a weird
pattern and this is one of these things
that’s where I have absolutely no idea
what the evolutionary benefit now or
they might actually not be even be an
evolutionary benefit yeah just restart
this and just do it exactly the same
thing well what will happen yeah let’s
find out I guess yeah so she talked
about in the last episode like when you
when a pix thing is born like it gets a
little bit of a random brain so it’s not
the randomness doesn’t only come from
the fact that I positioned them
differently it also comes from the fact
that there’s just inherent randomness in
right in the universe like both this one
on the real one things on this one not
much is happening yeah and sometimes you
just have to spray more before we see oh
there we go yes guys do something I
can’t see if they’re also forming
they’re forming it Patten as well aren’t
they but it’s not the same pattern yeah
yeah it’s G it strikes me how organic
this is like it it really is life oh
yeah it’s there’s no like computer like
it doesn’t feel like a computer at all
yeah I think I mean I’m very interested
in like the theory of life and
artificial life and
I mean there’s this one thought that
maybe maybe life or like you know a way
of thinking of life is uniqueness yeah
and each of these picks things are
unique yeah like they’re not none of
them are the same or I mean there might
be two of them there with a very small
likelihood that are the same but it’s
just one of these ways that maybe you
know in some form of philosophical sense
they’re like a little bit alive because
they’re unique but there’s no one else
in the universe of everything is the
same as this particularly pick thing
here really cool cold should be well
we’re we’re at thirty five minutes now
which is like a long episode and I think
that we should should wrap up because I
think that this was pretty amazing
actually
alright because I think that wait what
what okay let’s wrap up what did we
learn but what did we set out to learn I
kind of forgot it was something about
sleep we started out with wanting to
model sleep and then I think we didn’t
really get to one thing sleep but now
yeah went into modeling yeah yeah in in
that’s like in a yeah in an environment
where like there’s enormous amounts of
food yeah there’s a there’s a abundance
of food in this environment I’m not sure
if we learnt like anything on you know
philosophical level of like well is it
good or bad do you know it sure than
anything like that but yeah I was
surprised by how like I thought the
neural networks would be better at
learning the whole overrate thing or
maybe maybe well if maybe that was not a
problem actually if I change the rules
to twenty five right now
yeah right now it’s right yeah yeah
right now yeah they are pretty good at
figuring out the right amount of eating
actually yeah it might sometimes it’s
just a case of like spawning like you
said before spawning enough of them and
then
it changes and it seems like in this
case there’s yet another way of
organizing themselves I don’t know if
they’re forming a pattern now as well as
well or not but yeah it seems like
they’re also forming it a slightly
different pattern this time all right
you know what we’re gonna give like this
episode is pretty long so what we’re
gonna do is that we’re going to live
leave the audience we’re completing this
and to they and modeling the concept of
sleep and evolutionary terms I think
that would be pretty interesting like if
you can do that leave leave a link in
the comments to your pics Ling world
again you can find the link to this
fixing world in the in the episode
description and I also like if you’re on
the phone fan forum make sure you post
it there and I think that that is it if
you have any questions or thoughts about
this like what like like you have access
to the Creator here in the comment field
so like if you that’s very good a good
way of getting support for you stuff
make sure you play around with fixing
work because it’s amazing like these
these tools to learn machine learning in
a way that isn’t watching Andrew and
Gees tutorials worry like oh math
symbols I think that’s pretty great
yeah and I think it’s it’s cool that
people people don’t understand evolution
all that much like I it just this gives
me a completely different newfound view
of what evolution is yeah for sure I
think it’s you it also blows me away all
the time when I look at it and just just
like you know how evolution actually
works the whole concept of not having a
goal that they’re just competing with
each other is always like fascinating to
me and yeah oh that’s such a good lesson
in life in general that what like just
like why does this exist why do we do
this why – people like because it was
somehow
beneficial for them to do it like rather
it made it that behavior replicated it’s
exactly it’s not miss it because like
it’s fair or anything like things happen
because it evolved to do that or
necessarily did not might not even be
better it might just be better in this
specific context yeah so yeah that’s
yeah it might also be that the people
like the the pics links that maybe they
just develop like some kind of weird
behavior where they just go like this
like if they had like motor functions
and it would be like why is it doing
that well it might not really be related
to anything just that it was not
detrimental for them to do it and
someone develop it but at the same time
they also developed some other really
good ability and now like they they just
do these things and this actually
happens in nature as well you might have
like you know a sub optimal or again or
like some function in an animal or a
plant that is actually it’s not that
great but because they’re doing
something else really well yeah it’s
like it doesn’t matter yeah the pink if
like just the pinky toe like yeah like
that it’s just like does it yeah yeah
our appendix that is a much worth it’s
now proven that is doing something
useful but it’s the type of thing yeah
like this is just endlessly fascinating
and it’s such a such a cool project we
are like we’re done with demonstrating
the the pixel world but next week we’re
actually gonna have a talk more about
like projects in general like the having
like why build something like this like
you can’t make money of this I don’t
know maybe can’t like maybe it’s like or
at least that’s not where you did it no
it has no obvious monetization like and
it doesn’t even have like is this even
good yeah good way of learning about
evolution may be it may be this may be
innocent it just but it’s cool
it’s just really nice and we want to
like we’ve been talking about that a lot
and we’re gonna talk spend an entire
episode on that subject Ness next week
so please tune in them and I’m here from
the future again to remind you that if
feel like it’s finally time to learn Big
O and binary trees then you should go to
brilliant dude org slash FFF also in the
absurd description that is it
are you have if you’re new and I’m still
watch this far
you have just watched an episode of fun
fun function I released these every
Monday morning Oh 800 GMT but you will
forget that so you can click here to
subscribe or watch another one here
right now by clicking there I am mbj I’m Fred ignoring
Fredrik is back from last week we’re
actually recording these back-to-back
but and we are last week we looked at
pics link which is a project where kind
of like a neural network sandbox so
let’s be yeah it’s like an evolution
simulator where the fixings agents or
the creatures in it has been your
networks yeah it’s a game I might like
it’s a little game sort of we can it’s
kind of like a Minecraft for kind of
programmers that want to learn about
about neural networks and evolution so
no well if you like you don’t have to
watch that episode to to watch this one
body my cap so you can find it there or
there I will link it it’s also in the
episode description so what are we gonna
do today PJ from the future here to tell
you that today’s sponsor is brilliant a
problem solving website that teaches you
how to think like a computer scientist
if you are interested in learning
computer science fundamentals binary
trees quicksort big ole with that kind
of stuff instead of just passively
listening to video courses they have
interactive challenges with increasing
difficulty so it’s kind of like a game
with a sense of progression it’s very
satisfying use brilliant dot org slash
FFF when you sign up to let them know
that you came from here also if you are
one of the first 200 to use that link
you get 20% off the annual subscription
if that sounds interesting to you make
sure you check it out after the show
that link is also in the episode
description so what are we gonna do
today yes the last time we started
building a simple world and we started
giving the pics things some abilities
yeah and maybe for the people that
didn’t watch the last episode quickly
like what you do in this game is that
you build worlds and you put these
creatures into them so we started
building one of these worlds and the the
worlds are online this is where
editor so we’re gonna link this world
that we’re playing with right now in the
episode description so that you can
check it out and play around with it as
you as to watch this video basically
yeah so I think today since what we did
last time we started giving them the
ability to kill each other and also give
them this kind of weird ability to sleep
and do nothing
yeah and I think also humans have yes
it’s such a weird ability so we were
thinking maybe let’s explore sleeping a
little bit and see what we can do with
it I think we don’t have like really a
plan of no this this is like last
episode we have planned out a little bit
like figuring I would have a couple of
like we’re gonna do this and this and
this this is gonna be more like how you
would actually use pick sling yeah like
yes hey I have this like why do we sleep
how did that came to be an evolution
like it’s really strange really because
I’m reading this book now called why we
sleep which summarizes like the last 20
years of sleep research and it’s only
recently like we have figured out why we
sleep earlier we were just basically
dumbfounded like we are stick okay
animals sit down for eight hours and do
nothing like and they shut down
everything so they were super easy to
eat and we don’t know why and all
animals do it like there’s no like it
all animals that didn’t sleep they got
like terminated by evolution right so it
seems very beneficial but we don’t know
what so let’s see if we can make the
pics things sleep them and see if they
can somehow get any benefit from this
yeah or maybe they won’t know we might
not be able to do this like who might
fail so let’s try to do this let’s get
into it I think the first thing we’re
gonna do is that we’re just going to
open up the editor and take a look at
what we already have
alright let me pause this thing ah
they’re cool
I’m going to check this out like this I
wanna look inside the pics link I love
it yeah because we didn’t do that show
advanced yeah go for it
so here you can see okay so it has
abilities that we defined before like
the killing properties yeah why does it
have to
so actually what happens that is that
sometimes they randomize the abilities
that I have so in this case this guy
actually ended up having the same
ability twice no but you can actually
changed the second one to a sleep
ability oh and now he’s got two
different abilities cool and then it has
a bunch of properties like State and hat
and where it is and in the world yep and
it has position and it has brain inputs
right so this is where we get into the
whole neural network world yeah because
if you didn’t watch the previous episode
every single pixel in has its own little
own little brain has a neural network
inside of it exactly and they’re a
little bit like you know living in the
matrix that they only the way they
perceive the world is used through
numbers so their inputs like their eyes
or their senses it’s just a long list of
numbers yeah and the senses are
basically these not quite okay that’s
the properties and then down here or
actually the is the brain inputs and
some of them are some of the properties
some of them are things around them in
the world yeah we’re gonna see how we
define the environment of the pixels
right so all right so it takes like a
current state of the world and what he
sees and that’s the brain inputs like
eyes ears there and then you go like get
these brain outputs which is move and
focus and writing to memory
exactly so exactly he takes his inputs
processes that through his neural
network and improves produces basically
his choices or is what he wants to do in
this world the movie decides which
direction he wants to move and then at
the bottom you see their ability slots
that’s deciding which ability he wants
to use in the certain situation that
he’s in and if we keep going down we can
actually actually see his brain yeah
alright so this is yeah this is a neural
network basically this is a you know all
the weights for a neural network exactly
so the way neural networks works is that
you take the inputs and then you run
these mathematical operation
together with the weights yeah and then
that is what produces the output and the
cool thing is that these each little
weight here matters for the output yeah
so if you change one of these numbers
the behavior of the pics thing might
change drastically yeah and that’s the
way the the inheritance and the
evolution of this works because the
children each of like their weights
which of them might have a small chance
to actually change a little bit so like
when we say like all this brain like
these weights and we take this inputs
and get these outputs and we somehow use
this way is this very complicated math I
would say not super complicated it’s
there’s a few operations it’s matrix
multiplication and then you have basic
summation with the biases yeah and then
you have what’s called an activation
yeah and matrix multiplication is for
like I don’t have a math background like
you like it’s basically our matrix is a
multi-dimensional array exactly it’s
worth two-dimensional it’s kind of like
a table of numbers yeah and then you
just matrix multiplication it’s
basically just taking a matrix and then
another matrix and then multiplying them
which means like multiplying that value
with that value in a certain order
exactly exactly yeah you use regular
multiplication and addition and you
combine these numbers yeah so there’s
it’s basically just times it’s like it’s
not it’s like some kind of weird simple
no no it’s very simple and then the the
final thing uses what’s called a real
ooh is the fancy name exactly as the max
and at the end of these so it’s kind of
this thing where you take where you kind
of cut the number off and if it’s
negative then it becomes zero and if
it’s positive it stays the same okay so
basically is the way of normalizing the
numbers to a certain format sort of yeah
yeah and that’s basically the neural
network of these pics things all right
cool yeah there was a digression I just
wanted to like show that are they the
interior of this thing is it’s a bit
complicated but not at all that’s
complicated that you might think yeah
exactly
neural networks is one of those things
that sounds insane
when you
you hear about it like oh we’re building
a brain and moving on the brain and like
this axis norms and dendrite it’s really
a pretty simple math it’s very simple
actually and I would add one very
interesting things about neural networks
so neural network is it’s it’s like a
function right it takes inputs produces
outputs and it’s been proven that if you
have a large enough neural network it
can actually model any other function in
the world oh so the code was there
what’s called a universal approximator
and the reason this is interesting for
fixing world is because that means that
there is no behavior that these pixels
can have with their inputs but that they
can’t have like there’s there’s there’s
no limit to what they can do as long as
the neural network is big enough of
course they know that they’re quite
small but if it was big enough then it
could essentially do anything that is so
cool yeah it’s just an airline your
networks are insane it’s very
interesting such a little simple thing
that isn’t cool all right okay yeah
let’s get into it let’s give them let’s
give this a bit more detail yeah so I
think the first thing we’re going to do
is that we’re going to give them an
environment to interact with alright so
if you go ahead and hit add environment
on the Left yep
so now we’re getting to the editor and I
think then right we’re going to create
this something for them to eat so I
don’t know we can change the colors and
we can change you can see the masks up
here yeah and there’s a couple of things
that we can choose from yeah we’re going
from from function theme we should have
like Buddha but we don’t have that like
it’s not even okay yeah I suppose like
up yeah exactly
upload graphics area would be like well
but I do i gold nugget seems like it’s
closest thing sure let’s go with it
and then you guys should change the
color so you can zoom in on the right
here to see a preview of this
environment all right
oh this it’s not too far off actually
yeah it’s pretty good you can change the
min and Max colors to whatever you like
and then what we’re going to do is a
very simple operation with this
environment so each environment way it
works is that it’s it’s just a single
value for each cell in the world okay so
what we can do here is that we can
increase that value for each cell in the
world at each time step see so what do
you mean like there’s more bootlegging
one second it’s certain exactly yeah
we’re gonna grow pull out in in the
world of to fix things so if you go
ahead and hit Plus on each step then you
can change that set value to an
increased value now just increase I
guess yeah and I think we’re pretty
happy with that and then finally we want
to add one more thing here which is
gonna be a clamp yeah change this one so
the guys who never rises above a certain
value exactly so it can’t go above 100
or below and if you hit update now we’re
gonna get this in our world and we can
see it at now is standing on Clara its
bola and if you play here
we can actually see that right yeah okay
so alright cool
I wanna yeah all right I’m with you all
right so this is I know I need to change
yeah like are these colors good yeah I’m
good with the colors it’s gonna be fine
so like this goes from brown when it’s
like almost no bola and then like really
gold when it’s a lot of bullet okay so
the second thing I’m going to do now is
that we want to give the fixing some
kind of property that they can so they
can eat from the ground and transfer it
to their property you know like in the
game you might have cows or mana we’re
gonna give them something like this so
if you open up the editor again here
yeah they all actually lead to the same
kind of place but okay so you go into
properties oh so pics ling-su is
basically all of the pics links and then
like this years happens to be selected
no it’s more that this is kind of like a
prototype picks thing it’s like the
Sheep the Dolly the sheep of pegasi and
he only exists in the editor but then
you can take him and create many
instances of him in the world I see this
is actually where prototypal inheritance
sort of start to make sense in this
context exactly yeah all right cool yeah
let’s go to properties and then we
create a new property and you can give
it whatever name you want all right
we’re gonna use this to clicked Bullard
yeah okay so like it smarts because
we’ve become smart if you perfect we can
change the arm reproduce too let’s make
that a split what does that mean so if
one pic sync clicked it a lot of Buller
and then he reproduce then the child is
gonna get half of them and he’s gonna
keep my first time it’s a sharing
economy all right so yeah but that’s how
I don’t know that’s how inheritance
works in Sweden yeah cool and then
unless there’s a spouse like all right
and then let’s actually give him an
ability to eat eat these things then so
if you go to everybody’s yeah we let’s
actually change this whole thing that
would be last time so in certain if up
here yeah let’s change that into a
transfer up to yeah perfect
transfer to yeah right they’re similar
but and you can do it from so you can
change this focus fixing property if you
click that and then go environmental
pixels position environment InFocus I
know the top one air oh okay
and then you change the environment from
energy to the environment that you
create and then the Pixum property you
want to keep that now when you want to
change generally to smart and then you
can adopt it and then you can create a
few more of these guys all right cool
oh my god and you’re gonna hit play up
here and actually you can change this
alright we need to change this more oh
no this is the I picked the worst color
like it’s just hiding all the pics links
oh it burns there and what would be a
good color that is less I think actually
if you change the mask maybe to like a
neutral mask maybe grass for now and
then it’s gonna be a bit easier to see
no that’s cool that’s better so if you
go ahead and hit play and if you change
the extreme to maybe just hyper so we
can see now that in the background of
these pics links to grab the bar that we
defined is lower for some of them right
yeah they kind of like eating this and
then they are coming back up again yes
that’s how we built the environment but
and you can actually sue now again yeah
we can see it on these guys here yeah
and you can actually see these like
health bar something right but I have
filled up their health bars because
they’ve been eating from the ground all
right so it’s like and the health bar is
it like I guess at that points to some
default property so it’s if it’s just
one property that’s gonna be the health
bars or yeah actually you can you can
have multiple health bars okay if you
had more have properties than each of
them can actually be showing off how far
alright but they’re actually not using
this this but here we see like this
these just flying around and eating yeah
exactly but there’s no there’s actually
no point to the meeting right now they
are not they’re not using this for
anything
so it’s jumping to the editor again and
then we’re gonna actually make this
requirement that they have to eat to be
able to reproduce oh yes let’s jump into
abilities and then we’re gonna create a
new ability we can actually change the
name of this one that’s right alright
yeah which one so we can change the name
of the one we just created oh yeah
no no no no no yeah perfect yeah well if
you don’t know what that is
you’ll ask in the comments cool and
let’s give them one more ability to use
this bowler and reproduce yeah all right
yeah I guess like birth yeah and what we
want to do now is that you won’t even
use an if statement so yeah yeah and the
condition is gonna be that they have to
have eaten enough blood
otherwise they can’t reproduce oh yeah
if so if hang on maybe I can do this
if see yeah if any of the cup following
conditions true maybe okay no I can’t do
it this so you can see because I was
looking for like if a property is yeah
you were actually close to it so if you
go back here you can see that this first
one so I comparison oh okay
so like if pics link property smart is a
bomb like I don’t know 10 perfect then
it can reproduce no yeah okay I need to
delete this that because the den run
says it here this is perfect
let’s try this and you might have to at
this point we might want to restart to a
simulation because this is lot of yeah
yeah these don’t have these abilities
right exactly they want to have the
reproducibility as mouse to animate yeah
so it doesn’t change because it’s it’s
not a class based model it’s a prototype
based model so like they it will like
only new ones will be spawned from a
prototype it won’t change like the exist
exactly so if you change the code inside
of an ability then that will update if
yes the abilities a global but like as
if it’s assigned exactly exactly alright
brilliant
I’ll restart it and we’ll paint some new
page links whoops
click this paint paint pics link button
here now get some cool and then we can
see to the right here that they’re
spreading over here yeah so yeah these
are the they learn to they have learned
to eat or they were pics links that that
ate because of randomness and that means
that they they spread a lot more yeah Oh
safe home yeah I just I want to say
feature I don’t know what to say that’s
nice but there’s a bunch of problems
with this world like it’s such a simple
world for instance there is no way for
them to die right now know what just
well yeah exactly and that I guess that
creates a problem with a world where
things don’t die is the this is actually
an interesting has become like the
purpose of death like it becomes like it
becomes very stagnant after a while
exactly because there’s no developments
if you if you don’t have death they were
going really deep really but yeah if you
don’t have death in a world like this
then I don’t think that death is very
deep in the context of an evolutionary
system that like super basic yeah that’s
true that’s true but yeah like really
for them to evolve and become more
interesting you need to leave space for
like the next generation yeah and right
now they’re not really doing that so I
think that’s actually a probably good
step for us to to add some way for them
to die as well yeah perhaps if they eat
too much blue that’s all right so so
let’s see environment rule no no okay
abilities see how if I can do this
actually you were it’s an ability to die
well actually you were on the right tab
that rules might actually be a good way
to use because rules are kind of like
these global things that you can run
yeah and they always run they don’t the
pieces don’t choose to
or not yeah but I figured I could use
this as a no I could model this as an as
an ability I think all right yeah like
because now I want to use what I’ve
learned I’m sure there’s a better
version way to do it but I’m gonna like
do the ability overeater and so if this
like this also no it’s it’s bigger than
a fixing property smart okay we need to
change that to like fat instead it’s
above no I got it it’s about oh yeah
this we want a constant here yeah
where’s the constant oh it’s not a
constant it’s this oh this is not in
alphabetical order
already some special yeah there’s no
real good order at all right now it’s a
little bit of miss yeah but this is beta
very beta software yeah but it works
really well anyway it hasn’t crashed yet
well fingers crossed but yeah so if I
could say that it’s like 25 or something
yeah and then wrong die yep so it might
not necessarily like some fix links
because these are chosen by the pics
links like they do have free will
picking between the abilities so I won’t
always die when it hits 25 but it’s like
it’s likely to die after eating one yeah
yeah and I want to change this property
now to call it fat instead cold we’re
gonna have to just update them in all
right all the places where you use them
because it’s yeah it’s me it’s weakly
typed it does not matter it fat but it’s
generalised that it gives me like the
errors and because everything is running
all the time I suppose yeah
well actually the game get becomes post
when you open the editor but not here
okay inability overly snow yeah but
there we go debugging cool so I’ll let’s
restart the world and see yeah so yeah
so now we have like some I don’t wonder
if we get some pics links at overeat
here yeah I see that
so I can I can tell you right now that I
think there’s gonna be one problem with
this yeah
because as you create this and as an
ability they choose the wrong ability
yeah so if there’s there’s very likely
to they are very likely to reproduce
during the period between when having
enough fat I think you reproduce and I
think both that but it’s also that they
would have to choose to kill themselves
oh that that strain yeah so this strain
the breed of pick slings that or
choosing to kill themselves when they
overeat they will very quickly die out
it’s like he gave them a gun and you
made it so they can only shoot
themselves and of course that’s amazing
because I didn’t think of that because I
don’t think in in AI I don’t think it
your record I think in programming so I
just think oh the choice of course
that’s just a random random function but
no they will they will they learn
actually learn or they they won’t learn
it’s just that they genetically they
will learn exactly I mean the ones that
choose to kill themselves they won’t
spread so it’s just yeah that’s gonna be
enough using that one that’s oh that’s
actually such a good example I actually
died I learned something okay so yeah
okay so let’s that didn’t work we have
to hold on you can actually what you can
do is that you can actually copy this Oh
see if you right click little thing here
and only copy all right
and then now we can delete it but then
we have this in our clipboard right and
now I can create a rule yes all right so
if this cap paste here no actually but
like so it’s a little bit buggy still
but if you click the plus on the second
one on the one below oh and then click
the just click it normally like yeah
left click and then right click this one
and then click oh there room so a little
bit cumbersome guys it’s gonna become
better in the future
alright but run the following every step
ok
does this look right if the over pics a
property fat then wrong die exactly and
then I made it like this so now they
don’t have a choice to run this like
it’s always gonna run yeah so now they
better watch out for overeating alright
so like the old overeating rule yeah so
I love that little it’s like I love that
so let’s maybe restart it and see if we
can see what we can have happen alright
let’s see what happens and I I don’t
know what’s gonna happen because I’ve
never tried this either now that’s a
very quickly new experiment it seems
like they are not actually eating from
the NIE
anymore or not I won’t expect the shorts
happening yeah and they’re not
reproducing either they are not perhaps
if we catch some more because this is
something that we learned I learned from
the from the other episode that
sometimes you just need more six links
yes it’s all like how our life will like
life in the universe is very unlikely so
we needed basically an infinite universe
for life to happen
exactly so let’s paint a bunch more
fluffykins all right they seem very
reluctant to tewi reproducing so no they
say yeah exactly the scene no acts like
they seem reluctant to eat in the first
place right Yeah right right so why
might they not be eating all right let’s
and let’s have rules
so if pixie Papaji fat is above 25 then
run die
should we just increase the number or
maybe that’s okay so let’s say that they
have to beat 250 but it’s still weird
it’s still like it they they can
reproduce after 10 yeah right it’s kind
of odd all right it’s pretty science oh
now it’s all right and then a bunch of
them use died there as well yeah so they
like their ro oh yeah exactly they are
overeating so actually what you can do
is that you can you can actually lower
this number now yeah and it will affect
see if you run open up the editor and
make it 25 now it might be that they
have evolved enough that they will
survive oh that’s pretty cool so they
can now yeah so they’re now clever
enough to survive the harsh credit yeah
now a lot of them died out but some of
them are still surviving so presumably
babe yeah but I mean over time there
will I think that they will if not
everyone dies that means that some
strain well figure out how to not
overeat right yeah maybe all right yeah
because this is a kind of like not it
turns out that like maybe not eating too
much is a very tricky thing to do well I
think actually what they will never
learn I think I mean I mean it reflects
like the problem that we have in where
food is abundant like they have so much
food it just grows in men sleep fast so
they they their food is never scarce
here right all right so I mean it’s it’s
it’s kind of one of these things that
it’s actually not too hard for them to
learn because the rule is so simple it’s
like do I have too much food yeah don’t
eat any more so it’s usually yeah work
slots as you see like it also it seems
like they have a trouble finding the
balance though because no no breed here
is is dominant
that could be a breed that was very good
eventually at eating just the right
amount then they would completely
dominate exactly yeah but it’s not like
they’re there it seems to be like they
seem to be like not yeah you’re right
stop eating basically yeah it’s
interesting and stop reproducing
maybe we should punish them if they stop
doing like because it seems like when
they’re likely what you’re saying that
they’re just not doing anything right
now no exactly they’re not starving yeah
so we’ve perhaps you should add
starvation why do I notice that we are
completely like gone away from the sleep
thing here that will happen well yeah
all right actually one thing I can teach
you here is that if you click on one of
them if you use this tool and the other
one though
yeah the selector will pick this one so
we can see that he’s actually he’s got
only two abilities and he’s alternating
between them yeah and it seems that it’s
actually it doesn’t even have the eat
ability has only got the sleep and a
birth ability oh so okay I see so that
is that is tragic reminds me of my old
high school friends anyway it’s also
it’s very slow right now like in
summation there’s our that’s about like
when the Select window is open like it
becomes the simulation becomes much
exactly it slows down a lot because it
needs to read back information all the
time all right cool but what we can do
is that we can give them at least the
three abilities because now these guys
are just surviving by doing nothing and
kind of boring so you can you can do
that by opening up the auditorium all
right and then going into picks thing so
and then changing the ability slot count
here to three abilities like that okay
it’s here and then we can actually
configure exactly which ones they are
gonna come born with one we Oh games so
you can set set them up to have all
these three and then a final thing you
do is that we can change the mutations
abilities chance to zero oh okay so what
that means is that they’re gonna always
keep this one oh yeah otherwise there’s
a chance for them to like replace one of
them oh so they basically evolved out
like fika in the first place or
something
exactly all right cool
so I’m just gonna restore it because I
don’t remember anything
and it’s paint so many pics links and
see what happens hmm okay they seem like
oh there we go there’s something on the
address Jesus yeah somebody is like they
I figured out how to you yeah this is
what I would have expected to happen
actual right yeah yeah like they seem to
oh they’re clustering in a interesting
way as well you all right it’s a weird
pattern and this is one of these things
that’s where I have absolutely no idea
what the evolutionary benefit now or
they might actually not be even be an
evolutionary benefit yeah just restart
this and just do it exactly the same
thing well what will happen yeah let’s
find out I guess yeah so she talked
about in the last episode like when you
when a pix thing is born like it gets a
little bit of a random brain so it’s not
the randomness doesn’t only come from
the fact that I positioned them
differently it also comes from the fact
that there’s just inherent randomness in
right in the universe like both this one
on the real one things on this one not
much is happening yeah and sometimes you
just have to spray more before we see oh
there we go yes guys do something I
can’t see if they’re also forming
they’re forming it Patten as well aren’t
they but it’s not the same pattern yeah
yeah it’s G it strikes me how organic
this is like it it really is life oh
yeah it’s there’s no like computer like
it doesn’t feel like a computer at all
yeah I think I mean I’m very interested
in like the theory of life and
artificial life and
I mean there’s this one thought that
maybe maybe life or like you know a way
of thinking of life is uniqueness yeah
and each of these picks things are
unique yeah like they’re not none of
them are the same or I mean there might
be two of them there with a very small
likelihood that are the same but it’s
just one of these ways that maybe you
know in some form of philosophical sense
they’re like a little bit alive because
they’re unique but there’s no one else
in the universe of everything is the
same as this particularly pick thing
here really cool cold should be well
we’re we’re at thirty five minutes now
which is like a long episode and I think
that we should should wrap up because I
think that this was pretty amazing
actually
alright because I think that wait what
what okay let’s wrap up what did we
learn but what did we set out to learn I
kind of forgot it was something about
sleep we started out with wanting to
model sleep and then I think we didn’t
really get to one thing sleep but now
yeah went into modeling yeah yeah in in
that’s like in a yeah in an environment
where like there’s enormous amounts of
food yeah there’s a there’s a abundance
of food in this environment I’m not sure
if we learnt like anything on you know
philosophical level of like well is it
good or bad do you know it sure than
anything like that but yeah I was
surprised by how like I thought the
neural networks would be better at
learning the whole overrate thing or
maybe maybe well if maybe that was not a
problem actually if I change the rules
to twenty five right now
yeah right now it’s right yeah yeah
right now yeah they are pretty good at
figuring out the right amount of eating
actually yeah it might sometimes it’s
just a case of like spawning like you
said before spawning enough of them and
then
it changes and it seems like in this
case there’s yet another way of
organizing themselves I don’t know if
they’re forming a pattern now as well as
well or not but yeah it seems like
they’re also forming it a slightly
different pattern this time all right
you know what we’re gonna give like this
episode is pretty long so what we’re
gonna do is that we’re going to live
leave the audience we’re completing this
and to they and modeling the concept of
sleep and evolutionary terms I think
that would be pretty interesting like if
you can do that leave leave a link in
the comments to your pics Ling world
again you can find the link to this
fixing world in the in the episode
description and I also like if you’re on
the phone fan forum make sure you post
it there and I think that that is it if
you have any questions or thoughts about
this like what like like you have access
to the Creator here in the comment field
so like if you that’s very good a good
way of getting support for you stuff
make sure you play around with fixing
work because it’s amazing like these
these tools to learn machine learning in
a way that isn’t watching Andrew and
Gees tutorials worry like oh math
symbols I think that’s pretty great
yeah and I think it’s it’s cool that
people people don’t understand evolution
all that much like I it just this gives
me a completely different newfound view
of what evolution is yeah for sure I
think it’s you it also blows me away all
the time when I look at it and just just
like you know how evolution actually
works the whole concept of not having a
goal that they’re just competing with
each other is always like fascinating to
me and yeah oh that’s such a good lesson
in life in general that what like just
like why does this exist why do we do
this why – people like because it was
somehow
beneficial for them to do it like rather
it made it that behavior replicated it’s
exactly it’s not miss it because like
it’s fair or anything like things happen
because it evolved to do that or
necessarily did not might not even be
better it might just be better in this
specific context yeah so yeah that’s
yeah it might also be that the people
like the the pics links that maybe they
just develop like some kind of weird
behavior where they just go like this
like if they had like motor functions
and it would be like why is it doing
that well it might not really be related
to anything just that it was not
detrimental for them to do it and
someone develop it but at the same time
they also developed some other really
good ability and now like they they just
do these things and this actually
happens in nature as well you might have
like you know a sub optimal or again or
like some function in an animal or a
plant that is actually it’s not that
great but because they’re doing
something else really well yeah it’s
like it doesn’t matter yeah the pink if
like just the pinky toe like yeah like
that it’s just like does it yeah yeah
our appendix that is a much worth it’s
now proven that is doing something
useful but it’s the type of thing yeah
like this is just endlessly fascinating
and it’s such a such a cool project we
are like we’re done with demonstrating
the the pixel world but next week we’re
actually gonna have a talk more about
like projects in general like the having
like why build something like this like
you can’t make money of this I don’t
know maybe can’t like maybe it’s like or
at least that’s not where you did it no
it has no obvious monetization like and
it doesn’t even have like is this even
good yeah good way of learning about
evolution may be it may be this may be
innocent it just but it’s cool
it’s just really nice and we want to
like we’ve been talking about that a lot
and we’re gonna talk spend an entire
episode on that subject Ness next week
so please tune in them and I’m here from
the future again to remind you that if
feel like it’s finally time to learn Big
O and binary trees then you should go to
brilliant dude org slash FFF also in the
absurd description that is it
are you have if you’re new and I’m still
watch this far
you have just watched an episode of fun
fun function I released these every
Monday morning Oh 800 GMT but you will
forget that so you can click here to
subscribe or watch another one here
right now by clicking there I am mbj I’m Fred ignoring
Please follow and like us: