Episode Transcript
Transcripts are displayed as originally observed. Some content, including advertisements may have changed.
Use Ctrl + F to search
0:02
Hello and welcome to the big Story. Two
0:04
point oh hello, this
0:07
is the first episode of the second season
0:09
and we are your hosts. I am
0:11
Anjali and I am Prateek.
0:14
So Anjali, what's
0:16
new in the season? What is so two point
0:18
about it. So you know how our news
0:20
culture is quite noisy and
0:23
there are so many headlines that come every
0:25
day, this is happening, that is happening. So
0:28
exactly what is happening, that
0:30
is where the story comes in. So
0:33
just like season one, we will
0:35
get experts on the podcast to answer
0:37
all of our burning questions. But
0:39
in this season since we wanted to spend
0:41
more time on longer discussions,
0:44
we are moving to a fortnightly
0:46
series instead of a daily show and the
0:48
episodes are longer as well. We
0:50
have very free wheeling,
0:53
you know long, long discussion with
0:55
our experts so that we get the more
0:58
nuanced and well rounded up that
1:00
is all about the show. But let's talk about
1:03
this episode before we
1:05
go further. I just want to make an appeal
1:07
to our amazing listeners. Please
1:10
listen to and check out all the future
1:12
episodes of big story, all the previous
1:14
episodes of the big story on your
1:16
favorite audio streaming platforms like apple
1:18
podcast, Spotify gone adios everywhere.
1:21
We are on the Queen's website, we're on the Queen's
1:23
Youtube channel. We have taken over the world
1:26
and also check out our other podcasts.
1:28
Yes, we have a movie
1:30
and tv review show. We have
1:34
and we have a show on indian politics called
1:36
C assets. So to check them
1:38
out as well. Right? So
1:41
for our first episode we sort of looked around
1:44
and we try to find
1:46
out what is the one thing that is creating
1:48
headlines and
1:50
you might have guessed it. It is,
1:52
it's artificial intelligence,
1:55
machine learning. So
1:58
there is a lot of curiosity about how
2:00
it works, what it can do, is it
2:02
good? Is it bad? There is a group
2:04
of people who believe that it is going to change the
2:06
world for the better. And
2:08
then there is another group of people saying
2:10
that all is not take
2:12
our jobs to replace
2:15
humans, robots will kill us, we
2:18
will come back. So there's a lot
2:20
of talk and so we wanted
2:22
to answer a lot of noise, like
2:24
genuine, genuine noise probably
2:27
generated by But
2:30
yeah, so we thought, let us
2:32
get to experts from
2:34
different walks of life but
2:36
connected to Ai and
2:38
ask them some questions that
2:41
I would say we have and
2:43
a lot of our listeners would have as well, like we start
2:45
from very basic and then we want
2:47
to go into some, you know, deep questions
2:50
about how it will impact the society at large.
2:52
So Anjali, what were
2:54
some of the questions that you have for our
2:56
guests? So I think I would
2:58
like to start by understanding how does that work?
3:01
Like in very simple language, how
3:04
does it work? What is the
3:06
thing behind there also this emergence
3:09
of these new tools like
3:11
gPT dolly
3:13
and lens are in, you
3:15
know, very common use
3:18
cases like art and essay
3:21
writing and all of that. So it's actually
3:23
a really interesting, you know,
3:25
I think to deep
3:27
dive upon key AI
3:30
and Art co relation
3:32
right? And for
3:35
that, like we have
3:37
our first guest who is
3:39
Jemima claims he is a Youtuber
3:42
and he's also a graphic designer of
3:44
working real human graphic designer,
3:46
not an a a graphic designer. He works
3:48
at weapons. And
3:50
we want to talk to him about his
3:52
journey with art in general
3:55
and how he sees this
3:57
encroachment of AI in
3:59
the space of art. Interestingly
4:02
enough, he actually made a Youtube video about
4:04
this recently where where he sent
4:07
an AI generated design to a client.
4:09
So that
4:12
would be interesting to explore
4:14
what was that? Like what was the whole
4:17
vibe of the clients? Did they know
4:19
that they find out? So that was really nice.
4:21
So we just want to get
4:23
a peek into how
4:25
does the creative world look at ai
4:28
what are the kind of problems that
4:30
AI is solving for them? Do
4:32
they like it? Do they not like it? And
4:34
at the same time we really want to zoom out
4:36
and talk about a more macro view
4:38
at AI for which we have
4:40
Doctor who is
4:42
a professor at iit Bombay and
4:45
who works extensively in
4:47
aI policy. He has
4:49
spoken about this in a lot of places.
4:51
So, I think he'd be the perfect fit for
4:53
us to understand what
4:55
a policy around here should look like, what
4:57
is really doing to the world.
5:00
And should we be
5:02
worried about it's going to be so interesting to have an artist
5:04
and a turkey talk about. All
5:07
right. So let's jump right into the corners. Let's
5:09
jump right into it. Welcome. Sammy. Welcome
5:12
And hello, I want to
5:14
start today's discussion, first of all
5:16
by talking about ai
5:22
ai artificial intelligence
5:25
Dalil bad.
5:28
Let us you know, hold back a
5:30
bit and discuss what
5:32
artificial intelligence is and how
5:35
it works in a sort of simple
5:37
language. Right? But I was thinking instead
5:39
of Anupam telling us
5:41
what it is because he has been working
5:43
closely with the eye and he knows the internal,
5:46
you know, workings of it. How
5:48
about all of us tell
5:50
him what we think Ai
5:52
is and how it works. And then he
5:55
takes our class and we're like, okay,
5:58
we will do our version of
6:00
it. And he'll give us marks who
6:03
wants to start, I am at the bottom of
6:05
the of this pyramid. So I
6:07
but no, no, we'll do we'll do we'll do
6:10
it alphabetically. So Anjali, like
6:12
I said, I'm at the bottom of this pyramid. So my
6:15
definition is going to be the most
6:17
basic. So how I think social
6:19
intelligence works is like a
6:21
told a while back, it is just a
6:23
very sophisticated coding
6:25
where you feel into
6:27
a program, all the information
6:30
that that is on the internet. And
6:32
you ask it to identify
6:35
patterns. And then
6:37
when given a prompt with
6:39
the identification of those patterns
6:41
create a new result. Now
6:43
this can be done for Yeah,
6:51
so helping off of what she is
6:54
saying. I think artificial artificial
6:56
intelligence, how I understand is
6:58
is a way to, as the name
7:00
suggests, artificially create
7:02
the pattern of thinking of her human
7:05
brain I would say because Hamas
7:09
that becomes sort of our data
7:12
observed and then
7:14
we recognize some patterns
7:20
process information
7:24
Intelligence generate
7:27
over the years. It gets more sophisticated
7:30
and that's how
7:32
I understand it in a more most basic
7:34
sense. I mean what do you think? Yeah, obviously
7:37
you guys went a little too technical
7:39
I think I feel will
7:41
correct me if I'm wrong but Ai is
7:43
basically pattern recognition.
7:45
Like it's very good in recognizing patterns.
7:48
That is what Ai is for me in my
7:50
profession because I'll only talk about my
7:52
profession here but more
7:54
or less like when we talk about jobs and all everything
7:57
and if you look at it from
7:59
a bird's eye view, then I think Ai
8:01
is a tool that will make us
8:04
as a species more productive in the future
8:06
and will make our lives easier. So
8:09
that's what Ai is according
8:11
to me. Okay, so
8:14
so first of all, first
8:17
of all, thank you for having
8:19
me here and
8:21
uh thank you for coming.
8:23
So it's very interesting to talk to you
8:26
a number uh
8:31
definition Angelica
8:36
actually
8:37
actually because
8:46
let me try to explain, first
8:49
of all instead of explaining
8:51
ai Kahuta. I want to start
8:54
with a I can write
8:56
because there is a lot of cultural
8:58
opinions on what it is and I think
9:01
we need to cut a little
9:03
bit there. So the first
9:05
thing you have to understand that artificial
9:08
intelligence is a misnomer or rather
9:10
a market term, it
9:13
has pretty much nothing to do with intelligence
9:16
which is where protic, your
9:18
definition flies out of the window.
9:20
There is no neuroscientists
9:24
on earth right now.
9:26
Who knows why
9:28
are we conscious survey? Are we really
9:30
intelligent and
9:32
hence there is no technology which can
9:34
artificially replicate intelligence
9:43
outward patterns exist,
9:48
record any
9:53
machine which can autonomously
9:55
make decisions is called a nail. And
9:57
generally the technology which is being
9:59
used in 95 96%
10:02
of ai is machine learning. Machine
10:05
learning how to, I would say angelica,
10:07
definition more or less character. Machine
10:09
learning data.
10:12
Machine decision
10:24
decision. Later classification.
10:30
Machine orange
10:39
classification machine learning or
10:50
who are quite fruit drawing.
10:52
Bannock is um generative
10:54
viable daily
10:57
charge gpt a generative machine
10:59
learning types but fundamentally
11:02
all machine learning algorithms,
11:05
they eat a lot of data
11:07
which is called training data.
11:09
So sorry first
11:13
are they new like things like dolly and
11:15
charge. Gpt.
11:16
No they are not, they're not.
11:20
I think 2nd 3rd versions
11:22
of like dolly
11:24
actually
11:25
more more than more than 2nd
11:27
3rd versions of that technology particular
11:30
brand names. But
11:38
so it is basically what is
11:40
called in the technical field a
11:43
language model exiled
11:45
language english
11:49
or patterns
11:53
or patterns is
11:56
a language model. Language
11:59
models, basic technology
12:02
nanotechnology. Earlier two thousand's
12:05
major technologies exist but
12:08
joe hardware software
12:12
but Gpus market manager
12:14
of hardware Nous software checked
12:17
earlier research
12:20
or Joey I just
12:23
ca laboratories methodology which
12:25
was something we were working on
12:27
in our fields. They
12:29
suddenly started to stop
12:32
remaining as scientific projects
12:34
and became market commodities
12:36
which started to be converted into products.
12:39
I would say that was a man.
12:41
I mean we were trying to solve some
12:43
specific problems but
12:45
as a I became commercialized
12:48
people also started to realize
12:50
instead of trying to solve specific
12:52
but hard problems, it is much
12:55
easier to throw lots
12:57
of data models, quantity
12:59
has a quality of its own or
13:03
it's a research
13:06
career or what
13:15
problem are they trying to solve? I
13:28
want to I want to pause you for a second because you
13:31
bring up an interesting point now
13:35
this and since we were talking about things
13:37
like dolly and mid journey and
13:39
stable diffusion. Shamim Ur
13:42
a working professional right. You work
13:44
with graphics on a day to day basis. So
13:46
how do you think have
13:48
these tools sort
13:51
of infiltrated in your
13:53
industry and casually and probably problem
13:55
discussed, correct problem
13:57
exists. Tools
14:00
solve career in your industry specifically
14:02
in my industry at this point
14:04
it's making us more productive.
14:07
Like I still remember in
14:09
my initial days, 2007, 8
14:11
at the time of seven was
14:13
I learned in like the very
14:15
like vintage version there. If you
14:17
have to remove people from a background,
14:20
it was a very big task like it was like a full
14:22
day task. Now it's just one click.
14:25
If I have to make a brochure
14:27
then if I'm using five
14:29
images as such, then
14:31
I can create a brochure in one hour, two or
14:34
maximum three years at the time
14:36
that those three years were three days.
14:38
So as a designer
14:41
I'm being more productive, my team
14:43
is being more productive and if you
14:45
talk about all these designing software, they are becoming
14:47
like intelligence
14:50
is not what they are but for us they
14:52
are becoming intelligent in a way that
14:54
they give us solutions easier
14:56
and faster. Do do you think they seem intelligent
14:59
to us looking
15:01
at it as a layman coffee, truly
15:04
intelligent Like they they might not be intelligent
15:06
in the way we are intelligent but
15:09
they seem pretty into
15:14
background removal. I think also
15:17
because for so long people have
15:19
been removing background that is
15:21
something that the software has also been learning
15:23
that when somebody wants to remove a background
15:25
this is the distinction that they are seeing
15:27
between foreground background and
15:29
after we just discussed
15:32
after all of this data has been collected
15:34
now they've developed a thing where
15:36
one click and the software knows
15:39
what do you think is the foreground and the background
15:41
if you see a new update will come
15:43
and that software will add a new
15:46
to background removal. Background extension
15:48
is just one option. Suppose it's
15:50
very difficult to tell you but you
15:52
get multiple options in a very short
15:54
amount of time. But I I don't
15:56
I don't feel it's it's
15:58
a threat because see jobs
16:01
will change the perception
16:03
of jobs will change. I cannot
16:05
be rigid key. No no I will I will
16:07
do the best work. A I cannot do
16:10
the way I will do and I'll
16:12
only do it myself then I will
16:14
be kicked out because other design will come he'll use
16:16
AI and he'll do
16:19
multiple things in the same amount of time.
16:21
Is something like that happening. He
16:23
has. No
16:26
no I haven't seen because it's
16:28
all about productivity. Right. Who wants to work
16:30
till four in everybody is
16:33
a. I being used very prominently.
16:35
Yeah I'll tell you
16:37
one live example. I'll tell you suppose
16:39
you have a holiday
16:42
client. Okay suppose you are working with a client
16:44
who sells holidays. It's not possible for
16:46
us to go and click a hut in the mountain picture
16:48
all the time. So what we do will
16:50
create an image like that in a I suppose
16:53
and then if necessary we'll try to
16:55
regenerate it because we we
16:57
don't have inspiration before right? But the
16:59
Ai has as said no the
17:01
Ai has been fed so many
17:04
images with a mountain in the heart.
17:07
He'll give us options because
17:09
Ai work is still very patchy. Like I
17:11
I don't see it very like
17:13
fine fine. Okay so we'll
17:15
watch this image and we'll get
17:17
inspiration of the lighting and set up and how
17:19
the mountains and our this thing. But
17:21
we'll create it manually so
17:24
it will look better for it will look better
17:26
and it will be in our control.
17:29
So that's what Ai
17:31
is being used now so
17:33
we can create a safer image. Like it
17:35
won't get cooperated because we are not stealing the image
17:38
we are constructing saying you are
17:40
currently using Ai as a starting
17:42
bouncing board of the
17:44
first idea mildew. And then we
17:47
can because you
17:50
also had made a video about this where
17:52
you said that you gave an
17:54
Ai generated design to an actual client.
17:57
So talk to us about that key
18:00
battle. What kind of brief did you get
18:02
and how was that whole experience
18:05
of you know actually going
18:07
through it and and
18:10
what was that like see how
18:13
it works like and first of all, why did you do did you just
18:15
do it as an experiment? Yes.
18:18
Yes, yes. It was an experiment because we
18:20
know okay. We know when we are teaching something.
18:22
No we are throwing
18:24
things like in the dark we don't know
18:26
what will stick. So you show
18:28
what you can do for them And they'll
18:31
say yeah we can do this also. So you
18:33
do things to get rejected
18:35
sometimes if you're lucky then something from
18:37
that only will get selected okay.
18:39
But like 99% of time
18:41
though it's like okay,
18:44
okay these are very good. But we had
18:46
this in our mind and this is the hook point
18:48
in the business, right? If they
18:50
if you get to that point then you have the job
18:53
and those two logo options.
18:55
No, that was my way of
18:57
like reducing the what
18:59
you call beating because if I tell the initial
19:02
exactly if I did,
19:04
you know, I don't know quite
19:07
possible, likely reject. But
19:09
at least the client actually
19:12
yes, exactly. Because the
19:14
best thing about Ai in our businesses like
19:17
you're not stealing anything. It's not
19:19
cooperated yet because
19:21
I will talk about today.
19:23
What is the idea behind stealing
19:25
it because the air generated
19:27
options also know I did some tweaking
19:29
there also it was not properly. I
19:32
we showed it to them and
19:35
if they like you. Okay this is the go
19:37
one and we need 10 poses
19:39
of this one. At least we have one
19:41
basic post. So we'll do more changes and
19:43
you're saying Ai helps in you know, cracking
19:46
that first step. Yes. There will
19:48
be one graphic designer who's better in air
19:50
and there will be one graphic designer who is really who wants
19:52
to do things the old way.
19:55
Okay, okay. It's a simple digital
19:57
camera and the camera, the transition was
19:59
very, very crooked. Okay. The
20:02
old people who were like real camera
20:04
users, they were very, they didn't want
20:06
it to go to digital camera,
20:08
but who did, who did go to the digital
20:10
camera? They survived. Right.
20:12
So this is what it is. The technologists
20:15
will be evolving and you need to
20:17
write the way basically you need to be updated.
20:19
That's that's what I feel I
20:25
was listening. I have
20:27
one question for Shamim in your
20:29
video. Also you mentioned that
20:31
sometimes what an Ai lacks
20:34
is context. So
20:36
and I think this is very cultural also
20:39
that when you have an indian client
20:41
you have the entire cultural
20:44
background that this client is coming from.
20:46
And so you know you
20:48
might not get it in one go what
20:50
the client wants but you have certain
20:53
context which a lot of times
20:55
because the air doesn't know who this person
20:57
feeding in the prompt is
20:59
might not be able to give
21:02
a more accurate
21:05
image. So how does that
21:07
work? If you can explain more about context
21:09
in graphic communication,
21:14
see your question. No, I I
21:16
want to give it to because your question
21:18
answer is hidden is
21:21
Jarvis possible and if
21:23
Jarvis is possible then is Skynet
21:25
possible, Jarvis and Skynet
21:27
as possible before he answers this, I just
21:29
want to interrupt and explain
21:32
to our listeners what Skynet actually is.
21:34
So a terminator movie,
21:36
the Kyonggi Skynet is like
21:38
the villain of that film and it is like
21:41
a I would say which
21:43
gains self awareness and which becomes
21:45
human and at
21:47
the end it decides that
21:50
humans are the biggest
21:52
enemy in the world and they mission
21:55
is to you know kill humans.
21:58
And essentially Skynet is the example
22:00
that is used to illustrate ghetto
22:04
hockey a human, this
22:09
is like the first time
22:12
in popular culture that we at the
22:14
scale. So all of this obviously irobot
22:16
or go India go
22:18
robot shankar androgyny, that
22:22
movie's premier Ai basically
22:25
humans. So um
22:27
yes is Jarvis and
22:30
Skynet possible. Yeah
22:33
then then you have your
22:34
answer again, I used the
22:37
I I used the word when
22:39
I started to describe what Ai
22:41
is and isn't right, I use the word
22:43
and I used it very specifically I
22:46
said think of ai as a parrot,
22:49
Jarvis is
22:51
not a parrot right? A parrot doesn't
22:53
know what it is saying, A parrot just
22:55
says things, it has heard
22:57
patches from things it has heard
23:00
but I want to answer this
23:02
in a bigger detail, you know like
23:04
because you guys are having this conversation and I was
23:06
listening very closely to
23:08
try to sort of so there
23:11
are again things which
23:13
we have to be very clear about
23:15
Shamim, you used the word
23:17
or rather the phrase that
23:20
Ai is coming right
23:22
and you
23:26
would like to make an analogy. Technology
23:29
never comes. People choose
23:31
to use technology and the people
23:33
who have power and money get
23:35
to dictate which technology is
23:37
used in waterway technological
23:39
part three nautical technologies coming
23:41
another phrase we keep hearing is genie's
23:44
out of the bottle. Yeah, it's
23:46
not a genie, it can't be out of the
23:48
bottle. So bad. There
23:51
is a technology, it is
23:53
used in particular ways
23:55
which way's will become popular
23:58
and which ways will not become popular
24:00
depend a lot on
24:02
ai industry or control what
24:05
use cases they fund, What use cases
24:07
the research, what use cases they
24:09
develop an
24:12
individual level. Sure Arabic
24:14
professional to achieve skills
24:18
survived but at
24:20
a macro level and
24:28
as a citizen facial
24:34
recognition tech are
24:37
identified privacy technology.
24:43
So remember at the end of the
24:45
day her technology wider
24:48
context apart, policy,
24:51
political, economics,
24:54
politics or economics. So Tianna
24:56
who control the technology and
24:59
the second thing you guys were talking about
25:02
a plot bad, correct and Anjali said something
25:04
which was actually correctly segmentation
25:07
job, bad background removal. Computer
25:10
vision that problem is called segmentation,
25:17
segmentation, possible segmentation
25:21
millions of times important
25:28
point. Whatever
25:34
value any eye generates
25:36
that value is
25:38
coming out of training
25:41
of data which was at some point
25:43
made by human beings.
25:44
So doesn't this then bring the question
25:46
of ownership
25:48
exactly here
25:50
the question comes at no AI can
25:52
exist without
25:54
data. A lot of data
25:57
and it is
25:59
a shame that right
26:01
now that data is kind
26:04
of being stolen Internet
26:07
Italia. It's that right. I'll give you one
26:09
example. There is a very
26:11
famous facial recognition
26:13
tech company in America evaluation.
26:16
Billions Clearview Clearview
26:19
dataset. They
26:21
just went on the internet on
26:23
America digital law enforcement website
26:27
under trials, photos, Internet,
26:31
social media, facebook, photo
26:34
media natalia right.
26:38
Company company.
26:46
So yeah, ke
26:50
hame jo bad career gives the job
26:52
maker surprise society. It's
26:56
easy. You can't leave
26:58
them to people
27:00
who are going to make money from it. You
27:03
have to have that discussion at
27:05
a societal or at a policy level
27:10
data protection, europe
27:15
had a law, it is called G. D.
27:17
P. R. You might have heard of it. It's a very
27:20
strong law is locked
27:22
in europe. Your data
27:24
cannot be even if it is on the internet, some
27:27
company can't come and randomly.
27:28
So I'm just going
27:30
to pause you here. So in some way
27:32
would you think that what
27:35
Shamim did with his experiment
27:38
that using somebody else's like using
27:41
ai generated images or even I
27:43
have my profile picture as
27:45
many lands and generate Corinthian.
27:47
I use a lot of a yard. So
27:49
would you consider that unethical
27:52
at some levels.
27:54
As a policy academic.
27:57
I try to stay
27:59
away from questions of
28:01
individual ethics, ethics
28:03
are not something an individual can decide.
28:06
Si Shamim when
28:08
he did that Shameka
28:11
demand met Tonia anarchy. I am using an
28:13
Ai ai millions of images. Million
28:16
images does
28:25
in the past
28:29
but as
28:34
an individual but
28:37
company
29:06
I actually had another question
29:08
to ask. The Ai
29:10
tools are using the databases that
29:12
are available on the internet virtually
29:15
for free of cost. Right. So
29:17
then is
29:19
there a problem with the entire
29:22
money making strategy of
29:28
they have collected the data from the
29:30
internet without giving any money to
29:32
the people to collect or
29:34
fruits. Data model train
29:38
a Prada problem.
30:00
You had that in our culture
30:03
and when I say our culture I just wrote me in
30:05
indian culture. I also mean american culture.
30:07
There is a false narrative or a
30:09
false narrative here. Regulation
30:12
innovation or year
30:15
problem here in America
30:18
or India and our country's key.
30:20
We don't have any
30:23
regulation on data or
30:26
even data protection
30:29
data protection bill. Hell
30:31
on him. But even an
30:33
imperfect law would have been progress
30:36
but I'm sorry to survive now.
30:38
We don't have any law to protect
30:40
your on my
30:41
data. Progress and what will happen if
30:43
a law like this comes
30:46
home ari daily life
30:48
depends on what depends
30:50
on what is written in the law. Like
30:52
right now last
30:54
joe draft data protection card
30:58
criticism law,
31:01
citizens sorry, responsibilities
31:04
but state sorry responsibility,
31:07
Jessica
31:08
again, Shamim should
31:10
be king of
31:12
like professor
31:22
science or humanities. Would
31:35
you be right now If you wouldn't have pivoted
31:37
to policy. Would you be working in
31:39
one of these companies now? Yeah
31:41
I would have been working in Silicon Valley in
31:45
2019.
31:54
I wanted to go
31:56
back to that thing like because after
31:59
listening to my, I'm completely
32:02
with and we are on the same page where
32:04
we talk about laws and policies because whenever
32:06
I meet somebody like who wants to be
32:09
a graphics and who sees me, who meets me
32:11
somebody I know I'll tell them our
32:13
professional, our profession is not just a job, it's
32:15
a superpower because all the fake
32:17
news, all the hysteria and all everything that
32:19
is done by people like us.
32:22
Okay and now there's
32:24
all this like
32:26
what you call Deepfakes and now
32:28
the fake news is exactly
32:30
defects and everything. This is done
32:32
by Ai. You don't put
32:34
a law in breathing right? You put
32:36
a law in cutting a tree because that's something
32:39
is dangerous, breathing is not dangerous. I
32:41
don't have problem with Ai. Okay, I have a problem
32:43
with the hysteria that is being created
32:45
against a you
32:48
will eat your job is dangerous this
32:50
and that. So I'm going back to the same
32:52
how dangerous is let's okay,
32:54
let's talk about key danger in
32:56
its worst form, what is the
32:58
harm they can do? And since we
33:00
are recording it on a friday the 13th
33:02
let us go,
33:03
Okay, let's
33:06
let's actually talk about the danger of
33:10
the answer is a very
33:12
contradictory answer to the patients,
33:17
yes Ai is extremely
33:20
dangerous. No. Skynet is not possible
33:22
because it is not dangerous in the way people
33:24
think of its danger then a common
33:26
person thinks of AI is dangerous.
33:49
You
33:50
make eyes also stupid
33:52
act, you've taken all the
33:54
power away from
33:55
it. Yes because but
33:58
I would also say this very stupid
34:00
thing is extremely dangerous in very
34:02
stupid ways. A problem here
34:04
was stupid, basic, basic,
34:09
basic, basic
34:11
examples how dangerous he is
34:13
America joe
34:21
building America
34:26
program technically
34:30
what is the chance? Criminal
34:37
judges or juries? Common
34:41
sense, shoulder case, black
34:53
community but
35:08
but problem
35:18
or
35:27
was this, what year was this
35:29
has been going on for decades.
35:31
This is a huge problem in America
35:34
example, amazon
35:37
warehouses to
35:39
amazon has moved over from
35:41
purely online presence and going to
35:43
brick and mortar industry of a warehouse. Banaras
35:46
jamming carried warehouses,
35:51
amazon, they tried to
35:53
replicate their whole
35:56
hiring process by converting it
35:58
to machine learning. Machine learning
36:03
problem bias
36:10
people, machine learning, machine
36:15
learning women, Black,
36:45
okay, similarly, but
36:59
generally in
37:10
the army police
37:13
departments, Delhi punjab
37:16
assam, Hyderabad Chennai
37:19
police departments, facial
37:22
recognition tech used to start
37:25
machine learning machine learning
37:30
probabilistic up
37:36
a I say inaccuracy nickel,
37:39
it is not a technical
37:41
fault, inaccurate,
37:44
Neykova inaccurate,
37:48
inaccurate right about decision
37:53
the idea. Okay, inaccurate recently,
38:02
ex Chief justice sheriff
38:04
bob, he announced that
38:06
he has decided and the Supreme Court
38:09
judges think indian courts make
38:13
resources, procedure,
38:27
procedure simplify justice
38:29
be a content justice be
38:31
a product of justice,
38:34
A product now
38:41
machine learning, Machine learning, sorry,
38:48
jobs, jobs spear
38:51
member vodka, intelligent
38:55
creative jobs. Economic
39:01
50 60% jobs, manual
39:04
laborers, especially especially
39:11
like India these jobs
39:13
cannot be replaced because the
39:15
worth of a human being in India is very
39:18
low. Let's be very clear
39:20
about jobs, jobs,
39:24
jobs are but never
39:31
color job correct
39:37
companies, companies
39:39
mining or
39:43
jobs, jobs, jobs
39:46
employee to employee. You
39:49
have an entire economic
39:52
in this country where
39:54
increasing automation and increasing
39:56
machine learning has enabled platforms,
39:59
the platforms in a machine learning and
40:02
because of the popularity
40:04
and profitability of these platforms as
40:06
a profitable dot com company
40:10
will be lost because of that.
40:12
And because of no policy
40:14
on gig work whatsoever.
40:16
Our entire workforce
40:19
is slowly changing into gig work and
40:21
this is extremely dangerous,
40:23
short long
40:37
overall jobs.
40:44
But the biggest problem also
40:47
I would want to get Shamim into this
40:49
key. Even
40:51
us when we were talking about it,
40:54
we were only considering
40:56
all of these higher
40:58
level jobs like creatives and all of that.
41:01
But the actual jobs, we
41:03
aren't talking about those kind of jobs,
41:05
right? And when we also don't associate
41:07
them with Ai Ai
41:11
S C jobs could replace Karenga especially
41:17
after listening to after listening to
41:20
my whole point of
41:22
view has changed this problem that has
41:24
said has never come came to my mind.
41:26
Exactly yeah. The question coming
41:28
to my mind is that key, what
41:30
we should do because mom said key individual
41:33
contribution is very less so
41:35
that's why we need laws and policies
41:38
but how we can
41:39
help like at an individual
41:41
level do what is best
41:44
for you at an individual level
41:46
encourage famine because he has a platform.
41:48
You should encourage people to learn these skills
41:50
so that at an individual level.
42:01
But as
42:06
a, as a policy professor as
42:08
somebody who looks at the entire society
42:11
masses of individuals individuals
42:15
hama english, advanced
42:18
degree police.
42:33
Exactly, sarkar
42:35
policies, laws
42:38
at the end of the day, they
42:40
also come from the popular mass
42:56
popular topic local
43:03
insurance, cancel job
43:06
jobs are jobs
43:09
they call her cell jobs but
43:14
online content creator Hamara
43:24
come bath,
43:30
low quality content, high
43:32
quality content. So
43:34
the point I'm trying to make especially with
43:36
Sammy because Sammy has a large audience
43:39
on youtube right? Loco
43:41
encouraged debates,
43:45
just scholarly debates. Academy
43:47
academy a measure debates, policymakers,
43:53
conference fairness, accountability,
43:55
transparency, fact, conference ai
43:59
or policies or actual
44:08
danger technologies
44:11
related her
44:13
actual danger, boring sensation
44:22
gradually. Exactly
44:29
I was just saying because
44:31
you may be
44:43
in the same
44:45
the same way medical
44:47
field maybe like er
44:51
careful because medical
44:54
it is said it will be the most
44:56
useful
44:58
right now. Yes,
45:00
you should be So simple
45:02
reason. India publishes
45:05
a lot of data for free. Some open
45:07
government data whose website,
45:09
52% downloads I think
45:12
health data, health
45:15
data, jokey health
45:17
data, machine learning
45:19
Alana hottest field commercially
45:24
there are big companies
45:28
wings, telemedicine telemedicine
45:32
remotely bad chat vodka
45:34
through doctor. So
45:36
Covid fella basically
45:38
doctors training as far as technology
45:44
and science
45:49
methodology. Techno solution is
45:51
um techno solution is um
45:53
is that bad idea that
45:56
any social problem can be solved with
45:58
technology Dr Valerie
46:09
problem, chat
46:14
board media, stable
46:20
internet connection, smartphone
46:22
in english. Problem
46:28
solved problem DR
46:32
problem solved colleges.
46:38
Education don't call dr medical
46:46
problem already.
46:56
America makes research
46:59
american medical data,
47:01
hospitals machine learning
47:04
use cases patients
47:10
patients have
47:15
to be very patient a
47:25
malnourished, poor or black
47:27
local data time machine
47:31
decision to medical decision.
47:33
There are a lot of time they were medically
47:35
incorrect decision a
47:39
devil's advocate wala. Point comes
47:41
where people say that
47:43
the exclusive
47:53
technology
48:01
main difference healthcare
48:03
healthcare. So
48:11
then
48:12
would you would you advocate
48:14
for removal of things like
48:16
AI in matters of health
48:18
care and education or more?
48:20
I would advocate for
48:22
I would advocate for no
48:25
I would advocate for rational decision
48:27
making on who is owning that
48:29
rate of O.
48:31
A. I. Ca. Use seer of
48:33
profit making kelly. Individuals
48:36
and private companies decide career yet
48:38
as a society. Hm Sabbatical decided
48:41
correct. Use examples
48:44
of where popular
49:06
election useful
49:11
now but
49:16
marlo you
49:31
can't as a customer ask for
49:33
the company to change its policy but
49:35
as a citizen you can ask the government
49:38
to change its policy services
49:55
and citizenship
50:01
to sort of conclude as
50:04
a policy or where government
50:06
comes in ai is to
50:08
regulate for what and
50:10
for what not AI should
50:12
be used. That
50:14
is where the government should come in and
50:16
it should clearly dictate the use
50:19
of AI is okay. So that
50:21
would also cover all of our creative
50:23
fields created exactly
50:27
what in your opinion would be okay use
50:29
for a
50:30
more than okay as I would say that
50:33
first of all this proceed this process
50:35
to decide what the government should do.
50:37
You should not just hand it to the government
50:39
like that. It
50:45
has to be a democratic process right
50:50
down to
50:59
the public is
51:01
response dolly is
51:05
document I think
51:10
with this chat also it was
51:13
our objective. Yes
51:15
we want to talk about dolly and chad
51:17
Gpt and all of these are
51:19
you know certainly nice
51:21
popular issues but as
51:24
I got Amerco views
51:28
the car and even from your
51:30
understanding of it that these are not
51:32
the biggest issues that we should
51:34
be worried about. I think it's episode
51:37
titled We are worried about the wrong.
51:39
Oh yeah, I
51:42
would like to add one thing is that yes,
51:44
as I mentioned in my video also.
51:46
Okay. Somehow I feel
51:48
key we are being targeted like
51:50
the creative people are being targeted. Everybody
51:52
saying your job will go, your job will
51:54
go today. I understood what AI
51:57
is and how AI is dangerous.
51:59
Exactly. So I want to thank you guys
52:01
thank you for doing this with
52:03
me because it was such a privilege because
52:05
um you just opened my mind
52:08
truly like I'm
52:10
feeling good. Also that the way I use
52:12
AI is not the most dangerous one.
52:14
Don't you think this is also
52:16
like a self romanticization?
52:19
A lot of even creative people are thinking
52:22
that homeless victims it
52:27
I don't think that is the case. We
52:29
are the most privileged
52:31
people. Um it's as a toy
52:33
used we have the option
52:36
to choose it to use it or not use the term.
52:38
This is a playground for us right
52:40
now. Exactly. But year
52:43
options taxi driver because
52:48
Uber has out competed the taxes from
52:50
the market. So that is very important.
52:52
This point of our
52:55
ability to confuse
52:57
the actual reality with our reality
52:59
are very small
53:01
group of elite people who consumed
53:07
but I'm very happy protect
53:10
and Angela that you invited me my
53:12
subcommittee criticized Sada
53:16
ham academics problem but
53:22
very few of us talk to the common
53:24
masses. Some paper papers
53:26
have papers, there
53:29
are very few platforms, they have
53:33
issues seriously but okay so first of
53:35
all I would really like to thank, went for
53:38
in take me to give me a small
53:40
chance to translate
53:42
some of these issues to the common
53:44
language. I hope I was a little
53:47
successful at that.
53:48
Thank you. So
53:51
in closing I would just like to ask
53:54
um what is the first thing because obviously
53:56
the recommendation and I would be endless.
53:58
What is the first thing that people
54:00
who have just heard this chat should
54:02
go and read up
54:04
about. So there is
54:06
a paper I would
54:08
like people to read. The
54:10
paper is called on
54:12
the dangers of stochastic parrots,
54:15
stochastic simply means probabilistic.
54:17
This paper was written by a bunch
54:20
of researchers in the
54:22
U. S. It is a very nice introductory
54:25
paper to understand the
54:27
dangers of language model, language
54:29
model charge deputy, anything that
54:31
generates language. Also,
54:33
the paper is written I think in a very
54:35
easy to understand english
54:37
which anybody can read. You don't need to be
54:39
a machine learning professional. This would be an
54:42
interesting paper to start off with
54:44
very realized that these
54:46
artifacts are not operating
54:48
in a vacuum that they have a social
54:51
impact and what those social impacts
54:53
can be. So this could be an interesting paper
54:55
to begin. I wouldn't say this is you
54:58
know the end of the conversation. But
55:00
the beginning of the conversation
55:02
for somebody, I am
55:04
not the first person who calls Ai parrots
55:08
is conversation with
55:12
one thing that I will carry for
55:15
the next my life is a
55:17
I. Tota. The starting question
55:19
we started with was a I. And art
55:21
and creative job van
55:24
logic hold Kotecki
55:27
creative and
55:30
where the point where we ended
55:35
actually justice
55:38
system, labor market.
55:49
So I would like to close
55:51
with you as well and going back to our
55:54
starting questions. What
55:58
do you think after you've heard
56:01
all of this key your art
56:03
club exist exist, what
56:06
I personally feel after this long
56:08
conversation and after learning so much from them
56:10
is that key. We are being targeted
56:13
because we can be targeted because
56:15
some people want a i in everything
56:17
without any regulation and
56:20
we are the easiest people to talk because
56:22
they'll get people like me who will say is
56:24
making a life easier because
56:30
even this actually is so
56:33
interesting that you say, because I consider
56:35
myself a very, you know, tech savvy
56:37
person every time some
56:39
google assistant comes.
56:41
I really, I'm excited, Kara,
56:44
that's why as soon as ai dolly
56:46
all of this came I was very excited and
56:48
obviously I am not alarmist about
56:50
it. I
56:54
had a very favorable view of
56:56
it. So I
57:00
think advertisers
57:08
or any newspaper Mitnick alarmist
57:11
article article
57:15
article because I was talking about some
57:17
actual issues. But Mark,
57:21
I had reproduce
57:23
the results which had read
57:25
in some McKenzie report somewhere
57:30
academic organization that it's a
57:32
company or a company. Yes,
57:40
sensationalism, but or
57:45
boring imagination.
57:53
I wouldn't say
57:59
sensationalism
58:16
trying to do with this podcast.
58:19
The actual issue is
58:22
like much deeper and honestly speaking,
58:24
boring but love Yeah,
58:27
interesting. Which
58:29
is sad. But I think we
58:32
tried our level best to make it an interesting
58:34
conversation and thanks a lot
58:37
for that. Very
58:38
happy to talk to you guys. Also very happy
58:40
to meet you. You look at the channel,
58:42
I find it very nice. So
58:44
I'm glad you are doing what you're doing.
58:46
Thank you. Thank you so much. Thank you. Okay,
58:49
so Anjali thoughts. First
58:51
thoughts more than more than exciting. Jasmine.
58:53
Yeah. What
58:57
what were your expectations before we
59:00
started? I really was expecting
59:03
just you know, tell us about how he's not
59:05
threatened by. I and
59:07
I'm talking about you should be it might
59:09
it might get so powerful that will take over
59:11
your job. But I think it went into a whole
59:14
other direction. Was I was
59:16
expecting like a modification
59:18
of a. Because usually when people
59:20
talk about AI in popular culture entity
59:26
but now my
59:28
opinion on AI has changed
59:31
so much after obviously
59:36
catchphrase of this episode
59:40
and every time, you know, I feel like
59:42
oh it's going to take over the world, it's going to do this and
59:44
that like the image of a cute
59:46
little at
59:51
the same time, I feel that we've discovered it's
59:53
not all that good
59:55
and that it solely depends
59:57
on what it is used for. And I think we've drawn
1:00:00
a very clear distinction between
1:00:02
using our Ai for art and
1:00:04
using ai to replace human
1:00:09
consciousness. I would say. It
1:00:11
was interesting the case study that
1:00:13
tom talked about the 100 trials in
1:00:16
the U. S. Where that
1:00:18
is scary but it's way more
1:00:20
scarier than all your irobot
1:00:22
type stuff and it is such a babe. But
1:00:29
okay, this can have real world implications.
1:00:32
It's definitely something that these
1:00:34
giant corporations are looking at
1:00:37
and we as citizens have to be
1:00:39
aware and we have to be worried about the right
1:00:41
a and listen
1:00:43
to more academicians. Yes.
1:00:46
And that just goes beyond I
1:00:48
think everything that is popularly
1:00:51
discussed, we like
1:00:53
we should not be listening to influencers talk
1:00:55
about it, but listen to academy,
1:00:57
just listen to the big story. Yes,
1:01:00
that's because we will
1:01:02
get you access to these
1:01:05
crazy, crazy intelligent admissions.
1:01:07
So I think that's it for this episode.
1:01:10
It was a great chat chat,
1:01:13
let us know in the comments, what you
1:01:15
thought about it, what you think about, why
1:01:17
do you think under
1:01:19
or what animal do you think it is
1:01:22
and share
1:01:24
this with your friends, share this with your
1:01:26
parents, share this with anyone,
1:01:28
anyone, anyone who talks about,
1:01:35
What are we going to talk about in the
1:01:37
next episode of the big story. So
1:01:40
after the grand opening
1:01:42
of Ai and art and
1:01:45
Policy, let's talk
1:01:47
about media and we want to talk about
1:01:49
a very interesting aspect of media
1:01:51
which is media trials. It's
1:01:53
like this unique combination of
1:01:56
media and the judiciary, what
1:01:58
role they play in each other's functioning
1:02:01
and we're going to break down
1:02:03
the entire concept of popular
1:02:05
media trials that you've seen in the past decade.
1:02:12
Okay, so, tune into the next
1:02:14
episode. We have two amazing
1:02:16
guests with us. So this was
1:02:18
Prateek and this and
1:02:21
this was the big story. Thank you guys,
1:02:23
Bye bye. The
1:02:26
Big Story is a quint original podcast
1:02:29
executive produced by Shelley Value. And this
1:02:32
episode was hosted, produced and edited
1:02:34
by practically do and Anjali payload and
1:02:36
it uses theme music from BMG production
1:02:38
music, A special thanks to our guests,
1:02:41
Dr Anupam Gupta and Sammy Markelis.
1:02:44
Yeah, you
1:02:47
were listening to the Quinns podcast.
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More