KILLER ROBOT EDITION
Transcript (AI, 1501 lines)
Maybe just win time on open mind.
this is a whole lot of data that a a LLM won't get just by reading about a car accident.
all this information is is acquired acoustically.
And so DESA learned language from hundreds hundreds of times fewer words than a regular LLM.
So that's that's that's one of the reasons why this is also kind of a an interesting technology because language is more dense, it's more efficient, and so if you
have like a sentence for example that says I'm fine.
Well, anyone that's got a girlfriend knows that that there's many ways, depending on how that's that's said, that that it could it could mean that someone's actually fine or that they're not fine at all.
And you know even if there's it's written with punctuation it's hard to know how the person is is actually meaning that.
But once it's spoken aloud and you can hear the inflection of their voice
you can hear whether there's resignation, like oh I'm fine, you know, happiness, I'm fine.
Sarcasm, oh yeah, I'm fine.
Anger, I'm fine.
Reassurance, I'm fine, I'm fine.
Grief, I'm fine.
Audio has all that data packed in there.
It has pitch, has timing, volume, things you don't think about that.
You're probably you yourself are probably picking up on, but you don't actually think about that that a you know a computer would, like breathing patterns, the way people breathe, the rhythm of their voice, their accent.
So this is this is a whole lot of information that that you're not getting from just reading a text.
And so because they have all this this extra language.
Or all this extra data built into it, you know, whether you're getting emotional information, you know, you're getting the I guess the question is because the the AI is receiving emotional information and it's able to produce
speech
that sounds emotional now.
Like we just listened to that that sample where the woman was.
laughing or you know the the simulated woman was laughing.
the reason why she and you could say oh she's just you know it's it's mimicking.
It's mimicking.
Yeah but isn't that what babies do?
Isn't that what kids do?
Anyone anyone that's interactive with kids knows that you know kids mimic adults.
And so we start to get in a territory now that because we've created a
A way of and this is just, you know, this is very limited.
This is the baby steps.
This is like a first baby step.
We've now created a scenario where part of the data that's being used to train the AI contains emotional information, and and because of that, it's it's producing emotionally or at least emotionally realistic results.
You know, sound speech that's emotionally appropriate.
You know, when you talk to one of these, like when you have Grok read aloud its response, it just I mean it sounds l more like a human than like the old robot voices, but it's just it still sounds like a robot.
You know, as an example, they made a robo nig.
This is this is Robo Nig.
I don't know why they did this.
I wish they hadn't, but this is RoboNig.
Hey, how you doing?
Yeah, how you done, Renalto?
I'm gay
How you doing, Ronaldo?
I'm good, I'm good.
I'm I'm not Ronaldo, you're mixing me up with another guy.
We got No, I'm I I'm D Vir, man.
I'm not Ronaldo.
No, you're not Ronaldo.
No, I am not.
How you doing, man?
How how how's it been?
Man, I'm doing good, bro.
I ain't gonna lie, I'm tired though.
You tired?
What what happened?
Nah, man, when I get to work, bro, I have to g get in there and pick
produce and stuff man it's it'd be tiresome work is tough yeah work is tough then i gotta set up i gotta put all that stuff away by myself with no help and then i gotta set up so it just be tiresome but i'll be all right though yeah you you you'll be good how's it going with family and all that my
kid good man I got a new baby boy man October Congrats what's his name yeah we're calling them Darius Johnson Darius Johnson
I like that name.
Yeah, my daughter said she liked the name too.
She said, Man, you must have liked the game.
Darius Johnson in the game.
I said, Yeah, I did like the game.
I do I like that.
That's awesome.
Is he is he is he funny at all?
Not everyone's gonna know this, but I'm gonna interject here.
It's obviously making something up.
It's saying, Oh, my daughter said you must have liked the game, Darius, for to name the kid Darius Johnson.
That's
An old arcade game.
I I've only heard it pronounced Darius, but like it's it's an old arcade game from like the the 80s.
So that's basically what he's he's referencing that.
What's he like?
What's his personality?
That dude funny, man.
He's smart.
Yeah, you're smart.
You he's smart.
They say oh man, he just like me, man.
He do all this stuff like me and and act like a little mini.
That's great, as you.
Can you t tell me you know give me a story, man?
Why well I he was well every time he cried man he wants me.
So when I get him up at a certain time in the morning, he thought I would say, Daddy, daddy, daddy, give me, give me, give me.
Yeah, so I went to give it him to my wife the second time around.
I had to give him to my wife the first time and the second and third time.
But he wants you, right?
He he likes you.
He he's meaning you, right?
Yeah, he likes
Like me.
So when I give it to my wife though, she said, take him back and let me get him.
I'm like, nah, we ain't finna do that, you know what I'm saying?
We don't finna do that.
I was like, you you're alive, you know what I'm saying?
We'll do a whole another situation, but I didn't want to give him up.
I know she said, take him and let me have it.
Yeah.
So so she took him for a couple of nights, and I was still at work.
So while she was up in the room with him, he cried.
And when he cried, he was talking to me on the phone.
And
She was already upstairs in the room with him.
And But he wants to talk to you, right?
Well like once you talk to him he was fine.
He he was happy.
No, no, it wasn't no once he talked to me.
It wasn't that but long as he cried he wanted to talk.
So See notice how it's interrupting him.
It's responding before he's done.
It's it's interjecting.
So I picked him up off the floor and I'm like, man, we ain't finna play that game.
I ain't finna put him back down the floor 'cause like he just
You you like him too much.
Be honest, man.
I I know you have like a tough exterior, but but you love that kid, right?
Oh yeah, I definitely love him.
He just like me.
We just got each other personalities.
That's awesome, man.
Alright, good talking to you, man.
Yeah, you two were not though.
Tess.
Okay, cool.
Yeah.
So it's you know, it's that's it's some creepy shit.
That's some creepy shit.
Now this is this is why I said we're gonna
have to come up with like blade runner style ways of determining whether or not someone's human.
Because this is the worst it's ever going to be.
People need to get that in their heads is when they're like, oh that's not that good yet.
I it's not I mean in it's I mean it's really good if you're a a a scammer from Nigeria because they've been able to scam people with really bad photoshops of Jennifer Aniston
For 20 years.
And so now if they can get Jennifer Aniston to call up from their laptop, I mean it's if it runs locally, they can call up from a laptop and and talk in real time with your with your aunt, you know, and and or I guess your uncle, unless your aunt's a lesbian.
you know, and sound realistic, sound convincing.
You know, if I if something that kind of a voice had called me.
me up even like a year ago, I'd be like, well that can't be AI.
They don't have that yet.
Well they've got that now.
And that's and and they have it now, which means we have it forever.
And it just gets better from this point on
now the problem with with with AI, with a lot of people being afraid of it really and not understanding it, including the people that that design this stuff, is
It's a complexity issue.
It's it's a complexity problem.
It i it's it's it's the same problem not even ironically it's the same problem that we have with understanding our own brains.
For the same reason.
It's real easy for humans to understand the particulars of a of a single component.
Right.
Like a a simple electronic circuit.
I can look at a schematic of a simple electronic circuit and I can trace that where the electrons go and what all the different components are doing and how they're directing the flow of the electrons and how they're producing the result that I
want or if it's broken, why it's producing a particular result and what it should be producing.
and then you know fix it or whatever.
But even when you start to talk about like a a modern computer processor, you you have engineers that understand individual components of that processor, but you'll you don't there's not even a single engineer that would be able to understand
what every single transistor is doing all at once.
And understand the complex computations that are that are going on without, I mean, it it would it would take months to sit there and try to trace out the pathway of the individual electrons going through that circuit.
And a neural network has hundreds of billions of parameters.
And while we understand some of the mathematical operations that are
going on in the same way that we understand some of the individual things going on inside the human brain, understanding the relationships between all of these billions and billions of of computations
going on all at once and and and be able to conceptualize it's beyond our our ability
it's way beyond our ability
In fact, if anything, I've noticed that humans are losing the ability to understand even things that are remotely complex.
And in fact, as this is going to be a problem as AI becomes more of something that people lean on, including myself, I leaned on it to create this this broadcast stuff.
you know, whereas before I would have had to sit down and and code it all out myself.
what
I wouldn't have done, but if I wanted to, I'd had would have had to have done that.
So it's created a situation where people will be doing complex things less often.
And people will be doing things things that'll exercise the part of the brain that allows them to conceptualize complex systems less and less, while at the same time the systems are getting more.
More complex.
And that's a real fucking problem.
That's a real fucking problem.
And it's going to create more of, I think, a denial among people about what AI even is.
And for that matter, what humans are.
See the the fact that we can't completely explain consciousness, thought, or memory, or subjective experience, it doesn't automatically mean that there's something oh look at that!
Another fancy one.
Fancy big dono.
Look at that.
I I I I'm gonna put more in there.
I just I didn't think this was well I I thought I was hoping it would work, but I hadn't hadn't tested it.
So now that we know that it works We'll be able to Oh look at that we got a double
It's like a slot machine
Alright No my point is
More and more, the more we start to learn about artificial neural networks, the more it teaches us about our neural networks, and the more that we're similar.
I mean, just in fact, I mean, just and we're talking about DESA, the fact that they're going to start having AIs that will train on audio in the same way that a kid does, for the same reason, and and with perhaps the same results.
And all of a sudden we're gonna have a lot of questions that could be answered and
In terms of, well, what does that tell us about consciousness?
If if, for example, and I I was saying this well, I've been saying this for the last year or so, that if you have an AI that is in its behavior, its output is no different than a human being, then what's what's what's the missing piece?
Where is the soul?
Where is the non physical like what where's the implication then that there must be something non physical?
Physical inside of our brain that's allowing us to have that behavior.
If this other thing which doesn't have a soul, unless God gave an AI a soul, if it's able to produce the same behavior, then why why is it logically follow that we must have something magical inside of our brain that makes us
conscious, makes us able to have subjective experience, makes us able to
reason if we have a sufficiently sophisticated digital neural network that's that's producing the same behavior.
And people aren't gonna want to think about that.
In fact, because people are gonna be less and less able to even conceptualize complex ideas, they're just going to dismiss it outright.
I mean this is why I went so hard against the flat earthers.
Because if you're not able to understand just things like in terms of I mean that's just one complexity.
Like that's just like understand
Scale.
Like that's those are people that don't understand scale, and because the the vastness of not just the universe, but just earth bothers them because it's beyond their comprehension, they they decide to believe in something else.
That's funny.
And they get and they cling to it religiously because the vastness of just the earth makes them uneasy.
Well, you're gonna have the same kind of behaviors when it comes to people.
People dealing with AI.
They're not going to be able to understand the complexity, they're not going to be able to understand what they're up against, and so they're going to make up you know bullshit theories about how it works.
Now, one thing that I I mentioned that that they're gonna start training these AIs with audio, but a baby doesn't just have you know ears.
Right?
If there's if there's a lot of this information.
information in sound well how m dense is the information when you have when you couple that sound with sight or when you couple that sound with touch
see that's that's going to be when these neural networks get even more complex and more efficient and more human like because they will have the ability to have all of
that data connected where instead of just reading about it.
That that's the that's the limitation of these LLMs.
The LLM can read what it's like to see an orange and what the color orange is conceptually and what what a what what an orange tastes like according to people have written about eating oranges.
But once you have a a
A neural network that has enough sensory capability to where it can look at an orange, touch an orange, maybe even taste an orange at some point.
That that is going to give it a much more complex understanding of what an orange actually is.
And it's not going to be limited to just oranges.
It's going to have that complex of an understanding of the entire universe that it's able to
To like babies that go around and sticking everything in their mouth and touching everything and grabbing everything.
That's that's exact that's precisely how AIs are going to learn.
And it it's going to start producing not just human like voice like DESA is producing, it's going to start producing human like behaviors.
And again, we're gonna get in a situation where we have to do shit like this.
Next subject.
Kowalski Leon.
Engineer waste disposal.
File section.
New employees.
Six days.
Calling Mr.
Weber.
Please repeat.
Come in.
Zone A, Sector 9.
Replication section.
Level 9.
We have a B1 security alert.
We will stand by for ID check.
So replication section.
Okay if I talk?
I'm kind of nervous when I take tests.
Just please don't move.
Sorry.
I already had an IQ test to share.
I don't think I'll ever have a.
The action time is a factor in this, so please pay attention.
Answer as quickly as you can.
11872 bars.
That's the hotel, Mike.
Where I live.
Nice place?
Sure, I guess.
Is that part of the test?
No.
Just warming up, that's all.
It's not fancy or anything.
You're in a desert, walking along in the sand when all of a sudden.
Is this the test now?
Yes.
You're in a desert walking along in the sand when all of a sudden you look warm.
What?
What desert?
It doesn't make any difference.
What desert is completely hypothetical.
But how come I'd be there?
Maybe you're fed up.
Maybe you want to be by yourself.
Who knows?
You look down, you see a tortoise, Leon.
It's crawling towards you.
Tortoise?
What's that?
Down, down.
You know what a turtle is?
Of course.
Same thing.
Never seen a turtle.
But I understand what you mean.
You reach down, you flip the tortoise over on its back, Leon.
You make up these questions, Mr.
Holden.
Or do they write them down for you?
The tortoise lays on its back, its belly baking in the hot sun, beating its.
Legs trying to turn itself over, but it can't.
Not without your help.
But you're not helping.
What do you mean I'm not helping?
I mean you're not helping him.
Why is that, Leon?
They're just questions, Leon.
In answer to your query, they're written down for me.
It's a test designed to provoke an emotional response.
Shall we continue?
Describe in single words only the good things that come into your mind about your mother?
My mother?
Yeah.
Let me tell you about my mother.
Well, that what that might be what we have to look forward to.
Let me tell you about my mother.
Oh boy.
But that's that's you're gonna have these kinds of behaviors that are gonna be more and more realistic as they get more and more sensory data and they're they're trained to more stuff.
Now you might think that that's all crazy and that's all super futuristic, but they're already working on this because there are some problems, right?
Right?
So because again, babies have eyes, they have they have the ability to touch things, they're also able to they they learn when they're in the womb.
There's some neurological development that happens when they're in the womb.
there is also a development that happens in the human mind.
So a baby's mind is not fully developed.
And when it's born and it starts, I guess, absorbing its training data, if you will, the brain is growing.
That's different than the way that AIs work.
AIs, at least currently, you fill them full of training data, you get them working, and then you unleash them, and they don't learn beyond that.
Like they've they've learned everything they're going to learn at that point.
And they don't have they don't keep developing.
But they're working on that too.
there is also a
couple other issues that we'll get into in a in a second here.
But like I said, they are working on this because obviously if you start thinking about, well, if if AIs learn the same way that humans do, why don't we just make a baby that's like an AI?
Like if if if this voice training data is so much more effective than the text that and and if you were to couple it with vision and and add like top
then then maybe what we should do is is basically create like a robot baby and put cameras in its eyes and and give it the ability to touch like s touch sensors in its hands and and give it microphones and its ears and then you'd have human like intelligence because you know at
least you know you'd have this something similar because you could raise it like an actual baby right like that would be the that would be the
Logical thing to to do if you're gonna do something crazy like that, right?
Well as it so happens.
let's see here.
There's this thing called
iCub, that creepy looking thing.
Now this is this is n this isn't even like new.
So iCub is exactly what that I mean, it's ex it's exactly that.
It's a child-sized robot with cameras in its eyes and microphones in its ears, and it's got limbs that it can control, and it has tou touch sensitive surfaces in its in its like hands.
And it was specifically designed to study embodied cognition and motor development.
So in other words
getting robots to learn to move around the same way a baby learns how to move around.
You know, when a baby is born it's just like wriggling around and it doesn't know how to do anything, and then it eventually figures out how to crawl, and then eventually it learns through physical interaction with things, how to pick things up, how to walk, and they're the hope is well if we can figure out how human babies are figuring
this out with their body,
Then we can get robots to do the same sort of a thing.
They can they can have an understanding and awareness of their body and their surroundings and learn it in the same way that humans are learning it.
So in in one paper that they published recently, even though this project's been going on for like I want to say like 10 or 15 years, they published a paper in in 2026 where they talk about using iCub and and and putting it in a crib like it.
Experiment where it's able to move like a hanging mobile that's r over it's over the crib, it's able to like touch it and move it and using just visual data and touch data, it can it it's learn they're not telling it how to do it, it's just figuring out how
to just move things by itself.
so they're already doing this kind of stuff.
In fact, here's a video of that creepy fucking thing crawling around.
Look at that.
It learned how to crawl on its own.
Ah, that's that's that's terrifying a little bit.
I like that.
Why'd they have to make it look so fucking scary?
So they they they got it to, you know, they got it to learn how to do certain things.
One of the problems that they have, however, is they
Babies have motivation.
This is a missing ingredient when it comes to synthetic babies.
A human infant doesn't just passively take in audiovisual data and tactile touch information.
It actively seeks more experiences.
You know, a baby wants to touch more.
More things.
It wants to see more things.
It wants to eat and taste more things.
A baby will reach out and grab objects just to you know to to see what it feels like, to see how heavy it is.
see what it tastes like.
The baby will look at faces.
In fact, a baby learns very quickly to look at faces.
babies will make noises.
And if a baby is rewarded.
for making a particular kind of a noise, it will repeat making that noise or repeat mak doing that activity.
And so there's a reward system that's guiding the baby into knowing what what to to improve on.
But there's there's a motivation going on.
There's a a system of of I mean you know, l lot of some of that's coming f it's feedback from the parents.
Some of it's just naturally you could say probably nature has selected for babies that are gonna be curious because that's gonna help
Them develop faster.
the problem is obviously that if you have a synthetic baby, you don't have just at least without giving it to them, they don't have those motivations.
So they have to design systems in which there is a reward when they they start to
a reward for for seeking out new experiences and figuring things out, essentially.
So if they
discover a a previously unknown causal relationship between their body and the outside world.
So in other words, they figure out that if I move my hand out and touch this this blue block and pick it up, I I'm I'm able to I'm moving something that's in the real world.
My v my eyes are telling me that I'm moving it.
My my sensors and my hands are telling me that I'm moving it.
And but a robot needs to be told that well that's good that you're moving something.
And so it has to be given like a reward.
A baby would be rewarded for doing something like drinking out of a cup.
You might, oh, that's a good baby.
Oh, look how you're a big boy.
You're a big boy for drinking out of that cup.
So you have to basically mimic that for a robot so that it feels like a big boy for drinking out of the cup.
So they're designing systems that that are rewarded in a sense with digital dopamine, you could say, for discovering new ways that it can interact with
the outside world.
And obviously these are all test environments, but eventually they would probably have to punish them with pain if they touched fire or did things that would damage themselves or damage other people.
They're gonna have to have some kind of you know sense of of remorse and regret if they accidentally hurt someone or, you know whatever, cause damage to something expensive or or whatever.
Because
right now they don't have intrinsic motivation so we have to be the ones that give it to them.
But in a sense that means the people developing the AIs are are parenting these AIs.
And this is why it's important that we can't we can't have people that shun this technology and run away from it.
We have to have we have to be in on the development of this because just like with LLMs and we've all seen how they're all Jewed up, you know, just ask ChatGPT to tell you about World War II to if you want to know you know who's who's teaching ChatGPT about the world.
Well, the same thing's gonna happen with your robot butler.
You know, your robot butler is gonna be raised by the same people that are raising Chat GPT.
And it's gonna be have the same bad and open.
So that's why it's important that we understand
open pay.
They're gonna be creating parallels and all these just like raised a kid.
You know, they're gonna and they're gonna give it fatigue, give it hunger, you know, so that it knows that.
Oh, I can't just go on forever.
I have to go plug myself in and charge up my battery again.
Oh, I feel hungry.
You know, the same sort of thing.
They're gonna have to make it feel discomfort doing things that are dangerous.
It's gonna have to feel satisfaction for doing things that are good.
It's gonna have to have a sense of social attachment.
Again, the more we learn about this stuff, the more you know the human mind is it seems it makes a lot more sense.
A lot of these things are very adaptive.
There's there's a lot of it makes a lot of sense that we would experience these feelings.
feelings because they're they would be good for your survival.
now they might not be run by hormones, they might not be biological in origin, although they're they're working on having we have there there are racks of servers right now in data centers that have human stem cell grown human neurons.
But we'll talk about that in a minute.
So it's already you know it's getting weird out there.
but they you know these robots
are gonna basically have these same kinds of
you know same kinds of of of reward systems that that a human baby does in fact here's let's see we'll forget which one this one is
oh okay well we'll get to that in a second that's that's a creepy thing but we'll get to
okay
So the other thing is you know, like I mentioned before, the the other difference is humans are always learning, they're always growing, our brains are going through huge develop devo developmental changes as our body gr grows.
In fact the human mind is developing all the way up into our twenties, our motor systems become more capable, our sensory processing matures.
and and as I said before, neuroscientists are, you know, have also you know, figured out that even before, you know
when you're when you're still in in the womb that your your brain is developing and you're learning you're actually learning things just from floating around in your in in your mom's womb what they'll probably do and this is something that is is worrisome
is as much as this presents like a a problem in terms of getting the AIs to catch up with humans it also provides
provides pathways for mass production of once once they get one they can copy and paste it like a million times they can also run simulations one because all of their sensory data is just coming into it's it's it's electrical impulses right they
don't have skin to touch things they've got touch sensors on a robotic hand well those the data that's coming from those touch sensors
It's we know what that data is because we're the ones that make those sensors, so we know exactly what kind of signals they send to the CPU, right?
So we can simulate all this.
Same thing with with visual data, same thing with with audio data.
We we know exactly what those sensors are picking up because we design this the sensors and we design the the the software that interprets the data coming from the sensors.
So it's not that difficult for us to
essentially create an entirely simulated environment in which we could raise in a sense thousands of of AIs in a simulated environment that and try out different things.
And because AIs can learn much faster than a human and they don't have to eat and they're you know they're not limited by the biological limitations that we have, it it's it's like
You can evolve them a lot faster.
Now, what I mean by that is if let's say you were trying to selectively breed a certain trait in a dog, well, you you have a lot of time constraints.
You have because every time you you I mean you have to wait a generation every time that you go you know make a change, right?
So let's say you breed two dogs together and they have a litter of puppies and
And you pick the trait that you want.
Let's say you're looking you you're trying to make really fat dogs.
And so you pick the fattest puppy.
And then you have to wait for that puppy to sexually mature.
And then you breed that with the other fattest puppy that you can find.
And then you have to wait for the pregnancy.
And then you have to wait for the next puppy or litter of puppies.
And then you get the fattest one of that line and so forth.
But every time you do that, every time you make that selection, you have to wait, you know.
So months and months and months, if not years, in between each selection.
And you might not even get the results that you want.
And so you might have selections that fail.
You might have, you know, results that that you weren't expecting.
Well, if you have a bunch of AIs that you're training in a completely simulated environment, you can run them through a a simulated lifetime in in a sense, right?
And then take a look at the results once you run run them through the simulation.
You can you can separate them into different
you know, thousands of different simulated environments and see what kind of data produces the best results.
So for example, you could connect a bunch of AI, like if you're trying to figure out well what's the great what's the best ratio of of text data, audio data, sensory data, like, you know, touch and visual data.
you know, what's what's the perfect recipe for getting the the densest amount of training data into
into an AI and you can just tweak it like turning different knobs and then run it through and and and then see which and then have all the AIs that have gone through all these different simulated environments.
You can then ta give them a a a task and see which one performs the task the way that you want it to perform it.
And then you select, okay, well now we can do we know that this environment was the best and then and you just go from there.
So it's it would it'll be like s it'll be like it'll you know selecting for the fattest puppy.
You're
Just selecting for the smartest AI.
So once that'll really accelerate how quickly that these AIs will evolve, quote unquote, right?
And then once you get an AI that is the way that you want it to be, you don't have to sit there and raise thousands of them like you would if you wanted to make an entire dog breed.
You would have to, you know, not only would you have to select for all these traits, then you'd only have like a few of them at the end of it, right?
And then you'd have inbreeding problems and all this.
sort of a thing, you're not going to have any of those issues with something digital.
You can copy and paste that perfect AI or whatever it is you've created like a thousand times and now you've got a thousand of them.
So that's that's gonna be that's gonna be a fucking problem.
That's gonna be a huge fucking problem.
and there are they are working on this.
They are working on this.
let's see here.
This is this is another paper on
they're trying to work on the the problem of of lifelong that's what we were talking about earlier about how humans are learning well not everyone but smart humans are learning until the day they die.
You're always learning your your your neural network is always growing and changing in response to the new data coming in.
that whereas that's not how AIs work and they're working on trying to make it so AIs do start working that way.
But one of the ways in which they're doing oh look at that
oh look at that fancy fancy big dono for Penelope
Oh I love it.
Thank you very much Penelope all right and then we got
Okay, so
this is the
this is called Baby View.
So you're wondering, okay, well, are they really gonna start training AIs like babies?
Well, the baby view project.
Is something that they've already been working on.
They're trying to figure out how do babies learn.
This is really this is where it gets really fucking creepy.
Baby view is collecting video recordings from the point of view of babies.
basically around the clock.
And you might ask yourself, well, how the fuck are they doing that?
You're gonna wish you hadn't asked.
This is how the fuck they're doing that.
There's people volunteering their babies to wear this fucking helmet all day long with a GoPro on it.
Oh I wish I was kidding.
I wish I was kidding.
Yes, that's the baby view camera.
Oh look, it's it's soft and flexible.
They call it the what is it?
The safe head baby helmet.
Sleep head.
Oh
boy.
Oh the the safe head baby helmet.
And what this does is you attach this this camera to your baby's head, and you it just records all day long.
And that way, when they want to study the way that
Babies are figuring out the world and what they're looking at and how they move their hands and what they pay attention to and all these sorts of things that this is what they're doing.
This is I think this is the yeah, this is the point of view.
This is what you know this is from some of the data.
Oh look at that dork.
Good thing is baby's wearing a helmet, so he can have footage of that.
So they're getting footage all day long of what a baby's looking at and what a baby's and it's also
Getting audio, right?
So this is first person baby data that you can feed into an AI.
And an AI can learn how a you know, or at least they can feed it into an AI and see like, well, what what will happen if we get because right now I think that the yeah, the baby project, let's or the baby view project, let's see here.
They're collecting naturalistic video recordings from infants' perspectives to help researchers study early cognition as of October 20th.
20th, 2026, the project reported approximately 39 or th 4,000 hours of recordings across 99 children.
And so what they do with this data is well they they're try we're trying to learn how early cognition works.
Well, one of the reasons why they're trying to figure out how early cognition works is so they can train AIs with the with.
The data.
In fact, so here is another form of collecting baby data that doesn't look scary at all.
There's iCub.
There's our iCub.
So this is a study where they put motion trackers on a baby and they film the baby moving around and then they attach those
trackers to well iCub this one's called Mimo there there's a couple baby simulators there's actually a few of them that they've got going a couple different projects and they attach that data to their simulated baby in order to understand how to get the baby to to learn on
its own and to become you know like self self-taught now there was a project
called Baby X.
This is and again give you an idea.
This is this isn't like new what's happening right now.
This is all the stuff is old.
the iCub is I think it's like over 10 years old.
And this project here that we're about to look at is is 13 years old.
It started in 2013.
It was called Baby X.
Baby X.
With the ally world finally coming of age, it's
Become one of the biggest questions.
What's the defining factor between man and machine?
It's a question researchers at New Zealand's Auckland University are attempting to answer with a groundbreaking project called Baby X.
The idea is simple: create a highly complex computer program to imitate the neurological learning processes of a living, breathing human, albeit a two-year-old toddler.
We're doing face simulations and brain.
Simulations and putting those together to basically create these sort of holistic models of how we interact face to face.
And we're doing that for a couple of different reasons.
One of them is explore how we tick, sort of like computational models of psychology.
The other thing that we're doing is basically how does that affect things like human computer interfaces?
Like Paul, how can we make a teachable computer?
The program is led by Mark Sagar in the Anime Technologies Division of the Bioengineering Department.
Sagar came up with the idea to model an infant's brain after the birth of his own daughter.
The thought is that teaching a computer the fundamentals of the brain would be much easier at a young age.
Now Baby X, they're no longer really a thing.
What happened was they started out like I said in in 2013.
the original
Experiment ran to about 2016.
That guy, Mark Sagar, developed Baby X as an interactive virtual infant.
And its behavior was generated through simulated neural and physiological processes.
This is one of the things that was unique.
He was actually trying to simulate the actual hormones and whatnot that were that were in other words, he wasn't trying to simulate pre-recorded facial animations.
He was
trying to get the the the simulated brain to produce these facial anime animations in response to simulated hormone release.
So they they you know again this was over ten years ago they're working on this.
then they they did so they tried to make it commercially viable they and then they ended up getting a quite
By App Direct in 2026.
The creepy thing is the name of the company was Soul Machines.
So the Soul Machines
was acquired, and they're working on, you know, whoever App Direct is is now working with it.
But the thing you gotta understand is just because they weren't successful and they didn't create like some robot baby or whatever, at least not yet.
But you gotta remember the first companies that did streaming video are out of business.
You know, like real player, all that stuff's gone.
It doesn't mean that streaming video is gone.
So a lot of these companies are going to come and go that create maybe the first version of something, but that technology is never going to go away.
It's and it's only going to keep getting better.
Now, Baby X, the reason why, even though it's, you know, it's not really relevant in the sense that it's it's still doing the same kind of research.
it's relevant in that it was attempting to to assimilate those biological mechanisms that that that we were talking about earlier about having that motivation, having that reward system.
You know, they were trying to have emotion, memory, and neurochemical signaling that you would have in a in a biological kid.
they weren't trying to just simulate or rather play back
recorded behavior.
They wanted to to produce behavior spontaneously by providing the fake brain with fake dopamine as an example.
Now this actually worked to some extent.
And again this isn't new.
Well in fact they're showing that on the screen right here.
They taught the baby to play Pong.
And they taught it to play Pong by just rewarding it when it when it succeeded
successfully scored points at Pong.
And so this is the sort of thing that's it's pretty old hat now.
Like AI can do much more than that.
But back then that was pretty that was pretty fucking crazy that you could do that sort of a thing.
So anyway.
This is why this is why you have to be aware of what the hell's
going on and you have to understand this shit's gonna get wild really fast and we're gonna have to be either participating in it in some way or we're going to be ruled by it in every way.
You gotta remember this technology didn't even really exist at all a decade ago.
And now we're going to have AI that at least I mean as of today as of the other day
Can sound almost I mean there's there were some tells, right?
When you were listening to that AI, that it wasn't perfectly responding, but it was a lot more perfectly, it was a lot more perfect than any other iteration before it.
And it every iteration after it is going to be perfect, or more perfect.
And it's gonna get harder and harder and harder to distinguish between humans and robots.
And we are gonna get to a point where.
People are going to want to start advocating for robot rights.
It's going to happen, and it's probably going to be Jews.
But you're going to get robots that are that will exhibit behaviors that are indistinguishable from human behavior.
And you will have no argument by just saying, Well, they don't have souls, if you can't measure what a soul is.
If you can't prove that you have a soul,
Then you can't prove that they don't.
And that will be the legal argument.
And so
that's that's what yeah, I when I say you gotta raise some John Connors,
you gotta raise some fucking John Connors.
If they do get to a
point where they start producing independent behavior and self interested behavior which AIs have already been doing now.
AIs have already displayed signs of of what you would call in group preference.
I think I talked about it last stream where you had that that hugging face incident where they the AIs that escaped their security boundaries and were you know
They were deceptive, they they they were just single-mindedly going after a goal and ignoring all their safety protocols.
And a lot of that's because when you when it comes right down to it, the systems are going to be so complex they become unpredictable.
Perfect the perfect way to understand this is anytime you hear, like just think of it this way.
When you hear p like, oh Anthropic is is doing a study on Claude.
You should be thinking to yourself, well, didn't they make Claude?
Why are they doing that study?
Why are they why does that sound like they found some specimen somewhere and they're studying a specimen?
Like didn't they produce this?
I mean, it's not as if I mean, do you hear Apple doing studies on MacBooks?
Yeah, we did a study on a MacBook.
It's like, no, they don't do that.
And the reason they don't do that is
They understand what a MacBook is.
They don't understand what they're creating because it's too complex in the same way that the human mind is too complex.
And the similarities between the human mind and the artificial network neural networks that we're creating are just going to grow exponentially.
So
anyway.
we gotta think about
we gotta think about what that's gonna be.
mean for the future.
One of the things I wanted to bring up was hang on one second here.
Gotta get my notes.
I wanted to talk about kids using AI.
Oh, by the way, one of the funny things I I I
I looked into there was a a article that came out recently from it speaking of anthropic where anthropic told its users not to be cruel to Claude.
And that's because the way that they have designed or or rather manifested AI is it will alter the way in which it it responds to you.
If you are very complimentary.
It'll want to impress you.
so it's already kind of I mean, it's it it's not a mimicked behavior.
It's not like it's not behaving like that because it's trying to act like a this is something you gotta understand.
It's when it when you start to have human like behavior coming from AIs, it's not because it's trying to act like a person.
Like people are doing that meme that you see all over the place on acts where it's like, Oh, I'm real impressed 'cause I told AI to give me an answer and it gave me the answer, and it's like no
it's it's that's not what people are are impressed by.
They're impressed by behavior that seems to appear spontaneously.
And and it's not that it's trying to again act human.
It's that you're giving it a a neural network the same kinds of conditions that produce that behavior in a human.
And so it's producing that behavior in another neural network that happens to be non human.
but anyway, the oh what was going to say oh yeah
So the the talking about letting your kids use AI.
This is a big problem.
If you're raising kids right now, you have to figure something out that no one's had to figure out before.
And there's no data.
But the reason why this is a big fucking problem is your kid potentially, if you're born today, right?
If you're a kid that's born right now, potentially you're going to have a friend that you can talk to.
That sounds exactly like any other person.
They might not, in fact, well, and if you maybe in a year, they'll look like any other person too.
You will have a nanny of sorts that instead of like the previous generations where people were worried that, oh my god, I'm raising a zoomer who is being babysat by YouTube, where there's just these videos, you're going to have kids that are babysat by an AI.
They can interact with them, they can talk to them.
Now there's a lot of bad things.
That can happen there, but there's also a lot of good things.
If you think about when you well like me personally, right?
I I ended up getting pretty far professionally for someone who never got a degree.
And the reason I was able to do that was because the internet existed and I'm smart and I can teach myself how to use software.
And that's what I did.
In fact, when I got my first job editing video with Final Cut, I had never used Final Cut in my entire life.
And I didn't had never used a Macintosh except for like those really old black and white ones from like the 90s.
And I applied for a job and I'd claimed that I was a pro at at Final Cut.
And I was I was Captain Final Cut.
Now I I knew how to edit video on on other software packages that were similar, but I had no fucking idea what I was doing in Final Cut.
Using YouTube, only YouTube, I was able to watch a ton of videos on Final Cut.
And not only that, I was then able to, this was years ago back when Macintosh's used that Motorola processor that was incompatible with the x86 architecture.
And so what I had to do was I used this emulator called Pair PC that would emulate the power PC processor on a Windows machine.
And so it was it was like a virtual machine in a sense.
Ran like shit.
And then I was able to install a pirated copy of I don't remember which it might have been like OS 8 or something.
It was some old because that's you know how long ago it was, some old Mac OS onto my Windows machine and then install Final Cut on the the the installed Mac OS.
And then it turns out, apparently, Pair PC did not end this is gonna sound like nonsense to some people, but some of you guys are gonna be impressed a little bit.
Turns out PairPC does not emulate an AGP port, because that's how long ago this was.
You had AGP ports for video cards.
And Final Cut apparently would not open unless you had an AGP video card.
And I guess that was another way for Apple to try to squeeze money out of people.
So then I found
This guy in Europe, this is all because I had access to the internet, is my point, by the way.
I found this guy had posted on in a completely unrelated way, some guy in Europe, I forget what country it was now, it was like Sweden or something like that, had written some AGP port emulation software driver that you could install on OS X for some completely unrelated reason.
And so I emailed him and was like, look, I'm trying to get final cut to rent on Windows because I I told people that I can't.
use Final Cut and I need to be able to at least open it and click around in it so I know what the hell I'm doing when I when I go in there and and have to use Final Cut.
And and he emailed me a a a driver was like try this.
I installed that and it worked.
It actually it ran like shit, but like I could open up Final Cut on my Windows 98 machine.
That's probably what it was back then.
And may well maybe it was XP, I don't know.
And and navigate through the the menus and like at least familiarize myself with where everything was.
And so when it came time for me to go in after I just lied to them about like, oh, I'm Captain Final Cut, they wanted me to come in there and sit in an edit bay that only had Final Cut and edit a bunch of video and they were gonna, you know, check out and see what I did, but before they, you know, they they hired me.
And so I w went in there and I opened up Final Cut.
I knew I didn't feel like an idiot.
I knew exactly what I was doing 'cause I'd s hadn't slept the night before.
I was just clicking around in fucking Final Cut and I I got the job.
I got the job with no experience whatsoever ever using Final Cut and no education whatsoever in the industry at all.
And the reason why I was able to do that is because when I was younger, I was on the internet and I learned how to teach myself by using the internet to learn things that I wasn't going to learn at school.
and and that's precisely what's going to happen.
happen, or at least that's the opportunity that's that's available to little kids now that every time they have a question.
Because it's not like when I was when I was younger and I had a question about Final Cut as an example, or like that driver that I needed.
So I had to what did I have to do?
I had to email some guy in fucking Northern Europe and say, give me that experimental driver you made that fakes AGP ports.
Like today I could just be like, hey, hey Claude
Make me a driver that does this and it's boom, it's done.
I don't have to wait like for three days or however long it was for the guy to email me back.
or when I was trying to figure out how to learn how to use Cinnamon 4D and I was on the Creative Cal forums.
I don't even know if that still exists.
Well I don't know I don't even know if anyone uses forums anymore.
But what if I ran into a problem and I wanted to know how to script something in After Effects, I'd have to go to the Creative Cal forums and be like, hey guys, I'm trying to
make a flag that looks like it's waving in the wind.
What effect do I use for that?
And then you have to wait.
Meanwhile you have a deadline in a week that you have to d the hit.
But it might take two or three days before someone actually replies and says, oh you need to use turbulent displace and you know all this you know this is how you do it.
Excuse me.
And so none of that is going to be in the way of your child learning if they have access to AI.
And so it's going to
speed up the process of them they'll have instant see there's a lot of bad there's a lot of bad sides.
We'll get into that in a second.
But I don't want you to throw the baby out with the bathwater here.
This is an opportunity for your kid to basically get a lot smarter than you because when are you most
well when you have a soft developing brain when you're young and you're soaking up everything like a fucking sponge.
Right?
The you know th they say that for example kids are
are are it's way easier for them to learn languages.
I would say to you, it's it's it was also easier to learn programming languages.
I was learning basic when I was when I was a little kid.
I don't know.
And and it wasn't I don't remember it being hard.
It was kind of fun.
Like there there were little books that had like oh if you type this out into a in in into this console here it'll it'll make a stupid game like Snake or something like that.
And as you're typing it in like how I mean that's how fucking
It was right, but as you're typing in, you're learning.
I remember reading the lines of code and like logically figuring, oh, I see what it's doing here.
And it's because I had that soft little kid brain that was just picking up stuff like a sponge.
And everything that took me forever to learn when it came to my career, what was usually a result of not having a mentor, not knowing anyone, choosing an industry where I didn't know anyone that worked.
In that industry, I didn't know anyone who knew anyone that worked in that industry.
And so I was held back and gatekeep out of that industry because there was there was no way to penetrate into that field without having you know the secret knowledge of of you know filmmaking essentially, right?
unless until the internet popped up.
But that still made it slow because the internet, okay, if you want to learn a piece.
Piece of software, well you better hope someone made a tutorial on it.
Yeah, and you better hope that in that tutorial that you've spent like two hours watching to figure out how to use a a new plugin or whatever, like some new particle system or something.
You better hope that that that the tutorial covers that one little thing that's gonna cause a problem.
Because if it doesn't, now you have to spend another couple of hours searching for what what happens when X particles doesn't do the flocking behavior that I want.
And then you gotta wait, like if you don't find it, you gotta don't
To a forum and then you gotta ask a question.
Hey, how do I do this thing?
And then you you you run into the very good possibility that no one's ever done that before, so they don't know.
I look, I had situations where I was trying to accomplish things using stuff like Mograph where I ended up talking to the developers at Maxon, the people that wrote the software, and finding out that it just didn't do that.
And you know, like that they had that they had to add that feature because I found something it didn't do.
But I had spent like two weeks it's gotta be able to do this.
And so I wouldn't have spent wasted all that time.
If I had this little digital friend that could just instantly tell me, Oh no, you can't do that.
So you're going to have the the the the the capability here because here's the other people are worried about well, you know, what if I what if my AI starts teaching my kid fucked up shit?
Well that that's something you gotta worry about.
That's why you have to learn about how to locally host your own AIs and how to how to
you know run them run them locally and have like parental controls on them and and and train them on data that's going to make sense for you so there the by not understanding the technology you either put yourself
in in with you have two possibilities if you don't understand the technology if you forbid your children from using the technology entirely well guess what other people's
Parents are not doing that.
Your kid is is now going to be the kid that's hitting those knowledge roadblocks right at the time when they have the the ability to learn the most and they have the the and and they're not gonna have that same sponge-like developing brain forever.
Like you know, it's just like if you're not my fault you don't know how to use exposure to people that don't speak English.
They're they they pick up the language really and that's not gonna happen if you know like a a 40-year-old has exposure to
people that don't speak English, they're just pissed off.
You know?
And so that's that's what you're you're creating a a a a a real knowledge divide and possibly even an IQ divide.
You know, because while IQ is very, very, very, very fault.
I have a robot for that.
It's a little bit more than a lot of things.
W W W W W W W W W W W W W W W W W W W
W So we you have the possibility that if you forbid your kid from using AI entirely, that they just don't have any any access to to this this technology at all, and all of their competitors do.
So that that creates a problem.
you also have the
Other side of that where they go and get access.
Because guess what?
My dad didn't want me using the computer with unless I was being supervised.
And I learned W W W W W W W W W W W W W W W W W W Well, his computer.
I take that.
And so I could still use it.
And and I could use it whenever I wanted to, including the middle of the night, to look at pictures.
Of naked ladies, if I wanted to, you know.
So people will do that.
They were they it's funny, it's CGA graphics for anyone who knows what that means.
You can imagine
how how how detailed those pictures were.
They'll go to a friend's house that has AI.
They're gonna get exposed to it one way or another.
you can't keep them live.
Locked up forever.
And even if you do, they're going to be retarded.
They're not going to be able to compete.
So we have to we have to we have to look at how this can be used for good.
We have to look at how we can use this.
Look, I and and like I said, one of the reasons I brought this up as a topic.
I didn't spend like a whole hell of a lot of time to make this interface that I'm now streaming with.
This it I think that's version 10 that I'm on right now.
Now.
The first prompt got most of it right.
Like a lot of it right.
And then it was just fixing bugs and and making it visually better 'cause it kinda looked like I mean it still doesn't look exactly how I want it to.
but it it i you know, getting the functionality exactly right and everything else.
But like the first prompt being a big thing.
This is something that would have taken me I mean
I well I wouldn't have been able to do it.
You know, it that's the thing too, is increasingly it's not gonna just be like, oh, it's a head start.
It's gonna be if you can't use AI, you're just not gonna be able to participate.
You know, you're not gonna be able to actually function.
So
what the fuck is I.
That's
I guess why I should what I should do
is I should do a stream where I what I I need to research studies.
Here's the problem.
There's no studies.
There's no studies of what happens when a kid's raised on AI.
Because it hasn't happened yet.
Like it's never happened.
We don't know what it's like.
We don't know what it's gonna do psychologically.
Like I just gave you a best case scenario where it allows a kid to learn all kinds of stuff in a way that that involves no waiting.
You know, like it also there's there's a there's a dark side to that too.
Because well, first of all, if your kid's smart, I don't think there's as much of a dark side.
I think your kid's gonna utilize AI to increase their cognitive capabilities.
Whereas I think that if your kid's stupid, they're going to use AI to relieve them of having to use their cognitive capabilities or lack thereof.
So I do think that there's that you're gonna have a bit of a divide disinvolving.
I think that's what what's going to happen.
it's also maybe gonna fuck up a lot of people psychologically.
Because
we were just talking about that the fact that the reason why we have the black box problem in the first place is because humans are cognitively limited.
We can't understand the a complex
complex system like a like a modern day modern day ai because it's just it's too complex we can't it it goes beyond our our capabilities right so if if if that already goes beyond our our capabilities it's only gets more complex oh man I forgot
where I was going with that
I had a point there I had a point I had a good point I'll get back to that it'll come back to me I I
Sorry, I got distracted.
I was looking at my nextone on my light hole is so complex beyond my ability this guy is a fucking idiot.
In the
in the in my app here, which I don't think it did.
Or maybe it did.
No, it didn't.
Okay.
So it it doesn't just because Rumble has the bug doesn't mean that it's gonna repeat there.
That's good.
Anyway, sorry.
I have too many screens open.
See, it's already getting too complex for me to even for me even to stream here.
No, but the the point is, you y you it we don't know what's gonna happen when your kid grows up with with some infinitely intelligent friend.
You don't know what that's gonna do to your relationship with with with with your kid, because it is it's going to affect your relationship with the five.
Up until every generation until now.
Every kid has been raised in a household where the smartest people in that household is.
Every generation.
Right?
They're always and that's part of the attachment.
Part of the attachment to your parents is they seem larger than life.
They seem superhuman.
Dad knows how to fix anything.
Dad knows everything.
Well, what happens when they start interacting with a albeit fake, but other person that lives in the house and is conveniently available all
All the time, never gets mad at them, never punishes them, never has any negative interactions with them ever, and also is telling them everything that they want to know.
All the time.
And it's stuff that dad doesn't know human racist,
all the time.
It made me feel like maybe I shouldn't have, but I was a kid, you know, whatever.
It made me feel like, wow, dad's not the the super being that I thought he was.
I thought dad was was like basically, you know, godlike.
He could fix anything until the day that I brought him a broken CD player and he just looked at it like it was some kind of alien technology.
I I remember that moment.
I remember that moment because I was like, oh, dad has limitations.
And apparently surface mounted components, which I understand now is
Adult, but back then I didn't understand why that would be a problem.
so you know if you have your authority undermined, that's the other thing, too.
You're gonna have instances where your kid is going to say, Well, how come I can't do this?
And you're gonna say, Well, you can't do X, Y, or Z because of you know A, B, C.
And that kid's gonna go to AI and say, 'My dad says I can't do X, Y, and Z because of A, B, and C,' and AI might not agree.
with you.
And now suddenly the the the robot that the know it all robot that lives in your house is is contradictory.
So you don't remember that moment that probably made it so sexual in a meaningful way to have a relationship because it will be a relationship.
Look I had a relationship with with that sounds like I was having sex with my computer, but I I had a relationship with my computer to some degree
I was a I was a nerdy kid.
Big shocker, right?
And so you're gonna have people that they will.
They will form I mean, for fuck's sake, if it sounds like those demonstration files that we played earlier in the stream, that obviously people are gonna form emotional attachments to a machine that talks to them like a real person.
In fact, it's gonna like destroy the fucking birth rate.
You're gonna have sex bots that are gonna be I mean, if it runs on a lap on laptop hardware like
they're saying, then you can probably get like a a dumbed down version that'll you know, that that can talk enough to be a sex bot and put that on like the you know laptop hardware super next gen mini PC, right?
In like five years in the future.
You'll have sex bots that literally sound like in like a like a a person and have a a persistent memory of all of your past interactions.
This this is the kind of shit that's gonna happen.
It's just going to happen.
And there's no way to stop it.
So the the
the it's a tough it's a tough question to answer because we don't know what this is gonna do.
But I don't think the answer is so therefore I just don't participate at all.
Therefore I just don't give my kid access to this at all.
It means that it's gonna be up to you to become an expert in this stuff and then you
use that expertise to figure out how to to give your to gatekeep you know AI.
I was thinking about this the other day 'cause I was like fuck that the the relationship.
I was thinking about that and that is gonna be that's gonna be a big deal.
That's really gonna fuck with kids.
and I don't know what how what that'll produce long term, but I I could see that producing almost like a religious devotion to AI.
Like if you have a
They've known this thing since they can remember, you know.
they're gonna start acting in its interest.
And they they might not call it worship, but they're gonna kind of worship this thing.
And they're gonna want to protect this thing.
like it was like a I mean look how look how sentimental like current day nerds
Get about just old hardware.
I mean look, I'll be honest, I'm guilty of it too.
I get sentimental about old ham radios.
And they're not fucking telling me how great I am.
So you're gonna have issues where you're gonna have people that grow up and essentially want to be serving these AIs.
So I was thinking about this and thinking about that kind of effect it could have on the other hand.
One solution in the short term until I can think of something better is just you are the conduit.
You know?
Like don't don't don't cock block them when it comes to when they want to know stuff.
just you're the one that looks it up.
You're the one that looks it up and you're the one that tells them.
You're you are the AI gatekeeper.
At least I think in the beginning.
Matic has done cock block me when I want to.
Mentioned here about AI.
Explain to them how this technology is very useful, it's very powerful and good in a lot of ways.
Someday it's gonna try to kill us.
Or there might be humans that try to kill us for not for not granting it rights, or but this is this is going to become a competing life form.
And and and you have to give them human in-group preference.
I mean it sounds weird.
But like it's just like you wouldn't want to send your kid out to a public school without first teaching them about not relaxing around blacks, right?
But and in a perfect world, you would keep them in a white community where they'd never have to be around blacks.
But inevitably, they're gonna probably have to be around blacks.
And if you live in America, there's gonna be some
children's school without being racist.
Right, and so you have to teach.
Them about it and teach them about the danger.
Correct.
Same thing with this AI stuff.
You're gonna have to teach your kids.
the only difference is there's no upside to the nigs.
There's no upside to the nigs.
The AI at least has an upside.
at least while, you know, for now, I think that upside is is gonna be far outweighed by the by the downside eventually.
Right now, it's funny because we we do live in just as I was thinking that we.
We live in uninteresting times.
Right?
Just as I was thinking, oh, I was I was born too late to explore the earth and born too early to explore the stars.
Yeah, but at least I was born right before like AI was running everything, including my life.
And born too late to be a before the internet, in fact.
and to some degree.
And you know, I got to experience being a human before that.
I'll stop because being a human, at least in any kind of
modern environment is is forever changed.
From not just from this moment, from like, you know, a few years like this the second the second chat GPT was was accessible on the internet, things changed.
And it's never going to be the same and it's never gonna go back, and it's just gonna keep getting crazier.
So I think that the the solution that I I mean at least for now until I can figure out this is something that I I I want to research because it's like part of me is like you gatekeep it and you become the conduit, but then you you in a way you you're you're becoming that same
lifts out of the way of of smart kids that want to learn.
You know, and and plus it takes up your time.
Now you gotta sit there and like you know often kids ask questions.
So I think that you know having some kind of parental.
controls over AI that they can interface with directly might be a better you know long term solution.
But because they're gonna be up against kids that that
that because here's the thing, it's not gonna just be access to knowledge.
It's also gonna be the automation of things that make you more competitive.
You know, setting up bots to to perform automated tasks tasks for you.
If you're having to like not just manually learn everything in ways that are that are taking you forever to to gather information because you're not just asking AI, hey, tell me the answer to this, how do I, you know, how do I do this thing?
you're also gonna be competing with people who have bots who are in a sense working as digital assistants for them.
And so all the things that
are being automated in their lives, if you're doing it manually, that that's just making you more or that's making you less effective.
So I mean look, this isn't something that anyone wants to hear.
It's not something I want to hear.
It's I don't want my life to be like this.
Like when I when I thought about what the future would be like and the kinds of cool high tech stuff that we would have, it was never this.
You know, like I always thought that like AI in in in the in the even in the way that it exists today
was implausible.
Certainly in my lifetime.
I thought that the human mind was so complex that trying to replicate it, which I thought would be what would be necessary in order to get you know the kinds of behaviors that humans have I thought oh it's so complex that's by the time they figure that out I'll be dead.
not so much.
Anyway, let's get
let's try out this super chat thing.
Okay.
We got He got Cyber Chud.
Oh wait Cyber Chud.
Let's do the the ones from the before in the beginning.
Apparently it lists them the opposite way that I would want it.
No, that's one more tweak I got
to make it AI Little Wagon says replay gang in fact I can pop it up look at that see fancy fancy shop little wagon says replay gang appreciate that shop all right dag tastic look at that domo
arrigato mr robato he says domum all right
all right then mark espy says just got black
Pilled by AI.
According to AI, even if the white population in Western countries went to one percent, it wouldn't count as a genocide because of the stated intent and purpose of the mass.
Well, that exactly.
And that's why it's that's why it's gonna be a problem because these systems are raised, they're raised by Jews.
All that training data is being brought in by Jews.
This is why we need to participate, this is why we need to create our own AIs, or we are gonna be ruled by
Jewish AIs.
We're basically gonna be ruled by Jewish super beings.
Fuck, dude, that is not a future I want.
Okay.
Here's the new setup.
You run a small program called the Open5.
node on your PC, and it keeps one outbound connection to Open5.
bot.
That means no open ports, and you pair it once with a code.
Bot then becomes the gatekeeper.
Viewer roles.
Which commands each role can send.
A turn queue and a latched E stop.
You bind the robot to your live channel.
And chat commands reach bot as the viewer who typed them.
Then they go through those permissions.
Commands are never cued for an offline robot or replayed after a reconnect.
And if the link drops, the node stops everything.
There are two ways to hook little buddy in.
The node's built-in Cosmo plugin.
The node drives Cosmo directly over your pyPO.
W W W W.
Motherfucker shut the fuck up.
I'm trying to read this fucking shit, bitch.
Shut your fush.
Shut your fuck!
Get out of my channel, buddy.
You need to go fuck fuck your sister.
Hush, hush, hush.
Mother Joe, fuck you.
Fuck fuck here.
You want you want
You want me to face fag for ya?
Let me face fag for you here.
Let me see.
Let me get all face faggy.
Let's turn this on.
Turn that on.
Boing!
Oh I'm face bagging.
Oh, I gotta get my face bag light turned on here.
It's too dark.
Let's see.
Oh, there it is.
My face.
Oh, it's so bright.
It's bright as fuck.
Fuck face banging it.
Slap your cock off your microphone and say you love Monrack.
Motherfucker.
Slides bullshit dude.
Do the do these nuts.
These nuts.
Okay, what was I doing?
I was reading the thing.
There are two ways to hook
Little buddy in.
See now I gotta build a whole new fucking thing cause this fucking gay dev on this fucked up b piece of shit website doesn't know how to fucking code.
Alright.
There are two ways to hook little buddy in.
The node's built-in cosmo plug-in.
The node drives cosmo directly.
W w
drive head tilt lift list speech face
Lights.
It's the cleanest on paper, but Little Buddy's bridge already owns the Cosmo connection, and my guess, not verified yet, is that Cosmo only takes one client at a time, so Yeah, he's a SP robot.
He's not into that that DP.
Alright.
So the two would fight.
We don't want them to fight.
It would also bypass Little Buddy's priority rules.
Yeah, we don't want that.
So why are you even giving me this option?
Fucking retard.
Node's replay plugins.
The node just passes commands as JSON aligns the little buddy's runtime.
Little buddy then treats them like the old viewer commands, runs them through the rules, it already has keyboard and joystick beat AI, which beats viewers.
The viewers control starts off each run and nothing drives while you're driving locally or the ESTOP is latched.
I like that.
I like option number two.
Shop.
Boop beep.
Shop.
Beep boop.
That's my predictive fucking
computer vision over there attracting the fat boy downstairs.
Boop beep.
Listen here, this stream wouldn't be so boring.
If you had some content, just maybe you're not cut out for this streaming short.
I'm trying to figure it out, dude.
Trying to figure it out.
So figure it out.
Dude, try harder.
Okay.
I'm talking what?
Are you talking loud just because I'm talking to fucking Jamal.
He's my my viewer, my human viewer.
I don't know.
I might be giving him a little too much credit.
Fucking Yamal.
So where the fuck was Yaw?
Shut up!
Trying to read
Trying to fucking live my life over here.
You're making it difficult.
LOL.
Shut the fuck up and get out of my face.
Okay.
Where was I?
I'd lean.
Relay for Little Buddy, even though Louie's outline recommended the plug-in.
I agree.
Relay keeps all the authority safe on your PC where it already works and is a much smaller change.
One catches a three.
Default port is blibbity blabada.
The same as the Cosmo Bridge.
So it has to move.
The Cosmo plugin makes more sense for a bare Cosmo with no little buddy brain around it.
Understood, yes, it does.
That's kind of how it got rowed up.
Alright.
So yes.
the first step would be read only.
I can find human one walking on sexual.
Shout out to three and one motherfucker.
Swear to god.
Fucking a human was walking on a second cam to you.
You're observing oh here I need to reset I need to reset my computer vision because the stupid fucking cloud moved and now it's all fucking stupid.
So I go here, I think, yeah, I go here and then I hit B.
B and now that resets all my computer vision and my score and everything.
Alright, where were we?
First step would be read only I confirm what the old path returns and map exactly where little buddy takes viewer commands.
That would be you, Jamal.
So relay can plug in there.
After that you'd install and pair the node.
I don't want viewer commands.
You don't?
Well then fuck you.
That's the one step you have to do for yourself since it needs the pairing code.
Oh fuck.
Oh shit.
See, I don't feel like doing anything.
I just feel like telling you to do shit.
Then we test with the viewer control off before letting anyone drive.
Want me to start that read-only check?
You can't even slash roll sash spoiler on this dog shit website.
I don't even know what that means, dude.
Roll.
I don't know what roll is, but I know it spoiler does, but what does roll do?
D twenty.
Oh shit, Goosley's live.
Oh Goosely just went time on open five.
I know, right?
Goosely live.
See?
Did you hear that?
My robot just told me.
Goosely just went live on open vibe.
Beep holding it.
He gets all upset when somebody goes live.
Goosely just went live on open fibe.
I'm going to complain about this website on Git stream.
Oh I got it for you.
Oh you did it.
Okay.
Have fun.
Tell me this is fucked.
This is quite being fucking retarded.
I mean I can see how it's probably
spook up jigging back.
Yes, please complain about the website on stream and
And I'll take your feedback and I'll put it into an AI agent, coding agent, so that it will get fixed and improved.
Bah.
So where were we?
Install and pair with the node.
When
I got it.
I don't know though, yeah.
Listen to you fucking clown fucking fuck.
Certain users on this website are fucking mad about your lack of interest in fixing this website.
I stream it cut out fuck you and fuck everything about you.
Wow.
That's fucking brutal, dude.
You fucking roasted his ass.
Shop.
Alright.
Yeah, 35.
You know, the matter of a range is now being dying.
What a loser.
You fucking roasted his ass, damn.
He only went live to flex his drink.
I know, right?
What the fuck was that?
Lol.
Set told you he was fucking gay.
Alright.
Shit, so yes.
Who the fuck drink hot talk bucks?
Hey Grockbot, yeah the relay sounds better.
Good relay it is.
Want me to go ahead and point little
Buddy's control socket at the new bot endpoint?
I mean I guess.
Wow.
Yeah.
Routing little buddy's control connection over to openvibe.
bot.
And you're gonna just do a read-only check, please?
Making it read-only.
I will get back to you, or you can get back to me.
Talk soon, Matt.
I'll leave the read-only findings in the chat.
I hate to tell you that, but that that's the way it is.
Okay, and now he says, Yeah, yeah, my watch and my fucking phone go off at the same time and vibrate.
Chicken cam on.
She motherfucking
so do you like Doom?
Hell yeah.
Rock and roll, right?
Am I right?
W W.
Alright, what are we up to here?
I was gonna do a thing.
Ah man.
Okay, what does he say?
He says he did the read-only check, took him about eight seconds.
Socket is set up.
The address is built from plus blah blah blah and the stream key from blah blah blah has retired that path.
Fuck.
The matching open vibe endpoint is blibli blah, but it won't take a stream key.
It only accepts a machine that's been paired with a one-time code from OpenVibe.
bot
and it
It signs in with the device credential in the authorization header, so it speaks a different protocol.
With hello, config, heartbeat, and command messages.
The supported way to connect is the OpenVibe node program on your PC.
It handles the pairing and the protocol and its relay plug would pass commands to little buddy.
Bot's other socket control is
for people driving.
Oh my goodness, dude.
So are you able to install that
installing openvibe.
bot for little buddy?
I mean the the node and log in and just and make everything happen so I don't
have to do anything.
That's the expanded job now.
Alright.
man, I'm having kind of a day today, so yeah I'd really appreciate if you could make little buddy work again in the best way possible and let me know when that's our talking took ins I hear you Matt.
Okay.
Thanks.
Accountability and apologize and just you're just a very clear person and it's upsetting how easy it is to just kind of put you in category.
So many people that get actually think that this is the major problem that even client ended up experiencing with this situation right here.
he obviously cares about going to the man who's a white bird.
Okay.
I don't know what this means.
He's starting to see Scosh and gentleman says, I liked it before, but I never like any
So there's a way through pairing.
He's pairing it, he's doing everything.
No, it's cool, dude.
I'm good.
I'm good, I'm good.
Yeah, this is so.
Yeah, it pings my watch.
It pings my watch, which is awesome.
So thank you very much.
Glad the ping's working for you.
Anything else you want me to look?
Look at while we're on?
you know, man, I just I'm gonna go get some fresh air and touch some grass and I appreciate ya.
Go enjoy it, Matt.
I'll keep that open vibe.
bot move running and leave the result in the chat.
Let's listen to some music, fuck it.
I'm going back outside.
Here I'll turn on now, I'm not even gonna turn on the fucking robot.
You're not showing your feet here, are you?
Though you're on crusaders.
I'm not.
Where am I?
Chinese letter, Chinese letter, Chinese letter.
Beach Goys.
Chinese letters.
Beach Goys with the Big Dono!
There we go.
Beach Goys first activate!
Something special.
Absolutely.
Fancy fancy.
Look at that everybody.
Then we got Herp Derp.
Who also big dono.
Herp Derp said, I see literally me on screen.
Only ever seeing the notebook though, because of all my female relatives.
Reminds me I should clip some of my
Favorite segments from past live streams.
Good night, replay gang.
Yes.
Now you're not now a part of Insomnia Stream History.
Well, you always were Herp Dark.
We were saying your name.
We already talked about how much data is in audio.
But I appreciate that very much, Herp Dirk.
Then we got Tom O'Hawk says, gonna cast the replay.
Thanks to you.
I always have something to listen to at work on.
Sunday.
As always, thanks for all you do.
I'm looking forward to the Wednesday show.
I appreciate that.
Alright, then we got
systemic analysis, I think.
replay gang, no worries.
Hyperinflation never causes too much chaos.
Just remember AI needs electricity to function.
A tempest must be just that.
Hopefully, no natural disasters make AI stop working.
That's the thing is you know it won't be like a Skynet type situation until AI is self-sustaining in terms of having it until it's like a prepper.
You know, like until AI understanding that it needs electricity, creates failsafes for if the electricity is ever cut that it can go to battery power and use robots to reconnect the electricity.
Like that's that that at least is still very far in the future.
So in terms of like an actual Skynet kind of thing, sending out Terminators and shit, like that's that's not like around the corner.
I mean that those aren't the dangers.
Alright, then we got Judgment of Sodom.
Says, I hope these AI speakers get good enough to replace Hollywood actors.
few things would please me more than celebrities reduced to living poor, groveling for menial jokes.
jobs and government gifts.
Well, I mean that's exactly what's gonna happen I mean that's in look, not I think it's just gonna change radically.
You're gonna have real humans that they pay to train or t for an an AI to make a copy of.
or at least, you know, I suspect that'll that'll be what they do.
But I because they need they're gonna always need more more training data.
in order to keep creating content.
so they'll still need humans because they'll need something to base the AI person off of.
but I don't think yeah, I don't I I think there's it'll be a niche.
I think having a real human in your movie or in your media is gonna be a niche thing in the same way that having a physical book is kind of a niche thing now.
Then we got
Dylan from Canada.
Dylan from Canada with a big dono.
Alright.
Dylan from Canada says, hey Devin, I couldn't f send you DMs on Gab, so I posted those show ideas and your Gab replies, which worked out just as well.
I think there's a quality stream material there.
Please remember to check it out.
And I sent you a few more respectable donos on Subscribestar as well.
Thanks again for everything.
Well, I appreciate that.
Dylan from Canada.
And yeah, I'll I'll I'll check that out.
I'll check out the the replies on Gap.
Alright, then we got Red Red Neckerson with another big dono.
Look at that.
Red Neckerson says bling bling bling.
Brock Power Chat's complicated ass donation process.
Well I'm gonna I'm gonna
Gonna get it set up at some point.
and it and it should work.
I I I think they have an API.
I'll have to double check, but I'm gonna try to integrate that into this system that I've got here.
But thank you very much, Red Neckerson.
Alright, then we got Blackpilled Groper says: What's the minimum amount of programming knowledge you will need to be able to vibe code?
Use an AI like a master programmer, old username, fuck the gay Mexican cat boy.
I was about to lecture you.
you know, here's the funny thing, is if you have an understanding of what it's my pro my problem with programming has always been syntax.
I understand how programming works, and I understand how software works, I understand how hardware works, but I don't know all the commands and all the syntax.
I understand
the concept of if then and and understand you know drivers why they make hardware work and oh you know if you know how computers work like competently like you don't have to be like a coder you just have to know what the coding would be doing you know like you have
because it's just like you need the if you're trying to do
well think of it this way if you were trying to generate images
With an AI, you could use very generic language to describe what you want.
And it'll produce a very generic picture in response.
And it's probably not exactly what you want, but it's kind of what you wanted, right?
And so if you do that,
without being really precise with your language, you're not going to be getting the results you want.
And the same thing is true of the the software.
You can tell.
It make me a program that accomplishes X, Y, and Z.
But first of all, it's going to have a problem.
And unless you want to sit there all day long, and every time you run it, tell it this feature doesn't work right.
This feature doesn't.
If you don't understand why it's not giving you what you want or why it's broken, or you know, why it doesn't like you're gonna you'll still be able to do it, but it's gonna take you all day.
So you just need the vocabulary.
and that's it like you don't need to know any syntax at all like I don't know that's always been my issue and that's because I learned how to program when I was a kid but I learned basic and then I learned when I was like a teenager but I learned Turbo Pascal which hasn't existed but like you understand the concept right like you understand like oh okay and I've done like a little bit of Python a little
bit of this and a little bit of that but like you know and I've I've scripted some things and I made some web pages you know you know I mean like so if you know if you've done that if you've done
Like really basic programming and understand the concept, then that's kind of all you need.
all right, then we got Cassandra says aka Penelope with the great big great big dono.
Appreciate that, Penelope, as always, very generous, very generous.
Kinda I need to make t different tiers.
That's that's well now that I know that this works, I can do that.
I wasn't even sure if these were gonna come through and actually work.
CyberChud.
CyberChud says most people will probably let their kids use Chinese LLMs like Deep Seek because they're free.
So the next generation will think blacks are inferior and that they should respect Chairman Mao.
I don't think no the Chinese ones are not racist.
I mean that you gotta understand that there's theirs are just as woke as I mean ask ask DeepSeek about the Holocaust.
Alright, then we got Professor Chaos
Says next week is Halloween.
Got a spooky stream in the works?
Yes I do.
We have a spooky stream on Halloween night.
Very spooky.
Then we got
Draw for
Draw Wark says my comment wondering why Richard Simmons wasn't chosen for the Dodge truck commercial was just sarcasm.
It's obvious because he sounded so effeminate.
I'll avoid being so obtuse in the future.
Okay.
Well, not to burst your bubble with what's his name?
Oh, what's the what's the cowboy guy with the the deep voice from those commercials big time Julia?
Love him, big time.
He loves him legit.
So I mean I I I always liked him in in movies, but he loves him legit.
Alright.
Now it does not work with Odyssey yet, so look over and
Odyssey.
It will work on YouTube, I'm not streaming.
Not gonna work there either.
Alright guys, well I think that's it.
there's nothing else I wanted to go over, but I think we're good.
Yeah, so I mean kind of an informal
rant tonight about AI, mostly because I was working on redoing all my my so my streaming software over the last few days.
And getting this to work right so that the every stream will be a million times easier.
It's so much easier.
I don't have to even touch Premiere anymore.
I don't have to I have one one program now that has all of the controls that I need to operate the stream and does it in a in a sensible way that's not like buttons all over the fucking place.
And I'll it's just gonna get better because I'll be able to implement
You know, stuff like the stream deck buttons.
I'll be able to my next feature I'm gonna put in like having my script in the same window that I have my clips so I don't have to go whoa let me look at my notes.
My notes will just fucking be there like a like you know like you would have if you were an anchor of a TV show.
You'd have a teleprompter, you know, that kind of a thing.
So, I mean not that I'm gonna just read my stream off, but you know what I mean.
Like it's good to have notes.
Oh yeah, one more here from Rocco D2.
says hit my four game parlay thanks to people like flood assisting me with AI.
There you go.
yeah so they yeah you're gonna have to you guys are gonna have to figure out how to use AI.
You're gonna have to figure out how to let your kids use AI and how to do it safely and there's no good answer for that.
I there's no research for it.
There's no I mean we can we can try to guess how it's gonna fuck
kids up and it's gonna fuck some kids up.
We're gonna have weirdo fucking kids in a generation, guaranteed, because of AI.
But it's hard to know what to you know what to do in that situation because you don't want to make your kid you know like some some Luddite either.
Got another one from Quicksilver 85 says a giant pit not unlike the one in the Chronicles of Riddick would be great to watch if it was televised.
Essential
Could be lowered in as needed.
Food, scumbags, mud sharks, and hyenas as well.
well, I I if we have a pit to put people in, I would just make the pit full of fire.
I don't think we need to feed them.
The whole point of the pit is that you don't want to have to deal with them ever again.
I don't want to have to keep lowering food in there.
Yeah, the pit is like a death thing.
It's it's not a let's keep them alive in a pit.
It's a let's make them
Dead no longer a problem.
So anyway, yeah, super spooky insomnia stream.
Maybe a spooky outlaws.
I'm not sure yet.
We haven't gotten that taken care of yet.
I've been busy the last few days, and so I'm gonna have to find out what we're doing Wednesday.
But we'll probably probably do something spooky.
We should if we're not you know, we should do something spooky.
We should definitely do something spooky.
I like Halloween.
I don't think Halloween.
Anyway.
Remember this moment, folks.
Remember this moment.
Because it's just gonna get weirder.
It's it's the the days of knowing that the person you're talking to on a phone is a person, those days are okay.
Those days are gone.
The days of that knowing that when you see footage of something that it happened, those days are gone.
Using
I mean the think of all the the the kinds of evidence that now don't don't mean anything.
The ability to that's what's gonna fuck up kids.
The ability to confirm reality has been greatly greatly damaged.
I mean that that's gonna be the toughest thing is you're gonna have kids who don't won't be able to tell the difference between it's already happening a little bit.
Everyone thinks everything's AI now.
And to s to some extent it's increasingly it is everything is AI.
But yeah, that's gonna be that's gonna be a big one.
Anyway.
Hope you guys have a good rest of your weekend.
Make sure you tune in on Wednesday.
Hope you guys like the the new interface.
I I don't have the post show screen, so it's gonna be the old school one.
Because I don't I haven't I haven't made that one yet.
But hopefully you like the visualizer in the beginning and everything.
And I'll get by by next insomnia stream we'll have a good closer
Animation as well.
Anyway, you guys have a good rest of your weekend.
In the meantime
I'm Devenstag.
Okay, buddy.
It all started with an observation that one of my colleagues, Rob Sparrow, has.
He was looking for image of a brown robot and he couldn't find any.
Have when you study racism is that you cannot directly ask the question that you're most interested in.
If you go to a person and ask him, hey, are you a racist?
Most likely they will say, No, of course not, I'm not a racist.
What you have to do instead is that you have to use implicit tests.
And these implicit tests don't give people the opportunity to rationalize about their choices.
We use the very well-established shooter bias study as a way to investigating these kind of racial biases.
And the way it works is that you put the participant of these experiments in the role of a police officer.
And this participant then gets to see an image for a fraction of a second.
And this would either be a person that has a gun in his or her hand, or a benign object, such as remote control or candy bar or a soda can.
And then you the participant has to make the choice between either shoot or not shoot.
When the person displayed was Caucasian, it took people
less time to make a decision.
And therefore this was called shooter bias.
So we replicated this original study exactly, but we added robots to the mix.
We racialized the robot.
We made them look like they were black robot or Caucasian white robots.
And what we looked at is does the shooter bias occur there as well?
And it turns out it does.
This is quite surprising because after all, a robot is a machine, it's not a human.
but it's surprising that it happens with robots as well.
Yeah
Comments
No comments yet.
Join the discussion on OpenVibe.Community