WEBVTT

NOTE
This file was generated by Descript 

00:00:01.686 --> 00:00:04.716
Susannah de Jager: Welcome to Ethics
and Innovation by Oxford Plus, a

00:00:04.716 --> 00:00:08.706
special miniseries hosted by me,
Susannah de Jager and sponsored by

00:00:08.706 --> 00:00:10.986
Equinox, Equitable Innovation Oxford.

00:00:12.591 --> 00:00:16.791
As artificial intelligence becomes
increasingly embedded with everyday life,

00:00:17.121 --> 00:00:21.441
ethical questions are moving from academic
seminars into boardrooms, government

00:00:21.441 --> 00:00:23.421
departments, and product design teams.

00:00:24.081 --> 00:00:27.901
Professor Edward Harcourt has
spent his career examining some of

00:00:27.901 --> 00:00:32.461
philosophy's most enduring questions
about responsibility, character,

00:00:32.521 --> 00:00:34.501
decision-making, and human flourishing.

00:00:34.971 --> 00:00:39.681
Today, as the Director of Oxford's
Institute of Ethics and AI, he

00:00:39.681 --> 00:00:43.011
is helping shape one of the most
important conversations of our time.

00:00:43.791 --> 00:00:47.661
In this discussion, we move beyond
questions of technology and ask

00:00:47.661 --> 00:00:49.791
deeper questions about human values.

00:00:50.241 --> 00:00:53.061
What does responsible
innovation actually mean?

00:00:53.361 --> 00:00:54.831
And how should we be measuring it?

00:00:55.281 --> 00:00:58.378
Can ethical frameworks keep
pace with technological change?

00:00:58.851 --> 00:01:02.751
And how do we ensure that increasingly
powerful technologies remain

00:01:02.751 --> 00:01:06.591
aligned with the interests of the
societies they are intended to serve?

00:01:06.992 --> 00:01:10.232
Edward, thank you so much for
joining today in your office.

00:01:10.732 --> 00:01:11.662
Edward Harcourt: You're very welcome.

00:01:11.975 --> 00:01:16.165
Susannah de Jager: When most people
think about ethics and AI, they're often

00:01:16.465 --> 00:01:22.025
looking at very broad questions thinking
about jobs, the economics, inequality.

00:01:22.445 --> 00:01:26.555
Your particular angle on it is much
more specific and actually very

00:01:26.555 --> 00:01:30.355
relevant to the fact that social
media just got banned - two days

00:01:30.355 --> 00:01:32.395
ago where, as we record in the UK.

00:01:32.895 --> 00:01:36.874
Just for anyone listening, can you give
a sort of brief outline of your subject

00:01:36.874 --> 00:01:39.231
matter expertise and what you focus on?

00:01:39.231 --> 00:01:41.461
Because I think it's
helpful to contextualise.

00:01:41.961 --> 00:01:42.241
Edward Harcourt: Right.

00:01:42.681 --> 00:01:48.727
So there is a tendency for people
to label, to slap the ethics label,

00:01:48.967 --> 00:01:52.717
on the bucket where they put any
difficult issues, whether it's,

00:01:52.957 --> 00:01:54.217
are we all going to be out of work?

00:01:54.217 --> 00:01:56.997
Are robots going to take over the world?

00:01:57.497 --> 00:02:01.427
Questions and anything that they find
very difficult to cope with and I

00:02:01.427 --> 00:02:04.247
don't object to that use of the word.

00:02:04.747 --> 00:02:10.017
But I try to use the word
slightly more specifically.

00:02:10.317 --> 00:02:16.747
And in the institute, we deal
with issues about AI ethics,

00:02:16.777 --> 00:02:22.017
issues to do with AI safety, and
issues to do with AI governance.

00:02:22.017 --> 00:02:25.771
So what is an appropriate
regulatory regime for AI?

00:02:26.191 --> 00:02:28.861
Who should be held liable
when things go wrong?

00:02:29.361 --> 00:02:33.591
What the right balance to strike
between regulation and innovation?

00:02:34.091 --> 00:02:36.731
All those big questions that
are on lots of people's minds.

00:02:37.231 --> 00:02:44.111
What I think is our intellectual USP,
if you like, our intellectual brand,

00:02:44.611 --> 00:02:49.361
is the following that, questions
about ethics and about the nature

00:02:49.361 --> 00:02:52.661
of human mindedness have been around
for a very long time in philosophy.

00:02:53.161 --> 00:02:56.841
But I think something that
they've suddenly become relevant

00:02:56.871 --> 00:03:00.471
thanks to the advent of AI in
the way that they weren't before.

00:03:00.531 --> 00:03:04.251
If you go back, 20 years or
something, people I told people

00:03:04.251 --> 00:03:05.001
what I was interested in.

00:03:05.001 --> 00:03:10.211
They say, "Wow, Edward, that's so
abstract, yawn." Now it's, "Oh, really?

00:03:10.211 --> 00:03:11.291
Ethics, the mind.

00:03:11.351 --> 00:03:12.371
Tell us all about it.

00:03:12.376 --> 00:03:18.891
" Because people have suddenly woken
up to the fact that there are other

00:03:19.391 --> 00:03:23.981
entities, or at least potentially other
entities cohabiting with us on this

00:03:23.981 --> 00:03:29.621
planet who might also have minds, who
might also be ethical subjects, and who

00:03:29.621 --> 00:03:34.801
might also edge us aside in all sorts
of important contexts in the context

00:03:34.801 --> 00:03:39.221
of psychotherapy, that labour market,
friendship, they might take the role

00:03:39.581 --> 00:03:44.671
that had previously only been occupied
by human beings and not just contingently

00:03:44.701 --> 00:03:48.301
occupied only by human beings, but
where the very question, "I want to get

00:03:48.301 --> 00:03:49.921
some advice about where to go to eat.

00:03:49.921 --> 00:03:51.391
Shall I ask a human being or not?

00:03:51.396 --> 00:03:52.831
" Well, who else would you ask?

00:03:53.191 --> 00:03:54.841
Now, you can ask a machine.

00:03:55.141 --> 00:04:00.901
Many people think that the beginning
and the end of AI ethics is training

00:04:01.401 --> 00:04:03.891
AI models to reason ethically.

00:04:04.371 --> 00:04:08.241
And I think that's not the
case for a variety of reasons.

00:04:08.241 --> 00:04:10.491
First of all, what is thinking ethically?

00:04:10.491 --> 00:04:13.851
There's a whole lot of ethical argument
that needs to be settled before you

00:04:13.851 --> 00:04:15.771
decide what to train the models on.

00:04:16.271 --> 00:04:19.511
And so in a way, many of the
very interesting cards have been

00:04:19.511 --> 00:04:22.631
dealt already by the time you
decide what to train them on.

00:04:22.841 --> 00:04:30.601
But the other point is that I think many
positive ethical values are realised by

00:04:30.601 --> 00:04:36.301
the AI ecosystem as a whole, not just by
the content of what models are trained on.

00:04:36.691 --> 00:04:39.271
And a key point here is pluralism.

00:04:39.771 --> 00:04:42.921
It's not a defect in a free society
that people have differences

00:04:42.921 --> 00:04:44.601
of view about how to live.

00:04:44.871 --> 00:04:48.411
We regard it as a positive feature
of a free society that people w have

00:04:48.681 --> 00:04:52.101
really quite deep-seated differences
about what constitutes a well-lived

00:04:52.101 --> 00:04:56.241
life, what constitutes a worthwhile
thing to pursue in life can live

00:04:56.241 --> 00:04:58.251
together without coming to blows.

00:04:58.251 --> 00:05:00.201
That's one of the things
that a free society does.

00:05:00.201 --> 00:05:01.941
It contains difference in that way.

00:05:02.441 --> 00:05:08.011
Now, if you pose the question, "Well, what
is the ethical code that I should teach my

00:05:08.011 --> 00:05:11.961
model?" You're forgetting about the value
of difference, but I think difference,

00:05:11.991 --> 00:05:14.681
pluralism, has a value in its own right.

00:05:14.936 --> 00:05:20.066
It has forced us to think about what,
if anything, is the unique contribution

00:05:20.066 --> 00:05:24.676
that the human presence makes to
the meaning or value of a situation.

00:05:24.676 --> 00:05:29.479
And I don't think that question is the
whole of AI ethics by any means, and

00:05:29.479 --> 00:05:32.869
it's no accident that this institute is
located in the philosophy department,

00:05:32.869 --> 00:05:36.829
which in turn is located in the
humanities division of this university.

00:05:37.129 --> 00:05:41.329
I want to make that the starting
point of our inquiries, not just

00:05:41.809 --> 00:05:45.139
let's think about machines and
their effects on us and our future.

00:05:45.559 --> 00:05:49.929
But what is it that the existence
of these machines forces us to

00:05:49.929 --> 00:05:51.909
realise is unique about ourselves?

00:05:52.409 --> 00:05:53.249
Susannah de Jager: And what is it?

00:05:53.682 --> 00:05:55.662
A big question in a small sentence.

00:05:55.667 --> 00:05:58.222
Edward Harcourt: Well, I'm not
sure that there's an extra thing

00:05:58.252 --> 00:05:59.722
that's unique about ourselves.

00:05:59.722 --> 00:06:05.732
But I think there is something unique
about the embodied human person

00:06:06.232 --> 00:06:12.952
that is not reproducible by, or not
substitutable by, anything else.

00:06:12.952 --> 00:06:15.292
However cognitively sophisticated it is.

00:06:15.642 --> 00:06:18.222
And there are various ways
of thinking about that.

00:06:18.432 --> 00:06:22.942
The first is human minds have
all sorts of bits to them.

00:06:23.442 --> 00:06:26.112
Our cognitive capacities is one of them.

00:06:26.142 --> 00:06:30.072
That's one of the things that makes
us so successful as a species that

00:06:30.072 --> 00:06:34.452
why we've accomplished all these
clever things, including inventing AI.

00:06:34.952 --> 00:06:41.242
But our intellectual capacities come
in a very complex package in human

00:06:41.242 --> 00:06:44.622
beings with the capacities to move.

00:06:44.862 --> 00:06:48.452
The fact that we're embodied and
we're therefore mortal and we're

00:06:48.452 --> 00:06:50.462
capable of suffering various kinds.

00:06:50.962 --> 00:06:54.262
The fact that we're subject to
appetites that we can't control

00:06:54.562 --> 00:06:59.182
and that are only problematically
related to our reasoning capacities.

00:06:59.452 --> 00:07:03.560
I can reason all I like, but it doesn't
stop me falling asleep when I'm tired.

00:07:03.810 --> 00:07:07.980
Plato talked about that two and a
half thousand years ago, and that

00:07:07.980 --> 00:07:12.620
comes with having a body, and also
the way in which human relationship

00:07:12.620 --> 00:07:14.360
can be mediated by our bodies.

00:07:14.360 --> 00:07:17.690
The fact that touching another person is
such an important way of communicating,

00:07:17.720 --> 00:07:22.070
not just about conveying messages or
sentences from one mind to another.

00:07:22.570 --> 00:07:27.600
So that's one way of thinking about
what we have and artificial minds don't.

00:07:27.660 --> 00:07:30.720
We have an awful lot beyond
our intellectual powers.

00:07:30.868 --> 00:07:33.928
And the other thing is, and this is an
example which is supposed to provoke

00:07:33.928 --> 00:07:39.648
thought rather than sweep all before
it, but dead human bodies have a very

00:07:39.648 --> 00:07:42.058
special significance in our lives.

00:07:42.058 --> 00:07:47.578
Sometimes, archaeologists will say
that we can detect the first presence

00:07:47.578 --> 00:07:50.598
of homo sapiens by two measures.

00:07:50.598 --> 00:07:52.758
The first is, do we dig up tools?

00:07:53.258 --> 00:07:56.588
And the second is, do we dig up
human bodies that have obviously been

00:07:56.588 --> 00:07:58.448
deliberately arranged for burial?

00:07:58.868 --> 00:08:05.728
And it's as if the desire to dispose
of the dead with solemnity and

00:08:05.938 --> 00:08:10.288
our humanity as our existence as a
species come as part of a package.

00:08:10.288 --> 00:08:13.138
And I don't think there is
any culture that does not

00:08:13.198 --> 00:08:14.878
solemnly dispose of their dead.

00:08:15.178 --> 00:08:17.758
And if you walk into a room, there's
a dead body lying on the floor.

00:08:17.758 --> 00:08:19.578
Woo, that's a big deal.

00:08:19.728 --> 00:08:21.678
Susannah de Jager: And in one
of your blogs you articulated

00:08:21.678 --> 00:08:23.508
that in contrast to a robot

00:08:23.508 --> 00:08:26.035
Edward Harcourt: A robot that
doesn't work is just rubbish

00:08:26.675 --> 00:08:28.065
and that's a huge contrast.

00:08:28.275 --> 00:08:33.145
And so I think embodiment is important
because it brings with it all those

00:08:33.145 --> 00:08:37.855
complicated capacities in the living human
being that go along with intellection.

00:08:38.355 --> 00:08:42.765
But embodiment is important even
after we cease to be alive in a way

00:08:42.765 --> 00:08:44.175
that it's not with artificial life.

00:08:44.565 --> 00:08:48.945
Susannah de Jager: So acknowledging
those differences as you frame

00:08:48.945 --> 00:08:53.353
them there, we're talking about an
agreement that there is a difference.

00:08:53.603 --> 00:08:58.113
We just spoke to Raphaël Millière,
your colleague, who was talking about

00:08:58.473 --> 00:09:02.633
that increasingly models that were
historically trained on large language

00:09:02.743 --> 00:09:08.803
data sets are now looking at vision
inputs and video content and even as

00:09:08.803 --> 00:09:11.757
he was saying that, I was thinking,
but there's the sensory element.

00:09:11.757 --> 00:09:12.777
You've gone even further.

00:09:12.777 --> 00:09:16.257
You're talking about a sense of
what it is to be human and to

00:09:16.257 --> 00:09:18.057
engage physically with one another.

00:09:18.487 --> 00:09:23.027
It's very hard to imagine a scenario
where actually all those gaps can

00:09:23.027 --> 00:09:29.347
be closed, and yet, people are
conflating human consciousness

00:09:29.847 --> 00:09:32.687
onto AI models even as they are.

00:09:33.187 --> 00:09:36.110
Why does it matter to draw
those two things apart?

00:09:36.610 --> 00:09:37.090
Edward Harcourt: Okay.

00:09:37.510 --> 00:09:38.950
Well, several thoughts there.

00:09:39.410 --> 00:09:46.470
The first is that I think it would be a
lot harder for people to engage with large

00:09:46.470 --> 00:09:51.710
language models with chatbots in the way
that they do, if they had not already

00:09:51.710 --> 00:09:54.020
got used to certain other technologies.

00:09:54.320 --> 00:09:56.120
The letter is one.

00:09:56.620 --> 00:09:58.390
The telephone call is another.

00:09:58.450 --> 00:10:00.340
The text message is a third.

00:10:00.820 --> 00:10:07.600
So the idea that, we can be presented
with another human being in the very

00:10:07.600 --> 00:10:15.440
restricted medium of written text or heard
text is something we've got very used to.

00:10:15.470 --> 00:10:21.060
But before writing, you'd think some
letters appear on a piece of paper.

00:10:21.060 --> 00:10:24.600
You think, "Well, what on earth is
that? Do you have to be in the actual

00:10:24.600 --> 00:10:29.130
presence of a person?" So we're used
to remote communication and that

00:10:29.130 --> 00:10:32.760
makes it far easier to imagine that
there's a mind behind a chat bot.

00:10:33.466 --> 00:10:35.656
Susannah de Jager: A lot of your
work is looking at childhood

00:10:35.656 --> 00:10:37.306
development attachment theory.

00:10:38.013 --> 00:10:43.113
In some of your blog work and your
articles that I looked at what jumped

00:10:43.113 --> 00:10:48.213
out at me is that children are used
to engaging in the way you described.

00:10:48.393 --> 00:10:52.413
They are used to that distance
interaction, but perhaps to a lesser

00:10:52.413 --> 00:10:53.953
degree than adults realise it.

00:10:54.326 --> 00:10:59.560
They are conferring human attributes
and understanding upon that interaction.

00:10:59.930 --> 00:11:01.740
What is the implication of that?

00:11:01.740 --> 00:11:02.160
Edward Harcourt: Okay.

00:11:02.690 --> 00:11:04.280
I've two things to say.

00:11:04.610 --> 00:11:11.090
I think the basic reason why things can go
wrong when children or indeed adolescents

00:11:11.090 --> 00:11:16.760
or adults interact with chat bots as
if they were real is that they aren't.

00:11:17.240 --> 00:11:20.300
So the interaction is
vitiated by a false belief.

00:11:20.800 --> 00:11:22.210
I'll come onto that in a second.

00:11:22.510 --> 00:11:26.690
But it's important to
recognise what I'm not saying.

00:11:27.110 --> 00:11:31.870
So many parents, and I think if
my children were still young, I

00:11:31.870 --> 00:11:33.190
would be having the same worry.

00:11:33.680 --> 00:11:37.440
Many parents are terribly worried about
how children's development will be

00:11:37.440 --> 00:11:44.220
affected by their excessive or overly
early interaction with small screens.

00:11:44.720 --> 00:11:50.460
And all I want to say about
that is, look, the evaluative

00:11:50.460 --> 00:11:55.440
significance of technological
change is very hard to assess.

00:11:55.840 --> 00:11:57.719
Susannah de Jager: That's not
as reassuring as I was hoping

00:11:57.719 --> 00:11:58.593
you were going to be, Edward.

00:11:58.799 --> 00:12:01.566
Edward Harcourt: Well, I'll come back
to the illusion point in a minute.

00:12:01.926 --> 00:12:03.336
But take writing.

00:12:03.666 --> 00:12:07.146
People will often say small
screens are rewiring our children's

00:12:07.146 --> 00:12:12.666
brains and no doubt they are, but
literacy rewired our brains, right?

00:12:12.666 --> 00:12:17.696
Because there are huge bits of our
brains that were not used before writing

00:12:17.696 --> 00:12:22.926
was invented and writing has made a
whole lot of things possible, but it

00:12:23.046 --> 00:12:25.266
also devalued certain other things.

00:12:25.746 --> 00:12:29.156
Going back to what I was saying a
minute ago, face-to-face interaction no

00:12:29.156 --> 00:12:32.786
longer has the significance that it does
post-writing because you can communicate

00:12:32.786 --> 00:12:33.926
with somebody when they're not there.

00:12:34.212 --> 00:12:38.282
I can imagine people thousands of
years ago saying, this is ruining human

00:12:38.282 --> 00:12:41.192
society, because you no longer have to
be in the physical presence of somebody

00:12:41.192 --> 00:12:42.302
in order to communicate with them.

00:12:42.302 --> 00:12:45.542
If the only thing you've got is
speech, that makes a certain kind

00:12:45.542 --> 00:12:47.552
of human interaction essential.

00:12:47.912 --> 00:12:49.832
It's optional post-writing.

00:12:50.332 --> 00:12:56.852
But I don't think any of us regret the
invention of writing because we're so used

00:12:56.852 --> 00:12:59.552
to the things that it has made possible.

00:13:00.052 --> 00:13:02.932
We think of it as a net
expansion of our capacities.

00:13:02.982 --> 00:13:06.612
Even though we can imaginatively cast
our minds back into a society in which

00:13:06.612 --> 00:13:10.422
people would experience the invention
of writing as it were on the cusp of

00:13:10.422 --> 00:13:14.862
its invention as a loss, what's going to
go missing if we adopt this technology?

00:13:15.362 --> 00:13:18.692
I don't think anybody wishes that
it'd never been invented because we're

00:13:18.692 --> 00:13:21.782
so much creatures of that invention.

00:13:21.782 --> 00:13:25.172
We've grown up valuing the forms
of social interaction which

00:13:25.172 --> 00:13:27.062
have been made possible by it.

00:13:27.312 --> 00:13:33.002
That's why it's very hard to assess the
evaluative significance of technological

00:13:33.002 --> 00:13:37.352
change, because even though it seems
dreadful while it's happening, the

00:13:37.352 --> 00:13:40.802
people who will tell its history will
be people who've been shaped by it.

00:13:41.052 --> 00:13:43.326
It's not saying everything's
fine with chat bots.

00:13:43.331 --> 00:13:44.681
It's not supposed to imply that.

00:13:45.041 --> 00:13:50.761
But it's a general reason to not, reach
straight for the panic button when we're

00:13:50.761 --> 00:13:53.221
in the face of some technological change.

00:13:53.781 --> 00:13:57.171
And in particular, saying that it'll
rewire our brains is not a good reason.

00:13:57.464 --> 00:14:01.694
I do worry about the excessive
use of small screens, excessive

00:14:01.694 --> 00:14:03.014
reliance on computers.

00:14:03.194 --> 00:14:09.414
I was giving a talk to some sort of 18
to 21-year-olds recently about chat bots

00:14:09.914 --> 00:14:15.194
and it became apparent to me that when
they talked about their social lives, what

00:14:15.194 --> 00:14:17.354
they meant was their online social lives.

00:14:17.674 --> 00:14:20.464
For awhile, I though, "Well, you
must been meeting in the common room

00:14:20.464 --> 00:14:24.454
or meeting in the bar or going for
a walk." No no, they were talking

00:14:24.454 --> 00:14:27.184
about social life is something
you do in front of a small screen.

00:14:27.604 --> 00:14:31.434
And I though, "I wouldn't like
that. What kind of a life is

00:14:31.434 --> 00:14:32.964
that?" But they've got used to it.

00:14:32.964 --> 00:14:34.314
They've still got friends and so on.

00:14:34.314 --> 00:14:37.524
So I think that the objection to
chat bots is totally different.

00:14:37.524 --> 00:14:39.594
It's not that social
life has been reshaped.

00:14:40.014 --> 00:14:42.774
It's that people have the illusion
of interacting with something real

00:14:42.834 --> 00:14:47.564
and it isn't and so people need to
understand that there isn't something

00:14:47.564 --> 00:14:51.494
on the other end of, as a friend is on
the other end of a text message or a

00:14:51.734 --> 00:14:56.074
telephone call, there isn't something
on the other end of the chatbot and

00:14:56.107 --> 00:14:57.640
people are victims of an illusion.

00:14:58.030 --> 00:14:59.950
Not saying it has no use at all.

00:15:00.310 --> 00:15:07.700
I got a rather moving message from a woman
in the Ukraine recently who is responding

00:15:07.700 --> 00:15:12.571
to that blog that you read about chat bots
saying, "Look, what if you've got nobody

00:15:12.571 --> 00:15:16.291
else to talk to because you're living in
a war zone and you're completely isolated?

00:15:16.291 --> 00:15:19.878
There's no other opportunity to interact
with your professional peers." And

00:15:19.878 --> 00:15:23.608
I said, "Well, yeah, that's a good
point and I'm very sorry to hear

00:15:23.788 --> 00:15:25.258
that's the situation that you're in.

00:15:25.468 --> 00:15:30.018
But of course, if it's a choice
of interacting with the bot or

00:15:30.018 --> 00:15:33.588
interacting with nothing at all,
go with the bot, but just don't

00:15:33.588 --> 00:15:36.078
believe that what you're talking to
is a person, A, because it isn't.

00:15:36.578 --> 00:15:40.548
But B, because if you think about it
for five minutes, you'll realise that

00:15:40.548 --> 00:15:45.408
there are all sorts of conversational
moves or all sorts of articulations

00:15:45.408 --> 00:15:49.518
in the relationship that are there
if you're interacting with a person

00:15:49.698 --> 00:15:52.398
and which are not there if you're
interacting with a bot and which are

00:15:52.398 --> 00:15:53.958
valuable in the case of a person.

00:15:53.958 --> 00:15:56.058
So a person can say, I'm fed up with you.

00:15:56.558 --> 00:15:58.298
I've had enough from you from one day.

00:15:58.598 --> 00:16:03.748
Enough of you and your woes, ask
me a question or just go to bed.

00:16:04.048 --> 00:16:07.212
Susannah de Jager: And you have
articulated a lot there, but you've

00:16:07.212 --> 00:16:10.902
actually articulated sort of three
different relationships there.

00:16:11.152 --> 00:16:13.612
One is with a chat bot, and
you've articulated some of

00:16:13.612 --> 00:16:15.022
the pros and cons of that.

00:16:15.392 --> 00:16:20.522
One is in person and you've done a
lot of work on how people typically

00:16:20.642 --> 00:16:24.782
and historically have developed
and how that forms character and

00:16:24.782 --> 00:16:27.512
personalities and social structures.

00:16:28.012 --> 00:16:32.192
But then in the middle you slightly
skimmed over a modern phenomenon, which

00:16:32.192 --> 00:16:34.322
is people socialising, not in person.

00:16:34.738 --> 00:16:38.658
And actually, I would love to
dig into that a little bit more

00:16:38.958 --> 00:16:41.948
because it is so much more common.

00:16:42.038 --> 00:16:47.748
We are seeing things like anxiety and
kind of sadness and feelings of being

00:16:47.988 --> 00:16:52.488
subject to the world rather than having
control of your own environment and you

00:16:52.488 --> 00:16:55.758
could argue that's to do with a lack
of development of a locus of control.

00:16:55.758 --> 00:17:00.998
And I want to know your opinion
on that middle ground because you

00:17:00.998 --> 00:17:04.088
and I are of a generation where we
would have socialised in person.

00:17:04.208 --> 00:17:05.768
It is less common.

00:17:06.398 --> 00:17:10.448
And the implications of
that seem to me very large.

00:17:10.748 --> 00:17:14.278
Edward Harcourt: I think that the
difference between, interacting

00:17:14.278 --> 00:17:18.985
with a real person and interacting
with something that is not a person,

00:17:19.695 --> 00:17:21.705
those are very different differences.

00:17:22.005 --> 00:17:26.885
And so as I say, the interaction
with the chatbot is viciated if

00:17:26.885 --> 00:17:29.795
you're under the illusion that it's
a real thing and you can't step

00:17:29.795 --> 00:17:34.775
back from the illusion as even small
children can step back from games and

00:17:34.775 --> 00:17:36.545
understand that it's just make-believe.

00:17:36.575 --> 00:17:39.425
That's one of the weird things
about people's interaction with

00:17:39.425 --> 00:17:43.015
chat bots is that at least some of
them seem incapable of doing that.

00:17:43.515 --> 00:17:49.745
Whereas I don't think it's great if
social life or the preponderance of

00:17:49.745 --> 00:17:51.965
people's social life is conducted online.

00:17:52.465 --> 00:17:58.035
But I feel hesitant about,
delivering an opinion that this

00:17:58.035 --> 00:17:59.955
is "it's not like the old days."

00:18:00.405 --> 00:18:06.945
So in lockdown, when teenagers
were only interacting with their

00:18:06.945 --> 00:18:13.165
peers via social media, I can see
that's a huge stressor, right?

00:18:13.665 --> 00:18:22.755
But if their relationships with one
another are both via social media and

00:18:22.765 --> 00:18:26.985
face-to-face, you see kids on the bus
going home together and they're looking

00:18:26.985 --> 00:18:31.719
at their phone and laughing and pointing
at stuff and you think, well they look

00:18:31.719 --> 00:18:36.219
like two friends having a good time,
not two friends who victims of the small

00:18:36.219 --> 00:18:41.076
screen and that's not to comment on
the ban about social media under 16s.

00:18:41.566 --> 00:18:46.186
And I think I could actually argue
that both ways, but, it seems to

00:18:46.186 --> 00:18:48.576
be many people's default view.

00:18:48.576 --> 00:18:51.576
I think it's going to be really popular
with lots of parents, even if it's not

00:18:51.576 --> 00:18:53.196
really popular with lots of teenagers.

00:18:53.696 --> 00:18:56.556
But I think there's a big difference
between, as it were the way in

00:18:56.556 --> 00:19:01.986
which small screens bed down in
face-to-face social relationships and

00:19:01.986 --> 00:19:05.106
what you would have if they replaced
face-to-face social relationships.

00:19:05.412 --> 00:19:09.732
Susannah de Jager: I interviewed earlier
Lord Lionel Tarassenko and interestingly

00:19:10.032 --> 00:19:14.962
he gave a speech as lots of people
actually wanted to do in the end on this

00:19:14.962 --> 00:19:16.882
subject, in the House of Lords on Friday.

00:19:16.882 --> 00:19:21.142
And clearly two days ago, as we're
recording, Kier Starmer announced

00:19:21.172 --> 00:19:23.602
the UK ban from Spring 2027.

00:19:24.342 --> 00:19:27.462
What he said on it that I though was
interesting was that he had changed

00:19:27.522 --> 00:19:32.957
his mind from thinking we should
use this as a carrot - to try and

00:19:32.957 --> 00:19:38.007
encourage companies to alter for the
positive their algorithms and the

00:19:38.007 --> 00:19:40.867
impact on cohorts, especially children.

00:19:41.367 --> 00:19:45.207
But that actually he had swivelled, having
seen some initial responses on that.

00:19:45.267 --> 00:19:48.837
Swivelled sounds rather capricious
actually that he had researched and

00:19:48.837 --> 00:19:52.677
changed his mind in a meaningful way
to their saying, actually, social

00:19:52.677 --> 00:19:56.987
media companies really seem to be
responding much more to having been

00:19:56.987 --> 00:20:01.617
given the ban and being forced which
was informative to me and I mentioned.

00:20:03.147 --> 00:20:06.267
This Oxford Plus miniseries
is brought to you by Equinox,

00:20:06.567 --> 00:20:08.157
Equitable Innovation Oxford.

00:20:08.937 --> 00:20:12.657
Equinox is a major regional partnership
established by the University of

00:20:12.657 --> 00:20:16.797
Oxford to drive inclusive innovation
and sustainable economic growth.

00:20:17.427 --> 00:20:20.487
Bringing together universities,
local authorities, government,

00:20:20.487 --> 00:20:24.747
industry, investors, and community
organisations, Equinox provides

00:20:24.777 --> 00:20:28.737
a unified voice for Oxfordshire,
championing innovation that delivers

00:20:28.767 --> 00:20:31.497
economic impact alongside social value.

00:20:32.007 --> 00:20:35.667
More than 80 organisations have signed
the Equinox charter, committing to

00:20:35.697 --> 00:20:39.507
atract investment, remove barriers
to growth, and ensure the benefits

00:20:39.507 --> 00:20:41.487
of innovation are shared more widely.

00:20:41.997 --> 00:20:46.887
By aligning Oxfordshire's world-leading
research, enterprise, and talent, Equinox

00:20:46.887 --> 00:20:51.717
aims to unlock nationally significant
growth, reinforce the county's position

00:20:51.717 --> 00:20:56.067
as a global innovation hub, and ensure
that success creates skilled jobs,

00:20:56.337 --> 00:21:00.777
attracts investment, and delivers
lasting prosperity across the region.

00:21:04.995 --> 00:21:10.305
Drawing you again on the impacts on,
as you said, on mental health, I'd

00:21:10.305 --> 00:21:17.185
love to understand a little bit more
how we are deviating without perhaps

00:21:17.185 --> 00:21:21.145
enough focus from what's optimal
for our children at the moment.

00:21:21.145 --> 00:21:25.925
There are these gaps from what we
know and the impacts it has, what

00:21:25.925 --> 00:21:30.315
we have that's a sort of working
live experiment, and what perhaps,

00:21:31.185 --> 00:21:36.005
those that study it such as yourself
know are bad for long-term outcomes.

00:21:36.335 --> 00:21:39.365
I'd love to understand a little
bit how you see the experiment

00:21:39.365 --> 00:21:40.853
we're running through those lenses.

00:21:41.353 --> 00:21:41.773
Edward Harcourt: Yeah.

00:21:42.023 --> 00:21:45.293
Well, I think we're running more than
one experiment at a time, aren't we?

00:21:45.773 --> 00:21:51.393
And I think the social, not withstanding
the fact that social media is

00:21:51.393 --> 00:21:54.613
an partly AI-driven technology.

00:21:54.913 --> 00:21:59.983
I don't think that the argument
about social media and the argument

00:21:59.983 --> 00:22:05.793
about the widespread deployment of
AI, let's say, in the workplace,

00:22:05.823 --> 00:22:07.473
is that they're not the same thing.

00:22:08.043 --> 00:22:12.423
My gut tells me that banning social media
for under 16s is a good thing to do.

00:22:12.723 --> 00:22:16.013
I think people should
have other things to do.

00:22:16.513 --> 00:22:22.303
One can imagine social media filling a
gap, that's left by having two working

00:22:23.793 --> 00:22:27.843
parents, parents' anxiety about the
safety of going out into the street,

00:22:27.915 --> 00:22:32.755
parents, I think, are much more content
if their children stay at home all

00:22:32.755 --> 00:22:34.225
day because they think they're safe.

00:22:34.255 --> 00:22:38.205
So what are you going to do, than
they were when I was a child?

00:22:38.885 --> 00:22:43.505
But my worries about the rollout of
AI in the workplace are different.

00:22:43.565 --> 00:22:48.205
So if I was a young person, I would
be worrying about the effects of

00:22:48.205 --> 00:22:53.325
AI among other things on the labour
market and which skills of mine

00:22:53.325 --> 00:22:54.795
are going to be useful in future.

00:22:55.155 --> 00:22:57.855
That's a totally different worry
from worrying about the effects

00:22:57.855 --> 00:22:59.505
of social media on mental health.

00:22:59.605 --> 00:23:03.975
Not that looking a future in the
eye in which you are worried about

00:23:03.975 --> 00:23:07.815
not being able to work would not
have an impact on your wellbeing.

00:23:07.875 --> 00:23:10.665
I think it has a very significant
impact on young people's well-being.

00:23:11.375 --> 00:23:14.485
What am I supposed to qualify myself for?

00:23:14.565 --> 00:23:16.125
How do I make myself useful?

00:23:16.545 --> 00:23:20.265
What skills constitute
preparing to be an adult?

00:23:20.795 --> 00:23:23.935
Susannah de Jager: It's interesting,
because if you dig into the more

00:23:23.935 --> 00:23:28.245
restrictive parenting that is pretty
broadly acknowledged to be the common

00:23:28.275 --> 00:23:32.184
model now, you will know this, but
those listening may or may not that,

00:23:32.684 --> 00:23:34.934
and correct me if this is wrong.

00:23:34.934 --> 00:23:38.324
But my understanding is that this
comes from there was a high-profile

00:23:38.324 --> 00:23:40.544
kidnap of a school child in America.

00:23:41.421 --> 00:23:44.931
In response to that, there was a large
scale manhunt and then after that, they

00:23:44.931 --> 00:23:48.081
started publishing on milk bottles,
pictures of missing children thinking that

00:23:48.081 --> 00:23:52.911
was a helpful thing to do, and therefore
what happened was that the perception of

00:23:52.911 --> 00:23:58.541
the risk of child abduction became much
greater than the probability of risk.

00:23:59.071 --> 00:24:01.831
I think that at the same time, it
conflated with obviously sort of

00:24:02.071 --> 00:24:05.701
propagation of the news in a different
way and social media and a whipping

00:24:05.701 --> 00:24:10.471
up  and so parents have been told that
this risk is much greater than it is.

00:24:10.971 --> 00:24:14.271
As a parent myself, I don't spend my
whole time thinking about the risk,

00:24:14.271 --> 00:24:20.241
but you are primed to be aware of it as
a risk along with others, by the way.

00:24:20.601 --> 00:24:25.241
And I wonder how we should be training
parents to realise that the harm they

00:24:25.241 --> 00:24:27.131
do is greater on the other side of that.

00:24:27.431 --> 00:24:29.252
Edward Harcourt: Goodness, I've no idea.

00:24:29.532 --> 00:24:34.142
Risk is a very difficult thing
to think about in a clear-sighted

00:24:34.142 --> 00:24:38.288
way because partly people's
attitudes to it vary a lot.

00:24:38.658 --> 00:24:44.272
A given individual's attitudes to
risk, what seems like a rational

00:24:44.272 --> 00:24:48.022
course of action, will vary a lot
depending on the size of the risk.

00:24:48.322 --> 00:24:51.823
And also because I think there
is this cultural pressure and I

00:24:51.823 --> 00:24:53.143
don't know where it comes from.

00:24:53.533 --> 00:24:58.823
I wish I knew more about
it to identify, rationality

00:24:58.823 --> 00:25:01.543
really, with being risk-averse.

00:25:01.793 --> 00:25:08.793
But as it were, taking out insurance,
not taking risks, being cautious in

00:25:08.793 --> 00:25:15.513
all sorts of ways, almost equal to
in thinking in this part of the world

00:25:15.513 --> 00:25:17.793
anyway, being practically rational.

00:25:18.312 --> 00:25:20.472
You're taking risks with
your health if you smoke.

00:25:20.762 --> 00:25:24.752
Nobody wants to die of lung
disease, but it doesn't go without

00:25:24.752 --> 00:25:26.312
saying that it's irrational.

00:25:26.842 --> 00:25:30.472
I am a fully paid up member
of the risk-averse party.

00:25:30.802 --> 00:25:36.102
I'm not a somebody who jumps into a
sports car or gets onto my bicycle

00:25:36.102 --> 00:25:40.072
without a helmet on because I want to
feel the wind on my forehead, I don't.

00:25:40.572 --> 00:25:42.880
Susannah de Jager: But that's fine because
that's what comes naturally to you.

00:25:43.570 --> 00:25:48.690
So coming back, how should one be
thinking about our children's engagement?

00:25:48.690 --> 00:25:52.910
What is it perhaps that we need
to be very protective of in our

00:25:52.910 --> 00:25:56.420
engagement with our children versus
their engagement with technology?

00:25:56.420 --> 00:25:58.833
I mean, If somebody's listening and
saying, "Well, this is all quite

00:25:58.952 --> 00:26:02.990
theoretical," what might be the more
practical thoughts that we could perhaps

00:26:02.990 --> 00:26:05.120
imbue them with from this conversation?

00:26:05.151 --> 00:26:09.175
Edward Harcourt: Well, one thing, and
this seems to be a real no-brainer, is

00:26:09.595 --> 00:26:16.885
regulation of toys that are AI-powered
seems to be quite ineffective.

00:26:17.155 --> 00:26:23.345
So even though you're not allowed
to expose children under 13, and

00:26:23.345 --> 00:26:27.785
this is not looking forward to
the social media ban to chat bots.

00:26:28.085 --> 00:26:32.135
There are toys, teddy bears, that
will talk back to you because

00:26:32.135 --> 00:26:34.737
they include AI-powered chips.

00:26:34.737 --> 00:26:38.156
Any child can have one of these because
they're bought for them by parents,

00:26:38.156 --> 00:26:41.506
but they're obviously geared to a
market for extremely young children.

00:26:41.506 --> 00:26:42.856
And I think that is not right.

00:26:42.856 --> 00:26:48.456
If there's some reason for not exposing
children under the age of 13 to AI, you

00:26:48.456 --> 00:26:51.276
shouldn't be able to get in through the
back door just because it's the teddy bear

00:26:51.276 --> 00:26:53.076
that's got the miniature chat bot in it.

00:26:53.576 --> 00:26:56.246
And I think that, that would be
an easy thing to correct for.

00:26:56.746 --> 00:27:01.576
Of course, if you've never been exposed
to AI and you're not interacting with

00:27:01.576 --> 00:27:07.676
it in your everyday, it's not clear
that a lesson in personal social and

00:27:07.676 --> 00:27:12.156
health education session at primary
school, in which you were taught to

00:27:12.156 --> 00:27:15.576
think of chatbots as not real, would
that make any sense to people who

00:27:15.576 --> 00:27:17.756
hadn't been exposed to chat bots?

00:27:18.026 --> 00:27:19.346
Well, I think it'd be good to have a go.

00:27:20.216 --> 00:27:23.806
Because you know that people
are going to graduate onto this.

00:27:24.046 --> 00:27:27.166
If they graduate onto it, then
the lesson will come too late.

00:27:27.496 --> 00:27:33.246
So I think asking little kids questions
like if you were talking to a machine

00:27:33.606 --> 00:27:36.796
and it was talking in the first
person, how do you think it feels?

00:27:36.796 --> 00:27:38.326
Do you think you can have feelings?

00:27:38.586 --> 00:27:41.886
Did you know that, your parents
can go on talking to these

00:27:42.096 --> 00:27:44.676
machines all night if they want to?

00:27:44.676 --> 00:27:46.056
Do you think the thing gets tired?

00:27:46.616 --> 00:27:51.176
In these lessons, kids do get taught,
"Have you ever had a pet who died?

00:27:51.436 --> 00:27:54.676
Has anybody else in your family died?"
And get them to think about death.

00:27:54.676 --> 00:27:57.136
"Well, do you think one of
these machines could die?

00:27:57.136 --> 00:28:00.226
And if so, what does it
mean if they can't die?

00:28:00.586 --> 00:28:03.826
Is that because they're like gods
that they live forever or something?

00:28:03.826 --> 00:28:05.266
Or is it, they're magic?

00:28:05.766 --> 00:28:06.246
No.

00:28:06.362 --> 00:28:10.252
Well, if they can't die, if they can't
suffer, can they really understand

00:28:10.252 --> 00:28:13.282
what you're saying when you try and
share something difficult with it?

00:28:13.782 --> 00:28:16.830
I think those are questions that
it'd be well worth asking children.

00:28:17.308 --> 00:28:19.688
Susannah de Jager: I think that your
point around what are the questions

00:28:19.688 --> 00:28:26.818
that for a child will bring that
gap between what they might think

00:28:26.818 --> 00:28:30.978
they're interacting with and the real
differences of no, there isn't a person

00:28:30.978 --> 00:28:35.208
on the other end of this, and therefore
you must view it, with some caution.

00:28:35.978 --> 00:28:41.898
If we take that as red that we want
to do that, taking a slightly more

00:28:41.898 --> 00:28:46.978
regulatory lens on it, what in your
opinion should we be lobbying for as

00:28:46.978 --> 00:28:51.748
parents or asking for from regulators in
order to do the other side of that coin

00:28:51.748 --> 00:28:53.848
and to actually be more restrictive?

00:28:53.878 --> 00:28:55.648
We've obviously got the social media ban.

00:28:56.098 --> 00:28:58.218
You were saying you could
argue on both sides of that.

00:28:58.228 --> 00:29:05.581
But, if we were to improve these things to
be safer for children or indeed adults to

00:29:05.581 --> 00:29:07.435
interact with, what might that look like?

00:29:07.745 --> 00:29:12.535
Edward Harcourt: Yes, I wouldn't
want to frame the potential

00:29:12.535 --> 00:29:20.555
downsides of interacting with chat
bots solely in terms of safety.

00:29:21.055 --> 00:29:26.595
So as I said in relation to the
woman from Ukraine, there can

00:29:26.595 --> 00:29:31.965
be cases where interacting with
a chatbot can be beneficial if

00:29:31.965 --> 00:29:33.615
there's no other human on the scene.

00:29:34.115 --> 00:29:39.585
Or somebody who's socially isolated
and not in a war zone or an adult

00:29:39.675 --> 00:29:46.245
or young person who can't get a
psychotherapy appointment and who

00:29:46.245 --> 00:29:50.115
needs to talk to somebody because
they're suffering in some way.

00:29:50.565 --> 00:29:55.245
So I'm not saying that interactions with
chatbots can't be instrumentally valuable.

00:29:55.245 --> 00:29:55.995
They can be.

00:29:56.495 --> 00:30:01.915
Nor do I want to say that the only thing
that can go wrong with them is that

00:30:01.915 --> 00:30:05.335
they're unsafe, for example, that they
could give rise to addictive patterns of

00:30:05.335 --> 00:30:06.895
behaviour, although I'm sure they can.

00:30:07.395 --> 00:30:09.435
Here are two other risks.

00:30:09.935 --> 00:30:14.375
The first is, and this is a much
talked about problem now, is

00:30:14.375 --> 00:30:16.145
the risk of cognitive atrophy.

00:30:16.645 --> 00:30:23.745
So I don't think that's exactly a safety
question, but if in your working life or

00:30:23.745 --> 00:30:30.995
your life as a student, it's so easy to
ask a chatbot, an LLM, a question rather

00:30:30.995 --> 00:30:32.765
than trying to find it out for yourself.

00:30:33.005 --> 00:30:37.044
If you go on and on doing that, the
huge risk is that you'll get out of

00:30:37.044 --> 00:30:40.554
the habit of answering those questions
for yourself and indeed of figuring

00:30:40.554 --> 00:30:43.074
out how to even address the question.

00:30:43.554 --> 00:30:46.104
So you acquire a dependency.

00:30:46.554 --> 00:30:49.634
What happens if there's a power
count and you can't ask that in a

00:30:49.814 --> 00:30:53.494
very boring kind of case and you
can't ask the chat bot anything.

00:30:53.994 --> 00:31:01.124
But also if you lose your powers to
question, lose your powers of open-ended

00:31:01.424 --> 00:31:06.869
inquiry, lose your powers to summarise
text and extract a view from it, you

00:31:06.869 --> 00:31:11.650
are bargaining away a lot of power to
somebody else because what comes out of

00:31:11.650 --> 00:31:15.460
the LLM is not under your control and you
may lose the capacity to criticise it.

00:31:15.700 --> 00:31:18.970
You may lose the capacity to take
critical distance on it and that,

00:31:18.970 --> 00:31:23.290
I think, is a huge risk that I
wouldn't put under the safety heading.

00:31:23.790 --> 00:31:27.860
The other is that, and I wouldn't
put this under the safety heading

00:31:27.860 --> 00:31:31.580
either, that if I'm right, that
there is something specially valuable

00:31:31.580 --> 00:31:37.030
about the human presence and that
artificial entities can't replace.

00:31:37.330 --> 00:31:42.860
That we'll become lonelier and lonelier
or there's certain sorts of human

00:31:42.860 --> 00:31:47.660
contact that we will unwittingly
deprive ourselves of, but that we

00:31:47.660 --> 00:31:52.000
need in order to thrive because we'll
think, "Oh, well, we don't really

00:31:52.000 --> 00:31:54.910
need a human carer in this situation.

00:31:54.910 --> 00:31:57.490
We don't really need a human
friend in this situation.

00:31:57.490 --> 00:32:01.900
We don't really need a human colleague
to bounce ideas off in this situation.

00:32:01.900 --> 00:32:06.080
It's cheaper or more efficient or more
productive to replace those people

00:32:06.080 --> 00:32:12.640
with artificial entities." Just as,
and this isn't an argument, a wholesale

00:32:12.640 --> 00:32:17.250
argument, against automation, but
just as people say that they suffer

00:32:17.250 --> 00:32:21.930
from social isolation because at
the checkout at the supermarket, you

00:32:21.930 --> 00:32:27.530
now don't talk to a real person, you
just put the barcode under a scanner.

00:32:28.030 --> 00:32:30.850
Obviously, if you're me and you interact
with people all day long, I don't

00:32:30.850 --> 00:32:34.000
really care what the checkout is like
at the supermarket, it's convenient.

00:32:34.300 --> 00:32:37.420
But if your only opportunity to
interact with somebody socially is

00:32:37.420 --> 00:32:41.070
when you go to the supermarket and
talk about the weather to the person

00:32:41.070 --> 00:32:42.630
at the checkout or say, "How are you?

00:32:42.630 --> 00:32:44.100
It's going to make a big difference.

00:32:44.160 --> 00:32:52.600
And I think that unwitting bargaining
away of opportunities for human contact

00:32:52.930 --> 00:33:00.380
that are non-instrumentally valuable,
but that mitigate the risk of social

00:33:00.380 --> 00:33:07.886
isolation, bargaining them way in the
name of efficiency or productivity is a

00:33:07.886 --> 00:33:12.780
huge risk of AI but not one that I'd bring
under the heading of safety, exactly.

00:33:13.150 --> 00:33:13.990
Susannah de Jager: No,
well it's interesting.

00:33:13.990 --> 00:33:17.910
You used the phrase "cognitive
atrophy" and below it I wrote

00:33:17.910 --> 00:33:21.360
"social atrophy" and what you've
just articulated is exactly that.

00:33:21.690 --> 00:33:25.650
And actually, it came up earlier in
the discussion, when you were talking

00:33:25.650 --> 00:33:32.480
about younger people interacting through
social media primarily and then you were

00:33:32.480 --> 00:33:35.000
talking about do chatbots get tired.

00:33:35.300 --> 00:33:38.960
There's a sort of further bridge
of, do we lose an ability to be

00:33:38.960 --> 00:33:44.800
patient with people who are frail
and demanding in their own ways?

00:33:44.910 --> 00:33:49.710
And actually, do we completely
lose the ability to have human

00:33:49.710 --> 00:33:53.847
connection that we just find
less satisfying in the immediacy?

00:33:54.077 --> 00:33:55.577
Edward Harcourt: I
absolutely agree with you.

00:33:55.643 --> 00:33:59.973
I think that a risk that is very
closely related to this risk of social

00:33:59.973 --> 00:34:06.923
isolation that I was describing is that
we will render ourselves less able to

00:34:06.923 --> 00:34:10.223
take part in human social interactions.

00:34:10.613 --> 00:34:12.983
Just think about the psychotherapy case.

00:34:13.403 --> 00:34:16.843
You might see a psychotherapist
for an hour a week, right?

00:34:16.843 --> 00:34:19.903
And pre-AI, that might
be a very special hour.

00:34:20.403 --> 00:34:22.683
But will it continue to seem special?

00:34:22.693 --> 00:34:24.793
Will you continue to concentrate on it?

00:34:24.933 --> 00:34:29.283
Will it mean the same to you if
you can ask the same questions to a

00:34:29.283 --> 00:34:32.883
chat bot, whether it's a specialist
chatbot or a general chat bot,

00:34:33.133 --> 00:34:35.453
potentially for 24 hours of the day?

00:34:35.903 --> 00:34:39.053
And I think the answer to that in
many people's case would be no.

00:34:39.053 --> 00:34:44.543
So we're disabling ourselves for ordinary
human contact by getting too used to this

00:34:44.543 --> 00:34:46.583
form of contact with artificial entities.

00:34:47.088 --> 00:34:48.498
This is a sort of a guess.

00:34:48.498 --> 00:34:54.578
It's a hunch because, as I said to
you earlier, I think it's very hard

00:34:54.578 --> 00:35:01.608
to evaluate the changes brought
about by technological innovation.

00:35:02.108 --> 00:35:07.048
But still, my gut feeling is
that we do need human contact.

00:35:07.398 --> 00:35:09.098
That is something that we thrive on.

00:35:09.385 --> 00:35:12.775
The co-presence of another
human is special to us and we're

00:35:12.775 --> 00:35:14.785
hardwired to see things that way.

00:35:15.285 --> 00:35:21.875
But that if we don't think very carefully,
we will bargain away opportunities for

00:35:22.295 --> 00:35:26.785
that sort of human contact, whether
it's the deliberate pursuit of human

00:35:26.785 --> 00:35:31.485
contact, like inviting somebody over
dinner or just casual human contact,

00:35:31.485 --> 00:35:32.985
like talking to somebody at the checkout.

00:35:33.235 --> 00:35:36.821
That territory will be seeded to
artificial entities in the name

00:35:36.821 --> 00:35:40.901
of efficiency or productivity
and then it'll be too late.

00:35:40.981 --> 00:35:43.441
Susannah de Jager: And if you, which
I'm sure you will have done, if you

00:35:43.441 --> 00:35:47.441
look at the characteristics of people
in blue zones some of them are diet

00:35:47.441 --> 00:35:50.051
and some of them are environmental,
but a lot of them are social.

00:35:50.271 --> 00:35:53.971
The social fabric, if you look at
a concept such as Ikagay, there's

00:35:53.971 --> 00:35:57.091
a lot to do with community and then
there's another one to do with purpose.

00:35:57.146 --> 00:36:01.206
And actually, I think that, the kind of
discourse around, well, if we get the

00:36:01.206 --> 00:36:05.316
robots and AI to do everything, then we
can go off and write poetry, which by the

00:36:05.316 --> 00:36:07.146
way, for some people would be amazing.

00:36:07.476 --> 00:36:10.006
I'm not disputing that, for
those that are really creative.

00:36:10.006 --> 00:36:13.326
But maybe it's just how one's
conditioned in a modern world and

00:36:13.326 --> 00:36:17.106
we all would be able to become more
in tune with our creative selves.

00:36:17.106 --> 00:36:21.386
But there is something about purpose
and productivity for a lot of people

00:36:21.506 --> 00:36:25.666
that if you were to say, "Here's a
universal income, go do nothing all the

00:36:25.666 --> 00:36:29.776
time or work out what the thing you can
do is." I wonder if it would actually

00:36:29.776 --> 00:36:34.736
be, beneficial or whether it would,
like people that retire and the death

00:36:34.736 --> 00:36:38.696
rates are extraordinarily high in people
that retire too fast and too speedily.

00:36:39.216 --> 00:36:39.256
Edward Harcourt: Yeah.

00:36:39.840 --> 00:36:40.934
You might well be right.

00:36:40.934 --> 00:36:45.130
I don't buy at all the idea that,
if we were suddenly liberated from

00:36:45.130 --> 00:36:50.380
work, we would be able to do be
creative beings in some other sphere.

00:36:50.380 --> 00:36:52.240
I think it'd be a very painful adjustment.

00:36:52.520 --> 00:36:53.690
Susannah de Jager: Edward,
thank you very much.

00:36:53.940 --> 00:36:54.870
Edward Harcourt: You're very welcome.

00:36:56.108 --> 00:36:59.498
Susannah de Jager: Thank you for listening
to this episode of Oxford Plus, hosted

00:36:59.498 --> 00:37:03.338
by me, Susannah de Jager If you wanna
keep up with all things Oxford Plus,

00:37:03.578 --> 00:37:08.888
visit our website, oxfordplus.co.uk or
sign up for our newsletter on Substack.

00:37:09.368 --> 00:37:12.248
Oxford Plus is a podcast
produced by Story 94.