Hello and welcome back to the ILO's Future of Work podcast.
I'm Sophy Fisher.
This programme is where we explore some of the key trends
that are shaping the world of work.
Few of those trends are as hot right now
or as hotly debated as the effects of artificial intelligence or AI.
What's brought this into the headlines is the launch of ChatGPT.
ChatGPT is a natural language processing tool driven by AI technology.
Two things seem to make it particularly interesting or scary,
depending upon your point of view.
The first is how well it seems to be able to mimic human language.
It can answer questions or write text or code.
The second is how quickly it has been adopted.
Within two months of launch,
it had attracted more than 100 million users.
It's clear from this that the impact of artificial intelligence
on the world of work is not something we need to plan for in the future.
It's here and it's now, and we need to deal with it.
To discuss how the world of work might do this,
I have with me two distinguished guests.
The first is Antonio Casilli,
who is a professor of sociology at the Polytechnic Institute of Paris
and is co-director of the DiPLab Research Group.
-Welcome, Antonio. -Thank you.
Joining us online is Stacie Haller.
Stacie is in Boston.
She is chief career advisor at resumebuilder.com,
and she has more than 30 years of experience in staffing, recruiting,
and career counseling.
Stacie, welcome to you too.
-Thank you so much. -Okay.
Let me start by putting the same question to both of you.
Massive amounts of hype around ChatGPT.
ChatGPT is one form of artificial intelligence.
We are told this time it is going to revolutionize the world of work.
Is this true or is this hype? Let me start with Antonio.
Well, first of all, thank you for not pulling the number of 'this intro
was written by ChatGPT.'
A lot of journalists tend to do it.
-Yes, this was written by a human. -I really appreciate it.
Well, anyways, my point is that basically there
is a lot of hype as you pointed out.
This hype is actually
produced or prompted by the very persons who produce ChatGPT.
OpenAI is a company that has to sell a product,
and this product is, of course,
destined to be used in several industries:
in training, in consulting, in software development.
Of course, they are interested in us
believing that it's going to revolutionize these sectors.
It's not a surprise because the OpenAI researchers have published
recently a paper that estimates the number
of jobs that will be transformed,
that will be exposed to OpenAI's technologies,
and they estimated that 80% at least will
have some exposure and 19% will have
at least 50% of their tasks entirely changed.
Now, of course, a lot of factoids, and this is typical, whenever there is
a major technological wave, this comes with these kind of prophecies.
10 years ago, back in 2013, we were confronted
with the emergence of deep learning, which was another type of AI technology.
That came with a famous Oxford report which was written by two economists,
Carl Benedikt Frey and Michael Osborne.
They estimated that 47% of existing jobs
would be wiped away by AI.
We are 10 years in and this is not happening,
despite a climate crisis,
a geopolitical crisis, and a pandemic crisis.
Okay, so we're all still here.
Yes, there is hype, but yes, there's also change.
Stacie, would you agree with that.
How would you put it.
I believe that we are ramping up along this road pretty quickly.
Our original survey in February showed
that one in four companies already have been replacing
some job positions and job skills with this.
In a recent survey that we did a few weeks ago,
that's ramped up.
We are seeing now that 9 out of 10
companies currently are hiring people
with this skillset. And organizations see bringing in this skill
as making them more competitive, making them more cutting edge.
I don't know where this is going,
and I think we've already seen the low-hanging fruit with coders
and more transactional-type positions being replaced already
in terms of writing job descriptions and interview requisitions.
Things that are very transactional are quickly changing.
In the two months between our surveys, we've seen this gain a lot of acceleration.
Now, where it will go, I think we all need to stay tuned.
Now, the world of work is constantly changing when technology is introduced.
We've seen this
since the Industrial Revolution and probably before.
When robotics were introduced, a lot of jobs changed,
a lot of the jobs that were destroyed were what you might call
blue-collar jobs, jobs in warehouses, in factories, and so forth.
Is that going to be the same this time or is it going to affect
the white-collar, the lawyers, the accountants,
the people with degrees as well?
Who's going to be worse and best affected by this?
In my opinion, it's definitely more on the white-collar side
because what this does is anyone who takes information,
shuffles information, puts it back out there
and even now can analyze it, that is now being adopted with the Chat.
It is more to your point in this stage of
our technical evolution, where in the past
it did affect more blue-collar workers in terms of robotics and automation.
Now, it is affecting more white-collar workers.
Now the hope is that as those jobs go away,
these organizations will be retraining them and redeploying these workers
in other areas of the organization and give them a chance
to get promoted and learn new skills.
-Antonio, what do you think. -Well, first of all,
I would like to highlight that there is a difference between robotics and AI.
Robotics is designed to perform physical tasks.
From that point of view,
it's only normal that over the centuries it had replaced human labour.
In the case
of these cognitive technologies that we are using today,
we've been using for a century already, they are designed
to work with human beings, so to create some kind of fruitful partnership.
Of course, they are always welcomed
with a salvo of fear, thrill, risks,
and this is also our own passion for scary stories as a culture.
Then on the other side, I'm also attentive to what Stacie just said
that the companies are now hiring prompt engineers.
This is true. I mean, they are.
They are doing it.
I would like to ask what's going to happen to those new professionals,
to those new workers in one year's time, two years' time
because companies have been hiring new and exotic types of workers:
web3 engineers, metaverse designers, blockchain architects.
Six months in,
they were making coffee for the other type of actual engineers, actual architects,
or on the other side, they morphed
into more traditional type of workers.
This is something that we have to take into account.
We have to be, of course,
cautious but we also have to not yield
to this kind of passionate embrace of technology.
What is AI good at, and what are humans better at.
[chuckles] Well, AI is good at computing, which is,
of course, obvious, because that's in the very definition of it.
Computing can encompass a number of activities
and what is striking and new with generative AIs
such as ChatGPT, DALL·E 2, Midjourney,
in several types of media, is
that these kinds of new AI solutions seem to be creative.
Now, what's the human forte,
the cup of tea for a human?
Well, humans are good at quick judgement calls.
They are good with common sense.
They are good with simple
and sometimes even obvious remarks.
How many feet do you put in a shoe.
Of course, one.
Even a two-year-old knows that, but an AI does not.
This is the kind of common sense that human are needed for,
and this is also the good and the bad news about the present labour market.
Because the new competencies, the new skills that are required
are not high-level skills, are really sometimes described
as low skills
or simple task-oriented skills like,
I don't know,
quickly saying if the image of an animal
is a cat or a dog or a tree or a car.
These skills are used to train the artificial
intelligence that now we use.
Stacie, would you agree with that.
Where would you see AI having the edge over
and where would you see humans having the edge in hiring?
Well, I do agree that wherever you apply AI,
you need someone to request information
and someone to check it when it comes out and massage it.
There will always be some human interaction here too.
I do agree
it is a different kind of skill set, even to be a prompt engineer.
Just to let you know, in our survey with prompt engineers,
they're mostly with companies that have 500 employees and higher.
That was the crux of our survey.
Those are the folks looking for prompt engineers
because they could take this application into all areas of a company.
They could work with HR,
they could work with marketing, they could work with finance
to see what skill sets could be generated quickly and faster through AI.
Smaller companies of maybe 100 people,
if they have someone sitting in their HR department,
they will want that person to know the skills.
At least that part of their job can be more automated and they could focus
on more strategic missions for the organization
and do more human interactions like onboarding and other things.
I do see a skill set,
a prompt engineer does not have to come out of the IT world.
I agree that it is a different skill set, and it depends
on the size of the organization and how it's going
to be implemented and affected throughout, but we're seeing across the board,
we have not come across one company that is not looking for this experience.
Most workers in the world are employed in SMEs.
That's 5, 10 employees max, maybe.
How is AI going to affect SMEs.
That's a very good question.
The very good question to provide an answer.
First of all, we have to say that there is a difference,
of course, we know between employment and labour.
Employment means that you are formally employed.
The fact that she mentioned, of course, is entirely true, it is also unevenly
distributed worldwide because the vast majority
of people who are employed are in Northern countries.
If you move to the Global South,
you have 20%, tops, of people who are employed,
usually by multinationals or international companies
or the public sector.
Now, this means that we have the vast majority
of the human genre that is now
simply working but not employed.
This is something that is largely affected
by these new technologies because if we look
at the need for human contribution to AI,
and if we see the development
of some services that we sometimes refer
to as crowd work or data work
that are necessary to produce these AI, well,
usually these workers are recruited among
the unemployed or informally employed,
informal workers of the Global South.
This is something that is changing entirely
the balance between employed workers and informally or
sometimes piece workers, sometimes [crosstalk].
-Self-employed, so on. -Self-employed, yes.
On the platform, yes.
It's not only a matter of SMEs or big companies now.
Is there, therefore, a danger that actually greater application of AI
in the world of work could actually increase the level
of inequality between the big companies who Stacie has been talking about,
who have 500 and a big HR department and so on,
and the bulk of the world of work, which as you say is SMEs,
-self-employed platform and so on? -Yes.
This is the typical argument of dualization of the labour market,
and this is something that has been observed
over the last 50 years.
This has clearly accelerated in the last 10 years because we have seen
a number of strange indicators economically speaking.
We have seen that despite this big increase
in technological innovation, the productivity is not growing
at the same level and not growing at all, actually.
If we look at the number of persons who are employed,
of course, I mean demographically, it is only normal
that now we have a bigger number of persons who are employed,
formally employed worldwide, but the number
of people who are informally employed or self-employed is increasing too.
This is creating a situation that is unprecedented,
and AI has played a big role in creating the situation.
Stacie, it seems from people playing around with ChatGPT
that these forms of AI really have
a problem with factual accuracy.
Now, isn't this actually going to be a big problem for companies
when it comes to hiring.
Not just with CVs, but with going back and checking references,
looking at past work, and so on. How are companies going to deal with that?
I'm hearing two different aspects of that question.
One, I'm hearing you say in the hiring process,
how will they know they're getting accurate information.
If AI is writing their resumes and their cover letters
and those kinds of things, are they getting to see the right person.
My response is, all that does is get you into the interview.
Once you're in the interview, very savvy hiring managers,
because this isn't new.
People have their resumes and cover letters written by other people before
AI, or they would go to job coaches or resume writing services.
In that regard, in the hiring process, it comes down to the interview process
and really vetting our candidates one-on-one.
Well, okay, I can provide probably a more oblique point of view
because it comes from my own professional experience.
I teach in a university, and as professors, we are confronted
with the same kind of problem, with the same kind
of risk of people cheating.
What we have decided, and at least I've been
rather successful in putting in place some experiments with ChatGPT is,
let's assume that people are going to use this tool that is afforded to them.
Let's work with them in order,
first of all, to teach them proper, fair,
and possibly not entirely biased use.
Then secondly, probably we have to change our own ways of working.
We are now back to an increasing need for face-to-face interaction.
This is going to be actually the kind of solution to many of these problems.
We will probably change our way of using face-to-face interaction
versus remote connections and remote communication.
Are we going to need regulation, and frankly, is it practical,
given how fast AI seems to be moving?
What do you think?
I think we're going to have to take a look at this.
I mean, we're at the very beginnings now, and in terms of company's adoptions,
from what we are seeing, it's ramping up very quickly.
What those regulations should be.
That is beyond my pay grade, but I do think we're going
to have to take a look at this because it's proliferating everywhere.
Antonio, what do you think?
Regulation, it's a thorny issue now because you have the very companies
who produce this technology that are calling for regulation,
but for effective regulation.
The adjective here is a scary one.
It's a scary part of it because when they say effective, they mean
a regulation that does not prevent them from doing whatever they want.
Usually, these companies don't like to comply with existing laws.
I am afraid that this
kind of regulation discourse can be a double-edged sword.
We have to, first of all, respect the law that already exists,
and the laws that already exist have to do with labour standards,
with security, with non-discrimination, with privacy.
For instance, this is a very good example,
this is something that has been happening in the last few weeks in Europe.
All of a sudden, we had a major bug with ChatGPT that was exposing
everybody else's chat history to all users.
This is the reason why, firstly,
the Italian and then the Spanish and the French,
and then eventually the EU data authorities decided
to gang up on ChatGPT and OpenAI
and ask them to comply with existing laws.
The interesting part is that for the first time
in the history of privacy protection since GDPR,
these data protection authorities said in filing their decision
that we have a problem because we do not know
how OpenAI trains the AI with private data.
That's the first time in at least to my knowledge,
where training is emphasized
to that extent. This is important
because we always consider that these kinds of tools,
these kinds of AIs are intelligent per se,
that they are born intelligent. They are not.
They are constantly trained and retrained and despite the fact that ChatGPT,
GPT means Generative Pre-trained Transformer,
it means that it was trained in the past, but the training never stops.
We are training the AI whenever we use it because whenever ChatGPT provides you
with a reply, you have a thumb up and a thumb down.
This is their way to prompt you to perform some free labour for them.
That is basically improving
their AI solution. -Right.
Now, we're almost out of time, so let me ask you both one final question,
which is, what should employers and businesses and indeed workers
be doing now to cope with the oncoming wave of AI?
Antonio, let me give you the first go at that.
Well, first of all,
do not use these AI as an excuse
to further create precarious jobs.
This is an actual risk.
AI is a splendid tool, is extremely useful.
The risk is that the number of tasks that each one of us
will be eventually prompted to perform
will increase exponentially
in the next few years because we now have to,
not only do our own jobs but also do the job of the prompt engineer,
and then anticipate the algorithm and then imagine, I don't know,
how to better format our texts and images and so on and so on.
These new skills are usually not paid,
and this is something that we see in the, well,
zero increase of the real salaries all the time.
This is something that has to be addressed by the regulators and by the employers.
Thank you. Stacie, final word to you.
What should businesses and indeed workers be doing now to get with this?
I think from a worker standpoint, this is going to be an important skill
to know just like when the proliferation of PCs and computers, everybody had
to learn MS Office and Word and everybody had to learn Excel.
I do believe that for workers to have the edge
in getting positions in today's world,
this is a skill you're going to want to have,
because companies are looking to hire that skill
in all areas within the company. And I think employers
have to stay on top of it too. And I agree that
we should just not be creating these jobs that may go away in the future.
I believe this is a skill that everyone is going
to have at some degree in their work life.
Just like we had to learn everything else, this is going to be another skill
that everybody's going to eventually have on their resume,
and companies have to watch how they apply it,
how they use it.
If we go in good faith to even the playing field,
that would be a good thing.
Stacie, Antonio, thank you very much.
I'm afraid that's all we have time for today.
My thanks to Antonio Casilli and Stacie Haller.
Please join us again soon for another edition
of the ILO's Future of Work podcast. Goodbye.