Elon Musk on HUMAN WILL vs AGI and Humanoid Robots...
Summary
TLDRIn this extensive discussion, Lex Fridman explores the future of brain-computer interfaces with Elon Musk and Neuralink's first human cyborg, showcasing impressive control over computers through thought. The conversation delves into AI's rapid advancements, the potential of humanoid robots like Tesla's Optimus, and the significance of real-time data in AI development. Musk emphasizes the importance of aligning human will with AI and the possibility of higher data rates improving this connection, hinting at a future where brain-computer interfaces might redefine human-AI interaction.
Takeaways
- ๐ Lex Fridman's experience in the Amazon rainforest led to a broadening of perspectives and an in-depth podcast with Elon Musk discussing Neuralink and its implications.
- ๐ง The first human cyborg, a quadriplegic patient, has successfully used Neuralink to interact with computers, play games, and potentially stream content, demonstrating significant advancements in brain-computer interface technology.
- ๐ฎ The patient achieved a score of 9.51 bits per second in a clicking game, showcasing the potential for brain-controlled interactions to rival traditional methods of input like mice and keyboards.
- ๐ฎ As AI models and neural interfaces improve, it's predicted that brain-computer interfaces will become more efficient and could replace traditional input methods in the near future.
- ๐ค The convergence of technologies like Neuralink and humanoid robots, such as Tesla's Optimus, opens up possibilities for rapid data transfer and new ways of interacting with machines.
- ๐คนโโ๏ธ Dr. Jim Fan from Nvidia highlights the importance of developments in robotic hands and the significant engineering effort required to replicate human dexterity in machines.
- ๐ The script discusses the potential of brain-computer interfaces and VR for controlling and training robots, indicating a growing overlap between AI, robotics, and human-machine interaction.
- ๐ก Elon Musk and other industry leaders are investing heavily in AI and robotics, suggesting a belief in the long-term viability and transformative potential of these technologies.
- ๐ Despite mainstream media suggesting that AI is overhyped and a bubble, insiders continue to invest and innovate, indicating a belief in the underlying value and future of AI advancements.
- ๐ AI has shown remarkable achievements, such as Google's AI system achieving a silver medal at the International Mathematical Olympiad and Nvidia's Eureka system outperforming human-written code for training robots.
- ๐งฉ The future of AI safety and human-AI alignment may involve technologies like Neuralink, which could increase the data rate between humans and AI, facilitating better understanding and cooperation.
Q & A
What was Lex Fridman's experience in the Amazon rainforest, and how did it relate to his podcast with Elon Musk?
-Lex Fridman had an experience in the Amazon rainforest that broadened his horizons, which he later discussed in a nearly 9-hour podcast with Elon Musk, covering various topics including Neuralink and the future of brain-computer interfaces.
Who is considered the first human cyborg with Neuralink, and what abilities has this technology granted them?
-The first human cyborg with Neuralink, known as 'patient zero,' is a quadriplegic individual who, with the help of Neuralink, can interact with computers, play games like Civilization and Mario Kart, and potentially even stream content.
What is the significance of the 'web grid' game in the context of Neuralink's technology?
-The 'web grid' game is a tool used by Neuralink to demonstrate the capabilities of their technology. It measures how quickly and accurately a user can click on targets, showcasing the performance of the first clinical trial participant who achieved a score of 9.51 bits per second.
How does the brain-computer interface technology differ from traditional input methods like keyboards, mice, and touchpads?
-The brain-computer interface technology allows for direct interaction with computers through brain signals, bypassing the need for physical input devices. This method is potentially much faster and more efficient, as it relies on neural signals rather than manual dexterity.
What are some potential implications of combining brain-computer interfaces with humanoid robots like Tesla's Optimus?
-Combining brain-computer interfaces with humanoid robots could lead to highly efficient and responsive machines capable of performing complex tasks. This synergy could revolutionize labor and various industries by enabling rapid data transfer and precise control through thought processes.
What is the current state of AI and robotics development according to Dr. Jim Fan, and what are some of the most exciting projects?
-Dr. Jim Fan, Nvidia's senior AI researcher, highlights projects like Nvidia's Project Groot as some of the most exciting developments in robotics. These projects are pushing the boundaries of what is possible with AI and robotics, including advancements in machine learning and autonomous operation.
What role do human hands play in the development of humanoid robots, and why are they significant?
-Human hands play a crucial role in the development of humanoid robots as they are responsible for a significant portion of human intelligence and manipulation of the world. The engineering of robot hands, such as those in the Tesla Optimus, is a major focus due to the complexity and dexterity required for real-world tasks.
How does the real-time data from Tesla cars and potential Tesla Optimus robots contribute to AI training and safety?
-Real-time data from Tesla cars and, in the future, Tesla Optimus robots, provides a vast and diverse dataset that can be used to train AI systems more effectively. This data helps in understanding cause and effect in the real world, which is essential for improving AI safety and decision-making.
What is the potential scale of production for humanoid robots like Tesla's Optimus, and what factors contribute to this scale?
-Elon Musk estimates that the potential scale of production for humanoid robots like Tesla's Optimus could be over a billion units per year due to their high utility. Factors contributing to this scale include advancements in AI, robotics, and the ability to generate and learn from vast amounts of real-world data.
How does the concept of 'play to win' apply to the development of AI, and what are the key components for success?
-The concept of 'play to win' in AI development means having the most powerful training compute, efficient use of data, and a strong human talent pool. It also involves having unique access to data and packaging the AI product effectively to ensure continuous improvement and alignment with collective human will.
What is the role of human will in the context of AI and human-AI alignment, and how might technologies like Neuralink help?
-Human will is considered a source of purpose and direction for AI. Technologies like Neuralink could help align collective human will with AI by increasing the data rate of human communication, thus improving the link between humans and AI and potentially enhancing AI safety.
Outlines
๐ง Neuralink's Breakthrough with the First Human Cyborg
In this segment, the script discusses Lex Fridman's experience in the Amazon rainforest and his subsequent podcast with Elon Musk, focusing on Neuralink's progress. The first human patient to receive a Neuralink device, a quadriplegic individual, can now interact with computers, playing games and potentially streaming, showcasing the device's capabilities. The script also compares the patient's performance in a clicking game to the average person's, highlighting the impressive control achieved through brain-computer interaction. The potential for future improvements in brain-computer interfaces is emphasized, suggesting they could surpass traditional input methods like keyboards and mice.
๐ค The Convergence of Robotics and AI
This paragraph delves into the overlap between robotics and AI, mentioning projects like Nvidia's Project Groot and the development of humanoid robots with advanced capabilities. The script discusses the importance of hands in robot design, the effort put into training robots, and the potential impact of intelligent humanoid robots on labor and economies. It also touches on the historical anecdote related to the song 'Money for Nothing' by Dire Straits, illustrating a disconnect between manual laborers and the leisure class, and ends with a commentary on the current hype and skepticism surrounding AI's potential impact.
๐ Inside Perspectives on AI Development and Investment
The script addresses the skepticism from mainstream news about AI being a hype or a bubble, contrasting it with the actions of industry insiders who are heavily investing in AI technologies. It highlights Google's order of a large number of GB 200 chips, indicating a strong commitment to AI development. The segment also discusses AI's achievements, such as Google DeepMind's AI system achieving a silver medal at the International Mathematics Olympiad and Nvidia's AI outperforming expert humans in creating reward programs for trading robots.
๐ง AI's Role in Material Discovery and the Future of Work
This part of the script focuses on AI's current capabilities in material synthesis, as demonstrated by Google's AI discovering new materials, and the potential of AI in workplaces. It includes a critique from a New York Times author who questions the significance of AI in improving mundane tasks like email writing. The script then presents Elon Musk's perspective on AI's potential role in aligning with human will and purpose, suggesting that AI could serve to fulfill human desires, even in seemingly trivial pursuits like using Tinder.
๐ง The Engineering Challenge of Creating Humanoid Robots
The script discusses the engineering challenges in developing humanoid robots, particularly the complexity involved in creating functional and dexterous hands. It compares the global vehicle production capacity to the potential scale of humanoid robot production, suggesting that the utility of humanoid robots could drive annual production into the billions. The conversation with Elon Musk explores the importance of the hand's role in human intelligence and manipulation of the world, and how reengineering this aspect into robots like Tesla's Optimus is crucial for their success.
๐ The Future of Communication and AI Safety
In this final segment, the script contemplates the future of human-computer communication, with a focus on Neuralink's potential to increase the data rate between humans and AI, which could improve alignment and safety. It discusses the limitations of current human output rates and the need for higher data rates to maintain a meaningful connection with advanced AI systems. The potential of Neuralink and other technologies to enhance human capabilities and ensure AI safety is highlighted, ending with a reflection on the importance of human will in guiding AI development.
Mindmap
Keywords
๐กAmazon Rainforest
๐กNeuralink
๐กQuadriplegic
๐กBits per second (BPS)
๐กBrain-Computer Interface (BCI)
๐กHumanoid Robots
๐กTesla Optimus
๐กAI Safety
๐กData Rate
๐กOptimus Pro
๐กFormula 1
Highlights
Lex Fridman's experience in the Amazon rainforest and his subsequent podcast with Elon Musk discussing Neuralink.
Introduction of 'patient zero', the first human cyborg with Neuralink, demonstrating the potential of brain-computer interfaces.
The first clinical trial participant of Neuralink achieving a 9.51 BPS score in a clicking game, showcasing the capabilities of brain-computer interaction.
Comparison of human performance in clicking games with the capabilities of the Neuralink patient, highlighting the potential of brain over hand control.
Elon Musk's vision for the future of brain-computer interfaces and their efficiency compared to traditional input methods.
The potential impact of combining Neuralink with humanoid robots like Tesla Optimus for rapid data transfer and enhanced interaction.
Dr. Jim Fan's perspective on the top three robotic developments, including Nvidia's Project Groot and its implications for AI.
The importance of hands in robot development and the engineering effort invested in creating dexterous robotic hands.
Nvidia's open-sourcing of projects like roboๅก่จ and mimic gen, indicating a trend towards collaborative AI and robotics development.
The role of humans in training robots through tasks that require dexterity and the potential for non-experts to contribute to AI development.
The economic implications of humanoid robots on labor and the potential shift in how tasks are performed and valued.
Insider investment in AI technology despite public skepticism, suggesting a belief in the long-term value of AI advancements.
Examples of AI achievements, such as Google DeepMind's performance at the IM mo math competition, countering claims of AI hype.
The role of AI in material synthesis, with Google's AI discovering new materials at a rate far exceeding human capabilities.
Elon Musk's thoughts on the future of AI and its potential alignment with human will through increased data rates and communication.
The concept of 'play to win' in AI development, emphasizing the need for the most powerful training compute and rapid improvement.
Musk's insights on the importance of real-world data for AI training, and the potential of Tesla's Optimus robots as data sources.
Challenges and considerations in scaling humanoid robot production, drawing parallels with the automotive industry's production capabilities.
The intricate engineering of robotic hands, their significance in the overall development of humanoid robots, and their comparison to human anatomy.
Neuralink's potential contribution to AI safety by enhancing the data rate between humans and AI, facilitating better alignment and understanding.
Transcripts
so Lex fredman recently went to the
Amazon rainforest had quite an
experience as far as we can tell and
after community of Nature and up
broadening his Horizons comes back and
does a podcast with Elon Musk that is
close to 9 hours long now a portion of
that is him talking to the various
people behind neuralink including
patient zero the first ever human cyborg
this is the person that this is the
first human patient of neuralink the
person that had brain surgery to get
neuralink installed he's quadriplegic
but but with the help of New link is
able to Interac with computers play
games like Civilization Mario Kart Etc
and potentially even do some streaming
in one of my earlier videos we took a
look at web grid so this is kind of
neural links little game that this first
clinical trial participant plays and
anybody can play it you can click Start
game and play it yourself and basically
he tracks how quickly you can click on
things so how accurately how quickly he
achieved a score of 9.51 BPS bits per
second for reference here's my score
13.41% second but I'm doing this on a
huge curved monitor with my Naga Razer
Mouse and I've played a first-person
shooter or two in the past now if you're
a 15-year-old kid playing these games
you're probably going to have a much
higher score but just for reference this
probably somewhere near the top 10 20%
of adult clickers let's say I'll leave a
link if you want to try yourself post
your score down below but his was 9.51
meaning that I'm maybe 40% faster SL
more accurate than someone that doesn't
have the use of his hands someone that
had to learn a brand new way of
interacting acting with the computer by
using his brain Powers basically
telepathy to aim and click that's
important to understand he's not using
his eyes to look at objects that's not
how it's controlled it's a whole new
different thing he describes it kind of
like the force from Star Wars you think
about it and then the cursor moves now
that's a sort of a small neural network
AI model that kind of is trained there's
different types of it and it picks up on
your brain waves to move the mouse
around so it's a brand new way of
interacting with a computer so this
person as an adult just within the last
year plus that he that he's had this
learned how to do this and he is beating
probably a good chunk of the human
population using their hands and a mouse
he's doing incredibly well at that all
things considered as the models improve
in how they track our brain waves to
click on the right stuff as the
researchers keep working with him to
improve the model as he keeps getting
better as the hardware keeps better
there might be a hardware limitation for
this I think it's very conceivable that
within the short foreseeable future we
will see brain computer interfaces that
are much better and much more efficient
at interacting computers than anything
we've had prior keyboards mice touchpads
voice interfaces all of that will seem
slow outdated very quickly we'll hear
from Elon Musk as to what he thinks
about where this is going it's
fascinating to hear but also we have to
take a look at what this kind of
technology means for humanoid robots for
example the Tesla Optimus what happens
when you kind of combine those two
technologies you have a humanoid robot
and you also have a brain chip capable
of very very rapidly transferring data
between your brain the big ball of
protein and fat floating in your cranium
between that and any computer like this
robot for example what does that mean
it's kind of mind-blowing to think about
it and Elon Musk will explain where this
is going now he's not the only one
that's swimming in this direction here's
Dr Jim fan in nidia's senior AI
researcher to me I would say the top
three most exciting robotic developments
right now are one Nvidia Project Groot
and everything that basically Nvidia is
cooking up very curious to see how
figure O2 how well that will perform
we're going to be seeing the release of
figure 2 I believe they said August 6th
notice the differences between figure
one and left figure two the new and
improved version with the fingers and
the Palms the hands if you think about
what proportion of resources and
engineering power and just effort is put
into these robots what percentage would
you say goes into the hands versus the
rest of the body versus the the battery
the power the vision the the brain quote
unquote the neural network Etc what
portion of developing something like
that do you think is just getting the
hand right is it 5% 10 20 the answer
might surprise you but notice with Dr
Jim fan project Groot we're beginning to
see a continuation of this pattern so
they're using Apple Vision Pro to do
what to basically control the robot to
generate various data this is sometimes
referred to as a Telly operation Dr Jim
fan here is saying till operation is
slow and timec consuming but we can't
afford to collect a small amount of data
so this is their way to collect data for
training the robots for improving their
abilities and in this video we'll take a
look at what Elon Musk thinks about this
approach and how it may be the same or
different to how they're training how
they're approaching this how they're
approaching this situation with t TLA
Optimus Nvidia is open sourcing some of
these very projects like rooc Casa and
mimic gen which is for robotic arms but
they'll also have another version for
humanoid and FiveFinger hands so this
idea of some sort of an overlap between
machine and humans is becoming closer
right brain computer interfaces
something like the Apple Vision Pro some
sort of a VR Vision that allows you to
see through the eyes of the robot either
to control it but also to help train how
to do certain things to gather data for
its training this will be becoming more
and more of a thing earlier this year
Brad atock and the figure robots so
they're in the Bay Area over there San
Francisco and the surrounding areas they
were hiring people to help them train
robots and I don't mean PhD AI
researchers I mean just people that were
pretty good with their hands had some
you know dexterity and stuff like that
if any of you listening maybe want to
get into the field of AI SL Robotics and
by the way those two fields are really
beginning to see a huge overlap getting
into into doing something like this
might be a great way to get your foot in
the door whether that's at figure Nvidia
any of the other robot startups that are
going to be popping up it's a relatively
new field like who can claim 20 years of
experience of controlling robots through
some sort of a virtual reality headset
probably not too many and at the same
time as Brad atock is saying here only
recently has time opened a window of
opportunity to scale billions of
intelligent humanoid robots this was not
feasible 5 years ago we fortunate to be
in 2024 the first year in history when
this is possible life is about to turn
into a sci-fi film robots fundamentally
change how we view labor a lot of any
country's economic power you need
laborers you need people working and
doing stuff right we need people
installing microwave ovens custom
kitchen deliveries we people moving
refrigerators moving these color TVs
right we can't have just everyone play
the guitar on the MTV I didn't know this
but this song Money for Nothing by Dire
apparently the lead singer was in some
electronic store in New York that had
TVs and refrigerators and MTV was
playing a music video of somebody
playing the guitar and one of the Dock
Workers one of the people working in the
store the delivery person this kind of
gruff New Yorker he basically was
complaining about the people being able
to live their life like that not knowing
he was right next to one of the people
that did that that that was their life
they were a rock star and I didn't know
this but these lyrics are literally his
Ty raade the store worker basically
complaining and yelling and I've deleted
some of the more offensive Parts here
but this is just him going off about the
rock stars and how they don't do any
work while he has to install and move
refrigerators I apologize for my ADHD
moment but I just love the fact that he
heard somebody just talking crap about
him just jotted down there at the store
exactly what he was saying and then just
saying it making it into a song that
probably made him some Untold Millions I
don't know something very poetic about
that in other news there's a lot of uh
kind of mainstream news I guess you can
say talking about how AI is just hype
and it's a bubble it's about to burst
it's over it's done it's BS as they say
I think this was a great way of putting
it people on the outside are saying AI
is all hype what are the people on the
inside doing very often times in
investing for example people keep close
track on what insiders are doing if
you're close to a company you have some
potentially Insider information and of
course there's laws about how you're
able to trade on that or whatever but a
lot of people pay attention to what
insiders are doing with the idea that
they probably know more than the people
on the outside the guy that's running
the hedge fund and meets with the CEOs
and the CFOs of a particular company
probably knows a lot more about where
the company's heading than some reporter
that's covering that and 30 other topics
and what the Insiders are doing is
they're doubling down the new Nvidia
thing that was announced the Blackwell
we've covered in the previous video but
notice this line right here Google for
example has ordered more than 400 ,000
GB 200 chips said the two people who
work on the chip that could be worth
well north of 10 billion and they don't
even know when they're going to get them
Google by the way has their own chips
the TPU the tensor processing unit and
is on Pace to spend about 50 billion
this year on chips and other equipment
property up more than 50% from last year
meta also is doing that Microsoft is
also doing that and as you see in a
second Ela musk is also doing that and
other things that are really showing
that they have no signs of stopping you
know here's New York Times saying press
pause on the Silicon Valley hype machine
right so not press pause on AI but press
pause on the hype machine and a little
picture of somebody walking head first
into a wall I'm guessing this is I'm not
sure what this is supposed to be but a
graph plateauing and a cup that says
Chad
gbt and the author is claiming that
we're not sure if AI is too stupid or
unreliable to ever be useful she gives
an example where Sam Alman promised to
unveil stuff that feels like magic but
it was just a routine update that makes
Chad gbt cheaper and faster this is not
true I'm not sure if she's trying to
deceive people or is uninformed but this
is not correct the thing that Sam Alman
was referring to was of course the
advanced voice mode which is as if
you've seen some of the clips is fairly
mindblowing I was laughing very hard at
some of its Antics does a baby computer
call its father what does a baby
computer say to its
father what
data see the old chat gbt what it would
do is it would translate my voice into
text and then feed that text into gbt 4
and there's a lot that gets Lost in
Translation when you do that because the
translation software is like kind of
dumb and so gbd4 because it could only
read the text was only getting like kind
of a compressed faximile of all the all
the nuances of what I what I had just
said that is a very important point that
I think a lot of people missed before
voice mode was just Underneath It All it
was it was just text it created the
answer in text and then another model
would read that text when you said
something it would convert your voice to
text and it would read that text to
itself to understand it and as I'm sure
you're aware a big part of our
communication isn't just the words right
the things that are captured in text
it's also tone of voice the inflection
the emotion we put into the voice a
million other things that could change
the meaning of what we're saying that
doesn't get captured when you just
transcribe the thing you just said and
of course when the AI model just reads
the text back to you it also misses tons
of potential little nuances of
communication they're just lost when
that happens the new omn model is
trained on all of it together it's
integrated Omni is the gbt 4 o the O is
Omni and Omni means all it's all of it
in one the routine update that makes
Chad GPT cheaper and faster if you click
on this this is The Verge so this is
open AI releases GPT 40 a faster model
it's free for all Chad GPT users this
completely misses the point of all of
this by the way since then we've also
had GPT 40 mini which is a huge leap
forward because we're getting the
capabilities of what a year ago two
years ago were these top-of-the-line
Frontier models at a fraction of the
cost they're much more efficient they're
cheap cheaper they're they're less
energy intensive but the point is while
the people reporting on AI from the
outside are saying that it's all fake
it's all hype but can't do anything at
the same time all the Insiders are
putting tons of money into acquiring the
necessary resources to continue
developing it so who's right who's wrong
do you think all these companies like
Microsoft Google meta open a Tesla XI
Etc do you think they're all going to go
bankrupt within the next 5 to 10 years
because they bet on some hype cycle of
AI that will never amount to anything if
I was a betting man I'd be betting on
musk and Corp and Jensen and Zuckerberg
Demi sabis and SAA nadela they don't
strike me as foolish people or do you
think that AI is most likely going to
end up like the Roomba the mediocre
vacuum robot that does a passable job
when you're home alone but not if you're
expecting guests by the way Google deep
mine recently used their AI models to
achieve a silver Metal Standard at the
IM mo one most prestigious worldwide
mathematics competition and by the way
saying it's got the silver medal I think
is kind of understating it because had
it gotten just one more point that would
have been gold it was one point away
from getting the top award at the top
math competition in the world and of
course Nvidia post their research where
Eureka it's their kind of system that's
built around GPT 4 when creating reward
programs for trading robots in
simulation in this case getting them to
do these pen spinning tricks well it
outperforms expert human written ones on
more than 80% of the tasks so these are
expert humans who write this code for a
living GPT 4 with a few modifications
beats 80% of them so AI is top 10 top
15% of worldclass human solving math
problems it's better than 80% of expert
level humans writing code for training
robots and as of last year Google demmin
was autonomously able to discover and
reproduce a massive amounts of new
neverbe seen materials this blue circle
is human experimentation this is how
much we were able to create using our
abilities and here's gnome or I guess
you would say gome right 20 times the
amount roughly the AI Model come comes
up with potential recipes for these
materials and then this little robot
that's sitting in a blast proof cage is
uh you know trying to recreate them as
as block post puts it the autonomous lab
created new recipes for Crystal
structures and successfully synthesized
more than 41 new materials opening up
new possibilities for Aid driven
material synthesis the New York Times
author here continues saying but in
workplaces where quality matters AI may
not make significant inroads one of her
big concerns is that a generation of the
brightest Math and Science Minds will be
spent on incremental improvements in
mediocre email writing what do you think
is she right or do you think that this
generation's brightest Math and Science
Minds will in fact be AI let me know in
the comments here's Elon Musk my name is
Wes rth and thank you for watching I
mean if the AI can communicate at
terabits per second and you're
communicating at you know bits per
second yeah it's like talk to a tree
well it is a very interesting question
for a super intelligent species what use
are humans I think there is some
argument for humans as a source of will
will will yeah source of the will or
purpose so if if you consider the human
mind as being essentially the there's
the Primitive limic Elements which
basically even like reptiles have and
there's the cortex the thinking and
planning part of the brain now the
cortex is much smarter than the limpic
system and yet is largely in service to
the lyic system it's trying to make the
lyic system happy I mean the sheer
amount of compute that's gone into
people trying to get laid is insane um
without the without actually seeking
procreation they're just literally
trying to do this sort of simple motion
and they get a kick out of it yeah so
this uh simple which in the abstract
rather absurd motion which is sex uh the
cortex is putting a massive amount of
compute into trying to figure out how to
do that so like 90% of distributed
compute of the human species is spent on
trying to get laid probably like large
perent yeah yeah there's no purpose to
most sex except uh hedonistic you know
it's just sort of Joy or whatever DOP
mean release um now what once in a while
it's procreation but for humans it's
mostly modern humans is mostly uh
recreational um and uh and so so so your
cortex much smarter than your lmic
system is trying to make lmic system
happy because lipic system wants to have
sex so um or want some tasty food or
whatever the case may be and then that
that is then further augmented by the
tertiary system which is your phone your
laptop iPad whatever you know all your
Computing stuff that's your tery layer
so you're actually already a cybor you
have this tertiary compute layer which
is in the form of your your computer
with all the applications all your
computer devices and uh and so in the
getting laid front there's actually a
massive amount of comp of digital
compute also trying to get laid you know
with like Tinder and whatever you know
yeah so the compute that we humans have
built is also participating yeah I mean
there's like gws of compute going into
getting late of digital compute yeah
what if AI was this is happening as we
speak if we merge with AI it's just
going to expand the compute that we
humans use pretty much to try to it's
one of the things certainly yeah yeah
but what I'm saying is that that yes
like what's is there a use for humans um
well there's this fundamental question
of what's the meaning of life why do
anything at all um and so if if if our
simple Olympic system provides a source
of will to do something um that then
goes to our cortex that then goes to our
you know tertiary compute layer then I
don't know it might actually be that the
AI in a benign scenario is simply trying
to make the human living system happy
plot what your neurons are doing is
distilling the concepts down to a small
number of symbols of of say syllables
that I'm speaking or key strokes
whatever the case may be so uh that
that's a lot of what your brain
computation is doing now there is an
argument that that's actually a healthy
thing to do or a helpful thing to do
because as you try to compress complex
Concepts you're perhaps forced to
distill the you know what is what is
most essential in those Concepts as
opposed to just all the fluff so in the
process of compression you distill
things down to what matters the most
because you can only say a few things so
that is perhaps helpful and you've also
said play to win or don't play at all so
yeah what does it take to win for AI
that means you've got to have the most
powerful training compute and your the
the rate of improvement of training
compute has to be faster than everyone
else or you will not win you your AI
will be worse so how can grock let's say
three that might be available what like
next year well hopefully end of this
year grock 3 for Lucky yeah how can that
be the best llm the best AI system
available in the world how much of it is
a compute how much of it is Data how
much of it is like post training how
much of it is the product that you
package it up in all that kind of stuff
I mean they will matter it's sort of
like saying what what you know let's say
it's a Formula 1 race like what matters
more the car or the driver I mean they
both matter um if if if you're if a car
is not fast then you know if it's like
let's say it's half the horsepower of
your competitors the best driver will
still lose on if it's twice the
horsepower then probably even a mediocre
driver will still win so the training
computer is kind of like the engine how
many this power of the engine so you
really you want to try to do the best on
that and you then um then how
efficiently do you use that training
compute and how efficiently do you do
the inference the use of the AI so
obviously that comes down to human
Talent um and then what unique access to
data do you have uh that's also plays a
plays a role you think Twitter data will
be useful yeah I mean I think I think
most of the leading AI companies already
have already scraped uh all the Twitter
data I know I think they have um so I
don't on a go forward basis what's
useful is is is the fact that it's up to
the second you know that's the because
it's hard for them to scrap in real time
so there's there's a an immediacy
advantage that Gro has already I think
with Tesla and the realtime video coming
from several million cars ultimately
tens of millions of cars with Optimus
there might be hundreds of millions of
Optimus robots maybe billions learning a
trous amount from The Real World uh
that's that's the the biggest source of
data I think ultimately is is sort of
Optimus Pro Optimus is going to be the
biggest source of data because it's
because reality scales reality scales to
the scale of reality um it's actually
humbling to see how little data humans
have actually been able to accumulate um
really say how many trillions of usable
tokens have humans generated where on a
non- duplicative like discounting spam
and repetitive stuff it's not a huge
number you run out pretty quickly and
Optimus can go so Tesla cars can are un
have to stay on the road uh Optimus
robot can go anywhere there's more
reality off the road and go off road I
mean like the Optimus robot can like
pick up the cup and see did it pick up
the cup in the right way did it you know
say pour water in the cup you know did
the water go in the cup or not go in the
cup to spill water or not yeah um simple
stuff like that I mean but it can do at
that at scale times a billion you know
so generate use useful data from reality
So Co cause and effect stuff what do you
think it takes to get to mass production
of humanoid robots like that it's the
same as cars really I mean Global
capacity for vehicles um is about 100
million a year and uh it it could be
higher just that the demand is on the
order of 100 million a year and then
there's roughly 2 billion vehicles that
are in use in some way so which makes
sense like the the life of a vehicle is
about 20 years so at steady state you
can have 100 million Vehicles produced a
year with a 2 billion vehicle Fleet
roughly um now for humanoid robots the
utility is much greater so my guess is
humanoid robots are more like at a
billion plus per year but you know until
you came along and started uh building
Optimus it it was thought to be an
extremely difficult problem I mean still
extremely difficult it's no walk in the
park I mean Optimus currently would
struggle to have to walk in the park I
mean it can walk in a park park is not
too difficult but it will be able to
walk um over a wide range of terrain
yeah and pick up objects yeah yeah I can
already do that but like all kinds of
objects yeah yeah foreign objects I mean
pouring water in a cup is not trival cuz
if you don't know anything about the
container could be all kinds of
containers yeah there's going to be an
immense amount of engineering just going
into the hand yeah the hand might be it
might be close to half of all the
engineering in the in Optimus from an
electromechanical standpoint the hand is
probably roughly half of the engineering
but so much of the intelligence so the
intelligence of humans goes into what we
do with our hands yeah the manipulation
of the world manipulation of objects in
the world and intelligent safe
manipulation of objects in the world
yeah yeah I mean you start really
thinking about your hand and how it
works you know I do all the time the
sensory in control of munculus is we
have humongous hands yeah so I mean like
your hands the actuators the muscles of
your hand are almost overwhelmingly in
your forearm so your forearm has the has
the muscles that that actually control
your hand um there there's a there's a
few small muscles in the hand itself but
your hand is really um like a skeleton
meat puppet and and and with cables that
so the muscles that control your fingers
or in your forearm and they go through
your the carpal tunnel which is like
you've got a little collection of Bones
and in a tiny tunnel that the that these
cables the tendons go through and those
tendons are what um mostly what move
your hands and something like those
tendons has to be reengineered into the
Optimus in order to do all that kind of
stuff yeah so like the current Optimus
um we tried putting the actuators in the
hand itself then you you sort of end up
having these like giant hands yeah giant
hands that look weird yeah and then they
they don't actually have enough degrees
of freedom and or enough strength so so
you realize okay that's why you got to
put the actuators in the Forum and and
just like a human you got to run cables
uh through a narrow tunnel to operate
the the fingers do you think neurolink
is in your view the the best current
approach we have for AI safety it's an
idea that may help with AI safety um
certainly not I wouldn't want I would
wouldn't want to claim it's like some
Panacea or that's a sure thing but I
mean many years ago I was thinking like
well what what would inhibit alignment
of human Collective human will with uh
artificial intelligence and the low data
rate of humans especially our our slow
output rate um would necessarily just
because it's such a because the
communication is so slow would diminish
the link between humans and computers
like the more you are a tree the less
you know what the tree is like let's say
you you look at a tree you look at this
plant whatever and like hey I'd really
like to make that plant happy but it's
it's not saying a lot you know so the
more we increase the data rate that
humans uh can intake and output then
that means the better the higher the
chance we have in a world full of agis
yeah we could better align Collective
human will with the AI if the output
rate especially was dramatically
increased
Browse More Related Video
Elon Musk: Neuralink and the Future of Humanity | Lex Fridman Podcast | Podcast Summary
La veritร sui CHIP NEL CERVELLO
Just Happened! Elon Musk Confirmed Tesla Optimus Gen 2 Produced At $10K Price To Work At GigaTexas
Are brain implants the future of computing?
Elon Musk - Where AI's Biggest Impact Will Be
How Smart Will AI Be in 2034?
5.0 / 5 (0 votes)