Artwork for podcast Tech Talk with Amit & Rinat
Impact of AI on Jobs
Episode 8113th March 2024 • Tech Talk with Amit & Rinat • Amit Sarkar & Rinat Malik
00:00:00 00:35:43

Share Episode

Shownotes

In this week's TechTalk, Amit and Rinat delve into the profound impact of AI on jobs. We discuss how AI is reshaping industries, leading to automation, and job displacement, but also creating new opportunities and roles. They provide insights into how individuals can adapt to this rapidly evolving landscape, emphasizing the importance of upskilling, reskilling, and embracing AI as a tool for innovation and growth. Tune in to learn more about navigating the intersection of AI and employment in the digital age.

Transcripts

Rinat Malik:

Hi everyone.

Rinat Malik:

Welcome to Tech Talk, a podcast where Amit and I talk about all things tech.

Rinat Malik:

We don't just talk about tech, we also talk about how the tech affects

Rinat Malik:

our society and all the other things related to everything tech.

Rinat Malik:

So today we are gonna talk about something that everyone's talking

Rinat Malik:

about impact of AI on jobs.

Rinat Malik:

with the last year 2023.

Rinat Malik:

We've seen a surge of many, many different types of AI being available to consumers

Rinat Malik:

to produce various types of content and help them augment their performance

Rinat Malik:

of anything that they want to do, and that obviously opened the discussion

Rinat Malik:

on how is it going to impact our jobs.

Rinat Malik:

Many people are thinking that what they do on a day-to-day

Rinat Malik:

basis will become obsolete.

Rinat Malik:

the optimists are thinking that there will be new jobs there are various

Rinat Malik:

weaknesses that with time we've started to figure out that okay, not everything

Rinat Malik:

can be replaced exactly with AI as we are seeing the AI scene evolve on a

Rinat Malik:

month by month and week by week basis.

Rinat Malik:

So yeah, there's many aspects to explore and talk about and, so it would be

Rinat Malik:

really fun topic to talk about with Amit.

Rinat Malik:

So what's your thought on this, Amit?

Amit Sarkar:

So I think it's a very relevant topic and, thanks,

Amit Sarkar:

Rinat for that introduction.

Amit Sarkar:

I think AI is quite relevant now and it's all around us.

Amit Sarkar:

A lot of tools are coming up and a lot of use cases are coming

Amit Sarkar:

up now with the advent of AI.

Amit Sarkar:

We talked about generative AI a couple of episodes back.

Amit Sarkar:

We talked about chat GPT, we talked about AR and VR.

Amit Sarkar:

So there are a lot of different aspects to this whole AI thing.

Amit Sarkar:

there is metaverse coming in, so there's a huge potential for this.

Amit Sarkar:

I think the surge of AI is predominantly because of the large language

Amit Sarkar:

models, LLMs, as they call it.

Amit Sarkar:

Machine learning has been there for a very long time, and we have been

Amit Sarkar:

using that for image recognition and mapping different images.

Amit Sarkar:

But I think with LLMs, it's just taken a whole new level with with

Amit Sarkar:

respect to AI reaching to the masses.

Amit Sarkar:

So what I feel is, and I've read this as a quote, is that AI will not replace you.

Amit Sarkar:

People using AI will replace you.

Amit Sarkar:

I.

Amit Sarkar:

So I think that's very apt.

Amit Sarkar:

So there are so many AI tools in the market and we ourselves have

Amit Sarkar:

been playing with a lot of AI tools.

Amit Sarkar:

And what we are trying to see is what these AI tools can do and what

Amit Sarkar:

can we do with these AI tools and what, what is the pricing like, what

Amit Sarkar:

can we do with the fee free tier?

Amit Sarkar:

What can we do in the paid tier?

Amit Sarkar:

So I am.

Amit Sarkar:

I'm currently paying for a tool called Descript on which

Amit Sarkar:

we are recording this podcast.

Amit Sarkar:

And it has this facility where it'll generate the transcript and we can

Amit Sarkar:

edit the video using transcript.

Amit Sarkar:

I think Rinat is using Chat GPT paid version, so there he can use GPT four

Amit Sarkar:

and then he can use images, he can use websites and everything to search

Amit Sarkar:

the internet and create content.

Amit Sarkar:

And we have used AI In generating transcripts, but also, like, I've used

Amit Sarkar:

it for a few articles for my newsletter.

Amit Sarkar:

I've created an AI influencer using AI.

Amit Sarkar:

And we have been using AI for the last one year, and it's really amazing.

Amit Sarkar:

So I think it's a very relevant topic and let's discuss further

Amit Sarkar:

about what it impacts on a day-to-Day life and how it affects our jobs.

Rinat Malik:

Yeah, very much so.

Rinat Malik:

when, when we talk about AI and jobs, this reminds me of a short video I've

Rinat Malik:

seen when chat GPT first came out at the end of Everyone was trying and,

Rinat Malik:

whoever was in whichever profession they were seeing how would chat GPT react

Rinat Malik:

or reply to that professional question.

Rinat Malik:

And there was a video.

Rinat Malik:

I remember a lawyer was doing a, a live in front of camera see what Chat

Rinat Malik:

GPT can do and it asks it to create a contract with particular various.

Rinat Malik:

legal terms, et cetera, and, some of those things as audiences who are not legally

Rinat Malik:

aware, we didn't understand those terms.

Rinat Malik:

But then Chat GPT wrote this contract, which actually exceeded the expectation

Rinat Malik:

of the lawyer who was doing it.

Rinat Malik:

And he was saying that not only he did what I asked it to do, it.

Rinat Malik:

Not only it did what I asked it to do but also it added some, relevant

Rinat Malik:

addendums and clause in that contract, which would be typical for that scenario.

Rinat Malik:

So now that's using common sense.

Rinat Malik:

it appears as though Chat GPT is using common sense and adding those things.

Rinat Malik:

And then, the, the next second he was very happy saying that, oh, my

Rinat Malik:

job would be really easy from now on.

Rinat Malik:

then the second afterwards, you see the realisation in his face and he

Rinat Malik:

realises that it's almost like he himself is not required anymore because

Rinat Malik:

he was a recent graduate from law school and he was like, okay, it can

Rinat Malik:

do everything I can do and better.

Rinat Malik:

So where, where does it put me?

Rinat Malik:

that realization is what people are having all throughout last year in

Rinat Malik:

2023, as they start using chat GPT more and more, as more people become aware.

Rinat Malik:

So even when you send.

Rinat Malik:

A very well written email, nowadays, people might not give you the

Rinat Malik:

credit or, or, the articulation that has gone into the email.

Rinat Malik:

They might actually not give you that credit for that

Rinat Malik:

because everyone's doing it.

Rinat Malik:

It's like a inflation of intelligence is what's happening right now

Rinat Malik:

and how does that affect jobs?

Rinat Malik:

It is affecting jobs, but I think what's happening is because of the hype of,

Rinat Malik:

oh, AI can do so many groundbreaking things, it's like day and night

Rinat Malik:

difference between 2022 and 2023.

Rinat Malik:

in terms of productivity.

Rinat Malik:

So companies also became quite hyped and thinking, there are

Rinat Malik:

so many ways we could use it.

Rinat Malik:

But actually on an enterprise scenario, there are many things to think about when

Rinat Malik:

you wanna implement a new technology.

Rinat Malik:

especially, I, I am primarily experienced in banking and finance,

Rinat Malik:

and there are regulatory requirements, there are various risks that you need

Rinat Malik:

to take into account and quantify how would you manage these risks?

Rinat Malik:

So for example, there has been talks that AI has been found to be biased in many

Rinat Malik:

ways, gender bias, race bias, et cetera.

Rinat Malik:

And what if a bank integrates AI to reply to a customer and these biases

Rinat Malik:

come through in that conversation, that would be a huge reputational risk for

Rinat Malik:

the bank or any organization as well as regulatory and compliance issue as

Rinat Malik:

well, because it appears as though that organization is treating people unfairly

Rinat Malik:

or differently based on their age, race, gender, So there are many, many risks.

Rinat Malik:

as useful as it may seem to make it work on an enterprise scenario

Rinat Malik:

is not just, okay, let's just buy subscription of Chat GPT, and we are set.

Rinat Malik:

it's not as easy as that.

Rinat Malik:

And sometimes you might even have to think about how can we dial down

Rinat Malik:

the intelligence, dial down the.

Rinat Malik:

bespoke responses, but always make sure it comes back with standardized

Rinat Malik:

set of responses as required, which is a bit of a dumbed down version

Rinat Malik:

of the solution people might think, but that is maybe what is required.

Rinat Malik:

So.

Rinat Malik:

just overnight transition of AI is gonna be doing my job from tomorrow is not

Rinat Malik:

realistic to think And also whether it can do everything to the degree that a

Rinat Malik:

human would is still very much questioned.

Rinat Malik:

So these are different perspective to think about when you think

Rinat Malik:

about the answer to this question.

Amit Sarkar:

Yeah, I think and you've raised a valid point about regulations

Amit Sarkar:

and regulatory requirements, but I think before we delve into job

Amit Sarkar:

security, I think our AI taking over.

Amit Sarkar:

I think I'll come back to the quote.

Amit Sarkar:

AI will not replace people.

Amit Sarkar:

People using AI will replace people.

Amit Sarkar:

So what that means, so it means that AI is augmenting you and it's augmenting

Amit Sarkar:

your productivity and augmenting anything that you're currently doing.

Amit Sarkar:

So suppose you take some time to write an email or suppose you take

Amit Sarkar:

some time to draft a, a contract.

Amit Sarkar:

So that now is augmented.

Amit Sarkar:

It means you can do the research far more quicker with AI.

Amit Sarkar:

Of course it's not a hundred percent accurate, but it

Amit Sarkar:

means you can do it quicker.

Amit Sarkar:

So it means you are already much better than someone who's doing it by himself,

Amit Sarkar:

just doing all the Google search.

Amit Sarkar:

so you are doing it much faster and you can create a early

Amit Sarkar:

draft using AI very easily.

Amit Sarkar:

So now, before we talk about the tools, we have to talk about these

Amit Sarkar:

large language models and what are the different aspects to current generative

Amit Sarkar:

AI which is actually generating different types of things for us.

Amit Sarkar:

So when we look at large language models, it is trained on text, and that text

Amit Sarkar:

is currently taken from the internet.

Amit Sarkar:

So it could be papers, it could be websites, it could be forums, it could

Amit Sarkar:

be video sites, et cetera, et cetera.

Amit Sarkar:

So, but large language models is predominantly, trained on text.

Amit Sarkar:

Now, this text data, it means that whatever is written on the internet,

Amit Sarkar:

say on Twitter or say on LinkedIn.

Amit Sarkar:

So it just captures it and trains itself on it.

Amit Sarkar:

So if there is bias on it, it means there is bias in the.

Amit Sarkar:

system.

Amit Sarkar:

So that means bias in training data translates to bias in the system itself.

Amit Sarkar:

So a lot of companies, what they try to do is they try to remove the bias.

Amit Sarkar:

They have humans to look at the training data, try to identify if

Amit Sarkar:

there is any bias, try to remove it.

Amit Sarkar:

But of course it's not possible because your dealing with billions

Amit Sarkar:

and billions of text lines of text.

Amit Sarkar:

Then we have images.

Amit Sarkar:

So we have mid journey, we have Dall-e, and they are trained

Amit Sarkar:

on different types of images.

Amit Sarkar:

So you have a big data set of images, and now you can generate new images,

Amit Sarkar:

especially for graphic designers, logo designers, website creators, et cetera.

Amit Sarkar:

So this is again, a.

Amit Sarkar:

Very different aspect now.

Amit Sarkar:

So suppose I want to create a thumbnail for my YouTube instead of creating

Amit Sarkar:

it myself I can ask AI to use certain images as input images and generate

Amit Sarkar:

a new image for my YouTube as a thumbnail or as a Instagram picture.

Amit Sarkar:

I.

Amit Sarkar:

The next thing is video.

Amit Sarkar:

So now there are tools that can generate videos for you.

Amit Sarkar:

So you give them a text prompt and it generates video.

Amit Sarkar:

And these videos are not highly realistic, but it can be highly stylized.

Amit Sarkar:

So suppose you want to create an animated video, but you are

Amit Sarkar:

not an expert in animation.

Amit Sarkar:

So you can create a text prompt, pay some money, and get these

Amit Sarkar:

videos created for you very quickly.

Amit Sarkar:

Without a lot of creative input.

Amit Sarkar:

So when the generative AI came, it was the age of democratizing creation.

Amit Sarkar:

So it means that anything that you want to create now or using ai, it's much

Amit Sarkar:

more faster, much better, and quicker.

Amit Sarkar:

But, but the thing is, it's not, going to replace you?

Amit Sarkar:

It's not at that stage yet because it is based on the training data.

Amit Sarkar:

So there is nothing new coming out of it because it is just

Amit Sarkar:

correlation from that data.

Amit Sarkar:

So if I ask it

Rinat Malik:

yeah, just, just a little bit of anecdote as you were saying.

Rinat Malik:

I just remembered something that AI falls over when it is

Rinat Malik:

fed AI generated training data.

Rinat Malik:

It just completely starts generating things that are not meaningful in any way.

Rinat Malik:

So if you keep feeding it data that's also generated by an ai,

Rinat Malik:

then it just can't work at all.

Rinat Malik:

So you might think that, from YouTube to everything is gonna be filled with

Rinat Malik:

AI generated content, but those content then can't generate new content anymore.

Rinat Malik:

So there has to be a steady influx of human generated content always in

Rinat Malik:

order for AI to work side by side.

Rinat Malik:

I.

Amit Sarkar:

So, that, that's a very good point.

Amit Sarkar:

Rinat, that AI being fed on data that's created by ai, so that's synthetic data.

Amit Sarkar:

So and a lot of companies are already thinking about how to create synthetic

Amit Sarkar:

data because you will run out of training data after a certain period of time.

Amit Sarkar:

If you have scanned all the internet over the last couple of years and that's

Amit Sarkar:

all you have, then you would have to create new data to create new models.

Amit Sarkar:

Of course, we are generating a lot of data.

Amit Sarkar:

But that may not be enough to train the new models.

Amit Sarkar:

So companies are looking on how to generate data by themselves so

Amit Sarkar:

that these models can be trained.

Amit Sarkar:

But of course, we'll have to be very careful about what there

Amit Sarkar:

is, what is there in the data.

Amit Sarkar:

But the whole idea of talking about this is that it's not

Amit Sarkar:

going to create something new.

Amit Sarkar:

It is going to take an input of what is already there and then

Amit Sarkar:

create something, a mixture of it.

Amit Sarkar:

So suppose I want to create a test policy.

Amit Sarkar:

I ask the test policy, create a test policy for me.

Amit Sarkar:

So it's not going to create a new test policy for me.

Amit Sarkar:

It's going to scan the, it's already scanned the internet.

Amit Sarkar:

It has got some of the data, and then it's just going to blabber out

Amit Sarkar:

some of the things that are already there in standard test policies,

Amit Sarkar:

that data is read on the internet.

Amit Sarkar:

So that is what it is.

Amit Sarkar:

But how it makes me better or more productive, say to my colleague is

Amit Sarkar:

that now I can create a test policy in few seconds, whereas my colleague

Amit Sarkar:

will have to think about it and draft a test policy and and like do a lot

Amit Sarkar:

of research in order to create one.

Amit Sarkar:

But for me, I asked Chat GPT, it created a test policy.

Amit Sarkar:

I just edited it.

Amit Sarkar:

So, that's the difference.

Rinat Malik:

Absolutely.

Rinat Malik:

And yeah, editing something rather than creating something from scratch

Rinat Malik:

is a lot quicker and You can focus on enhancing the quality of the final

Rinat Malik:

output rather than just creating the very crude skeleton part.

Rinat Malik:

'cause that takes a lot of time, but that doesn't really enhance

Rinat Malik:

the quality that just, the basic building blocks needs to be there.

Rinat Malik:

Time consuming.

Rinat Malik:

And now that part is taken care of.

Rinat Malik:

And now you can focus on more creative stuff how you can expand on the already

Rinat Malik:

created product or content, et cetera.

Rinat Malik:

And this reminds me of another interesting not study, but

Rinat Malik:

phenomena that we've seen in this.

Rinat Malik:

This is something I've seen in 2019 or 2020 before chat GPT or this

Rinat Malik:

popularized AI systems were in play.

Rinat Malik:

But we were, we've been working with AI way before that.

Rinat Malik:

the different types of ai, which are pattern identifying AI and then there was

Rinat Malik:

in medical sector I was working with an AI which would identify, look at MRI scans

Rinat Malik:

thousands and thousands, billions of MRI scans to, to train it and then identify

Rinat Malik:

whatever is wrong within those scans.

Rinat Malik:

the percentage of accuracy when after the AI model was matured with training data,

Rinat Malik:

It always Shifted between 88% to 93%.

Rinat Malik:

And at the same time, a very experienced medical professional doctor or a surgeon,

Rinat Malik:

their accuracy was also about at the same level, 90 to 95% or, or something.

Rinat Malik:

There is an overlap and, but they're pretty much on par.

Rinat Malik:

But where the difference came is if, medical professional worked together

Rinat Malik:

with the AI generated solution, then the accuracy shot up to 98%.

Rinat Malik:

So basically going back to your quote again you're not gonna be replaced by AI.

Rinat Malik:

You're gonna be replaced by people using AI because.

Rinat Malik:

That's the most lethal combination as well, because highest accuracy

Rinat Malik:

highest level of creativity and the highest level of speed you can

Rinat Malik:

also achieve in terms of generating whatever content or product or whatever

Rinat Malik:

output it is that you're looking for.

Rinat Malik:

So yeah, augmenting human performance is is what AI does

Rinat Malik:

best at this point in time.

Rinat Malik:

Obviously in future, there is talks about artificial general intelligence, although

Rinat Malik:

that is a vastly different concept.

Rinat Malik:

people do mistake the large language models as if it's thinking

Rinat Malik:

and understanding, but it's not.

Rinat Malik:

So I think we are still a bit far away from that, but you never know.

Amit Sarkar:

what I have noticed from using charge GPT over the

Amit Sarkar:

last couple of months is that it's not a hundred percent accurate and

Amit Sarkar:

sometimes it starts repeating itself.

Amit Sarkar:

So if you ask specific things, it can respond.

Amit Sarkar:

I.

Amit Sarkar:

But after you delve into it with much more questioning, then it starts failing.

Amit Sarkar:

And that's where you see the limitations of a system like this because it tries

Amit Sarkar:

to repeat the same thing and it's just trying to predict what the next word is.

Amit Sarkar:

it's not actually having any intelligence.

Amit Sarkar:

It's just trying to see that, okay, what's the next word?

Amit Sarkar:

What is the probability of that next word?

Amit Sarkar:

And I try to fit it to make sense of it, and that's how it works.

Rinat Malik:

I.

Rinat Malik:

Absolutely, and that's, that's another interesting thing

Rinat Malik:

that happens if you ask it.

Rinat Malik:

What's two plus two?

Rinat Malik:

It will very confidently answer four and it'll be right, but it doesn't know what

Rinat Malik:

two is and what plus is and what two is.

Rinat Malik:

It just knows.

Rinat Malik:

It just went through a lot of training data where it's seen two

Rinat Malik:

plus two is four, but if you ask it a random but a very simple arithmetic.

Rinat Malik:

An addition arithmetic, but two large random numbers, maybe 17,542

Rinat Malik:

plus 37,000 something something.

Rinat Malik:

It very likely will completely get it wrong because that's, it's very

Rinat Malik:

unlikely that it was fed on that exact training data multiple times.

Rinat Malik:

So it doesn't have anything to refer back to or, tell you what

Rinat Malik:

would be the next next word for it.

Rinat Malik:

So, that's how you can see where it breaks down.

Rinat Malik:

I think.

Rinat Malik:

after using chat, GPT and other LLMs, like Claude and Google Bard.

Rinat Malik:

And a few other ones as well.

Rinat Malik:

Much smaller ones I think I would come to a conclusion that even though it

Rinat Malik:

sounds like it's an overnight change from what we had before and after Chat

Rinat Malik:

GPT, but it's actually a really big step forward from google search engine

Rinat Malik:

because in 2022, we were already asking in in the Google search

Rinat Malik:

engine that, okay, how do I do this?

Rinat Malik:

And then it will find the answer and it will also try and, generate

Rinat Malik:

some question and answer for you from the result already.

Rinat Malik:

So it was doing that already Google.

Rinat Malik:

what is the job of a search engine is?

Rinat Malik:

It's basically by the time we reached 2022 or 2020, we had vast amount

Rinat Malik:

of information on the internet and.

Rinat Malik:

We needed a way to categorize it so that we can find the

Rinat Malik:

information when we need it.

Rinat Malik:

The, very particular information from that sea of everything.

Rinat Malik:

We just needed the answer to this one question that how do I, solve the blue

Rinat Malik:

screen issue when my BIOS is not working or some something when you're working.

Rinat Malik:

And then Google would find you that answer from all the things that he had indexed.

Amit Sarkar:

No, no.

Amit Sarkar:

Google will not find you the answer.

Amit Sarkar:

Google will give you a set of websites that it thinks has the

Amit Sarkar:

and Chat GPT gives you the answer because it doesn't give you a set of links.

Rinat Malik:

Yes.

Rinat Malik:

that's how it is to the user, right?

Rinat Malik:

It's appearing like that.

Rinat Malik:

But ultimately, what's happening is Google has indexed all of internet.

Rinat Malik:

But Chat GPT was, it doesn't index, it was trained on, on patterns of data,

Rinat Malik:

but those data is, is, is still there and it's basically bringing you up the

Rinat Malik:

right data and in a way that you don't have to click five links and, read

Rinat Malik:

five conversation between human beings.

Rinat Malik:

It just bringing you the specific information you need.

Rinat Malik:

And presenting it to you in a way that is most useful to you.

Rinat Malik:

it just a, a direct one-to-one conversation.

Rinat Malik:

So yeah, it is an overnight real big step forward.

Rinat Malik:

But if you think about it, it's a really big step forward from

Rinat Malik:

search engines to what it is now.

Rinat Malik:

obviously we can talk about the image generation and all of the

Rinat Malik:

other different content types after.

Rinat Malik:

But in terms of text generation, you're searching for something.

Rinat Malik:

Nowadays, most of the ways public are using chat GPT, if they have a

Rinat Malik:

question, they say, oh, give me the answer, and it gives you the answer.

Amit Sarkar:

Yeah, I think it's a valid point that, okay, the search

Amit Sarkar:

engines were used and find you the answer and charge strain on the data.

Amit Sarkar:

But if you look at say an example where I want to create a travel

Amit Sarkar:

plan for my family to go to South Africa on a holiday in November, I.

Amit Sarkar:

So, so if I go to Google search and I say, okay I want to have a travel itinerary

Amit Sarkar:

for a trip like this in November to South Africa, it'll gimme a lot of links.

Amit Sarkar:

I'll have to look at the different flight options.

Amit Sarkar:

I'll have to look at the hotel options.

Amit Sarkar:

I'll have to look at what places I want to go.

Amit Sarkar:

I'll have to look at how many days I have to spend in each place, et cetera.

Amit Sarkar:

If I ask the same question in charge GPT giving all the necessary details,

Amit Sarkar:

charge GPT is going to gimme an itinerary day by day itinerary.

Amit Sarkar:

It's going to break it down.

Amit Sarkar:

It's going to tell me that you take a flight from here to here.

Amit Sarkar:

And depending upon the weather in November, it'll tell me, okay, go

Amit Sarkar:

to this city or that city and do this sighting and that sight seeing.

Amit Sarkar:

So it's already done a lot of research for me, and now I just have to do those things

Amit Sarkar:

like book a hotel, look at the itinerary.

Amit Sarkar:

And it's not like I will plan my whole trip around the itinerary, but most

Amit Sarkar:

of the research has already been done.

Amit Sarkar:

And now if you take that into a job aspect, because I think we, we have, we

Amit Sarkar:

are, we have digressed a lot from the job.

Amit Sarkar:

So in a job a lot of people they have chat.

Amit Sarkar:

We chat a lot, say on Slack.

Amit Sarkar:

Then we send a lot of emails.

Amit Sarkar:

So email is where you do a lot of your work because you have to

Amit Sarkar:

communicate about different things.

Amit Sarkar:

Then you use some a project management tool where you'll

Amit Sarkar:

have to manage different tasks.

Amit Sarkar:

Or a task management tool, any, any task.

Amit Sarkar:

It could be task of development, task of testing, task of managing

Amit Sarkar:

projects, resources, budget, et cetera.

Amit Sarkar:

And then you would also try to do some presentations.

Amit Sarkar:

You will have to draft some contracts, et cetera.

Amit Sarkar:

Now, for all these use cases what I've, I've noticed is that chat gpt is very

Amit Sarkar:

good at giving you an initial draft.

Amit Sarkar:

So suppose I want to send out an email about about a new proposal.

Amit Sarkar:

Okay.

Amit Sarkar:

And I want to tell my boss that, okay, I think that we should be proposing this

Amit Sarkar:

company about our products like this.

Amit Sarkar:

And it's something there in my head, but I don't know how to articulate it.

Amit Sarkar:

And I'll have to do some Google search.

Amit Sarkar:

I'll have to see what others are thinking about this same product and

Amit Sarkar:

how do you promote things, et cetera.

Amit Sarkar:

Now, if I ask Chat GPT and I give it context, like, okay, I'm a person

Amit Sarkar:

who's a manager at this company that specializes in this sector.

Amit Sarkar:

And I'm now trying to draft an email to my boss, and we want to write a proposal

Amit Sarkar:

and we want to talk about this product.

Amit Sarkar:

Now give me an output and Chat GPT will give you a very detailed

Amit Sarkar:

output because you've defined the context, you've defined the scenario.

Amit Sarkar:

And Chat GPT will give you a very good output.

Amit Sarkar:

And on that output you can edit things, tweak things replace few things that you

Amit Sarkar:

don't like, and then just send it and also it can, write in a particular tone.

Amit Sarkar:

So it can be a very formal tone, it can be a very informal tone, it can

Amit Sarkar:

be a very casual tone, et cetera.

Amit Sarkar:

So that way you have the hard work is done.

Amit Sarkar:

So now you just have to focus on the final output, tweaking

Amit Sarkar:

it and sending it to your boss.

Amit Sarkar:

So, imagine.

Amit Sarkar:

A person who's doing it from scratch and someone who's using AI to do it.

Amit Sarkar:

So the person doing from scratch would take about 10 to 15 minutes

Amit Sarkar:

or maybe 20 to 20 minutes or 30 minutes to do the research, draft

Amit Sarkar:

an email with you, augmenting AI.

Amit Sarkar:

You write the prompt, it takes you about two minutes.

Amit Sarkar:

Then you get the output, you tweak it, edit it 15 minutes

Amit Sarkar:

max, so you have half the time.

Amit Sarkar:

So from 30 minutes you've come down to 15 minutes, and maybe you can do it even

Amit Sarkar:

in five minutes if you are good at it.

Amit Sarkar:

So most of Chat GPT now is about prompts.

Amit Sarkar:

what is the quality of your prompt and how can you get the best output

Amit Sarkar:

in the, in the least amount of inputs?

Amit Sarkar:

So suppose I want to get an output for a travel itinerary or an email.

Amit Sarkar:

I want it to give me the best possible answer in, in one or two prompts.

Amit Sarkar:

I don't want to write 10 different prompts.

Amit Sarkar:

And asking it for 10 different outputs, and then I collate all the

Amit Sarkar:

information and then I create something.

Amit Sarkar:

I won't have one single prompt and then one quick great

Amit Sarkar:

output, and then just use that.

Amit Sarkar:

So a lot of this is actually about writing good prompts.

Amit Sarkar:

So you get good outputs in the shortest amount of time because

Amit Sarkar:

you have to pay for these tools.

Amit Sarkar:

These are not free tools and a lot of companies, what they're now thinking

Amit Sarkar:

is, can we create an LLM for ourselves?

Amit Sarkar:

Can we use our company data, our product data, and train an AI

Amit Sarkar:

just for our specific things?

Amit Sarkar:

Because the thing is, the moment you start talking to chat GPT,

Amit Sarkar:

you are sending confidential information to Chat GPT servers.

Amit Sarkar:

Now that means that suppose you tweak in some information about your

Amit Sarkar:

company, some stock prices or anything, it's gone in the training data.

Amit Sarkar:

Of course, they won't be able to identify exactly but that data is now

Amit Sarkar:

in the training data because whatever you talk to them, it's using it to

Amit Sarkar:

train itself and getting better at it.

Amit Sarkar:

So that's the downside.

Rinat Malik:

Absolutely.

Rinat Malik:

it's a downside.

Rinat Malik:

But having known that this is what happens, you could protect yourself

Rinat Malik:

from it and see the power of it.

Rinat Malik:

So, the first scenario when you were mentioning that, yeah, you wanna

Rinat Malik:

draft an email and all of these.

Rinat Malik:

So if, imagine if the AI model had all the information.

Rinat Malik:

all, all the internal company information and for large companies with a hundred k

Rinat Malik:

employees, that's a lot of information in Slack, in teams, in all of these channels.

Rinat Malik:

And also knowledge bases.

Rinat Malik:

SharePoint is like, a slew of information for data protection, if

Rinat Malik:

you can have like a internal AI model.

Rinat Malik:

which has the base of understanding and responding to you like chat

Rinat Malik:

GPT, but then it's been trained on all the internal data.

Rinat Malik:

Then you could ask it, write

Rinat Malik:

this report.

Rinat Malik:

And then as a human, you would check that, that that report is actually fit for

Rinat Malik:

purpose and it quantifies the exact metric that you are looking for or your boss is

Rinat Malik:

looking for or your team is looking for.

Rinat Malik:

So you can then tailor make it but then the, generating the base

Rinat Malik:

report part is done in seconds

Amit Sarkar:

Yeah.

Rinat Malik:

and that's what makes chat GPT or AI models so much more powerful.

Rinat Malik:

In a way that it increases your productivity, but it does not

Rinat Malik:

make your human judgment obsolete.

Rinat Malik:

Not yet anyway.

Amit Sarkar:

And I think one of the things that I'm, I'm seeing even in my

Amit Sarkar:

own company is that a lot of effort is now being put into designing your own

Amit Sarkar:

LLMs or designing your own AI model.

Amit Sarkar:

The reason being that if you use Chat GPT you have to first pay for it.

Amit Sarkar:

You don't have control over the data.

Amit Sarkar:

You don't have control over the output, and the training data is quite vast,

Amit Sarkar:

and you want your training data to be very specific just for your industry.

Amit Sarkar:

So you don't want to want a AI model that uses the whole internet

Amit Sarkar:

as its training data base.

Amit Sarkar:

You just want the specific information needed for your sector,

Amit Sarkar:

for your industry, for your product, and train your model on that.

Amit Sarkar:

That's it.

Amit Sarkar:

You don't want, so suppose you're in the retail sector.

Amit Sarkar:

I don't want information about what do you say?

Amit Sarkar:

A travel company.

Amit Sarkar:

Okay.

Amit Sarkar:

I'm just giving a wrong example.

Amit Sarkar:

It can be related, but say defense company.

Amit Sarkar:

So retail has got no relation with defense.

Amit Sarkar:

We are not selling anything to defense companies, so I don't want that data

Amit Sarkar:

to be correlated or I don't want that unnecessary data or if I'm just

Amit Sarkar:

creating on certain types of images.

Amit Sarkar:

I don't want a different type of images.

Amit Sarkar:

So I'm, I've just given you example of email.

Amit Sarkar:

The other example could be logo generation.

Amit Sarkar:

So suppose there is a person who, who designs new logos or who designs new

Amit Sarkar:

new, new art for artwork, say book covers or podcast covers, et cetera.

Amit Sarkar:

Now that person's job can be done now much more quicker.

Amit Sarkar:

So you are not going to replace that person.

Amit Sarkar:

But imagine that person who was handling now 10 clients can handle

Amit Sarkar:

a hundred clients because that.

Amit Sarkar:

Can use AI to augment his or her productivity.

Amit Sarkar:

And that's just one example of logo generation.

Amit Sarkar:

The other example is

Rinat Malik:

So j just to also intercept a little bit on this example,

Rinat Malik:

'cause this is very interesting.

Rinat Malik:

So as you mentioned that the person who could generate 10 artworks before

Rinat Malik:

now can generate a hundred artworks.

Rinat Malik:

And now the immediate thought to our audience or anyone who's thinking might

Rinat Malik:

be that, okay, so now there was only ever a thousand logos to generate.

Rinat Malik:

Now, before a hundred people would be doing those thousand logo designs.

Rinat Malik:

Now only 10 people can do all the market, meet all the market needs, so there is

Rinat Malik:

no jobs for the rest of the 90 people.

Rinat Malik:

And that is gonna be a situation for the time being.

Rinat Malik:

if you put it on a graph, yeah, there is a dip.

Rinat Malik:

But what happens is now there is an influx more of a lot of other people who wanted

Rinat Malik:

to have logo designed but couldn't, or they wanted to have their own startup

Rinat Malik:

business idea, et cetera, couldn't.

Rinat Malik:

But now with all of these additional tools, they are now thinking, okay, let

Rinat Malik:

me start a side project as well as my job.

Rinat Malik:

Everyone's doing that.

Rinat Malik:

So now there is.

Rinat Malik:

When there was a thousand logo to be designed before now there are

Rinat Malik:

10,000 or a hundred thousand people are now asking for new logo designs.

Rinat Malik:

So after that initial dip, there will definitely be a dip

Rinat Malik:

where people will think, oh no 90 people are losing their job.

Rinat Malik:

But very soon after, there will be so much more demand because logo generation

Rinat Malik:

is now a lot easier and a lot cheaper.

Rinat Malik:

As a result, a lot more people would be interested in doing that.

Rinat Malik:

And then you would need a lot more people doing a hundred logos.

Rinat Malik:

So overall, The amount of content being generated is increased.

Rinat Malik:

The quality of content being generated is increased, but it then goes

Rinat Malik:

back to the same amount of people doing it just more efficiently.

Rinat Malik:

So that's how, it's in my mind, no one's losing their job.

Rinat Malik:

It's just that they're becoming more efficient and the demand is increasing.

Amit Sarkar:

and I think that's, that's a very good point.

Amit Sarkar:

Like the dip, there will be a dip, but it means that there will be

Amit Sarkar:

a new baseline of productivity.

Amit Sarkar:

So initially your baseline was a hundred images, now the

Amit Sarkar:

baseline is a thousand images.

Amit Sarkar:

So there.

Amit Sarkar:

The baseline will be changing.

Amit Sarkar:

It's not that, and, and jobs will some people will lose their jobs because their

Amit Sarkar:

baseline productivity is not matching the current standards because the current

Amit Sarkar:

standard is now you should be able to do a hundred images so a thousand images,

Amit Sarkar:

but you're able to do only 500 images.

Rinat Malik:

So that's interesting.

Rinat Malik:

So this is where we go back to your quote, that some people will be losing

Rinat Malik:

their jobs who did not start using AI.

Rinat Malik:

So people who use AI will replace people

Amit Sarkar:

who don't use AI.

Amit Sarkar:

And I think if, if you talk about all these things, we are

Amit Sarkar:

just talking about generation.

Amit Sarkar:

this generative AI is all about generating AI images, AI text, AI videos, AI sound.

Amit Sarkar:

I think the next big problem for any enterprise would

Amit Sarkar:

be, are we violating GDPR?

Amit Sarkar:

Are we violating any copyrights?

Amit Sarkar:

Because see, imagine this this training data is being used to generate

Amit Sarkar:

new text, but that text is already there and it might have copyright.

Amit Sarkar:

So suppose I want to write a book and I write a book on something very technical,

Amit Sarkar:

it is going to gimme something, but that something might already been copyrighted.

Amit Sarkar:

And if I try to use that, someone can throw a legal case at me and

Amit Sarkar:

say that, okay guys, you have used copyright text for your book.

Amit Sarkar:

That's not valid.

Amit Sarkar:

So be smart.

Amit Sarkar:

Don't just use whatever the training data is throwing at you.

Amit Sarkar:

Whatever the AI model is throwing at you, be smart and edit it.

Amit Sarkar:

So when it come to images, you would also see that animation or cartoonish

Amit Sarkar:

characters are far more better.

Amit Sarkar:

But if you try to generate real human beings, then you

Amit Sarkar:

would have a lot of issues.

Amit Sarkar:

Like arms are missing.

Amit Sarkar:

the proportion is not right limbs are missing, et cetera.

Amit Sarkar:

So I, and I have noticed this because I've created an AI influencer on Instagram,

Amit Sarkar:

and I have noticed this while generating the images that some images are morphed,

Amit Sarkar:

some, some images are not correct.

Amit Sarkar:

So even though you're using AI to generate these images, it's not a hundred percent.

Amit Sarkar:

So you might have to do some modifications.

Amit Sarkar:

And then same goes for sound.

Amit Sarkar:

you can generate a lot of new music.

Amit Sarkar:

you can train your model for, say, music inspired by Taylor Swift or say Dua Lipa.

Amit Sarkar:

But then you have to ensure that you're not violating any copyright

Amit Sarkar:

laws, because even in workplace, you can do whatever you want as long as

Amit Sarkar:

you're not violating copyright laws and you're not violating any GDPR laws.

Amit Sarkar:

Because if you are using confidential information off some other people,

Amit Sarkar:

then that just violates all these things and, your job will be at

Amit Sarkar:

risk, so be smart about it as well.

Rinat Malik:

Absolutely I think we've gone through many examples

Rinat Malik:

of how to use AI to augment your performance and how it impacts the

Rinat Malik:

society in terms of job availability.

Rinat Malik:

the main point of our talk today was that, it's not completely unfounded

Rinat Malik:

that there is a shift of how useful you can be in what way, to a company

Rinat Malik:

to keep your job, but then in no way is it a like a done deal where this is it.

Rinat Malik:

it's just an opportunity for us to upskill ourselves more, to make our

Rinat Malik:

jobs more interesting and more creative and expand upon more from the basic

Rinat Malik:

output that we can now easily generate.

Rinat Malik:

Now let's take this as an opportunity and rise to towards our new challenge

Rinat Malik:

and produce more, produce better, produce faster whatever it is that

Rinat Malik:

we are producing in whichever job.

Amit Sarkar:

be smart about using AI and definitely start using the AI

Amit Sarkar:

tools and start gauging what are the different tools in the market, how

Amit Sarkar:

you can use it and what benefit it is.

Amit Sarkar:

I think a lot of tools will surprise you initially, and once you

Amit Sarkar:

cross the o wow, this is magical.

Amit Sarkar:

I think once you cross that threshold, then you'll realize the limitations.

Amit Sarkar:

It's only when you start using certain tools in certain ways

Amit Sarkar:

that you realize the limitations.

Amit Sarkar:

But if you don't use it, you'll never realize it.

Amit Sarkar:

You'll always think it's magic if you use a tool once you think it's magic.

Amit Sarkar:

But the more you use the tool, you think it's not magic,

Amit Sarkar:

there is algorithm behind it.

Amit Sarkar:

you can tweak the prompt to make sure that the algorithm gives

Amit Sarkar:

you the output that you desire.

Amit Sarkar:

So you can do all that, but in the end, it's not magic.

Amit Sarkar:

It's just mathematics.

Amit Sarkar:

I've used a lot of tools and I've, I've attended a lot of sessions on

Amit Sarkar:

different AI tools and I can say that these tools are not that intelligent.

Amit Sarkar:

They can do something specific things really well.

Amit Sarkar:

So if you are involved in something very specific, like you

Amit Sarkar:

have to do the same thing every day, say you're just generating

Amit Sarkar:

logos, your job might be at risk.

Amit Sarkar:

In case you start using AI tools to generate more images.

Amit Sarkar:

So if you do a very specific task, your job might be at risk.

Rinat Malik:

specific repetitive job was at risk, even before as

Rinat Malik:

well because of automation anyway.

Amit Sarkar:

But now with ai, it's the creative aspect as well.

Amit Sarkar:

So suppose you write contracts and you write only specific

Amit Sarkar:

contracts, then it's at risk.

Amit Sarkar:

But if you write specialized contracts, which changes based

Amit Sarkar:

on need, then it's not at risk.

Amit Sarkar:

Tailored made, So I I think a lot of the AI tools in the market

Amit Sarkar:

now are catering to personalized experience, personalized solutions,

Amit Sarkar:

that's where the market is heading, and it's quite an exciting future.

Amit Sarkar:

if we all embrace AI rather than looking at it as a threat.

Amit Sarkar:

if you look at it as a, as a tool to augment us, think it's a,

Amit Sarkar:

it's a huge benefit for all us.

Rinat Malik:

I would echo what you said, Amit, that definitely use all the

Rinat Malik:

different tools that are out out there.

Rinat Malik:

Most of them have a free tier rather than the paid subscription.

Rinat Malik:

And you can see all the different ways AI can be used

Rinat Malik:

So it will open your eyes on, the limitation of it.

Rinat Malik:

And I have spoken to many people about this as well, and I'm surprised,

Rinat Malik:

obviously the initial hype of chat GPT was, like so many users faster

Rinat Malik:

than Facebook and everyone, but then.

Rinat Malik:

Not a lot of people uses tools like chat GPT as much as they could.

Rinat Malik:

Right?

Rinat Malik:

they've gone back to Yeah, the hype is finished.

Rinat Malik:

We do use it sometimes, but a lot of the times we don't even think about.

Rinat Malik:

even if you wanna reply to a very a sensitive text on WhatsApp my wife was

Rinat Malik:

having to do that and I said, Hey, why don't you use chat GPT on your phone?

Rinat Malik:

There is an app You can just literally copy and paste the text you received and.

Rinat Malik:

get, get your response written down in the correct tone that you want to so you don't

Rinat Malik:

have to overthink things and all of that.

Rinat Malik:

So many ways you can use it.

Rinat Malik:

Definitely would encourage everyone to explore, take some time every

Rinat Malik:

week to see, like, put, put yourself down for one evening and just search.

Rinat Malik:

On Google on different AI solutions.

Rinat Malik:

There is a website which I don't remember the exact name, but it's

Rinat Malik:

called, there is an AI for that.

Rinat Malik:

And there's thousands of different AI tools all cataloged over there.

Rinat Malik:

So you can see all the things that are possible.

Rinat Malik:

So definitely explore, know how to prompt in a way that you get predictable

Rinat Malik:

output and that way you can use it to your advantage and go a lot forward, a

Rinat Malik:

lot faster and make a big difference.

Rinat Malik:

So with that thought yeah we had a really nice conversation.

Rinat Malik:

Please do reach out if you guys want us to go through any particular AI tool and

Rinat Malik:

we would be happy to do that as well.

Amit Sarkar:

Thanks so much everyone.

Amit Sarkar:

See you next week.

Amit Sarkar:

Bye.

Rinat Malik:

Bye.

Chapters

Video

More from YouTube