Vídeos

Can AI really save the World? ft. Bill Gates




WHATCH IN YOUTUBE


Sitting down with Bill gates to see how AI can be also be used for something good! I spend a LOT of time trying to make my videos …
Bill hey hello good morning how are you
this is a pleasure so wait a sec where
am I what am I doing and why on Earth am
I sitting next to Bill Gates well I’ve
flown to Sagal in Africa to learn how AI
is being used to try and solve the
biggest world problems he’s hosted this
event called Grand challenges where he
gets together lots of people with ideas
on how to solve these problems and then
funds the projects that he thinks have
the most potential I have to say I have
always been very cautious about AI but
coming here the last couple of days I’ve
seen so many things where I’ve just been
like wow this makes sense so I read your
blog post online and you talked about
this idea that in your entire life
you’ve experienced two technologies that
you would count as revolutionary the
first was when you saw the graphical
user interface and then from what I
understand that kind of shaped a lot of
what Microsoft ended up becoming but
then the second was what you experienced
with AI what part of that made you so
convinced that that was the second there
has been progress in things like picture
recognition or speech recognition what
you might call fairly lowlevel things
well we didn’t have this Ani that could
could read and write I kept challenging
the people at Microsoft open AI with the
specifics like can you pass a test so
you set them this task of like solve
this biology AP paper it’ll probably
take them like years and they came back
in a few months that’s right and that
night we were like whoa now what was
especially cool about this is that the
reason Bill picked this bile Gap paper
is that it’s well a particularly tough
but also B it’s one that can’t be solved
by just regurgitating information which
at this point is all AI could do it
required critical thinking but that AI
it didn’t just pass it achieved the
highest possible score on it in fact I
think a lot of people in the last year
have had that experience of saying wow
what can it do and how do we apply it in
the vertical where we’re seeing it you
know most clearly is in coding whereas
you type it suggests things it helps you
with testing and already people are like
30% more productive I guess that brings
us to here so we’re in Sagal we’re at
this Grand Challengers event and you
know given what you know about how
powerful AI is I guess you’ve decided AI
is a big force that is going to be used
to also improve the world what things
are you most excited to use ai4 to make
a difference the Gates Foundation is all
about inequity you know including in
health and some work in education we’re
now focusing the foundation on saying
okay who are the innovators who are
thinking about tutoring African students
or providing Africans healthc care where
the shortage of doctors is worse here
than anywhere in the world and so if you
can have advice to a pregnant mother or
somebody who’s experiencing symptoms you
can improve the health system now you
probably have like a thousand questions
at this point I did too cuz like it’s a
big thing to even suggest that AI can
help to solve global inequality cuz with
most new technologies inequality
actually gets worse it’s usually the
wealthy that have access to it and can
afford it and therefore benefit from it
while those that can’t are left behind
and the Gap is widened so exactly how
does does that work well according to
Bill Gates there are four Avenues in
which AI can actually help to shrink the
gap for Farmer advice educational
tutoring patients directly using the AI
and healthcare workers using the AI so
those are kind of the four areas that
now we want to bring those innovators
together and help them so for Farmers
one key thing the AI can do is pull
together the information from weather
forast satellite images various sensors
and even specific news outlets to tell
them in simple terms when they should
expect a new crop disease outbreak so
that they can actually take preventative
measures for Education we’re very used
to the idea of the classroom being one
teacher 30 kids in front of them but in
a lot of less privileged areas it might
be one teacher to 100 kids and within
that there’ll be a vast range of
different capabilities so AI can be each
student’s personal tutor it can teach
them what they need to learn at the
exact level they need to learn at
effectively allowing the smartest kids
to not be held back and the people who
are struggling to not be left behind
plus with an AI you can ask unlimited
questions and not be worried about
judgment or holding the class up or for
another example we saw an amazing
example yesterday of textbooks not being
localized to other languages so a lot of
these underprivileged countries they
don’t have as good resources as we do in
the west the AI can not just create
local language textbooks but also create
animated serieses where people are
watching people who look like them and
sound like them with their intonation
explaining Concepts at their exact level
yeah the tutoring
capability you know particularly for
languages that in the past wouldn’t have
gotten the effort that English gets is
pretty phenomenal and of course it’s not
just text in text out now now you know
we’re taking avatars and having them
speak we have speech going in uh so the
additional modalities makes it even more
accessible and then for patients AI can
be a way for them to have some medical
support when otherwise they might have
none if it feels like hospitals in the
UK or the US are understaffed because
when you go in you have to wait hours
before getting seen then in some parts
of Africa turn their hours into weeks
the cues are so outrageous that no joke
every single year millions of people who
are completely savable are dying just
waiting to be seen but think about it if
every single patient had access to an AI
chatbot that was trained in medical
procedures then at the very least while
waiting for a doctor they could at least
take steps to stop things getting worse
plus the added layer for a lot of these
countries is they have very low literacy
rates so just handing out medical
leaflets or telling them to Google the
solution isn’t going to be very useful
but an AI chat bot that can actually
talk to them that absolutely is an AI
that you can talk to in your own native
language that basically does the
internet scouring on your behalf and
then just speaks the results back to you
the medical advice whether it’s your
mental health or your physical health
your diet you know the fact that you can
consult and you know get good advice on
health things that’s really world
changing do you reckon there’s like an
element of people feel a bit scared to
bring some things up with their human
doctors but with a chatbot they’re
actually they feel like it’s a safe
space you know with the chatbot you’re
not going to meet them at the grocery
store I mean it doesn’t go grocery
shopping so first people worried that
people would be less open with the
computer and you you run into a tiny bit
of that but in most cases like you’re
saying it feels like okay it’s private
to me now we need to make sure we
partition the data so we we deliver on
that promise but you know we can
absolutely say okay this is only private
to you and then you’ve got the
healthcare workers so many
underprivileged countries have a problem
here that they only have 5% of the
doctors required to actually serve their
whole population how do you fix this
well build them AI tools that can save
them on all their busy work AI that can
handle their clinical notes for them for
example AI that can recommend the best
course of action so a doct just has to
approve it instead of writing the whole
thing out an AI that can have a look
through every Radiology scan before the
doctor’s even seen it which would mean
that instead of the doctor starting from
scratch effectively every single time
they open up a new scan they’re actually
starting from a scan that’s already been
pre-checked with an AI having gone
through and highlighted every single
possible risk factor at which point all
the doctor has to do is to go through
those risk factors and then decide if
they agree or disagree with its
assessment AI that does this it actually
exists now and it’s already making
doctors fast and less likely to miss
things okay hopefully you’re starting to
see that the potential here is great but
we also have to be really careful with
AI I wanted to touch on bias cuz that’s
a question I have how do we avoid bias
in these new AIS that you’re trying to
help build it is fair to say that
getting data sets of disease prevalence
and things like that into the models
that’ll be important and that’s another
role for the foundation you know the
world’s best health data set is one we
fund at a group called intern AAL Health
metrics and so we need to make sure that
gets into these models so they’re able
to help people who aren’t just in the US
and yeah this is a really important trap
that we need to avoid when using AI it’s
very easy to be like well we’ve already
got chat GPT why don’t they just use
that in Africa why do they need to make
new stuff and yeah I mean aside from the
fact that the average healthcare worker
or farmer is going to really struggle to
afford the premium subscription to use
the best version of that but also most
of the investment in to AI across the
entire world right now is happening in
rich countries like the US now that’s
fine but what it means is that a lot of
the data that American chatbots are
being trained on is US data which can
distort the way that the chatbot
responds to situations with a US bias so
I was showing an example of someone in
Malawi in Africa who was using chat gbt
to try and figure out what illness they
had they typed in their symptoms they
said I’ve got a fever I’ve got dizziness
I’ve got headaches and chat gbt was like
well you have the flu that was incorrect
the the person actually had tuberculosis
but because the rates of tuberculosis
are so low in the US chat GPT didn’t
even consider it as an option now bear
in mind this doesn’t even need to be a
conscious bias for it to be a problem
like just for example if you say train
your AI on global Health Data you might
think you’re doing a fantastic job but
the problem is the Richer countries
report more Health Data which would mean
that as far as your AI is concerned the
problems in Western countries will be
far more heavily weighted and will
therefore become the focus so Che gbt
can be the basis on which these new AI
models are built but there is still
clear Merit to actually training those
models using local data the other
important question with all this AI
stuff is do you really want AI to be
making these big important decisions
instead of a human no not really I think
that would be a bad idea especially for
the time being while AI doesn’t have the
reliability or the accountability that
it needs which is why it’s really
important that all these tools that are
being built right now they’re being
built to focus on decision support a way
to help people make better decisions as
opposed to decision replacement making
those decisions for them but there is
still one big question mark if these AI
Tools in these underprivileged places
can be such a big help why has no one
made them yet so a lot of what you’re
seeing here is like a market failure or
in other words a situation in which
leaving profit maximizing companies to
their own devices has not resulted in
the best outcome if there was a demand
for a chatbot for students in
underprivileged backgrounds it kind of
would have been made already unless a
market failure was occurring and then
you’re coming in and being like we’re
going to fund projects exactly and so
philanthropy comine those things and get
involved and so AI to help Africans uh
that’s another example where we can
supercharge it the foundation will make
sure AI gets used for Africa that
there’s not you know a 10 or 20e delay
because the AI gets delivered through
the cell phone over the mobile network
so we don’t really need that much more
Hardware at the end user we need need
some data centers you know we have to
make sure that gets funded now some of
these innovators will figure out
business models but in education and
health philanthropy will be a a big part
of accelerating that and there’s two
sides to this philanthropy it’s both to
fund nonprofit companies like you
probably see a lot of but it’s also for
already profit making companies because
they’re the ones who actually have a
strong incentive to make better products
and also be efficient because they have
to worry about their own costs it’s just
that in a lot of cases these normal
profit maximizing companies they don’t
see any money in poor places like Africa
so they never even try to invest and
create solutions for them so in the case
of AI because the potential for AI is
just so massive you could have a
situation where the tools will all be
built by the West for the west and the
levels of global inequality will
Skyrocket to the highest they’ve ever
been and that is where philanthropy
comes in in the form of grants for new
companies who want to start up with the
objective of solving these problems and
in the form of subsidies for existing
companies to help give them an incentive
to refocus on not just the rich parts of
the world got a quick rapid fire round
for you okay what phone do you use I use
a Samsung fold oh use a fold yeah it’s
pretty nice uh nice what made you choose
that you know I’m I’m very PC Centric
but sometimes I like to read magazines
and news and so that gives me enough
screen I I’ll go to my PC when I’m
writing documents but for reading this
works very well look it what’s your
favorite app right now you know I use
Outlook a lot uh that’s my sort of
Center my kids criticize me for that one
Gadget that you can’t live without oh my
PC I’m I’m still you know I like a big
screen who’s your favorite
YouTuber
uh good
question
uh wow my favorite
YouTuber you know I watch Tennis there’s
poker there’s Bridge there’s you know UK
green it’s always just serving up just
such a variety of things your favorite
is the algorithm yeah I guess uh that’s
dangerous well thank you very much a
pleasure

Artigos relacionados

Deixe um comentário

O seu endereço de e-mail não será publicado. Campos obrigatórios são marcados com *

Botão Voltar ao topo