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,
Susanna de Jager and sponsored by

00:00:08.706 --> 00:00:10.986
Equinox, Equitable Innovation Oxford.

00:00:12.191 --> 00:00:15.011
Today I'm joined by
Professor Anne Trefethen.

00:00:15.761 --> 00:00:18.689
Anne is an applied computer
scientist specialising in

00:00:18.689 --> 00:00:22.109
high-performance scientific computing
and has spent her career at the

00:00:22.109 --> 00:00:23.969
forefront of technological change.

00:00:24.839 --> 00:00:28.619
Most recently, she served as the pro-vice
chancellor for people and digital at

00:00:28.619 --> 00:00:32.579
the University of Oxford, leading one of
the university's largest programmes of

00:00:32.579 --> 00:00:37.103
digital transformation before stepping
into a role as trustee of the Alan Turing

00:00:37.153 --> 00:00:42.493
Institute, the UK's National Institute for
Artificial Intelligence and Data Science.

00:00:43.093 --> 00:00:47.113
What makes Anne's perspective particularly
valuable in both of these roles is

00:00:47.113 --> 00:00:51.253
that she has experienced firsthand what
happens when technological progress

00:00:51.373 --> 00:00:53.353
outpaces institutional planning.

00:00:54.103 --> 00:00:58.693
As Oxford was developing its long-term
digital strategy, the rapid emergence of

00:00:58.693 --> 00:01:03.703
generative AI fundamentally altered many
of its assumptions, forcing a rethink

00:01:03.733 --> 00:01:09.073
of investments, priorities, and even the
nature of digital transformation itself.

00:01:09.433 --> 00:01:12.313
Today, we'll explore what leaders
can learn from that experience,

00:01:12.613 --> 00:01:16.843
how AI is reshaping education,
research, and society, and

00:01:16.843 --> 00:01:20.563
why questions of technological
sovereignty and national capability

00:01:20.923 --> 00:01:22.873
are becoming increasingly important.

00:01:23.560 --> 00:01:25.560
Anne thank you so much for joining today.

00:01:26.560 --> 00:01:29.750
For the purposes of this
conversation, give us a little

00:01:29.750 --> 00:01:31.370
bit of background on your career.

00:01:31.986 --> 00:01:32.826
Anne Trefethen: Yes, of course.

00:01:33.346 --> 00:01:36.526
I'm now a Professor of Scientific
Computing at the University

00:01:36.526 --> 00:01:41.626
of Oxford, but I don't have a
traditional academic career.

00:01:42.026 --> 00:01:46.676
I have worked in research on
algorithms pretty much throughout

00:01:46.676 --> 00:01:48.506
my career, but in different forms.

00:01:48.866 --> 00:01:52.296
So sometimes I've been
working with companies.

00:01:52.746 --> 00:01:58.806
Thinking Machines was one of the first
high-performance computing companies that

00:01:58.806 --> 00:02:03.576
sold massively parallel machines, and I
worked with them on parallel algorithms.

00:02:04.576 --> 00:02:09.656
I worked with software companies in
the UK, but I've also led research

00:02:09.716 --> 00:02:15.116
programmes, with what is now UKRI, but
worked with EPSLC and the other research

00:02:15.116 --> 00:02:20.876
councils and found myself at Oxford
setting up an interdisciplinary research

00:02:21.416 --> 00:02:28.746
centre, which is focused on enabling
new research in different disciplines,

00:02:28.746 --> 00:02:31.536
but based on an innovative technologies.

00:02:31.716 --> 00:02:32.856
So I'm a geek at heart.

00:02:33.856 --> 00:02:38.690
Susannah de Jager: Love it and you've
held a number of roles at the university.

00:02:39.110 --> 00:02:41.280
So I'd love to hear a little
bit more about that, please.

00:02:41.470 --> 00:02:48.340
Anne Trefethen: I came to the university
in 2005 to set up this interdisciplinary

00:02:48.340 --> 00:02:53.330
research centre, the Oxford E-Research
Centre, and I was the director of that

00:02:53.330 --> 00:02:58.280
for almost seven years and then moved
into the centre of the university

00:02:58.280 --> 00:03:01.700
to become the first university CIO.

00:03:02.060 --> 00:03:09.750
Where I led a large organisational
change to create a single central

00:03:10.110 --> 00:03:15.360
IT department, the governance around
that, to begin to put in place some

00:03:15.360 --> 00:03:20.010
of the cyber security that we needed,
which was sorely lacking at that time.

00:03:20.790 --> 00:03:25.380
And I did that, I guess I was a CIO
for probably around five years and

00:03:25.410 --> 00:03:29.880
in the meantime, I became a Pro-vice
Chancellor at that time responsible

00:03:30.210 --> 00:03:34.860
for the gardens, libraries and museums,
which was an absolutely fabulous job,

00:03:34.890 --> 00:03:40.180
wonderful colleagues, and from there
became the Pro-vice chancellor for people.

00:03:40.220 --> 00:03:45.080
That was everything to do with pay,
pensions, and people in general.

00:03:45.450 --> 00:03:48.120
Again, a really fascinating job to have.

00:03:48.510 --> 00:03:50.850
And then the Pro-vice
Chancellor for digital.

00:03:51.153 --> 00:03:54.850
So as I say, I'm a geek, so I was the
right person, I think, to lead on that.

00:03:55.070 --> 00:04:00.620
As we came out of the pandemic, it was
clear that the university needed to

00:04:00.650 --> 00:04:07.490
invest in its digital capability in the
infrastructure and everything around that.

00:04:07.820 --> 00:04:13.130
So the then Vice Chancellor asked if
I would lead on that and so I took

00:04:13.130 --> 00:04:18.510
a slight shift back to my roots in a
way to begin to take forward a, what

00:04:18.510 --> 00:04:20.920
we called, digital transformation.

00:04:21.920 --> 00:04:24.780
Susannah de Jager: And you and I have
spoken about that in the time period

00:04:24.780 --> 00:04:31.223
that plan was formulated, AI was a small
consideration and part of it, absolutely,

00:04:32.113 --> 00:04:35.203
but that almost by the time that piece
of work was finished, you realised you'd

00:04:35.203 --> 00:04:37.363
been overtaken by events around you.

00:04:37.543 --> 00:04:38.443
Anne Trefethen: Absolutely.

00:04:38.743 --> 00:04:44.683
So we began work on it in 2022,
and it took some time to work with

00:04:44.683 --> 00:04:46.603
colleagues across the organisation.

00:04:46.933 --> 00:04:51.673
I think it's fair to say people
were exhausted after the pandemic.

00:04:51.763 --> 00:04:56.203
It wasn't a time to bring
about lots more change.

00:04:56.623 --> 00:05:01.543
It was a time to reflect and
understand what we needed to do

00:05:01.633 --> 00:05:05.458
to become more effective, and how
we could use these technologies to

00:05:05.458 --> 00:05:08.518
support our mission in the university.

00:05:08.878 --> 00:05:14.398
So we took our time on creating the
strategy, and that was published

00:05:14.398 --> 00:05:19.532
in 2022 And of course, just
about then, ChatGPT popped up.

00:05:19.742 --> 00:05:24.152
Now, of course, we were
aware of AI in many forms.

00:05:24.242 --> 00:05:27.992
I mean, we're now very focused
on large language models, but AI

00:05:27.992 --> 00:05:32.702
has been part of the technology
shift for quite a long time.

00:05:33.002 --> 00:05:35.672
So as you say, it was
a part of our strategy.

00:05:35.672 --> 00:05:40.562
The strategy covered everything
from our digital infrastructure and

00:05:40.562 --> 00:05:46.242
then had pillars sitting on it that
focused on education, on research,

00:05:46.242 --> 00:05:50.532
and on administration, and then
a layer of innovation across top.

00:05:50.532 --> 00:05:52.602
So we wanted to be
looking out to the future.

00:05:52.842 --> 00:05:57.462
And really, at that time, AI
was part of that innovation.

00:05:58.452 --> 00:06:04.072
A year later, and it had become
evident that we had to rethink

00:06:04.072 --> 00:06:05.392
how we were doing things.

00:06:06.202 --> 00:06:09.112
Was it really that we had to
rethink what we wanted to do?

00:06:09.712 --> 00:06:13.472
But the old assumptions were gone.

00:06:14.005 --> 00:06:17.555
That wasn't just about those
pillars around research and

00:06:18.105 --> 00:06:21.115
education, but simply everything.

00:06:21.115 --> 00:06:27.025
So we had plans for some large new
administrative systems around finance

00:06:27.025 --> 00:06:33.565
and HR and it became clear as time
progressed that the assumptions around

00:06:33.565 --> 00:06:39.265
what those would look like in a few
years time were really no longer valid.

00:06:39.565 --> 00:06:43.555
So we had to step back and
begin to think about how do we

00:06:43.555 --> 00:06:45.835
approach this in a world with AI?

00:06:46.525 --> 00:06:48.895
What are the right
investments to be making?

00:06:49.555 --> 00:06:53.815
And having invested in some of
these areas, it does take a bit of

00:06:53.815 --> 00:06:56.035
courage to say, "No, let's stop.

00:06:56.095 --> 00:07:00.775
Let's rethink." I don't think anybody
yet fully understands the impact

00:07:00.805 --> 00:07:04.675
that we're going to see and where
we're going to see it and where we

00:07:04.675 --> 00:07:12.595
can best benefit from these tools and
technologies versus where we need to be

00:07:12.595 --> 00:07:18.535
more careful, where we need to introduce
more governance, and that's an area

00:07:18.535 --> 00:07:20.995
where, again, we've had to rethink.

00:07:21.415 --> 00:07:24.745
When we set off with the strategy
and we recognised we were

00:07:24.745 --> 00:07:26.305
going to be investing in AI.

00:07:26.995 --> 00:07:31.825
We tried to have a fairly light
touch governance framework

00:07:31.855 --> 00:07:35.845
because the university is a
place that has enough committees.

00:07:36.565 --> 00:07:40.225
We don't need more governance,
more committees, was my thought.

00:07:40.675 --> 00:07:44.815
So let's make sure that the education
committee can oversee what's happening

00:07:44.815 --> 00:07:46.495
in education, et cetera, et cetera.

00:07:47.185 --> 00:07:52.585
But the pace of change is such that we
have had to introduce more governance.

00:07:52.585 --> 00:07:58.105
We've had to introduce new ways of
thinking about governance and making

00:07:58.105 --> 00:08:04.855
these tools available and how we do
that and what we do and how we have to

00:08:04.855 --> 00:08:11.615
listen to the voice of, essentially,
as many of our community as we can.

00:08:12.365 --> 00:08:16.885
I think the other thing to note
is that the use of AI and the

00:08:16.885 --> 00:08:21.295
benefits of it are very different in
different areas of the university.

00:08:21.715 --> 00:08:24.205
We're a microcosm of society, really.

00:08:24.755 --> 00:08:28.702
In some areas of research,
it's massively impactful.

00:08:29.072 --> 00:08:30.222
In others, less so.

00:08:30.572 --> 00:08:35.402
In education, fabulously useful
for students in some areas.

00:08:35.612 --> 00:08:40.082
Others where we're having to say,
"That's not the right approach." But

00:08:40.082 --> 00:08:41.402
we're still figuring all that out.

00:08:42.402 --> 00:08:47.884
Susannah de Jager: Yeah and it's a big
ecosystem to be grappling with while this

00:08:47.884 --> 00:08:49.474
pace of change is going on around you.

00:08:49.474 --> 00:08:54.004
I've spoken to even rather small,
quite simple businesses, who have said

00:08:54.004 --> 00:08:57.408
they raised for a two-year strategy,
and they've already realised that

00:08:57.418 --> 00:09:01.915
the money they raised will last them
five years, potentially, because they

00:09:01.975 --> 00:09:06.131
can use AI so expeditiously that it
completely changes things and that's

00:09:06.581 --> 00:09:08.321
in a pretty straightforward business.

00:09:08.321 --> 00:09:11.891
So I can only imagine the complexity
that you are grappling with.

00:09:12.630 --> 00:09:17.190
Leading a big organisation through
this changing period, how would

00:09:17.190 --> 00:09:19.620
you advise other people that
are grappling with these issues?

00:09:19.620 --> 00:09:23.261
What are the kind of core frameworks that
you think people need to be considering?

00:09:23.520 --> 00:09:26.016
Anne Trefethen: Yeah, maybe I
should say a little bit more about

00:09:26.096 --> 00:09:30.956
how we approach things at the
university in terms of beginning to

00:09:30.956 --> 00:09:33.663
roll out technology to everybody.

00:09:34.353 --> 00:09:38.463
I think first of all, to say that
we did do a number of pilots.

00:09:38.463 --> 00:09:45.993
So beginning back in 2023 we expanded
those pilots enormously in 2024 and

00:09:45.993 --> 00:09:51.903
it was because at that point, it
became clear, absolutely clear, that

00:09:51.903 --> 00:09:54.363
this technology was not going away.

00:09:54.393 --> 00:09:59.253
This technology was here to stay and
part of what we needed as a university

00:09:59.583 --> 00:10:05.463
to both be using but teaching and
engaging in the conversation where what

00:10:05.463 --> 00:10:07.533
do we want this to be in the future?

00:10:08.073 --> 00:10:09.633
I think it's really important.

00:10:10.553 --> 00:10:15.023
This isn't a technology that's just
being built in these frontier labs.

00:10:16.023 --> 00:10:21.933
It is something that I think we all need
to be involved in the conversation about.

00:10:22.323 --> 00:10:23.403
How do we use it?

00:10:23.403 --> 00:10:24.963
How do we want to see it developed?

00:10:25.323 --> 00:10:28.053
How do we want it as
part of the education?

00:10:28.893 --> 00:10:35.673
Back in 2023, we had those at the
bleeding edge, such as myself, who are

00:10:35.673 --> 00:10:41.963
always first to try, and it was being
used in the university in a ad hoc way.

00:10:42.173 --> 00:10:47.423
Research groups who had recognised the
value had adopted it, not necessarily

00:10:47.483 --> 00:10:52.133
in a secure way because we didn't
really understand at the time.

00:10:52.545 --> 00:10:54.593
Susannah de Jager: And this is, I
think, a really interesting thing.

00:10:54.593 --> 00:10:59.464
Just sorry to interrupt, but you've got
an issue that the university doesn't

00:10:59.464 --> 00:11:05.644
have the luxury of waiting to put in a
governance structure because events will

00:11:05.644 --> 00:11:09.514
overtake you both externally, but also
people will just start adopting things.

00:11:09.934 --> 00:11:13.864
And so you're having to grapple
with that reality that people are

00:11:13.864 --> 00:11:17.504
going to start using these tools
and if you're not there with them,

00:11:17.654 --> 00:11:19.084
it just won't have any governance.

00:11:19.324 --> 00:11:21.794
So it's dealing with
an imperfect situation.

00:11:21.854 --> 00:11:26.944
Anne Trefethen: Absolutely and for some of
my colleagues, this was really difficult.

00:11:27.084 --> 00:11:27.624
Susannah de Jager: I can imagine.

00:11:27.864 --> 00:11:29.134
Anne Trefethen: It was really difficult.

00:11:29.254 --> 00:11:36.674
Some, when we began to roll out
availability to ChatGPT Edu for all of

00:11:36.674 --> 00:11:44.084
the students and all staff, we had some
colleagues respond in a very visceral

00:11:44.084 --> 00:11:48.794
way because they felt this was not what
an academic institution should be doing.

00:11:48.914 --> 00:11:49.214
Susannah de Jager: Yeah.

00:11:49.674 --> 00:11:50.726
Well, it's iterative.

00:11:50.814 --> 00:11:56.331
It's much more startup mindset than it
is traditional academic institution.

00:11:56.871 --> 00:11:58.313
But here, the two are meeting.

00:11:58.791 --> 00:11:59.331
Anne Trefethen: Exactly.

00:11:59.871 --> 00:12:03.871
What we couldn't stop, we couldn't stop
our students using these technologies.

00:12:04.071 --> 00:12:05.031
They were already doing it.

00:12:05.091 --> 00:12:10.941
We knew already that there were
over 5,000 of the university had

00:12:10.941 --> 00:12:17.267
used their Oxford email address to
create an account, a ChatGPT account.

00:12:17.354 --> 00:12:18.074
We knew that.

00:12:18.674 --> 00:12:23.174
So if there are over 5,000 that have
done that, we know there are many

00:12:23.174 --> 00:12:26.924
more who are using it with their
own email addresses, et cetera.

00:12:28.064 --> 00:12:33.651
So it was happening, but it was happening
in, as I said, first of all, an unsecure

00:12:33.651 --> 00:12:37.281
way, but also not in an equitable way.

00:12:37.581 --> 00:12:43.221
Students who could afford to have licences
or who could were getting very good tools

00:12:43.221 --> 00:12:47.911
and other students didn't have access
to that same capability and for the

00:12:47.911 --> 00:12:53.751
university, that also became an issue of
how do we ensure that there is equitable

00:12:53.781 --> 00:12:57.231
access or as best that we can make it.

00:12:57.731 --> 00:13:00.131
For me, there was a key moment.

00:13:00.491 --> 00:13:03.761
It was in 2024, as I say,
we'd done some pilots.

00:13:04.691 --> 00:13:09.491
We were talking to companies,
Microsoft, OpenAI, and others,

00:13:09.911 --> 00:13:14.361
around what we might be able to do
as a platform for the university.

00:13:15.361 --> 00:13:20.491
And I got invited to an
OpenAI executive briefing.

00:13:20.701 --> 00:13:23.131
Now, I don't always go to those
things because they're rather

00:13:23.131 --> 00:13:28.181
fluffy generally and it's not
necessarily a good use of time.

00:13:28.571 --> 00:13:31.931
But on this occasion, I decided I would
go along to see what they had to say.

00:13:32.791 --> 00:13:39.353
Just about a month before, we'd
had Geoffrey Hinton and his

00:13:39.353 --> 00:13:41.423
lecture was a real eye-opener.

00:13:41.483 --> 00:13:47.453
So his lecture was around, will
artificial intelligence become

00:13:47.453 --> 00:13:49.463
more than biological intelligence?

00:13:49.853 --> 00:13:55.307
And one of the startling things that
he said was he had always thought that

00:13:55.307 --> 00:14:00.257
it would take a century or so to get to
super intelligence and he now believed

00:14:00.257 --> 00:14:02.237
we would be there within a decade.

00:14:03.237 --> 00:14:07.737
He had a very downbeat
vision of the future.

00:14:07.844 --> 00:14:12.531
This is a man who knows
these things very well.

00:14:12.741 --> 00:14:19.717
So I'd been mulling that over and then
I went to the OpenAI executive briefing.

00:14:20.717 --> 00:14:25.307
They were essentially showing us the
capabilities of what was to become

00:14:25.667 --> 00:14:29.357
GPT-4.0 that was going to be announced.

00:14:29.859 --> 00:14:31.929
I think it came out in May 2024.

00:14:32.646 --> 00:14:37.446
The capability beyond
ChatGPT-3 was astonishing.

00:14:37.446 --> 00:14:45.456
It had multimodal capabilities, images,
voice and I could see immediately that

00:14:45.486 --> 00:14:53.066
we were progressing at a pace that I had
never seen before and the other thing

00:14:53.066 --> 00:14:59.256
that was very notable in talking to
the engineers from OpenAI was that they

00:14:59.316 --> 00:15:06.456
themselves were excited and startled
by the progress, the capabilities,

00:15:06.516 --> 00:15:09.366
of the models in such short time.

00:15:10.556 --> 00:15:13.516
That really was an eye-opener for me.

00:15:13.756 --> 00:15:17.756
That evening, I was lucky enough to
have dinner with Sam Altman after

00:15:17.756 --> 00:15:23.666
the event and I mentioned to him
Jeffrey Hinton's view of a decade.

00:15:24.416 --> 00:15:28.016
And he said "No" he
though it was five years.

00:15:28.946 --> 00:15:34.273
I did ask him "Well, what are we doing
about this? Should we be concerned?"

00:15:35.023 --> 00:15:39.373
And his view was if I was asking
him, was there an existential threat?

00:15:39.493 --> 00:15:45.643
He couldn't say, "Absolutely no. There
isn't." But that they are releasing

00:15:45.643 --> 00:15:50.683
the models iteratively after testing
and I can see that's happening.

00:15:50.713 --> 00:15:56.533
We've seen very recently requests
slow down the development.

00:15:56.676 --> 00:15:59.413
Susannah de Jager: And yesterday, as
you and I were discussing, Sam Altman,

00:15:59.443 --> 00:16:04.293
as we record this, posted this article
in the FT talking about exactly this.

00:16:05.023 --> 00:16:09.883
Anne Trefethen: I think we have known
for some time that we really need to have

00:16:09.913 --> 00:16:12.433
international collaboration on these.

00:16:13.093 --> 00:16:14.923
These go beyond boundaries.

00:16:15.309 --> 00:16:20.539
So it is interesting that he is
now one of the voices that is that

00:16:20.539 --> 00:16:23.509
is now asking for us to do that.

00:16:24.139 --> 00:16:28.729
And it's interesting too that
just a few days before Anthropic

00:16:29.279 --> 00:16:31.589
was suggesting that we slow down.

00:16:32.039 --> 00:16:37.819
Of course, it's some time ago now that
there was a open letter signed by many

00:16:37.819 --> 00:16:41.209
hundreds of AI leaders to slow down.

00:16:41.209 --> 00:16:43.459
But the genie is out of the bottle.

00:16:44.059 --> 00:16:46.549
The capabilities are already out there.

00:16:46.549 --> 00:16:54.973
So whilst I can understand this
desire now to let society catch up, I

00:16:54.973 --> 00:16:56.593
think we've got to do more than that.

00:16:56.643 --> 00:16:58.743
We've got a lot of open models.

00:16:59.223 --> 00:17:04.443
There are models now that are simply
distillation of these very large

00:17:04.553 --> 00:17:09.663
language models and what that means is
that they have very much of the same

00:17:09.663 --> 00:17:14.314
capability, not all the same capability,
very much the same capability, but

00:17:14.314 --> 00:17:17.774
able to run on much smaller systems.

00:17:18.374 --> 00:17:23.624
So we've really got to, I think,
collectively be thinking about

00:17:24.634 --> 00:17:29.604
their use, how we use them, where
we use them, and we've got to do

00:17:29.604 --> 00:17:36.064
the education, the AI literacy, to
enable that at a very broad level.

00:17:36.124 --> 00:17:37.714
It's not just for the leaders.

00:17:37.714 --> 00:17:40.084
It really is, I think, a collective issue.

00:17:40.444 --> 00:17:47.644
And that was why, I thought, I believed,
the university had to grapple with this.

00:17:47.644 --> 00:17:49.534
We have to be part of that conversation.

00:17:50.194 --> 00:17:54.097
And without, I don't want to say
experimentation with our students,

00:17:54.097 --> 00:17:57.277
it's not experimentation, but
without those conversations about

00:17:58.147 --> 00:17:59.527
what should we use this for?

00:17:59.527 --> 00:18:01.597
How can we inform what we're doing?

00:18:02.347 --> 00:18:08.467
Is this going to help us in how we
develop critical thinking or is it not?

00:18:09.407 --> 00:18:18.627
How do we remain resilient and able to
take things forward in an age where we

00:18:19.017 --> 00:18:21.387
perhaps are becoming dependent upon AI?

00:18:22.227 --> 00:18:24.147
Where there's a lot of misinformation?

00:18:25.097 --> 00:18:29.807
So I think these are really societal
issues that need to be grappled with

00:18:30.257 --> 00:18:35.347
at the university as much as anywhere
else and perhaps more because we are

00:18:35.767 --> 00:18:38.197
creating the next generation of leaders.

00:18:40.126 --> 00:18:42.646
Susannah de Jager: This Oxford Plus
miniseries is brought to you by

00:18:42.646 --> 00:18:45.136
Equinox, Equitable Innovation Oxford.

00:18:45.916 --> 00:18:49.636
Equinox is a major regional partnership
established by the University of

00:18:49.636 --> 00:18:53.776
Oxford to drive inclusive innovation
and sustainable economic growth.

00:18:54.406 --> 00:18:57.466
Bringing together universities,
local authorities, government,

00:18:57.466 --> 00:19:01.726
industry, investors, and community
organisations, Equinox provides

00:19:01.756 --> 00:19:05.716
a unified voice for Oxfordshire,
championing innovation that delivers

00:19:05.746 --> 00:19:08.476
economic impact alongside social value.

00:19:08.986 --> 00:19:12.646
More than 80 organisations have signed
the Equinox charter, committing to

00:19:12.676 --> 00:19:16.486
atract investment, remove barriers
to growth, and ensure the benefits

00:19:16.486 --> 00:19:18.466
of innovation are shared more widely.

00:19:18.976 --> 00:19:23.866
By aligning Oxfordshire's world-leading
research, enterprise, and talent, Equinox

00:19:23.866 --> 00:19:28.696
aims to unlock nationally significant
growth, reinforce the county's position

00:19:28.696 --> 00:19:33.046
as a global innovation hub, and ensure
that success creates skilled jobs,

00:19:33.316 --> 00:19:37.756
attracts investment, and delivers
lasting prosperity across the region.

00:19:41.862 --> 00:19:47.167
It brings us very naturally onto this
next part of your journey in this space,

00:19:47.167 --> 00:19:51.474
which is that you're also a trustee
of the Alan Turing Institute, and you

00:19:51.474 --> 00:19:55.794
sound wonderfully well prepared for
the issues that they are facing at a

00:19:55.794 --> 00:19:58.854
national level because it's very similar.

00:19:58.854 --> 00:20:01.914
You have the pace of company
adoption of individual adoption.

00:20:02.034 --> 00:20:07.434
Indeed, I imagine in some government
areas, adoption with governance and

00:20:07.434 --> 00:20:09.324
regulation effectively lagging that.

00:20:09.424 --> 00:20:12.874
There's this sort of push-pull
between, we want to get it right, but

00:20:12.874 --> 00:20:16.594
it's going to have to be iterative
and we're going to have to make,

00:20:17.254 --> 00:20:22.084
ultimately probably, some mistakes
and be able to pivot because otherwise

00:20:22.084 --> 00:20:23.434
you're just even people in a vacuum.

00:20:23.750 --> 00:20:26.930
I'd love to hear about some of
the conversations and the work

00:20:26.930 --> 00:20:29.883
that you're contributing to
at the  Alan Turing Institute.

00:20:30.380 --> 00:20:30.540
Anne Trefethen: Yeah.

00:20:30.747 --> 00:20:36.837
I think one of the things that I would
note is that no one Institute can do

00:20:36.837 --> 00:20:40.377
everything that's needed to be done and
I think the Alan Turing Institute is

00:20:40.377 --> 00:20:45.087
therefore in a great position as our
National AI Institute to be a convener.

00:20:45.837 --> 00:20:51.327
To be bringing together those developers
from frontier labs, for those researchers

00:20:51.357 --> 00:20:57.657
in the universities who are looking
at not only how these technologies are

00:20:57.704 --> 00:21:01.847
changing the way we approach science, but
they're also looking at what comes next.

00:21:01.847 --> 00:21:04.097
What are the next AI models?

00:21:04.407 --> 00:21:09.237
You may have seen just last week an
announcement the government investing

00:21:09.237 --> 00:21:16.317
in two new AI labs looking at the
future of models and the Turing have

00:21:16.317 --> 00:21:22.317
a role there to help understand where
those might help us in the future.

00:21:22.317 --> 00:21:28.916
But the Turing has been around now
since 2015, before this generation

00:21:28.916 --> 00:21:34.546
of AI and we're just at a pivotal
moment where we are really

00:21:34.644 --> 00:21:36.832
focusing on national resilience.

00:21:37.282 --> 00:21:42.712
We have a new Chief Executive, George
Williamson, who is brilliant and he

00:21:42.712 --> 00:21:50.692
just arrived in May full of ideas and
vision and we're beginning to develop

00:21:50.692 --> 00:21:57.382
a new strategy for the Institution
with the help of the researchers

00:21:57.382 --> 00:22:01.282
and staff at the institution,
but also with our stakeholders.

00:22:02.122 --> 00:22:07.532
I think it's interesting to think about
what is resilience in this time of AI?

00:22:07.951 --> 00:22:11.056
It encompasses many elements.

00:22:11.596 --> 00:22:15.616
So we need to have a sovereign
capability to be resilient.

00:22:16.156 --> 00:22:21.142
We can't, as we saw recently new
models being made available and then

00:22:21.142 --> 00:22:23.632
being switched off to international.

00:22:23.932 --> 00:22:29.752
If we are only dependent upon models
and capabilities being developed

00:22:29.992 --> 00:22:32.392
elsewhere, those could disappear.

00:22:32.392 --> 00:22:34.565
We need to be able to
be resilient to that.

00:22:35.255 --> 00:22:38.945
That may come from creating
our own capabilities.

00:22:39.395 --> 00:22:44.435
Not likely to be developing our
own sovereign models, but possibly.

00:22:44.999 --> 00:22:51.209
AI is permeating everything, be it
cybersecurity, be it defence, and

00:22:51.229 --> 00:22:56.549
we need to be able to protect our
critical infrastructure where we

00:22:56.549 --> 00:23:02.984
can see that AI is being used by
adversaries and we need to be able to

00:23:03.044 --> 00:23:04.994
defend our critical infrastructure.

00:23:05.624 --> 00:23:12.594
But also, it's about ensuring
that we have the human capacity.

00:23:13.074 --> 00:23:19.704
So training, ensuring that we are ensuring
we've got enough people who understand

00:23:19.704 --> 00:23:27.474
these technologies and where they meet
these very critical application areas.

00:23:28.224 --> 00:23:30.234
It's not just about the models.

00:23:30.264 --> 00:23:35.964
It's about how do they enable
us in these critical areas.

00:23:36.294 --> 00:23:42.534
So the Turing is focusing now on
taking forward that national resilience

00:23:42.564 --> 00:23:48.864
capability, working with our stakeholders,
but reaching down into the models, being

00:23:48.864 --> 00:23:53.884
able to understand them, understanding
what they bring and where we need to

00:23:53.884 --> 00:23:59.174
be aware that others will be using
them and I know that George is very

00:23:59.174 --> 00:24:04.454
interested in misinformation and how
can we build a capability that's helping

00:24:04.454 --> 00:24:09.194
the nation to deal with the fact that
there is so much misinformation that

00:24:09.194 --> 00:24:13.003
is now also fueled by AI capabilities.

00:24:13.567 --> 00:24:16.033
Susannah de Jager: I had a very
interesting conversation for this

00:24:16.033 --> 00:24:20.743
miniseries with Philippa Webb from
the Oxford Institute of Technology and

00:24:20.743 --> 00:24:26.343
Justice and she was talking about just
the volume of these kind of national

00:24:26.343 --> 00:24:32.233
level attacks, and that I think she was
saying 77% of them are originating in

00:24:32.353 --> 00:24:35.413
Russia, Iran, North Korea, and China.

00:24:36.170 --> 00:24:37.140
It's happening.

00:24:37.216 --> 00:24:37.546
Anne Trefethen: Yeah.

00:24:37.966 --> 00:24:39.306
It absolutely is and you're right.

00:24:39.306 --> 00:24:44.256
I think in many elements of what
we're trying to do in working with

00:24:44.396 --> 00:24:49.016
AI, and I would say that AI is now a
collaborator and not necessarily a tool.

00:24:49.433 --> 00:24:55.166
There is always a tension and it's
getting the balance of that tension right.

00:24:55.686 --> 00:24:55.926
Susannah de Jager: Yeah.

00:24:56.329 --> 00:24:59.509
You were touching upon, the training
of our own models not creating

00:24:59.509 --> 00:25:01.759
dependencies and having true resilience.

00:25:02.059 --> 00:25:06.253
There seems to be quite a defined
kind of two schools of though in this

00:25:06.253 --> 00:25:07.963
domain from the conversations I've had.

00:25:07.963 --> 00:25:11.713
And one is that we are going to have
to train our own foundational models in

00:25:11.713 --> 00:25:14.143
order to not have those dependencies.

00:25:14.353 --> 00:25:17.503
And the other is that they are going to
become so commoditized that even if one

00:25:17.503 --> 00:25:21.863
gets switched off, you'll have another,
and therefore you're not very dependent

00:25:21.923 --> 00:25:25.393
and you have some resilience just by
as I say, they've become commoditized.

00:25:25.813 --> 00:25:31.553
And that perhaps the UK with resources,
but finite resources, should be

00:25:31.913 --> 00:25:36.293
focused more on the applications
and our particular areas of

00:25:36.473 --> 00:25:43.803
specialisation and using that as our
technological leverage in this space.

00:25:43.823 --> 00:25:44.580
Do you have a view?

00:25:45.570 --> 00:25:51.830
Anne Trefethen: So I think in terms of
the very large-scale language models that

00:25:51.890 --> 00:25:57.171
are being invested in now, that's not
something we are likely to be able to.

00:25:57.201 --> 00:25:59.901
I'm not sure we've got the
energy in the country to do it.

00:26:00.561 --> 00:26:02.956
Susannah de Jager: Got to get First Light
Fusion online before we can do that.

00:26:03.956 --> 00:26:04.226
Anne Trefethen: Yeah.

00:26:04.306 --> 00:26:04.946
Oh, we're close.

00:26:05.296 --> 00:26:06.386
It's only 10 years away.

00:26:08.442 --> 00:26:11.535
In ideal world that's
maybe what we'd like to do.

00:26:11.972 --> 00:26:15.332
You know, there are discussions
across various countries about doing

00:26:15.332 --> 00:26:20.712
that collaboratively and maybe that
will be approach that will be taken.

00:26:21.132 --> 00:26:26.592
But I think in some of the nationally
important areas and I include education in

00:26:26.592 --> 00:26:30.732
that myself, although that's not an area
that the Turing Institute is focused on.

00:26:30.782 --> 00:26:35.402
But in some of these areas, we need
to have models that we can trust and

00:26:35.402 --> 00:26:39.452
to be able to trust them, we need
to know what were they trained on?

00:26:39.532 --> 00:26:40.822
What are the data?

00:26:41.422 --> 00:26:46.415
I mentioned earlier the possibility of
using distillation or a distilled model.

00:26:46.415 --> 00:26:50.548
So a model that is based on these
large language models, but it's

00:26:50.548 --> 00:26:54.598
distilled, and then we can use
something called post-training to

00:26:54.598 --> 00:26:57.868
be able to focus on specific areas.

00:26:57.868 --> 00:27:03.358
And I think, for me, that's the most
likely way we will make progress quickly

00:27:03.788 --> 00:27:06.538
in that it will require less power.

00:27:06.998 --> 00:27:12.858
But it will allow us to have models
we can trust in specific areas.

00:27:13.008 --> 00:27:15.768
That may be used for national
security, for instance.

00:27:15.858 --> 00:27:19.518
But I think, again, there is a
balance to be made and as the

00:27:19.518 --> 00:27:25.061
technologies are moving forward they
take less power than they used to.

00:27:25.581 --> 00:27:31.641
So we may find in a year's time
that whole equation has shifted.

00:27:32.251 --> 00:27:36.702
One thing I would mention is that
being a mature person in this…

00:27:37.692 --> 00:27:39.972
Susannah de Jager: I think
ChatGPT would tell us seasoned.

00:27:40.207 --> 00:27:41.247
Anne Trefethen: Seasoned, okay.

00:27:42.372 --> 00:27:43.722
Yeah, I've seen many seasons.

00:27:44.472 --> 00:27:49.872
It used to be that I would go to
computer companies as an algorithm

00:27:49.872 --> 00:27:51.492
developer, as a software developer.

00:27:51.852 --> 00:27:54.582
I would go and visit them and
they would show me a roadmap of,

00:27:54.882 --> 00:27:56.822
three, four, five years even.

00:27:57.162 --> 00:28:01.011
The chip design, the network
architectures that they were working on.

00:28:01.251 --> 00:28:05.813
So they could predict pretty
well what we would be seeing for

00:28:05.813 --> 00:28:07.643
sale in three, five years time.

00:28:07.793 --> 00:28:11.033
You go to these companies now,
they're simply not able to do that.

00:28:11.273 --> 00:28:13.763
They're simply not able to
really give you a roadmap

00:28:13.763 --> 00:28:16.883
beyond, three months I would say.

00:28:17.223 --> 00:28:23.463
So it's very difficult in that environment
to be able to say with any confidence at

00:28:23.703 --> 00:28:26.763
all of what approach one should be taking.

00:28:26.763 --> 00:28:30.863
I think we just have to do the
best that we can and for me at the

00:28:30.863 --> 00:28:35.963
moment, it is building on distilled
models to have a very focus.

00:28:35.963 --> 00:28:38.303
So it's kind of in between
what you were saying.

00:28:38.722 --> 00:28:42.982
I think the other interesting thing
that you can see that supports that

00:28:42.982 --> 00:28:47.632
trend is that if you look at what
the hardware people are beginning to

00:28:47.632 --> 00:28:52.551
develop, it is the kind of desktop that
is powerful to run your own models.

00:28:52.996 --> 00:28:57.557
That's going to lead to a commoditization,
I think, as you suggested.

00:28:58.087 --> 00:28:59.987
Susannah de Jager: So there's
really two elements here.

00:29:00.017 --> 00:29:03.557
One is the sort of protective
resilience side of it and then

00:29:03.557 --> 00:29:05.297
the other is the opportunity.

00:29:05.297 --> 00:29:10.817
And I think what's quite exciting is that
the UK has lots of sovereign data sets and

00:29:10.947 --> 00:29:15.682
you and I were kindly introduced by Lionel
Tarasanko, and he talks about, obviously

00:29:15.712 --> 00:29:19.042
NHS data, BBC data, the Met Office.

00:29:19.522 --> 00:29:24.622
And so there's quite a big opportunity
here for the UK to turn some of those

00:29:24.832 --> 00:29:28.642
data sets into really valuable assets.

00:29:28.692 --> 00:29:34.406
Anne Trefethen: Yeah, absolutely
and that is where we can begin to

00:29:35.036 --> 00:29:40.546
develop models that, if you like,
exhibit our values rather than

00:29:40.576 --> 00:29:47.086
another nation's values and those data
sets are incredibly valuable to us.

00:29:47.116 --> 00:29:53.056
And I think are what we would use to be
able to build our sovereign capability.

00:29:53.106 --> 00:29:57.606
My colleague at the Turing just
published a really interesting

00:29:57.606 --> 00:29:59.886
blog on this on sovereign AI.

00:29:59.886 --> 00:30:01.956
How can we create resilience?

00:30:02.496 --> 00:30:07.836
And that's one of the things that
he picks up too, that as we begin to

00:30:07.986 --> 00:30:13.626
develop models we trust, so they may
have this base of a distilled language

00:30:13.626 --> 00:30:15.396
model that we take from elsewhere.

00:30:15.696 --> 00:30:18.606
We will see whether we start from scratch.

00:30:18.946 --> 00:30:25.473
But absolutely, those are valuable
assets that will be able to give us

00:30:25.473 --> 00:30:30.423
models that we can trust in education,
that we can trust for our national

00:30:30.803 --> 00:30:33.233
resilience, security, defence.

00:30:33.653 --> 00:30:38.603
We don't want to be building things
that are absolutely critical that need

00:30:38.603 --> 00:30:46.193
to be able to help us take decisions
quickly on essentially the ground

00:30:46.193 --> 00:30:48.633
that we don't understand or know.

00:30:48.978 --> 00:30:49.278
Susannah de Jager: Yeah.

00:30:49.728 --> 00:30:53.328
And again, in another conversation
for this miniseries, I spoke to Audrey

00:30:53.328 --> 00:30:58.788
Tang, who was the digital minister for
Taiwan, and she was very much saying we

00:30:58.788 --> 00:31:05.118
can bound models to train on data sets
and that serve the communities we want

00:31:05.118 --> 00:31:06.928
them to and the values we want them to.

00:31:07.208 --> 00:31:07.937
Anne Trefethen: I agree.

00:31:07.937 --> 00:31:11.417
And we need to learn from that, I think.

00:31:11.517 --> 00:31:18.567
My colleagues at the university have set
up AI Education Oxford University Hub.

00:31:18.567 --> 00:31:23.667
Which is bringing together they've
got over a hundred different groups

00:31:23.717 --> 00:31:30.297
internationally that are from
policymakers, from educators, from

00:31:30.297 --> 00:31:34.197
different institutions that are
doing research, to try to learn

00:31:34.197 --> 00:31:39.027
from each other on approaches
that we can take in understanding.

00:31:39.537 --> 00:31:44.574
This is focused on education through
schools and higher education.

00:31:45.564 --> 00:31:50.551
Because these tools are being rolled
out everywhere and some approaches

00:31:50.551 --> 00:31:54.841
will win and some won't and we need
to collectively understand that.

00:31:55.111 --> 00:31:59.161
And the only way we can do that is
to have these international forums

00:31:59.431 --> 00:32:00.657
that we're beginning  to share.

00:32:01.497 --> 00:32:03.837
Susannah de Jager: Earlier you used the
expression, the genie out of the bottle.

00:32:04.407 --> 00:32:07.167
Which makes us feel a little bit
sort of the whole thing's cantering

00:32:07.167 --> 00:32:08.247
and we're tripping into it.

00:32:09.147 --> 00:32:12.744
Are you optimistic or
more pessimistic, Anne?

00:32:13.547 --> 00:32:15.797
Anne Trefethen: I am by
nature an optimistic person.

00:32:16.372 --> 00:32:22.541
I have to say that whilst I was
leading the move into bringing AI into

00:32:22.541 --> 00:32:27.941
the university, I did always tell my
colleagues, " I have drunk the Kool-Aid."

00:32:28.841 --> 00:32:35.441
Just to let them know, I recognise in
myself that I see a lot of good in this,

00:32:35.441 --> 00:32:38.531
but I do recognise that it isn't all good.

00:32:38.531 --> 00:32:41.171
So I am optimistic.

00:32:41.411 --> 00:32:45.491
I think these articles by the
likes of Sam Altman and others.

00:32:45.838 --> 00:32:50.278
There is a voice now saying let's
work together and understand these.

00:32:50.278 --> 00:32:51.808
We recognise there are dangers.

00:32:51.808 --> 00:32:57.918
But, when you look at the science advances
that we will be able to make using these

00:32:57.918 --> 00:33:04.328
technologies the capabilities that we will
have, it's difficult not to be optimistic.

00:33:04.958 --> 00:33:05.632
But how about you?

00:33:06.413 --> 00:33:07.613
Susannah de Jager: Ooh, interesting.

00:33:08.213 --> 00:33:12.533
I'm a layperson, so I can't profess
to have the same amount of knowledge

00:33:12.533 --> 00:33:13.523
in the domain that you have.

00:33:13.523 --> 00:33:17.993
But in these series of conversations
that I have been having, I have

00:33:17.993 --> 00:33:21.623
come away reassured that the
discourse we are often fed in

00:33:21.623 --> 00:33:25.203
newspapers, which is quite panicked.

00:33:25.353 --> 00:33:27.373
It's happening to us and this
is going to end the world and

00:33:27.373 --> 00:33:28.563
we're all going to lose our jobs.

00:33:29.313 --> 00:33:34.343
I have felt that people that are closer
to it, both in practise and policy and the

00:33:34.343 --> 00:33:40.103
technological side and within institutions
that will set regulation, I've come away

00:33:40.103 --> 00:33:41.993
thinking, no, there are things we can do.

00:33:42.443 --> 00:33:48.893
It can be constrained and that actually
those listening who are perhaps more

00:33:48.893 --> 00:33:55.173
in my camp need to be adding our voice
to pressure to make sure that is the

00:33:55.173 --> 00:34:00.573
case because it's not a good enough
to just say the horse has bolted.

00:34:00.723 --> 00:34:02.823
There are things we can do and should do.

00:34:03.573 --> 00:34:05.954
But that makes me optimistic
broadly, I would say.

00:34:06.954 --> 00:34:07.234
Anne Trefethen: Good.

00:34:07.794 --> 00:34:07.954
Yeah.

00:34:08.390 --> 00:34:14.810
When you look at it, every generation
has had a technological shift.

00:34:15.021 --> 00:34:17.511
Perhaps not at this pace.

00:34:18.231 --> 00:34:23.631
One of the startling things to
remind ourselves is that when ChatGPT

00:34:23.961 --> 00:34:30.141
was made available in 2022, within
five days, it had a million users.

00:34:30.191 --> 00:34:33.161
Within two months, it
had 100 million users.

00:34:33.371 --> 00:34:36.971
There is nothing that
has moved at that pace.

00:34:37.661 --> 00:34:42.791
But the point you just made,
this isn't happening to us.

00:34:43.791 --> 00:34:51.121
We can be part of, we need to
engage with, and it isn't just

00:34:51.121 --> 00:34:52.801
for those people in the labs.

00:34:53.341 --> 00:34:56.851
These types of conversations,
I think, make a difference.

00:34:57.851 --> 00:35:02.681
Susannah de Jager: And so where do
you hope we are in 10 years time?

00:35:03.681 --> 00:35:04.071
Anne Trefethen: Ooh.

00:35:04.631 --> 00:35:05.231
Gosh.

00:35:06.231 --> 00:35:08.012
Could we say two years time?

00:35:08.172 --> 00:35:10.252
Susannah de Jager: Fine, scrap that.

00:35:10.372 --> 00:35:12.552
Where do you hope we will
be in two years time, Anne?

00:35:12.857 --> 00:35:15.437
Anne Trefethen: I hope that we
are happily having a similar

00:35:15.437 --> 00:35:22.247
conversation having learned a great
deal more, having an evidence base.

00:35:22.947 --> 00:35:25.767
As a scientist, I want an evidence base.

00:35:26.171 --> 00:35:32.861
I hope that it is showing that we
have learned how to work with AI

00:35:33.281 --> 00:35:36.251
in a positive, constructive way.

00:35:36.821 --> 00:35:43.491
That we have a next generation of students
who are not afraid of the future, who are

00:35:43.491 --> 00:35:47.391
curious about the future, who are engaging
with the future, but are not afraid.

00:35:47.811 --> 00:35:54.441
At the moment, I think we are creating
fear by saying all the jobs are going,

00:35:55.131 --> 00:35:57.171
when actually the jobs will be changing.

00:35:57.171 --> 00:36:02.511
So I hope in two years time, we'll be
in a position where we'll have more

00:36:02.571 --> 00:36:10.911
of an understanding of that impact
and one that allows the younger

00:36:10.911 --> 00:36:17.301
generations to be less fearful and
to be able to know how to adapt to

00:36:17.301 --> 00:36:18.501
the future that we're going to have.

00:36:19.022 --> 00:36:21.441
Susannah de Jager: And I think
that that resonates so much.

00:36:21.441 --> 00:36:24.021
You know, I touched earlier upon
companies saying they're being

00:36:24.021 --> 00:36:26.871
able to expedite their plans,
potentially hire fewer people.

00:36:27.351 --> 00:36:30.501
But the part that I didn't say
is that what they're realising is

00:36:30.501 --> 00:36:33.351
that it expedites their plans so
much that they're not hiring the

00:36:33.351 --> 00:36:34.781
people they thought they might.

00:36:35.011 --> 00:36:39.721
But they do need to hire other people
to catch the kind of amplification

00:36:39.861 --> 00:36:43.311
of whatever their product or their
plans or their go-to-market is.

00:36:43.617 --> 00:36:48.657
I completely agree with you that we're
seeing the first tranche and even now, I'm

00:36:48.657 --> 00:36:53.817
already hearing about the second, we don't
yet know what that will quite look like.

00:36:53.967 --> 00:36:58.587
But there are going to be jobs created
by the speed and I, like you, I'm

00:36:58.587 --> 00:37:01.757
very hopeful that we're just going
to have more information on how to

00:37:01.767 --> 00:37:04.797
train our children to be those people.

00:37:04.833 --> 00:37:05.382
Anne Trefethen: Yes.

00:37:05.652 --> 00:37:09.102
As you think about it, so a lot of
the entry-level jobs are the ones that

00:37:09.102 --> 00:37:16.842
are going and we both know that when
you go into an entry-level job, you're

00:37:16.842 --> 00:37:21.462
not just there learning the skills for
that job, you're learning the skills

00:37:21.462 --> 00:37:24.372
of how to work within an organisation.

00:37:24.702 --> 00:37:28.452
You're learning the skills
of how to interact with other

00:37:28.452 --> 00:37:30.552
people in a professional way.

00:37:30.852 --> 00:37:35.172
You're learning the skills of
how to manage time, resources.

00:37:35.952 --> 00:37:40.242
One of the concerns I have is
without those entry-level jobs

00:37:40.912 --> 00:37:44.962
our students, as they leave, are
not prepared for that next level.

00:37:44.992 --> 00:37:49.972
So we need to think about how do
we help them prepare so that these

00:37:49.972 --> 00:37:54.892
new jobs that are coming up, they
have the skills already to be able

00:37:54.892 --> 00:37:57.082
to take into the organisations.

00:37:57.498 --> 00:38:00.962
I think it also is going to
impact the education we give to

00:38:00.962 --> 00:38:07.832
the students as they prepare to
go out into the AI-enabled world.

00:38:08.718 --> 00:38:09.578
Susannah de Jager: It's fascinating.

00:38:10.298 --> 00:38:12.677
Anne thank you very much
for this conversation.

00:38:12.707 --> 00:38:13.637
I've really enjoyed it.

00:38:14.347 --> 00:38:14.977
Anne Trefethen: Thank you.

00:38:16.365 --> 00:38:19.755
Susannah de Jager: Thank you for listening
to this episode of Oxford Plus, hosted

00:38:19.755 --> 00:38:23.595
by me, Susannah de Jager If you wanna
keep up with all things Oxford Plus,

00:38:23.835 --> 00:38:29.145
visit our website, oxfordplus.co.uk or
sign up for our newsletter on Substack.

00:38:29.625 --> 00:38:32.505
Oxford Plus is a podcast
produced by Story 94.