0:00
that’s what we hear coming in from IT companies across Bengaluru and the rest of the country 25 is when AI agents
0:06
clock in Big question here is can AI really take over jobs these agents aren’t flawless Some of them are in the
0:13
development stage They have their own car crash wars [Music]
0:22
Hey Hi Madul Uh how you how you be doing first of all sir thank you so much for
0:28
taking out time and being here So how you been doing how’s life treating you uh thanks for inviting me D Uh life’s
0:36
good All uh all fun and uh entertainments happening Uh how’s life
0:41
at you life’s been very great with me also So Nadul uh instead of me uh giving
0:49
out your introduction let’s do things differently Why don’t you tell your own story what
0:56
you are doing how you get there how’s been the journey actually and yes
1:03
audience would love to hear straight away from you Sure Uh so I’ll give you a quick brief of what I do Uh I work as
1:10
chief AI officer at top view uh my job is mainly helping clients and my
1:16
internal D leverage AI and uh get the most efficient version that we can using
1:22
the advanced in artificial intelligence So I’ve been uh in I have experience of
1:28
roughly eight and a half years I have experience of roughly eight and a half years Uh I I serve as advisor to one of
1:38
AI one of AI’s largest GitHub repositories I have been invited speaker
1:43
at University of Oxford Tata Motors Google DES NASCOM and others I have I
1:50
also do write uh and have been read by 150,000 plus times I’ve mentored close
1:56
to 200 IIT Bombay and Karapur graduates and especially in terms of AI interview
2:04
preparation and uh we’ve been a little active on government of India’s AI web
2:09
portal and at one point they did they did recognize me as one of top contributors in the space So that’s a
2:16
very quick brief of what I do uh I graduated from NIT Mega in 2016 uh and
2:21
then did work as software consultant for two years and I got lot of hands-on on
2:28
IoT there and was parallely working with ML and eventually that’s when I found
2:33
that okay this space might be extremely interesting for me and then eventually switched uh to
2:40
ML uh before Chad GPD lot of work that we were doing was on hardcore machine
2:47
learning where you take a lot of sensor datas predict if the vehicle is going to fail or not or what the electric say the
2:53
consumption would be post GPD I think AI became the main street and then the
2:59
stories in news and everywhere was quite inspiring nul whatever you are doing whatever you have done basically so now
3:07
there is lot of AI AI AI nu Now think of me as a newbie a layman I
3:14
barely have any knowledge about the AI I don’t even know the difference between the AI and the Wi-Fi let’s say And a lot
3:21
of the audience also don’t know this thing So Nadul what exactly is AI
3:27
according to you okay Uh so I’ll take this I’ll just take a step back and give
3:33
you an explanation of what AI is uh initially or in software uh what used to
3:39
happen was we would give a set of rules and then an output would come So
3:46
let’s say uh you have a you have a robot and the bot is supposed to go and it’s
3:54
not supposed to go and touch a hot surface because if it touches there’s a short circuit Let’s assume a
4:00
hypothetical scenario right so in uh traditionally past decades this is what
4:06
the software was about where we inspect what should be done and that was being done by the software So we were
4:13
providing the logic in AI this role is reversed What happens is we leave that
4:20
complication to the networks or to the system to understand So in our example
4:26
we would leave the bot okay go explore all the surfaces in this loop and based on its exploration it would come up to
4:33
the fact that okay touching this surface is causing some short short circuit in my interface So essentially uh earlier
4:41
what was happening was you are providing logics to the system and it was deriving some results Now we provide the inputs
4:49
and the results and let the system arrive at what the logics are So something that uh we observe but we
4:57
don’t understand why it’s happening is that we don’t understand why it is happening or the system is too complex
5:03
is AI it’s the understanding of mapping what is happening with what was the
5:09
input quite interesting and you laid out in very simple words Nadul uh so Madul
5:16
one thing about the AI so most of the use case I would tell you so nowadays
5:22
there are so many wannabe entrepreneurs You see in India now everyone want to
5:28
you know start their own business or scale their own business The whole economy is revolving around the
5:34
businesses businesses starting something new basically of their own and now because of the AI uh things seem to be
5:43
much easier So earlier building a building up a website or building up an applications were way tedious way
5:51
complex that could require a lot of time and also a lot of funding So do you
5:56
think uh this gap has been filled by the AI now things has become much easier for
6:03
the new entrepreneurs as well as the existing entrepreneurs to you know start something new or the scale the existing
6:10
ones So how do you see this thing yes uh so I I would break this into two uh
6:16
subpart the first is AI is definitely increasing our efficiency so previously
6:25
what was possible let’s say a website development that was possible in 5 days leveraging AI we should be able to do
6:31
that in matter of hours so that is what AI is enabling all of us to do be it
6:37
market research be it coding be it understanding some complex phenomena AI is doing all of this it’s helping us
6:44
with efficiency improvement at an extremely good scale and uh I would I
6:49
would go on to say that this is 20 to 80% of efficiency improvement but then there’s also second
6:56
segment which AI is enabling which is making the businesses 10x and uh this is where this is a space
7:04
where new entrepreneurs or business owners are now actually looking at how can they leverage their uh current
7:12
business or their current understanding or a new idea marry it with AI and make the 10x shift and this is not the 80%
7:20
improvement or reducing my work by 20% This is entirely new segment where AI
7:26
becomes a primary workforce rather than a human work I agree with most of the things you’ve said Now AI is doing a lot
7:34
of things we never imagined we would do like you know building a website at just
7:39
a couple of hours You see now now I have seen many models you know they are
7:45
building applications through just a few text prompts So I would just you know write a few texts and I want a
7:51
e-commerce website to sell my these XY Z products and you could just build a
7:56
application in just couple of minutes this things he never imagined this would take a lot of time as we spoke right so
8:05
now do you think uh the software developers and also the porters uh who
8:10
are working in the industry would they be replaced in a s like soon time because because of the rise of AI only
8:17
the top you know one or two% uh of the people of the from the industry would be
8:24
able to survive how do you see see these things okay very interesting question uh
8:29
and I have a slightly different overview as opposed to what you see in newspapers
8:34
Uh one is of course there’s no denial of the fact that a developer who’s not
8:40
really leveraging AI is is on a very high risk of being replaced That’s
8:46
that’s my personal take Having said that how I see about this industry is uh for past couple of
8:52
decades what happened was there were larger software and because the cost of creation of the software was so high uh
9:01
it was difficult for every company to create their own software and everybody was leveraging a larger software Now
9:08
because the cost of creation has gone incrementally lower I believe every
9:14
company is going to have their own personalized software which cater to
9:20
just upper crust or just soft instead of us leveraging one umbrella which caters
9:25
to upper crust also soft also google So I my personal belief is that software
9:33
per person uh if there’s a unit that will go higher uh and hence we will need
9:40
more software developers But having said that uh I also believe that the amount of work
9:48
done by each software developer will also multiply over the couple of months
9:53
or couple of years if it has not already be So if I was writing say 1,000 lines
9:58
of code per day this would eventually go to 10,000 lines of code per day without
10:04
me spending 10x the amount of time I spend just leveraging AI But I also
10:09
believe that the amount of software I leverage will also go incrementally high So this will be a very uh demand supply
10:17
very interesting demand supply game uh but I believe that all software developers that get uh exposed to AI
10:25
will will have a sufficiently higher chances of surviving or creating newer
10:31
value addition society right so there is this adaptability basically so you know when
10:39
internet came in the computers came in those who are able to adapt the system
10:44
was the ones who survived so yes I think here also So they will be in the game of
10:50
adaptability Those who adapt the AI and take the leverage of it will be able to survive and grow definitely Uh so one
10:58
more thing uh so India is a hub of it basically information technology there
11:05
are lot of engineers there are lot of you know students like bright students
11:11
bright young minds uh who is into this speed So recently uh we saw deepseek
11:19
coming out from China There are multiple models chat GPT Google MIM9 from US
11:25
coming in and there are very much and just naming a few that is coming from US
11:30
Europe or I say China but uh something revolutionary uh we haven’t seen coming
11:36
out from India despite the pool of this talented youngsters why why I’m not able
11:42
to understand why India is lagging behind in this AI revolution
11:49
um Right Yeah In fact uh last month I was in conference with uh few few of the
11:56
key decision makers and uh few of the thought leaders in this space and uh our
12:02
conversation was around this that there are there are two ways to look at this one is let’s say be deep be it GPT
12:10
series be it plot series or Gemini all of these models why why is China able to
12:15
create a model and India is not able to uh this one bigger challenge is the capital intensive nature of this
12:22
application to train a decent model this is a potential millions of dollars of
12:28
investment which uh openai has been doing for quite some time in India we
12:33
have not seen that space yet u having said that uh this one uh famous celebrity entrepreneur who has go on
12:41
gone on to Twitter and asked that okay I am ready to fund creation of a new large language model so that’s one piece
12:48
second is as Indians uh I think we are very adaptive and in fact Indians are
12:55
relatively on the better side of leveraging this AI we not creating newer
13:01
models but on the leveraging AI side we’re doing a lot of work so in terms of Indian youth uh I think we would be
13:09
sooner or later we would be that population who has understood how to leverage these models fastest and
13:14
develop applications on top of it So uh yes model creation should also happen in
13:20
in India and which will eventually happen because the kind of language dynamics we have in the country nobody
13:26
else has So an external uh force would never be able to create an application
13:32
that suffices our need without actually using our uh youth But at the same time our it’s not
13:39
uh completely hopeless situation Our youth or our engineers are also being
13:45
rammed up to understand this technology and leverage them to deliver a other product value
13:52
I agree and I can you know relate this thing So in most of the things most of the sectors India might not be the first
13:59
mover but uh sooner you know Indians are you know very good at grasping or very
14:06
good at adapting the things So so yes I think this is going going to be same in
14:11
the game of AI So so now one more thing that I wanted to
14:17
understand So about the economy of India So if you see Indian economy is being a
14:23
service-driven economy We have shifted to a service-driven economy wherein most
14:29
of our revenues is coming from outsourcing of it So most of the uh US
14:34
and Europe companies they outsource their uh basically daily tasks and all
14:40
those things to Indian companies because they found it uh relatively uh cheaper and efficient also So now because of
14:48
this AI uh the rise of AI now that AI can do and automate a lot of things Do
14:55
you think uh AI will become sooner a option or alternate to India for these
15:02
US and Europe companies how do you see these things uh right Yes Uh very very interesting
15:10
insight actually group Uh it’s it’s a bit difficult question to answer in
15:16
terms of u evaluating if you should actually give out the truth or not Uh
15:22
because it might come as a very harsh wakeup call But uh I’ll try I’ll stay
15:28
true to the nature of my work So I’ll give you the truth honest answer uh is in lot of services that we provide be uh
15:36
call center services be bo and customer ticket services AI is going to uh change the
15:44
landscape and this landscape change would be extremely drastic in nature So
15:50
uh and I can give you an example of a client we work with a team of 40 odd folks uh setting the job was to make
15:58
phone calls and get some feedback about the services that they have already delivered This job a 40 member team uh
16:06
was replaced by AI agents that do the exactly same job They would make an uh
16:12
voice call They would get the feedback and a human agent would not be able to
16:17
make 10,000 calls today My my air does that It can take the feedback It can log
16:23
It can do sentimental analysis It will perform end to end process While for
16:28
human it’s difficult job They have to take a call They have to be on the call for let’s say three three and a half
16:33
minutes They lock the entries They have to learn English to make proper ticketing system or make proper entries
16:39
into my system So this entire cycle was replaced by agents Uh having said that
16:46
uh the disclaimer is here that those 40 odd folks who were not unemployed they were shifted to other departments where
16:52
they add a lot more value Um and this is not just one case that you see uh we we do see a lot of these cases where uh
17:00
jobs that do not require lot of human intelligence involved are are being
17:06
questioned or are being made lot more efficient with AI
17:12
Right And uh so that’s one sorry why yes
17:19
yes so that’s one piece where we’re talking about customer service agents and others u regarding other IT services
17:27
that we’ll leverage uh I think uh yes we might see a shortterm disruption in the market
17:35
because uh so potentially let’s say if India was creating a lack websites per
17:41
day Now lot of these might not be needed because everybody now knows how to create a website But uh in longer span
17:50
the newer ideas will be gone So things that were not possible 2 years back are
17:56
now possible now So newer tangents of uh softwares will come into picture which
18:02
will eventually need some workforce to uh get this job done Initially maybe lot
18:09
of US and European folks would be able to suffice the need but once the volume becomes higher uh this should come back
18:16
to India right so even earlier also we used to
18:22
get a lot of you know calls from these agencies but now everything is getting
18:27
replaced by AI so we get calls from AIS right uh so yes I agree a lot of things
18:32
are shaping up and in short term there may be a disruption and this industry which may cause a lot of people to lose
18:40
their jobs for a shorter period of time I agree eventually the things will get
18:45
settled down but uh so as you said this is a wakeup call for this industry and
18:51
not only for this industry I think AI will dis will cause disruption in many other industries So I wanted to
18:58
understand so there are many multiple sectors now that you are working in this industry so you have a better
19:03
understanding of that So uh which sectors so now in in even in the
19:09
healthcare sector I see a lot of you know new things new inventions coming in
19:14
So which sectors you think will get most disrupt disrupted many people will lose
19:19
their jobs and do and on the other side I think uh AI will also create new jobs
19:24
So there will be a need of new people as well So there are many sectors where there will be a creation of new jobs So
19:31
how you you just uh tell me like which on the one part which sectors will lose most of the jobs and getting most
19:37
disrupted and on the other uh part which you’ll get benefited from this So
19:44
yes sure uh so I’ll start with sectors I think or I as of now see lot of traction
19:51
in is uh one is healthcare for sure uh what has happened in healthcare is lot
19:58
uh not many transformative changes have come into healthcare sector for past couple of years uh as opposed to lot of
20:05
other industries that were that were being transformed every other year So for example it it has to keep changing
20:11
at a very rapid pace and if you look at the pace of healthcare uh it was not
20:16
exactly very uh far So that is one space where I see a lot of things changing uh
20:21
and in fact a lot of projects that we are doing uh are changing the way healthcare is being perceived uh be it
20:28
diagnostic where the reliance on doctor is extremely high or be it
20:34
post care uh where because the number of medical professionals are limited the
20:41
amount of care a patient should have got was uh was lesser than what they should
20:47
get So AI would enable us to give every
20:54
patient the amount of time they should have gotten by reducing uh be it
20:59
documentation jobs for the doctors or be it looking at the patient history identifying what is the current scenario
21:06
and then evaluating If an AI agent can do that for you the p uh the doctor saves say 15 minutes and doctor can
21:13
spend 15 more minutes with the patient to understand what they are feeling about Uh and of course we also see a lot
21:19
of transformative changes where we see uh jobs that can be made extremely
21:25
simple in healthcare where be documentational job be clinician diagnostic uh we or be predictive
21:32
analysis just giving suggestions to the doctor instead of making decisions as of
21:37
now Um another field where I see and uh my personal opinion is all of the fields
21:43
which were thriving because the knowledge was centric Uh let’s let’s
21:49
take an example of a lawyer A lawyer who has read thousands of cases and multiple
21:55
books This lawyer is now replaceable via AI because the information retrieval has
22:02
become extremely easy Now what is different uh or what stands as a
22:09
deciding factor is their experience their information uh their information
22:15
about how to tweak the system how to interact with the system but knowledge retrieval uh that piece is no longer the
22:23
most uh more piece for that job so that’s another uh legal segment is where I see
22:30
a lot of change going to happen healthcare is one legal is one Uh I also see finance I also see insurance or and
22:39
all of the industries that were typically paper and knowledge based will
22:45
see lot of transformation Uh I would not go on to say that they will see lots of jobs but uh I think the way they are
22:52
working today and the way they will be working 3 years from now would be drastically
22:59
And on the flip side where do you see like new jobs will be created there will be new opportunities
23:06
Um right so on this side I think one example is
23:12
uh and I am I’m just thinking about if I should give this example but uh if you
23:19
look at statistics one of the most uh one of the most promising product that came out of this chat GPT was AI
23:26
girlfriend where and the but if you look at it from the underlying uh phenomena
23:33
point of view what’s happening is AI is making onetoone connection extremely
23:38
easy It might be we crave human to human connection but what’s happening is that
23:44
people are actually uh using apps like AI girlfriend or boyfriend interacting
23:50
with them to get this uh onetoone homey feeling uh and essentially what we are
23:57
or where I’m coming to is I see a lot of increase in personalized services that
24:03
will grow significantly up so any form of services where you can which you were
24:11
until now catering to a segment Now if you cater to an individual that would go up I also see uh defense as a sector
24:20
where we’ll see a lot of newer things coming into picture Now if you look at uh Russia
24:26
Ukraine we or any other recent combat incidences we’ll
24:32
see AI being leveraged or technology being leveraged in a way which was never
24:37
been imagined before So be it you uh leveraging drone to identify where should they drop the artilleries or
24:45
identify which are the places where the terrorists are hiding or where the
24:50
opponent’s soldiers are hiding A lot of these things would change for defense is one personalized services are one is
24:56
where I believe lot of changes happen like I agree there’s AI in
25:03
alternate things a lot of people are using AIS to to get rid of their loneliness I talked to this on the later
25:10
part Uh first uh you spoke about a lot of knowledge jobs and operations jobs
25:16
getting replaced lot of paper documents and all those things So here comes a
25:21
reliability part So you know a lot of your legal documentation let’s say your
25:26
knowledge you are asking so you mentioned about the the legal sector right so you need some sort of
25:33
information uh but instead of you know going to your lawyer or someone you ask the chat GPD if I have done this you
25:40
tell me what will be the implications or etc But is it even a reliable because mo
25:46
like there are many cases uh where you know this orders speak false pause like
25:52
give you a pause information that do very confidently that this is the truth you should believe but that is not it So
25:58
this is uh misleading many people So do you think this is even reliable uh at
26:04
least for now if not now do you think in future that AI will be 100% reliable
26:09
that you can you know just trust on the AIS so how do you see this
26:15
right uh so AI making this falsified information or confidently speaking
26:22
madeup information is what we call hallucination in in technical do so uh
26:28
what’s uh what’s going to happen or as of now what I believe is that all of you
26:35
can’t make AI liable for anything So let’s say Chad GPD gives you a response
26:41
for a medical diagnosis or it gives you a legal documentation or it makes let’s
26:46
say as a CA uh instead of CA you you go to Chat GP ask it to create some amendments or some draft it does that
26:54
but you can’t hold AI liable for that So as of now uh how I see this is all of
27:00
this goes through a supervision of a human individual and this is what we
27:05
call human in the loop uh is basically leveraging these technologies passing
27:11
them via human because someone has to be uh liable for
27:17
the decision that has been made and it can’t be charged for now So uh the
27:23
liability will be on the person who has signed it and they would be the person
27:28
who’s uh responsible for leveraging AI Having said that it’s their duty to make
27:34
sure whatever they are informing is correct in nature U now coming to the
27:39
second part of your question is do I see AI being perfect at one point of time
27:45
it’s a little technical uh problem to understand because this what we call AI
27:52
is a network of billions of nodes and there’s lot of randomness to this So do
27:59
I see models being non-h hallucinating at certain point of time uh reduced hallucination yes we see from
28:07
GBD 3.5 to current 03’s version the hallucination had reduced drastically
28:14
but do I see this going zero in near future no And I may be wrong
28:20
Um but there are other ways we currently leverage AI to reduce the hallucination
28:28
uh just to give you an example let’s say you uh we create a legal draft from the
28:34
AI agent or chat gi this draft now when created I can have an a new newer AI
28:41
agent let’s say claude anthropics cloud or Google’s gemini and ask it to just
28:47
dissect this piece and make sure that everything it is writing is correct so uh instead of relying on one model we
28:55
use ensemble or collection of models to make sure that one is generating and
29:00
other one is criticizing it Right Right This is interesting
29:07
basically So yes uh and on the now uh the later thing the second thing that
29:13
you mentioned about the you know AI girl trick or this thing basically now there
29:19
are a lot of people you have you must have gone through a lot of studies that there you know more than 50% who are
29:25
facing this loneliness thing they are they feel lonely and just to get a company now there is there is something
29:32
called AI now on the Instagram as well what they have seen this AI agents multiple characters has building uh and
29:40
people just talk to them So nowadays I see uh people talking more to this AI
29:46
girlfriends than their actual bites basically So there is more information
29:52
to this AI agents uh AI models about you what you do what you eat what you do
29:58
where where you are going basically So everything they know about you so there’s this privacy thing So the this
30:05
privacy is now no more left because this models knows everything and you don’t
30:10
know what is going to happen in future how it will react AI models what what it will do so is it like this is it might
30:18
be a threat basically this is what I am assuming what how do you thread
30:26
um right in fact uh privacy in AI is one of those segments which has which is
30:33
currently under lot of research And uh there’s no perfect answer to this
30:38
uh as of now at least uh there are two open-ended questions that we see in AI One is e ethics and the other one is
30:45
privacy But uh regarding privacy what I can say is uh now let’s take instead of a personal
30:53
example where I am using an app to enact as or mimic as a real friend Instead of
30:58
that let’s take an example where I am taking the AI and making a lot of
31:04
business decisions on top of it Now given I’m exposing this business data this is the question I get to uh I have
31:12
to answer every other day to every potential partner that we interact with is how safe is my data and
31:19
uh with with the current providers that we have uh lot of things you can ensure
31:25
that my data will not be used for training But this again comes uh there are only I
31:31
guess there only a certain percentage of the population who’s informed enough to fill these
31:37
settings otherwise almost everyone is using GPD without uh privacy mode
31:43
enabled So that data is being used to train the models in general and it’s
31:49
only that because so much of interaction is happening with these models the models are becoming better their
31:55
understanding is becoming better of uh the population in
32:01
general but privacy yes it remains a bigger concern and uh I think the way to mitigate is by educating people on how
32:08
to turn off these features how to turn on this case like just to you know I will get an idea
32:17
So like as you said uh there are many people they don’t turn off their settings of privacy and there is a lot
32:24
of information being passed through these AI models and AI models train themselves through these kind of
32:30
information So there is a every day there is a there is a information that is flooding to this model If you if you
32:37
name big AI models there are multiple like like you know millions and billions
32:42
of users who are using these models on everyday basis uh and a lot of information going to that So how bad uh
32:50
this can go basically you know in terms of everything you say AI models training
32:56
themselves through all these datas how bad a situation can go because I see a lot of threat to this thing because if
33:02
we cannot uh you know take a control of these AI models which we obviously don’t have so how bad it can go and how what
33:10
what amount of threat it can give to humans
33:16
um right and in In fact uh even if we were not so there’s one that we that
33:24
interact with let’s say charg this information being used to train the model um that’s one piece u and it’s not
33:32
as of now it’s not being extremely personalized so let’s say what droo interacts with GPT is the same
33:39
information goes into the same stream where madul interacts with GPD and the same information is being used to let’s
33:44
say potentially train a model but those are not two different models That’s one different that’s one model that is
33:50
learning all of this But uh if I was to take a very pessimistic scenario and
33:56
analyze what’s happening uh as we initiated our conversation AI understands things that we don’t uh and
34:04
the capability of it to uh find
34:10
underlying patterns that are not very evident on the surfaces can be leveraged
34:16
in uh lot of ways that you and I potentially can’t talk about Let’s just
34:21
take this to other extreme where we give arms to this AI agent or weapons to this
34:29
AI agent and let and let’s assume that this with all of the information it has
34:34
with the entirety of web being understood by it with uh millions of
34:40
conversations being used to train this model and if somehow AI gets free access
34:49
to do whatever it want it potentially actually will uh create lot of chaos and even it
34:57
doesn’t have the feeling it might just be trying to optimize for a certain objective and there might be other
35:04
chaoses that can come into the system and this is where lot of uh western
35:10
countries are spending time on to understand where should we restrict AI u
35:16
should weaponized AI be allowed should AI decision making be allowed especially
35:22
in terms of let’s say hiring person firing person or making financial decision So uh lot of research is
35:28
happening on this front uh and pessimistic scenario Yeah this can be very uh science fictional movie where
35:34
everything gets operated by AI Having said that we humans are extremely smart
35:39
very good at putting in guard rates So let’s see who comes out ahead be putting
35:44
heart rates or letting air run in right because at the certain limit you
35:51
will have to you know put in put a restriction on that I don’t know how we
35:56
would do this but yes because right now AI is just into our mobile phones and
36:02
you know laptops and everything but I’m you know sooner there will be an AI robot uh in my house you know doing all
36:09
the things which I say you know grasping all the things about me my family and
36:15
everything So this can go worst if we don’t put a restriction of this I completely agree to this Uh or but last
36:24
thing I wanted to understand and this is uh something very serious which I think so AI is advancing at you know insane
36:32
speed uh but as we uh speak initially right so there are there is a lot of
36:37
capital which is you know there to train AI models and also there is a lot of uh
36:43
energy which we require to train these AI models right uh so now that we are
36:49
getting you know dependent overly dependent on AI models so sooner as we spoke there are multiple sectors which
36:56
so there are multiple jobs which will be replaced by these AI models and companies now sooner the economy
37:03
companies will also become dependent on AI to do the lot of stuffs So so as we
37:09
grow as we uh as we go further this dependence will keep on rising but uh
37:16
this will have that implication on this uh environmental side because you know
37:22
right now also there is lot of energy which this AI models is consuming and now at this scale how much energy that
37:29
they will be requiring right so what do you think because once we go in there is
37:34
no U-turn I’m assuming So right Yeah So yes
37:41
No no I I completely agree with your point uh that the energy energy consumption uh because of AI is is going
37:48
to go exponentially higher as we keep leveraging them Um and in fact uh this
37:54
is going to become a national concern sooner or later given the amount of
38:00
energy we consume for this AI models How do we how do we derive so much of energy
38:05
how do we come up with how do we get so much of energy that’s first piece and second is the ecological impact of uh
38:13
this energy consumption uh so yeah that’s that is then u this is how I
38:19
think the current problem is that we are ending up spending a lot of creating
38:24
major foot global footprint in terms of uh leveraging AI by wasting water or be
38:31
carbon emission this is happening but I think the hope here is that uh sooner or
38:37
later we’ll get lot more efficient ways to create energy or generate energy
38:42
because AI is going nowhere Uh so the other way is to how do we identify places where we can generate lot of
38:48
energy from Uh so now all of the giants will have to think in terms of how do we
38:55
generate own energy uh so infrastructure will become one of the challenges uh for
39:01
AI and uh apart from energy I think this is a very well-known problem about GPUs
39:07
that everybody needs this graphical processing units to train their AI models but hardware becomes a limitation
39:15
here because the number of CPUs that are available in the market and I foresee that similar to GPUs electricity will
39:23
also be a similar problem where how do you get so much of energy to train the models how do to get so much of energy
39:28
to you keep using those more
39:34
like so this AI game uh if we don’t know where we are heading towards but
39:39
obviously we are going somewhere so sooner we will get to know where we are going to ditch so all right mad it was
39:47
very lovely interacting with you I personally got a lot of insights to my questions which I definitely don’t
39:54
wanted to ask to b this So yes it was a quite an insightful session Nadu thank
40:00
you for joining in thank you for kicking off your time So yes thanks for inviting me Doo It was
40:07
fun session It was nice to interact with you and especially your questions coming
40:13
from all across the spectrum to identify where we can see the impact how how do
40:18
we leverage this what should be protected at this It was a refreshing conversation Thanks Bobby
