The content proposes a paradigm shift in learning and career advancement, advocating for a "top-down" problem-solving approach augmented by AI, which bypasses traditional foundational learning and emphasizes practical demonstration of skills.
Mind Map
Click to expand
Click to explore the full interactive mind map • Zoom, pan, and navigate
Hello, I'm Tony Co, the host of Talkit
Global. In today's session, we're going
to explore how we can accelerate our own
growth using AI.
When you hear the story of Gabriel
Peterson, who dropped out of high school
in Sweden, self-studied AI, math, and
coding solely through generative AI
chat, and eventually became an AI
researcher at OpenAI in his mid20s. You
can sense that there is such a thing as
AI native learning and it's not just
about being good at prompting. We've
always been taught to build everything
step by step from the basics. But
Gabriel chose the exact opposite path.
Start with the problem, dig downward,
and pull in the knowledge you need. A
top- down learning approach.
Fixing bugs, asking questions, diving
deeper, the fundamental naturally
follow. In the AI era, perhaps the
people who touch the problem first are
The way I think people learn the fastest
is by what you would call like a like uh
um top down approach, right? You'll
probably learn faster if you start with
a problem and then you can read about
everything required to to to start
solving the problem and then you find
more problems and you read about those
and then you go down to like the the
core of the problem, right? is you start
with actual task and you go down. But
that's extremely rare way to learn like
in school everyone has this mindset
right of like okay we need to start with
the foundations we need to start like if
you want to work with machine learning
like you can forget about doing any
machine learning for the first like four
years right it's like math and then and
then you have like matrix
classifications you have linear
algorithm you have all these things that
build up and then you have the simpler
ML that's like super outdated you have
like you know linear regression all
these things that are still used partly
but it's like it will take you very long
time until you get to like production
grade ML
Why is this? Well, it's extremely hard
to scale the top down approach because
that requires like a teacher always
being there for you. It requires you
being able to know exactly what piece of
thing you need to learn at any point of
time. Well, if you do bottom up, you
know, okay, first you always learn this
and then you always learn this
>> and it's it's much easier to scale. It's
extremely inefficient. And now with
Shachvt, all this changes like this will
change. People say education will change
all the time, but I can barely take
universities seriously that don't teach
HGBT as a part of their curriculum. It's
like actually insane that this is not
like a a course that's taught from like
2 years old. Like suddenly foundational
knowledge universities don't have like
um a monopoly on on foundational
knowledge anymore. You can just get any
foundational knowledge from from from
shipb and people haven't really
internalized how top down problem
solving works. They will always tell you
things you know like oh but you'll never
actually understand the problem. you'll
never actually blah blah blah. And this
is not true. You start with a problem,
you recursively go down. Like if I want
to learn machine learning, I ask
Chashiv, okay, what project should I do?
Write a project for me. I have bugs. I
start fixing the bugs and then things
work. And from there, I start with a
specific part of the machine learning
problem. Like, okay, uh what happens
here? Can you explain to me with
intuition why this module here makes the
model learn? And it will explain to you.
And then oh, it uses matrix
multiplication and linear algebra, you
know. Okay. How do they work? What's the
math intuition behind this? like show me
like make up a couple graphs to really
make me get an intuition for this part
of ML and then suddenly you have all the
foundational knowledge like it doesn't
need to go bottom up anymore and this
shift will will yeah I think this shift
will like fundamentally change how
education is done and you just continue
to ask the model constantly until you
really understand and when you
understand you can just tell the model
okay this is my understanding of this is
this completely correct and then you
you'll also start learning about all
these like small tricks you can do right
like explain this concept like I'm 12
years That one is really good. It will
start like super easily like imagine
you're in a bookstore and you can
imagine the embeddings being the
different books in the store and then
you can imagine you know all this and it
will connect everything that has to do
with AI to like real world concepts
which makes it really easy to to reason
about for for someone like like me.
Gabriel emphasizes finding the gaps in
your knowledge and digging in until you
reach that aha moment. It means asking
AI to explain things more precisely and
more directly and reconstructing your
learning by exposing every intermediate
step in the code. How often you can
create these click moments
that becomes a new metric of learners in
The skill that's required here is
knowing what gaps you have in your
knowledge. like see you have an AI model
or like whatever else you want to learn
and understanding when you don't really
understand the part it's actually pretty
hard to do like it's something you need
to train up and practice on yourself
like wait a second do I really
understand this part and then so that's
one signal you need the second signal is
when you start asking questions you need
to have a really strong signal for when
it clicks when you're like ah there it clicked
clicked
>> okay just understand like fundamentally
why this thing is as it is how would
someone else learn how to learn with AI
>> this is a very good question I mean
first of all just change like the
misconception of AI being used to do the
work for you to instead you know use the
AI to explicitly help you learn like you
you don't you don't just use it to get
work done you actually learn from it I
mean the moment you just switch that
mindset which seems still fairly
uncommon but it's becoming more and more
common all the time you have most of the
things to to get there right and and
then to become really good first of all
like I said like know when you have gaps
in your knowledge understand what it
feels like when you fundamentally grasp
something and you know you you you
constantly come up with all these hacks
like uh you you'll notice
will respond in a fairly standard way
and your way of learning is probably not
exactly what it responds like because it
wants to you know make sure everyone has
a good experience
>> but you probably want it to respond in
another way. I very often tell it for
example be extremely direct and
concrete. Always show me all the
intermediate states and the shapes of
the code you produce. make sure to to
make sure I have like a really intuitive
understanding of why it happens. And if
you're unsure, make sure you show me
options and like what others have tried
and why this works and why something
else didn't work. And you start becoming
good at like asking these questions that
give you the aha moment. Like as fast as
possible, you want to get to the aha
moment. Yeah. Like the first time you
understood linear algebra or the first
time you understood what back
propagation works, you probably had a
very clear like, oh wow, it finally
clicked. and to chase these clicks and
to make them appear like as frequent as
possible, right? That's like kind of
your utility function.
The era of listing specs in an interview
is over. Gabriel stresses just one
thing, the proof that you built
something yourself. A demo that anyone
can understand at a glance, one that
makes people think within 3 seconds, ah,
this person can actually build things.
That simple proof carried more weight
than educational words, and it explained
Gabriel, a high school dropout, far
better than any resume ever could. The
number one thing I recommend to people
is making a demo that is super super
simple. It's actually really hard to
make a good demo for a lot of reasons.
Everyone thinks it's hard because they
need to make a demo that is hard and
they don't have the skills. This is very
not true. You can make very simple like
you don't need that much code knowledge
to make a really cool cool demo. The
hard part of making a demo is making
sure that people understand why you can
code within 3 seconds. You know you have
like 100 like applicants for something.
If you apply with one link and they
press the link and you know you have one
shot, right? Like making sure you build
a really cool demo where people
understand what they're looking at,
which is really hard, and where people
understand that you're a really good
engineer, which is really hard, but then
you're there. I mean, that's all they
want to see. I mean, companies just want
to make money. You show them how to make
money, that you can code, and they'll
hire you. And then you might say, "Oh,
but they only hire people with degrees."
Well, yeah, because literally no one has
ever showed them that they can do their
work. They're like, "Oh, I had these
internships." And the interviewer would
be like, "Okay, what did you do there?"
Oh, I streamlined pipelines and made
things 30% more efficient. And usually
like, uh, okay, well, that tells me
literally nothing. Okay, what what else?
What have you done? Oh, I went to
Harvard. I have the best grades. Well, I
still don't know if you can do the job,
right? Oh, but I have all this
extracurricular. I was debate champion.
You start going on about all these
things that your parents will tell you,
people around you will tell you. Nothing matters.
matters.
Lastly, your likes and subscriptions
greatly empower us to continue creating
Click on any text or timestamp to jump to that moment in the video
Share:
Most transcripts ready in under 5 seconds
One-Click Copy125+ LanguagesSearch ContentJump to Timestamps
Paste YouTube URL
Enter any YouTube video link to get the full transcript
Transcript Extraction Form
Most transcripts ready in under 5 seconds
Get Our Chrome Extension
Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.
