TRANSCRIPT 🔗
WEBVTT
00:00:20.574 --> 00:00:23.675
so welcome to another episode of cloud
00:00:23.736 --> 00:00:24.815
unplugged um
00:00:25.785 --> 00:00:26.184
Today,
00:00:26.865 --> 00:00:28.426
I believe we promised that we were going
00:00:28.446 --> 00:00:31.268
to talk through how to roll out AI
00:00:31.669 --> 00:00:32.149
safely.
00:00:32.768 --> 00:00:34.651
This has come up quite a lot in
00:00:34.731 --> 00:00:36.332
our existing customers and people we've
00:00:36.351 --> 00:00:37.091
been working with.
00:00:38.615 --> 00:00:42.137
The term shadow AI has been coined where
00:00:42.798 --> 00:00:45.561
people are secretly using AI to their own
00:00:45.621 --> 00:00:46.462
personal account.
00:00:48.404 --> 00:00:49.645
And obviously that's a bit of a concern
00:00:49.685 --> 00:00:51.767
for a lot of businesses because they don't
00:00:51.786 --> 00:00:54.128
really have any control or oversight.
00:00:55.570 --> 00:00:57.792
And they also just want to start to
00:00:57.932 --> 00:00:59.253
use it for their business.
00:00:59.374 --> 00:01:00.875
They obviously want to start to use it
00:01:00.915 --> 00:01:01.676
for coding.
00:01:02.137 --> 00:01:03.317
They want to start to use it just
00:01:03.357 --> 00:01:04.037
to improve
00:01:04.718 --> 00:01:07.021
efficiencies and content that they're
00:01:07.040 --> 00:01:08.283
trying to create, et cetera.
00:01:08.462 --> 00:01:08.862
However,
00:01:10.364 --> 00:01:12.686
how they approach it and where do you
00:01:12.727 --> 00:01:15.569
begin is often the challenge.
00:01:15.971 --> 00:01:18.192
So we've broken things down in this
00:01:18.253 --> 00:01:22.858
episode to kind of go through all of
00:01:22.878 --> 00:01:24.879
the different pillars that we consider all
00:01:24.900 --> 00:01:26.281
the prime ones.
00:01:26.301 --> 00:01:27.683
I think there are six in total.
00:01:28.906 --> 00:01:30.108
then we'll talk through a bit of an
00:01:30.128 --> 00:01:32.471
approach and a process around it and like
00:01:32.512 --> 00:01:34.915
how and where do you begin in relation
00:01:34.935 --> 00:01:38.099
to those pillars um but before we crack
00:01:38.140 --> 00:01:41.704
on to that um obviously i just salman
00:01:41.906 --> 00:01:43.408
how are you doing before we obviously get
00:01:43.447 --> 00:01:43.847
into that
00:01:44.628 --> 00:01:45.427
I'm doing very well,
00:01:45.507 --> 00:01:48.009
busy implementing these techniques so
00:01:48.209 --> 00:01:50.911
everybody can have a safer time with AI
00:01:51.090 --> 00:01:52.971
and the models they want to use.
00:01:53.712 --> 00:01:55.231
John, looking a bit different today.
00:01:55.832 --> 00:01:57.052
I see a bit of a tan going
00:01:57.132 --> 00:01:57.272
on.
00:01:58.394 --> 00:01:59.334
Were you away?
00:02:00.073 --> 00:02:00.775
Where were you, John?
00:02:00.915 --> 00:02:03.436
Tell us a little bit more.
00:02:03.695 --> 00:02:06.897
It's just my genetics, actually, Salvan.
Load full transcript
00:02:08.177 --> 00:02:09.437
I've not actually been anywhere.
00:02:09.497 --> 00:02:09.677
No,
00:02:10.057 --> 00:02:13.419
I was in Spain in Costa Brava area.
00:02:14.360 --> 00:02:16.882
which is like the Catalan area of Spain.
00:02:16.901 --> 00:02:18.182
There's a place called La Scala.
00:02:18.203 --> 00:02:19.104
I don't know if you've ever heard of
00:02:19.144 --> 00:02:23.826
that, but it's south of Girona,
00:02:24.147 --> 00:02:27.931
which is near Figueres, basically.
00:02:27.971 --> 00:02:29.592
But yeah, very, very beautiful.
00:02:30.372 --> 00:02:33.215
Lots of nice beaches.
00:02:33.675 --> 00:02:35.195
A lot of kite surfing on some of
00:02:35.216 --> 00:02:35.856
the beaches there,
00:02:35.877 --> 00:02:37.617
because I think they get some wind.
00:02:38.479 --> 00:02:40.300
So yeah, I've caught the sun.
00:02:40.379 --> 00:02:42.081
It was thirty-eight.
00:02:42.722 --> 00:02:45.544
thirty eight degrees yeah just casual
00:02:45.563 --> 00:02:48.205
thing because it wasn't hot just a couple
00:02:48.224 --> 00:02:50.225
of degrees above uk to be honest like
00:02:50.306 --> 00:02:53.508
you know maybe a couple any any talk
00:02:53.627 --> 00:02:55.908
any talk of ai out there in in
00:02:55.989 --> 00:02:59.471
spain john and holiday no talk of ai
00:02:59.510 --> 00:03:01.793
no because they're all like kind of um
00:03:02.332 --> 00:03:04.033
traditional snorkeling
00:03:05.080 --> 00:03:08.100
fishing boats, a bit of tourism,
00:03:08.140 --> 00:03:11.820
lots of restaurants, you know, type vibes.
00:03:11.860 --> 00:03:13.842
Very chilled, relaxed,
00:03:13.942 --> 00:03:16.421
non-technology based vibes, which is good.
00:03:16.542 --> 00:03:19.622
It was good to kind of switch off.
00:03:20.163 --> 00:03:21.182
You went for detox.
00:03:21.383 --> 00:03:24.043
John went for detox and came back for
00:03:24.062 --> 00:03:26.204
the next visit.
00:03:26.223 --> 00:03:26.683
Yeah, I stayed.
00:03:26.723 --> 00:03:28.243
My friend there,
00:03:28.264 --> 00:03:31.965
she's opened a grocery shop in a little
00:03:32.004 --> 00:03:32.365
village.
00:03:34.039 --> 00:03:37.403
And then her partner has got a restaurant
00:03:38.506 --> 00:03:39.706
in La Scala itself.
00:03:39.847 --> 00:03:42.050
So there was a lot of trying loads
00:03:42.090 --> 00:03:44.152
of things I don't normally eat,
00:03:44.332 --> 00:03:46.656
but being super polite and be like, oh,
00:03:46.695 --> 00:03:47.657
wow, this is really...
00:03:49.109 --> 00:03:51.611
uh but really be like okay i don't
00:03:51.671 --> 00:03:53.332
normally eat things like this but actually
00:03:53.353 --> 00:03:56.496
it was very good so out of politeness
00:03:56.515 --> 00:03:58.318
i was forced to try new things which
00:03:58.358 --> 00:04:02.562
is good excellent okay so tell us tell
00:04:02.602 --> 00:04:04.864
us john when people approach you and they
00:04:04.883 --> 00:04:07.265
say i want to roll out ai safely
00:04:07.286 --> 00:04:09.367
and securely just so i have control over
00:04:09.427 --> 00:04:12.050
what is being run on my estate and
00:04:12.091 --> 00:04:14.693
people not accessing any of these tools uh
00:04:15.173 --> 00:04:17.175
without any guardrails so they don't get
00:04:17.235 --> 00:04:17.555
hacked.
00:04:17.714 --> 00:04:19.156
You've seen so many issues that are
00:04:19.175 --> 00:04:20.057
happening recently.
00:04:20.858 --> 00:04:22.459
Things are getting called out that you
00:04:22.478 --> 00:04:24.800
don't even know are getting used and
00:04:24.860 --> 00:04:25.641
leaking data.
00:04:25.661 --> 00:04:27.302
You're leaking your sensitive information.
00:04:28.423 --> 00:04:30.564
What are the steps we have to take
00:04:30.785 --> 00:04:32.526
in the pillars that you talk about to
00:04:32.567 --> 00:04:34.687
make sure you roll out AI safely in
00:04:34.708 --> 00:04:35.468
your organization?
00:04:36.850 --> 00:04:37.110
Yeah,
00:04:37.189 --> 00:04:39.891
so I guess I'll kind of go through
00:04:39.951 --> 00:04:41.353
each pillar first.
00:04:41.434 --> 00:04:41.553
Yeah.
00:04:42.461 --> 00:04:46.730
so obviously the beginning bit is the
00:04:47.230 --> 00:04:50.519
identity and access and device management
00:04:50.538 --> 00:04:51.761
um so
00:04:53.735 --> 00:04:54.355
You know,
00:04:54.375 --> 00:04:56.396
do you have control of the devices that
00:04:56.576 --> 00:04:58.276
your, I guess,
00:04:58.396 --> 00:05:01.177
end users are kind of working on?
00:05:01.577 --> 00:05:02.899
Is there any controls over them?
00:05:02.959 --> 00:05:03.778
Do you have audit?
00:05:04.079 --> 00:05:05.740
Are you approving the applications that
00:05:05.759 --> 00:05:06.540
get installed?
00:05:08.641 --> 00:05:12.062
And can you tell whether or not people
00:05:12.081 --> 00:05:13.682
are doing things on their devices that
00:05:13.702 --> 00:05:15.584
they shouldn't be or that something maybe
00:05:15.624 --> 00:05:18.584
has got installed that maybe is malicious
00:05:18.644 --> 00:05:19.324
by accident?
00:05:19.404 --> 00:05:20.646
And do you have the audit of all
00:05:20.685 --> 00:05:21.266
those things?
00:05:22.367 --> 00:05:23.228
So that's one lens.
00:05:23.269 --> 00:05:23.709
So it's like,
00:05:23.749 --> 00:05:25.649
how are you managing the devices
00:05:25.689 --> 00:05:26.269
themselves?
00:05:27.250 --> 00:05:29.391
So that is even without AI, right?
00:05:29.411 --> 00:05:31.812
So that's just good practice anyway,
00:05:31.994 --> 00:05:32.694
right?
00:05:32.733 --> 00:05:33.494
Good practice,
00:05:33.553 --> 00:05:35.355
but a lot of businesses may have not
00:05:36.055 --> 00:05:39.757
felt a need necessarily to have full
00:05:39.778 --> 00:05:40.317
device management.
00:05:40.358 --> 00:05:41.918
They might have a different policy for
00:05:41.959 --> 00:05:42.839
different teams.
00:05:42.899 --> 00:05:45.321
They might have all
kinds of reasons,
00:05:45.360 --> 00:05:46.502
legitimate reasons that
00:05:47.346 --> 00:05:50.129
may have been cost-saving initiatives um
00:05:50.149 --> 00:05:51.550
that might have done it more for
00:05:52.190 --> 00:05:55.273
developers or people that are more closer
00:05:55.293 --> 00:05:57.153
to infrastructure because there's more
00:05:57.214 --> 00:05:58.935
risk because that's where the data is yeah
00:05:58.975 --> 00:06:00.757
they're closer to the general day-to-day
00:06:00.857 --> 00:06:02.437
data applications for customers but
00:06:02.959 --> 00:06:04.860
actually we do you want to treat it
00:06:05.420 --> 00:06:07.682
as one universal thing now that everyone's
00:06:07.742 --> 00:06:09.483
using ai um so
00:06:10.903 --> 00:06:12.685
That's a very specific thing.
00:06:12.805 --> 00:06:15.127
And there are tools that you can install
00:06:15.146 --> 00:06:16.867
that the cloud providers provide.
00:06:17.928 --> 00:06:19.470
They're not all equal.
00:06:19.709 --> 00:06:21.130
I think Azure is a little bit more
00:06:21.170 --> 00:06:21.552
improved.
00:06:21.591 --> 00:06:22.651
So they have...
00:06:24.244 --> 00:06:26.185
basically a browser DLP,
00:06:26.204 --> 00:06:30.908
which data protection and it can kind of
00:06:30.949 --> 00:06:34.711
detect in the tabs for Chrome or Edge
00:06:36.033 --> 00:06:36.833
or Firefox.
00:06:37.012 --> 00:06:39.574
It can do detection on whether or not
00:06:39.595 --> 00:06:42.637
you're doing malicious intent with data,
00:06:42.677 --> 00:06:43.939
potentially IAUR.
00:06:44.764 --> 00:06:47.326
copying and pasting things from documents
00:06:47.466 --> 00:06:51.548
into chat gpt that's you shouldn't be um
00:06:51.767 --> 00:06:53.129
and it can kind of detect those things
00:06:54.410 --> 00:06:56.850
um then you have the users themselves you
00:06:56.870 --> 00:06:59.312
know how does your identity look how are
00:06:59.353 --> 00:06:59.692
people
00:07:00.333 --> 00:07:01.913
logging into your systems,
00:07:03.074 --> 00:07:04.276
have you centralised that,
00:07:04.696 --> 00:07:09.879
are you using single identity, MFA,
00:07:09.920 --> 00:07:12.081
all kind of the standard aspects,
00:07:12.141 --> 00:07:14.242
least privilege, good RBAC controls,
00:07:15.322 --> 00:07:18.444
conditional access, et cetera, et cetera.
00:07:19.810 --> 00:07:21.550
Then you've got the control plane itself,
00:07:21.730 --> 00:07:24.072
which is a little bit ambiguous calling it
00:07:24.132 --> 00:07:25.093
a kind of control plane,
00:07:25.132 --> 00:07:27.553
but that's something that sits in front of
00:07:28.375 --> 00:07:28.814
the model.
00:07:29.615 --> 00:07:31.476
Some call it a gen AI gateway.
00:07:31.896 --> 00:07:34.418
So basically a way of obviously going
00:07:34.458 --> 00:07:36.119
through into the model itself,
00:07:36.540 --> 00:07:39.281
a proxy almost.
00:07:40.161 --> 00:07:41.963
And in there you can then put controls
00:07:42.463 --> 00:07:43.504
inside of the proxy.
00:07:43.543 --> 00:07:45.125
So obviously you can do token limits.
00:07:45.845 --> 00:07:47.565
You can do fail back routing to another
00:07:47.605 --> 00:07:48.867
model in there.
00:07:48.906 --> 00:07:53.550
You can do content safety filters and
00:07:53.589 --> 00:07:54.290
prompt shields,
00:07:54.310 --> 00:07:56.012
and you can obviously hook those things in
00:07:56.812 --> 00:07:59.494
as well into additional stuff like Purview
00:07:59.533 --> 00:07:59.874
as well.
00:08:01.975 --> 00:08:03.536
And then you get into data governance,
00:08:03.896 --> 00:08:06.057
which is the classification of your data.
00:08:06.959 --> 00:08:08.939
Do you have things labeled properly to
00:08:09.000 --> 00:08:10.521
know whether it's sensitive or not?
00:08:11.961 --> 00:08:14.303
And therefore, again, certain services
00:08:15.629 --> 00:08:15.930
Obviously,
00:08:15.949 --> 00:08:18.512
when you kind of categorize the data,
00:08:18.533 --> 00:08:19.533
it will then know, okay,
00:08:19.553 --> 00:08:21.136
this is sensitive information here.
00:08:21.156 --> 00:08:22.137
It's already labeled.
00:08:22.997 --> 00:08:26.322
I need to take precaution on that.
00:08:26.401 --> 00:08:28.485
Encryption, et cetera, et cetera.
00:08:29.485 --> 00:08:31.208
Then you've got the platform foundations,
00:08:31.608 --> 00:08:32.889
which are just general information.
00:08:33.783 --> 00:08:36.864
cloud foundations, landing zones,
00:08:36.943 --> 00:08:39.224
networking, private networks,
00:08:39.264 --> 00:08:41.504
internal endpoints, that kind of stuff.
00:08:41.784 --> 00:08:43.105
And then you've then got the monitoring
00:08:43.125 --> 00:08:44.326
and audit.
00:08:44.385 --> 00:08:46.505
Do you know if somebody is doing something
00:08:47.466 --> 00:08:48.086
malicious?
00:08:48.427 --> 00:08:49.407
Can you detect it?
00:08:50.147 --> 00:08:51.746
Are those events coming into a central
00:08:51.787 --> 00:08:52.147
place?
00:08:52.167 --> 00:08:53.768
Do you have all the right monitoring and
00:08:53.788 --> 00:08:56.729
alerting around those services so that you
00:08:56.749 --> 00:08:58.349
can detect something when it does happen?
00:08:59.326 --> 00:09:00.606
Even after you've prevented everything in
00:09:00.626 --> 00:09:01.687
those pillars, like we've said,
00:09:01.706 --> 00:09:03.467
you still need to know there's still risk.
00:09:04.327 --> 00:09:05.528
Do you have the right oversight?
00:09:07.089 --> 00:09:07.328
Yeah.
00:09:07.349 --> 00:09:08.730
Do all those things make sense to you,
00:09:08.769 --> 00:09:09.710
Salman?
00:09:09.769 --> 00:09:10.289
Absolutely.
00:09:11.770 --> 00:09:12.671
It seems like, John,
00:09:12.711 --> 00:09:15.152
that we're not reinventing the wheel.
00:09:15.312 --> 00:09:17.052
Is that fair to say that these are
00:09:17.072 --> 00:09:19.013
the practices that you currently follow
00:09:19.113
--> 00:09:19.514
anyway?
00:09:20.494 --> 00:09:21.774
These are good practices just to keep
00:09:21.793 --> 00:09:23.894
yourself secure and keep yourself
00:09:23.934 --> 00:09:25.375
compliant to a bunch of these things.
00:09:25.975 --> 00:09:29.096
But I guess because of these problems with
00:09:29.297 --> 00:09:30.837
accessing the AI models,
00:09:30.977 --> 00:09:33.359
it's a bit exacerbated because people see
00:09:33.759 --> 00:09:36.559
issues a lot more nowadays than perhaps
00:09:36.580 --> 00:09:37.279
they were in the past.
00:09:37.320 --> 00:09:39.620
But it seems like we're not talking about
00:09:39.740 --> 00:09:40.301
anything new.
00:09:40.821 --> 00:09:42.341
Perhaps some implementation is a bit
00:09:42.402 --> 00:09:44.503
different because we never had to worry
00:09:44.543 --> 00:09:45.423
about tokens before.
00:09:45.822 --> 00:09:47.344
And now we're thinking, oh,
00:09:47.384 --> 00:09:48.644
maybe in the proxy we can add some
00:09:48.663 --> 00:09:52.346
limitations of the cap of a token can
00:09:52.405 --> 00:09:52.806
go in there.
00:09:53.265 --> 00:09:54.927
But is that fair to say that we're
00:09:54.947 --> 00:09:56.149
not talking about something?
00:09:56.188 --> 00:09:59.812
Yeah, I don't think the concepts are new,
00:10:00.893 --> 00:10:03.294
but I would say that the user behavior
00:10:03.375 --> 00:10:04.696
is completely different.
00:10:05.017 --> 00:10:10.441
So before you would have to share
00:10:10.761 --> 00:10:15.206
sensitive business data would have been an
00:10:15.245 --> 00:10:18.148
intentional act to maybe email it to
00:10:18.187 --> 00:10:19.769
yourself or something like that.
00:10:20.559 --> 00:10:23.261
But today people are obviously using it to
00:10:23.322 --> 00:10:24.884
help them be better at their job or
00:10:24.923 --> 00:10:28.186
maybe to give insights into something that
00:10:29.606 --> 00:10:31.869
helps them produce a report or maybe scan
00:10:31.889 --> 00:10:33.630
things quickly for them on their behalf.
00:10:33.671 --> 00:10:36.592
So they now are like where you would
00:10:36.613 --> 00:10:39.315
have been really that would be more about
00:10:39.375 --> 00:10:43.097
sharing the assets externally for
00:10:43.138 --> 00:10:43.778
different reasons.
00:10:43.798 --> 00:10:45.860
You're now sharing them legitimately to
00:10:45.900 --> 00:10:47.402
help yourself be better at your job.
00:10:48.086 --> 00:10:49.706
Hence it being called kind of shadow AI.
00:10:49.725 --> 00:10:51.307
So those things are a little bit different
00:10:51.366 --> 00:10:53.586
because it means you've got to be quite
00:10:53.647 --> 00:10:56.048
good at really almost blocking.
00:10:56.668 --> 00:10:57.668
That's kind of what you really want to
00:10:57.687 --> 00:10:57.868
do.
00:10:57.988 --> 00:11:01.389
You want to block any other models outside
00:11:01.428 --> 00:11:03.429
of your own approved models so that
00:11:03.610 --> 00:11:06.350
everyone is forced to basically go through
00:11:06.370 --> 00:11:08.211
your central control plane, i.e.
00:11:08.270 --> 00:11:09.370
the Gen AI gateway,
00:11:09.811 --> 00:11:11.912
and they can't bypass it and go to
00:11:11.951 --> 00:11:13.672
some other model or use their own personal
00:11:13.692 --> 00:11:14.792
account or bypass.
00:11:15.653 --> 00:11:17.133
And that's interesting.
00:11:17.153 --> 00:11:17.273
Yeah.
00:11:18.279 --> 00:11:20.120
And the risk here is what, John?
00:11:20.160 --> 00:11:21.100
What are we trying to prevent?
00:11:22.961 --> 00:11:24.923
Are we saying that, oh, as a company,
00:11:25.062 --> 00:11:26.484
we can only use models that are provided
00:11:26.524 --> 00:11:27.063
by Claude,
00:11:27.124 --> 00:11:28.325
but we don't trust any of the models
00:11:28.345 --> 00:11:29.644
that are provided by Grok?
00:11:30.885 --> 00:11:33.047
What are we trying to protect here?
00:11:33.486 --> 00:11:34.868
Are we trying to protect IP?
00:11:34.947 --> 00:11:37.048
Are we trying to keep ourselves safe
00:11:38.629 --> 00:11:39.009
online?
00:11:39.090 --> 00:11:40.510
Or what is the thing that we're trying
00:11:40.530 --> 00:11:41.110
to protect here?
00:11:42.169 --> 00:11:43.172
I guess many reasons.
00:11:43.192 --> 00:11:46.538
I guess one could be you don't necessarily
00:11:46.577 --> 00:11:48.100
have an approval for that model.
00:11:48.181 --> 00:11:48.922
It could be that
00:11:49.601 --> 00:11:52.504
the sovereignty of the model,
00:11:52.524 --> 00:11:54.485
like where it operates in, you know,
00:11:54.745 --> 00:11:57.605
is not under the jurisdiction of the EU
00:11:57.686 --> 00:11:58.787
or, you know,
00:11:58.907 --> 00:12:03.370
under your own legal kind of rules and
00:12:03.409 --> 00:12:04.090
regulation,
00:12:04.570 --> 00:12:06.350
and therefore people sending sensitive
00:12:06.390 --> 00:12:09.592
customer data outside of your jurisdiction
00:12:09.773 --> 00:12:12.394
into other countries into the model is
00:12:12.434 --> 00:12:13.315
obviously not approved.
00:12:14.816 --> 00:12:16.636
It could be that you're installing,
00:12:16.697 --> 00:12:17.998
people are installing things
00:12:18.772 --> 00:12:21.134
to integrate into third-party tools to
00:12:21.153 --> 00:12:24.275
make it easier i.e um i don't know
00:12:24.416 --> 00:12:27.698
google drive and sharepoint and you know
00:12:27.759 --> 00:12:31.400
so you're connecting up um into say it
00:12:31.421 --> 00:12:33.043
could be anything could be like claude
00:12:33.182 --> 00:12:36.924
desktop or co-pilot you know and you're
00:12:36.945 --> 00:12:38.846
connecting things up third-party systems
00:12:39.407 --> 00:12:41.149
that then have access to lots of different
00:12:41.168 --> 00:12:44.451
data um and i think the difference between
00:12:44.471 --> 00:12:46.893
a human and ai is you know
00:12:48.028 --> 00:12:50.409
to do a task,
00:12:50.591 -->
00:12:52.110
you'd probably be quite intentional of
00:12:52.171 --> 00:12:52.971
where to look.
00:12:53.552 --> 00:12:54.312
So you'd be like, okay,
00:12:54.393 --> 00:12:56.333
I know this information's stored over
00:12:56.374 --> 00:12:58.195
here, or I'd probably quickly navigate.
00:12:58.595 --> 00:12:59.135
With AI,
00:12:59.275 --> 00:13:01.256
it's probably reading everything to figure
00:13:01.277 --> 00:13:03.378
it out.
00:13:03.418 --> 00:13:05.119
So to know where to go,
00:13:05.298 --> 00:13:07.080
it's had to process every other piece of
00:13:07.120 --> 00:13:09.042
data to then figure out the right answer.
00:13:09.721 --> 00:13:12.003
And obviously that's very different to a
00:13:12.083 --> 00:13:12.423
person,
00:13:12.443 --> 00:13:14.424
but it's now acting on behalf of you.
00:13:14.884 --> 00:13:16.666
So it's acting as if you're doing it
00:13:16.866 --> 00:13:18.326
because it's using your identity.
00:13:19.628 --> 00:13:20.729
And obviously that's a lot of risk.
00:13:21.429 --> 00:13:24.130
So what you're saying is that if a
00:13:24.191 --> 00:13:26.773
person is on their machine and they're
00:13:26.793 --> 00:13:28.913
trying to, let's say, look at a reporter,
00:13:29.955 --> 00:13:32.155
identify whatever task they're doing the
00:13:32.296 --> 00:13:35.398
ai could potentially read whatever's on
00:13:35.418 --> 00:13:36.739
their machine they could have sensitive
00:13:36.778 --> 00:13:38.820
data on their machine and feed it to
00:13:38.840 --> 00:13:40.659
the ai model which is what we don't
00:13:40.679 --> 00:13:42.041
want to do is that correct we're trying
00:13:42.061 --> 00:13:44.302
to protect the the sensitive information
00:13:44.322 --> 00:13:47.124
that you have from reaching the ai model
00:13:47.144 --> 00:13:49.664
so perhaps you can use it uh to
00:13:49.725 --> 00:13:51.865
train itself and get better is because i'm
00:13:51.885 --> 00:13:53.047
trying to get to is that a bad
00:13:53.067 --> 00:13:56.567
thing is this about i guess two things
00:13:56.607 --> 00:13:58.009
you could you could decide that for
00:13:58.089 --> 00:13:59.090
certain operations
00:13:59.815 --> 00:14:02.256
you want a much cheaper model.
00:14:03.197 --> 00:14:04.477
So that could be one reason.
00:14:04.498 --> 00:14:05.418
So you're like, actually,
00:14:05.499 --> 00:14:07.000
for this type of day to day stuff,
00:14:07.039 --> 00:14:08.581
I don't want to be sending loads of
00:14:08.620 --> 00:14:09.841
data to the fable,
00:14:10.162 --> 00:14:11.722
most expensive model for these types of
00:14:11.743 --> 00:14:12.323
activities.
00:14:12.984 --> 00:14:13.725
So yeah, there's that.
00:14:13.784 --> 00:14:14.784
And then obviously you don't want to be
00:14:14.804 --> 00:14:16.807
necessarily getting a load of random data
00:14:17.307 --> 00:14:19.668
and then shoehorning it into a model
00:14:21.529 --> 00:14:23.311
because that's just what you've got access
00:14:23.331 --> 00:14:23.410
to.
00:14:23.471 --> 00:14:24.731
So not just on your device,
00:14:24.772 --> 00:14:25.232
it could be
00:14:26.057 --> 00:14:26.918
You might be overprivileged.
00:14:26.937 --> 00:14:29.399
You might be an admin of your estate.
00:14:29.458 --> 00:14:30.980
You might be a drive admin on SharePoint.
00:14:31.000 --> 00:14:31.960
The whole thing, yeah.
00:14:32.240 --> 00:14:34.120
You've got access to absolutely every
00:14:34.160 --> 00:14:35.140
piece of information,
00:14:35.360 --> 00:14:37.942
and then you're using AI on your machine.
00:14:38.162 --> 00:14:38.562
Technically,
00:14:38.581 --> 00:14:40.582
then it is also getting access to
00:14:40.643 --> 00:14:42.443
absolutely everything through your own
00:14:42.464 --> 00:14:42.844
permission.
00:14:43.864 --> 00:14:44.403
All right, cool.
00:14:44.764 --> 00:14:45.323
Tell me, John,
00:14:45.544 --> 00:14:47.725
how do I get a device that is
00:14:47.845 --> 00:14:48.365
secure,
00:14:48.725 --> 00:14:50.446
the first of your pillar's device and
00:14:50.525 --> 00:14:50.926
endpoint?
00:14:51.265 --> 00:14:52.826
What do I need to do there to
00:14:52.846 --> 00:14:55.668
get myself safely into this AI land?
00:14:57.038 --> 00:14:58.739
uh well i think the first thing to
00:14:58.778 --> 00:15:02.620
do is if you are wanting to use
00:15:02.659 --> 00:15:06.561
it for your business um the high high
00:15:06.581 --> 00:15:10.441
probability is outside just coding um
00:15:10.480 --> 00:15:11.561
you're going to want it to have access
00:15:11.581 --> 00:15:14.341
to files and information right it's it's
00:15:14.721 --> 00:15:17.182
only valuable when it can start to process
00:15:17.201 --> 00:15:19.182
things on your behalf so you don't have
00:15:19.202 --> 00:15:20.802
to process it all to then give suggestions
00:15:20.863 --> 00:15:22.643
or commentary or improvements or new
00:15:22.702 --> 00:15:25.364
content but so you then need to really
00:15:25.423 --> 00:15:26.004
understand
00:15:27.017 --> 00:15:29.078
what is the current data access governance
00:15:29.158 --> 00:15:30.720
that's already in place?
00:15:30.779 --> 00:15:33.142
Have we shared everything with everyone by
00:15:33.221 --> 00:15:35.063
default?
00:15:35.104 --> 00:15:37.426
Because you might not even know what is
00:15:37.466 --> 00:15:39.267
really shared with you and what you really
00:15:39.287 --> 00:15:40.668
have access to because you're not
00:15:40.928 --> 00:15:42.730
necessarily going through every single
00:15:42.769 --> 00:15:44.230
file on the drive to see,
00:15:44.270 --> 00:15:45.993
can I or can't I access this stuff
00:15:46.013 --> 00:15:47.313
that I should or shouldn't be able to
00:15:47.394 --> 00:15:47.813
access?
00:15:49.375 --> 00:15:50.836
So you need to obviously do a bit
00:15:50.876 --> 00:15:53.318
of a data access governance process,
00:15:53.379 --> 00:15:55.541
produce a report, look,
00:15:56.563 --> 00:15:58.264
quite quickly at whether there's over
00:15:58.304 --> 00:16:01.187
permissive permissions in there and that
00:16:01.206 --> 00:16:02.147
you've got all the right going around.
00:16:02.167 --> 00:16:05.249
So that'd be the first thing to do
00:16:05.288 --> 00:16:06.809
before you start to roll it out.
00:16:06.970 --> 00:16:10.030
And then sort that out quickly and then
00:16:10.071 --> 00:16:10.892
look at identity.
00:16:10.951 --> 00:16:13.513
Similar thing in your RBAC controls,
00:16:15.014 --> 00:16:16.495
your identity and access management.
00:16:16.575 --> 00:16:19.496
Is that quite strong and a good posture
00:16:19.517 --> 00:16:21.138
there?
00:16:21.557 --> 00:16:22.697
What's RBAC control, John?
00:16:22.738 --> 00:16:24.840
Can you just please explain what is RBAC
00:16:24.879 --> 00:16:25.200
control?
00:16:26.024 --> 00:16:27.445
Well, you've heard it here first, guys.
00:16:27.985 --> 00:16:29.285
He doesn't know what RBAC is.
00:16:30.686 --> 00:16:31.225
Please, John.
00:16:31.486 --> 00:16:32.846
I'm a little bit in shock.
00:16:32.907 --> 00:16:36.269
What is RBAC, John?
00:16:37.048 --> 00:16:39.190
Role-based access control is what it is.
00:16:39.809 --> 00:16:41.671
Okay.
00:16:41.711 --> 00:16:42.130
Basically,
00:16:42.912 --> 00:16:44.773
what access are people allowed to have
00:16:44.832 --> 00:16:45.592
access to?
00:16:46.212 --> 00:16:47.614
What's the permission set?
00:16:48.193 --> 00:16:49.115
But not just that, as in
00:16:50.250 --> 00:16:53.234
are you doing short-term access?
00:16:53.413 --> 00:16:55.275
Are you making people escalate privilege
00:16:55.296 --> 00:16:55.976
when you need to?
00:16:56.518 --> 00:16:57.057
That kind of stuff.
00:16:57.097 --> 00:16:59.279
But also, have you...
00:17:00.293 --> 00:17:03.155
Are you blocking maybe certain GOs?
00:17:03.515 --> 00:17:05.097
Would you expect somebody to be logging in
00:17:05.137 --> 00:17:05.837
from China?
00:17:06.259 --> 00:17:08.040
You know, that kind of stuff as well,
00:17:08.080 --> 00:17:08.361
right?
00:17:08.381 --> 00:17:10.343
So it's not just the permissions.
00:17:11.203 --> 00:17:12.345
Yeah, okay.
00:17:12.365 --> 00:17:13.546
Not to pick on China, obviously,
00:17:13.566 --> 00:17:15.166
because it's got a...
00:17:15.347 --> 00:17:16.788
Rumor is it's got a very cheap model
00:17:16.868 --> 00:17:20.092
that's quite competitive nowadays.
00:17:20.112 --> 00:17:21.252
And there's Kimi K.T.
00:17:21.272 --> 00:17:22.034
that came out, John,
00:17:22.054 --> 00:17:22.734
a couple of days ago.
00:17:24.142 --> 00:17:26.083
on par with fable according to the tests
00:17:26.383 --> 00:17:29.184
and uh i have to confess i did
00:17:29.224 --> 00:17:31.746
try it it did work at times and
00:17:31.786 --> 00:17:33.666
at times i did get some errors perhaps
00:17:33.686 --> 00:17:37.749
everybody was trying it so uh yeah my
00:17:37.769 -->
00:17:39.789
machine doesn't block the use of kimi
00:17:39.809 --> 00:17:41.431
k-three at the moment john i'm just
00:17:41.471 --> 00:17:45.212
letting you know that wow okay uh so
00:17:46.584 --> 00:17:49.244
Got a real life security risk on the
00:17:49.265 --> 00:17:49.684
podcast.
00:17:49.724 --> 00:17:50.645
Is that what we're saying?
00:17:50.705 --> 00:17:51.945
Is this a confession?
00:17:52.066 --> 00:17:53.507
I'm not saying it's a security risk.
00:17:53.527 --> 00:17:54.807
I'm just saying that I tried it, John.
00:17:55.587 --> 00:17:56.969
But luckily on my machine,
00:17:57.009 --> 00:17:58.489
I don't have any sensitive information.
00:17:59.269 --> 00:18:01.431
Apart from some photos and podcasts.
00:18:01.971 --> 00:18:04.932
We're very equal opportunistic here,
00:18:04.951 --> 00:18:05.692
though.
00:18:05.792 --> 00:18:06.673
That's our stance.
00:18:06.732 --> 00:18:09.753
So we don't like to get into geopolitics.
00:18:09.794 --> 00:18:10.515
We're all about...
00:18:11.315 --> 00:18:13.957
open open and transparent geo uh
00:18:14.517 --> 00:18:19.299
approaches um so yeah the us whatever all
00:18:19.339 --> 00:18:21.141
for we're all for technology right that's
00:18:21.300 --> 00:18:23.442
what we are technology yeah awful
00:18:23.461 --> 00:18:25.784
technology where we just uh you know
00:18:25.804 --> 00:18:28.025
that's what we're doing techno optimists
00:18:28.105 --> 00:18:29.925
john that's what we are we try all
00:18:29.965 --> 00:18:30.426
technology
00:18:32.111 --> 00:18:32.391
But yeah,
00:18:32.451 --> 00:18:34.632
so I guess once you've done the
00:18:34.652 --> 00:18:35.373
assessments,
00:18:36.653 --> 00:18:38.913
once you have fixed your identity,
00:18:39.134 --> 00:18:41.234
once you've assessed as well your current
00:18:41.515 --> 00:18:42.974
kind of foundations,
00:18:43.015 --> 00:18:44.175
where's your maturity,
00:18:45.635 --> 00:18:48.538
have you got a good control over maybe
00:18:48.557 --> 00:18:50.498
a central VPN that people are routing
00:18:50.518 --> 00:18:53.358
towards,
00:18:53.419 --> 00:18:56.601
some form of governance over the device
00:18:56.661 --> 00:18:57.040
routing,
00:18:58.153 --> 00:18:59.758
Can you block people from going to
00:18:59.818 --> 00:19:02.144
specific websites?
00:19:02.506 --> 00:19:03.228
That type of stuff.
00:19:05.755 --> 00:19:07.435
And have you got all the right audit
00:19:07.935 --> 00:19:11.576
and logging into like some central kind of
00:19:11.616 --> 00:19:13.837
control place that you can do security
00:19:13.877 --> 00:19:16.439
reports on and central policies that you
00:19:16.459 --> 00:19:17.439
could easily roll out?
00:19:18.878 --> 00:19:19.940
That's what we call kind of the
00:19:19.960 --> 00:19:20.339
foundation.
00:19:20.359 --> 00:19:21.961
So they're the first things that get all
00:19:22.000 --> 00:19:22.800
of that in place.
00:19:23.101 --> 00:19:26.162
And once that's in place and the devices
00:19:27.021 --> 00:19:28.903
and you've ticked all those things off,
00:19:29.942 --> 00:19:31.344
then you're kind of safe to be like,
00:19:31.403 --> 00:19:31.624
right,
00:19:31.683 --> 00:19:35.186
let's start to think about the model and
00:19:35.287 --> 00:19:36.347
the gateway that we're going to put in
00:19:36.367 --> 00:19:36.867
front of it.
00:19:36.928 --> 00:19:38.348
And then who should have access to it?
00:19:38.429 --> 00:19:39.369
And are we going to roll it out
00:19:39.390 --> 00:19:40.230
team by team?
00:19:41.550 --> 00:19:42.912
And are we going to get the right
00:19:42.932 --> 00:19:43.772
telemetry there?
00:19:43.873 --> 00:19:45.253
Are we going to report on the token
00:19:45.394 --> 00:19:45.773
usage?
00:19:47.555 --> 00:19:50.017
Have we got good security over it in
00:19:50.037 --> 00:19:50.538
the gateway?
00:19:50.557 --> 00:19:51.657
Are we going to use Purview?
00:19:51.698 --> 00:19:54.500
Have we looked at the sensitivity and then
00:19:54.539 --> 00:19:55.800
start to do it that way on a
00:19:55.840 --> 00:19:56.981
case by case basis?
00:19:58.726 --> 00:20:00.468
And maybe you want kind of approved things
00:20:00.508 --> 00:20:03.150
like what are the things in the catalog
00:20:03.710 --> 00:20:04.371
that you're allowed to use?
00:20:04.391 --> 00:20:06.071
I was mentioning the connectors, right?
00:20:06.332 --> 00:20:08.134
What are you allowed third party-wise to
00:20:08.173 --> 00:20:08.954
connect into?
00:20:10.435 --> 00:20:12.538
Other specific things that we approve,
00:20:12.758 --> 00:20:12.897
i.e.
00:20:12.938 --> 00:20:14.138
SharePoint.
00:20:14.159 --> 00:20:15.799
We'll let you use the MCP for SharePoint.
00:20:15.859 --> 00:20:18.102
Maybe we'll let you use GitHub MCP,
00:20:19.663 --> 00:20:20.663
et cetera, et cetera.
00:20:21.525 --> 00:20:23.507
And then you'd control that rollout so
00:20:23.547 --> 00:20:25.028
that you can't just use whatever you want.
00:20:25.387 --> 00:20:27.269
were obviously managing it on a
00:20:27.308 --> 00:20:28.369
case-by-case basis,
00:20:28.410 --> 00:20:30.250
like you would policies and other
00:20:30.310 --> 00:20:30.971
exemptions.
00:20:31.051 --> 00:20:35.673
You'd safely want to control that.
00:20:35.773 --> 00:20:36.295
After that,
00:20:37.423 --> 00:20:40.284
You're then into obviously kind of
00:20:40.403 --> 00:20:42.105
generalizing it more.
00:20:42.164 --> 00:20:44.486
So then it would be looking at it
00:20:44.705 --> 00:20:46.546
through the lens of applications.
00:20:47.307 --> 00:20:49.326
You want to reuse all of this because
00:20:49.386 --> 00:20:50.307
it's universal.
00:20:51.988 --> 00:20:53.669
You'd have all the right controls that in
00:20:53.749 --> 00:20:54.749
place to then be like, right,
00:20:54.769 --> 00:20:56.390
let's get the applications doing a Gen
00:20:56.430 --> 00:20:56.710
Tick.
00:20:57.471 --> 00:20:58.971
Let's
use the same gateway.
00:20:59.071 --> 00:21:02.232
Let's use that control plane.
00:21:02.272 --> 00:21:03.233
Let's use Purview,
00:21:03.253 --> 00:21:04.134
which we're already using.
00:21:04.173 --> 00:21:05.513
Let's make sure that we've got all the
00:21:05.554 --> 00:21:06.894
reporting on the sensitive data.
00:21:06.914 --> 00:21:10.855
And then you'd start to maneuver into more
00:21:10.915 --> 00:21:11.816
expansive areas,
00:21:11.836 --> 00:21:13.655
which are application-centric things,
00:21:14.517 --> 00:21:16.297
reusing all the same governance and
00:21:16.336 --> 00:21:17.817
processes that you've already got in
00:21:17.857 --> 00:21:18.137
place.
00:21:19.458 --> 00:21:20.258
Does that make sense?
00:21:21.018 --> 00:21:21.778
It does make sense.
00:21:21.818 --> 00:21:22.598
It does make sense.
00:21:22.878 --> 00:21:23.179
It does.
00:21:24.971 --> 00:21:28.093
And so that's why we're about to stop
00:21:28.613 --> 00:21:31.595
your device from using the new China
00:21:31.615 --> 00:21:31.815
model.
00:21:31.855 --> 00:21:34.256
I believe we're about to roll that out
00:21:34.336 --> 00:21:35.435
on the MDM now.
00:21:35.516 --> 00:21:36.856
So if you've got anything running,
00:21:36.876 --> 00:21:39.678
you might want to quickly get the
00:21:39.718 --> 00:21:41.978
information out now before we block it.
00:21:44.464 --> 00:21:46.465
So I do need to go back to
00:21:46.486 --> 00:21:46.826
a point.
00:21:46.925 --> 00:21:50.087
You keep mentioning the use of Purview in
00:21:50.347 --> 00:21:51.689
the governance part.
00:21:51.749 --> 00:21:53.250
That's just an example of a service in
00:21:53.309 --> 00:21:53.809
Microsoft.
00:21:53.829 --> 00:21:54.970
Yeah, absolutely.
00:21:55.029 --> 00:21:56.931
Maybe can we go a little bit more
00:21:56.971 --> 00:22:00.432
into how do you use Purview to make
00:22:00.472 --> 00:22:03.314
sure that you've done the right level of
00:22:03.374 --> 00:22:03.855
governance?
00:22:04.115 --> 00:22:05.275
Can you go a little bit more into
00:22:05.295 --> 00:22:05.434
that?
00:22:07.017 --> 00:22:10.163
So Purview, just as an example,
00:22:10.986 --> 00:22:13.131
offers data lineage protection.
00:22:15.278 --> 00:22:18.881
It can keep track of all the
00:22:18.941 --> 00:22:20.682
transformation of information and data and
00:22:20.701 --> 00:22:22.824
where it's gone through.
00:22:22.923 --> 00:22:25.085
It also has these classification concepts.
00:22:25.184 --> 00:22:26.645
I think it comes out of the box
00:22:27.506 --> 00:22:30.587
with a load of default classification and
00:22:30.647 --> 00:22:31.128
rules.
00:22:32.189 --> 00:22:33.930
And you can add to those rules yourself,
00:22:33.970 --> 00:22:34.570
essentially,
00:22:34.631 --> 00:22:36.991
of how can it understand the data or
00:22:37.031 --> 00:22:38.873
your data and the sensitivity around it.
00:22:38.952 --> 00:22:39.192
Yeah.
00:22:41.152 --> 00:22:42.093
And so, yeah,
00:22:42.113 --> 00:22:45.875
I think it's based on aspects of what's
00:22:45.934 --> 00:22:46.595
it called?
00:22:47.335 --> 00:22:48.215
The Apache.
00:22:49.797 --> 00:22:50.517
Forgotten the name,
00:22:50.557 --> 00:22:52.137
but basically there's an open source
00:22:52.178 --> 00:22:53.479
project for data lineage.
00:22:53.538 --> 00:22:54.858
I think they've all wrapped that.
00:22:54.919 --> 00:22:56.759
I think Amazon and others
00:22:58.955 --> 00:22:59.615
Is it Atlas?
00:22:59.635 --> 00:23:00.435
Apache Atlas?
00:23:00.455 --> 00:23:01.096
Can't remember.
00:23:01.257 --> 00:23:01.497
Anyway.
00:23:01.517 --> 00:23:02.636
Yeah, that's fine.
00:23:03.136 --> 00:23:04.357
So all the same concepts,
00:23:04.438 --> 00:23:05.438
all open standards,
00:23:05.557 --> 00:23:07.919
nothing specifically bespoke about them.
00:23:07.979 --> 00:23:09.499
It's just that they wrap these services
00:23:09.638 --> 00:23:11.519
and you can then roll them out and
00:23:11.539 --> 00:23:13.400
then have extended it into additional
00:23:13.421 --> 00:23:15.740
things, obviously like the DLP stuff,
00:23:15.760 --> 00:23:18.541
which is the in-browser extensions that's
00:23:18.561 --> 00:23:20.042
then feeding back into Purview.
00:23:20.063 --> 00:23:21.923
So they're connecting things into Purview
00:23:21.962 --> 00:23:23.604
to then do the governance and control in
00:23:23.644 --> 00:23:24.663
that for the data layer.
00:23:25.546 --> 00:23:27.165
So to summarize,
00:23:27.666 --> 00:23:29.267
what we're saying is in order for you
00:23:29.287 --> 00:23:32.667
to roll out AI safely and securely in
00:23:32.688 --> 00:23:33.688
your organization, of course,
00:23:33.708 --> 00:23:34.627
there's always risks,
00:23:34.688 --> 00:23:37.269
but to reduce the risk of rolling out
00:23:37.848 --> 00:23:39.549
AI with governance,
00:23:39.750 --> 00:23:40.869
you have to do a few things.
00:23:41.490 --> 00:23:43.590
Make sure your devices and endpoints are
00:23:43.631 --> 00:23:45.371
controlled by the bits that you mentioned,
00:23:45.471 --> 00:23:48.652
use MDM Intune, have compliant devices,
00:23:49.192 --> 00:23:50.673
then make sure people have the right
00:23:50.772 --> 00:23:51.153
access.
00:23:52.472 --> 00:23:52.692
you know,
00:23:52.732 --> 00:23:54.394
you mentioned RBAC role-based access
00:23:54.413 --> 00:23:54.773
control,
00:23:55.413 --> 00:23:56.555
ensure that they have the right level of
00:23:56.694 --> 00:23:57.015
access,
00:23:57.095 --> 00:23:59.516
the right level of information on the
00:23:59.556 --> 00:23:59.895
machine,
00:23:59.935 --> 00:24:02.557
just so the models don't have way more
00:24:02.576 --> 00:24:03.917
information than they're supposed to have,
00:24:04.097 --> 00:24:06.019
and then have what we call the AI
00:24:06.098 --> 00:24:07.019
gateway, i.e.
00:24:07.038 --> 00:24:07.759
the proxy,
00:24:08.200 --> 00:24:09.559
in which you
can control and have
00:24:09.599 --> 00:24:11.921
guardrails and what the model can access
00:24:11.980 --> 00:24:13.761
from your machine, perhaps, you know,
00:24:13.801 --> 00:24:16.282
and token management and token limits and
00:24:16.323 --> 00:24:17.243
routing, you can do that.
00:24:17.763 --> 00:24:19.104
And then what we're talking about is
00:24:19.183 --> 00:24:20.484
making sure that, you know,
00:24:21.025 --> 00:24:22.606
You have the right data governance in
00:24:22.646 --> 00:24:22.946
place.
00:24:23.468 --> 00:24:25.088
You restrict the content discovery.
00:24:25.469 --> 00:24:27.250
You have the encryption in the right
00:24:27.290 --> 00:24:27.590
place.
00:24:28.372 --> 00:24:30.193
And then the good platform foundations
00:24:30.213 --> 00:24:31.154
that already exist.
00:24:31.634 --> 00:24:32.615
You have private endpoints.
00:24:32.655 --> 00:24:34.738
You're controlling everything correctly.
00:24:34.758 --> 00:24:35.999
You have the right keys in place.
00:24:37.000 --> 00:24:39.421
and after doing all this you always still
00:24:39.500 --> 00:24:41.501
need to make sure that you're monitoring
00:24:41.541 --> 00:24:43.001
you're auditing and logging what's
00:24:43.021 --> 00:24:44.762
happened because if something happens you
00:24:44.803 --> 00:24:46.644
perhaps want to go back and see what
00:24:46.663 --> 00:24:48.904
happened and also notify you when
00:24:48.924 --> 00:24:50.825
something wrong happens right so have
00:24:50.865 --> 00:24:53.385
these six things in place and then we
00:24:53.425 --> 00:24:55.126
think about how do we roll it out
00:24:55.146 --> 00:24:56.606
which is uh that's what we do is
00:24:56.626 --> 00:24:58.048
that correct john is that what we're doing
00:24:58.208 --> 00:24:58.887
That's pretty much it.
00:24:58.948 --> 00:24:59.087
I mean,
00:24:59.127 --> 00:25:02.509
you don't need to I think using a
00:25:02.828 --> 00:25:05.490
project to prove it out first as a
00:25:05.509 --> 00:25:06.670
bit of a lighthouse thing.
00:25:08.250 --> 00:25:09.931
I think it's probably that's how I would
00:25:09.951 --> 00:25:10.371
do it first,
00:25:10.411 --> 00:25:14.172
because you'd want to validate how
00:25:14.211 --> 00:25:17.832
effective and efficient not just the
00:25:17.892 --> 00:25:18.952
technical implementation is,
00:25:18.972 --> 00:25:21.252
but your own processes around it.
00:25:21.833 --> 00:25:24.153
Because if from a process perspective,
00:25:25.099 --> 00:25:25.440
You know,
00:25:25.799 --> 00:25:28.121
if the team's already saturated with BAU
00:25:28.260 --> 00:25:30.741
and you're trying to add more work into
00:25:30.781 --> 00:25:32.823
the teams to then be managing,
00:25:33.323 --> 00:25:36.025
onboarding different projects into AI and
00:25:36.464 --> 00:25:37.984
putting all the content filtering in
00:25:38.025 --> 00:25:41.426
place, maybe inside of the gen AI gateway,
00:25:41.446 --> 00:25:41.686
you know,
00:25:41.727 --> 00:25:43.367
what filters and protections do we need
00:25:43.407 --> 00:25:44.048
for this team?
00:25:44.627 --> 00:25:45.008
You know,
00:25:45.347 --> 00:25:47.730
is it still are the same same filters
00:25:47.750 --> 00:25:49.530
still applicable from the lighthouse team
00:25:49.550 --> 00:25:50.510
into the next one, et cetera?
00:25:50.530 --> 00:25:52.071
You're iterating and trying to understand
00:25:52.132 --> 00:25:53.813
from project by project where the risks
00:25:53.893 --> 00:25:54.012
are.
00:25:54.613 --> 00:25:55.693
So I guess you do need to look
00:25:55.733 --> 00:25:55.993
at like,
00:25:56.034 --> 00:25:57.815
how is it actually operationally going to
00:25:57.855 --> 00:25:58.476
come together?
00:25:58.576 --> 00:26:00.376
Who's accountable and responsible for
00:26:00.396 --> 00:26:00.576
what?
00:26:01.356 --> 00:26:04.058
you know is our device management already
00:26:04.118 --> 00:26:06.819
functioning really well and quickly or is
00:26:06.839 --> 00:26:09.842
this gonna cause scaling problems you know
00:26:09.882 --> 00:26:11.303
into the business when we start to
00:26:11.323 --> 00:26:12.683
obviously roll this so you do need to
00:26:12.723 --> 00:26:14.285
be thinking operationally as well as
00:26:14.325 --> 00:26:16.546
technically on both sides to get it right
00:26:16.846 --> 00:26:18.287
and doing a bit of assessment on that
00:26:18.807 --> 00:26:20.208
um but yeah from a technical perspective
00:26:20.228 --> 00:26:21.788
they're all the things you spot on they're
00:26:21.808 --> 00:26:23.851
all the things that you do need to
00:26:23.871 --> 00:26:25.892
think about um as well yeah
00:26:27.186 --> 00:26:30.496
If I'm a CIO or a CTO, John,
00:26:30.516 --> 00:26:32.319
which I'm not, or COO,
00:26:32.401 --> 00:26:34.266
or even a CEO, which you are, John.
00:26:34.286 --> 00:26:34.586
Yeah.
00:26:37.432 --> 00:26:40.934
what is one message that you should give
00:26:40.974 --> 00:26:43.516
to them for the safe rollout of ai
00:26:43.556 --> 00:26:45.758
because one one thing is yes we have
00:26:45.778 --> 00:26:47.338
all these things all the things that you
00:26:47.358 --> 00:26:50.902
have to do to to monitor this is
00:26:50.942 --> 00:26:52.603
for the usage but also the people who
00:26:52.623 --> 00:26:54.523
are implementing these policies you know
00:26:54.544 --> 00:26:56.325
like the it department or where it might
00:26:56.365 --> 00:26:58.946
be things are moving so quickly new models
00:26:58.967 --> 00:27:00.228
are coming out every time everybody's
00:27:00.248 --> 00:27:02.589
saying oh i need the claude fable five
00:27:02.730 --> 00:27:03.891
right it's just come out i want it
00:27:05.063 --> 00:27:06.183
What do you say to them?
00:27:06.263 --> 00:27:08.846
How do you first roll out safely?
00:27:09.468 --> 00:27:10.888
Secondly, stay on top of it.
00:27:11.390 --> 00:27:13.070
So, you know, people who
are using this,
00:27:13.592 --> 00:27:14.792
they don't stay behind.
00:27:15.153 --> 00:27:17.415
But what is the one thing that they
00:27:17.435 --> 00:27:18.076
should focus on?
00:27:18.115 --> 00:27:18.576
A couple of things.
00:27:18.636 --> 00:27:20.999
I know we talked about these six pillars,
00:27:21.160 --> 00:27:24.163
but what are the couple of things you
00:27:24.202 --> 00:27:26.865
should tell them in this podcast, John?
00:27:27.625 --> 00:27:27.865
Uh...
00:27:29.326 --> 00:27:30.746
i don't know about a couple of things
00:27:30.806 --> 00:27:33.366
i think the main thing is to understand
00:27:33.386 --> 00:27:35.567
the objective of it first as a business
00:27:35.907 --> 00:27:39.189
you know are you wanting to roll it
00:27:39.368 --> 00:27:42.210
out because you've got shadow ai going on
00:27:42.490 --> 00:27:44.009
and that feels like a high risk so
00:27:44.029 --> 00:27:46.590
it's more of a risk goal how do
00:27:46.611 --> 00:27:48.311
we mitigate the current risk or
00:27:49.511 --> 00:27:51.332
is it that you're wanting to roll it
00:27:51.633 --> 00:27:54.813
out because we're wanting to reduce our
00:27:54.853 --> 00:27:57.413
operational overheads currently in our
00:27:57.453 --> 00:28:01.234
business or is it the development side is
00:28:01.275 --> 00:28:03.994
it speed of coding or is it application
00:28:04.055 --> 00:28:05.736
centricity are we trying to roll it out
00:28:05.756 --> 00:28:07.236
because we want to get to new markets
00:28:07.455 --> 00:28:09.696
or wanted to improve our product features
00:28:09.737 --> 00:28:09.876
so
00:28:10.777 --> 00:28:12.498
You can't what I wouldn't suggest is
00:28:12.538 --> 00:28:14.240
saying all of it, please.
00:28:14.380 --> 00:28:15.380
Obviously, I know that,
00:28:15.721 --> 00:28:17.321
but it's probably trying to work out as
00:28:17.362 --> 00:28:17.961
a priority,
00:28:18.041 --> 00:28:22.905
which is the first thing to prove for
00:28:22.965 --> 00:28:23.566
AI first,
00:28:24.487 --> 00:28:25.907
rather than trying to look at everything
00:28:25.928 --> 00:28:27.108
that it can do and then trying to
00:28:27.148 --> 00:28:28.150
boil the ocean with everything.
00:28:29.284 --> 00:28:30.744
I think then the principles of all these
00:28:30.785 --> 00:28:32.945
pillars are contextual then to the goal
00:28:32.986 --> 00:28:34.166
and the objective.
00:28:34.406 --> 00:28:36.468
How deep you then go in those pillars
00:28:36.929 --> 00:28:39.269
is contextual to what the real ambition of
00:28:39.289 --> 00:28:39.890
the goal is.
00:28:40.290 --> 00:28:40.671
Therefore,
00:28:40.691 --> 00:28:42.432
you don't need to necessarily start
00:28:42.971 --> 00:28:44.692
getting into all of the depth of each
00:28:44.732 --> 00:28:46.394
pillar straight off the bat,
00:28:46.654 --> 00:28:49.195
and then it takes a year or whatever
00:28:49.296 --> 00:28:51.076
insane amount of time to roll it out
00:28:51.116 --> 00:28:52.298
safely across the business.
00:28:53.260 --> 00:28:54.682
And no one's getting to see any value
00:28:54.761 --> 00:28:55.522
out of it at all.
00:28:55.603 --> 00:28:56.923
Basically, you've diminished your value.
00:28:56.943 --> 00:28:59.204
So start with something very clear and
00:28:59.244 --> 00:29:01.527
concise that frames everything and then
00:29:01.567 --> 00:29:03.448
choose that lighthouse customer and then
00:29:03.508 --> 00:29:05.569
prove those pillars on a shallow front
00:29:06.631 --> 00:29:09.393
first without loads of depth and then
00:29:09.432 --> 00:29:11.814
build the depth over time as you then
00:29:11.854 --> 00:29:14.476
maybe buy another objective off would be
00:29:14.536 --> 00:29:15.477
my advice.
00:29:16.339 --> 00:29:16.559
Okay,
00:29:16.579 --> 00:29:19.280
so this is excellent advice because if you
00:29:19.320 --> 00:29:22.441
try and implement everything in the first
00:29:22.500 --> 00:29:24.142
place, it might just take you months.
00:29:24.321 --> 00:29:25.622
I think that is what I was trying
00:29:25.642 --> 00:29:26.241
to get to, right?
00:29:26.261 --> 00:29:28.083
So your advice, John,
00:29:28.282 --> 00:29:32.423
to people is pick one use case and
00:29:32.604 --> 00:29:35.304
implement for that first rather than
00:29:35.324 --> 00:29:37.384
thinking about everything under the sun,
00:29:37.444 --> 00:29:39.444
operations, development,
00:29:39.484 --> 00:29:41.846
just pick one use case and then implement
00:29:41.925 --> 00:29:42.026
it
00:29:42.605 --> 00:29:43.486
See how that goes.
00:29:44.067 --> 00:29:47.567
Because you have to iterate quick over
00:29:47.667 --> 00:29:48.067
this.
00:29:48.428 --> 00:29:49.048
I know, John,
00:29:49.087 --> 00:29:52.028
we keep joking about how these models are
00:29:52.068 --> 00:29:54.529
next token predictor and sometimes they're
00:29:54.690 --> 00:29:57.490
not that good at producing the information
00:29:57.510 --> 00:29:58.172
that you require.
00:29:58.652 --> 00:29:59.412
But at the same time,
00:29:59.471 --> 00:30:01.053
if you're not using these models,
00:30:02.272 --> 00:30:03.053
you'll be left behind.
00:30:03.073 --> 00:30:04.173
Yeah, you will.
00:30:04.213 --> 00:30:05.294
Yeah.
00:30:05.334 --> 00:30:07.515
But I think because there's so much to
00:30:07.615 --> 00:30:07.694
it.
00:30:09.182 --> 00:30:09.781
Like we're saying,
00:30:09.862 --> 00:30:11.722
I think because it assumes a certain level
00:30:11.782 --> 00:30:13.262
of current maturity,
00:30:13.782 --> 00:30:15.242
like you were saying at the beginning,
00:30:15.884 --> 00:30:17.084
these are not new concepts.
00:30:17.124 --> 00:30:19.483
Device management isn't a specific AI
00:30:19.924 --> 00:30:22.424
concept that's come off the bat of rolling
00:30:22.464 --> 00:30:22.924
AI out.
00:30:23.244 --> 00:30:25.105
Data governance isn't a new thing.
00:30:25.645 --> 00:30:27.465
Identity
and access management isn't a new
00:30:27.506 --> 00:30:27.766
thing.
00:30:28.006 --> 00:30:29.807
So I think what it is doing, though,
00:30:29.886 --> 00:30:31.586
is probably challenging a lot of
00:30:31.646 --> 00:30:33.407
businesses on their current maturity
00:30:33.448 --> 00:30:33.907
posture.
00:30:34.567 --> 00:30:36.588
in those pillars already you know where
00:30:36.709 --> 00:30:38.148
are you in relation to this and it's
00:30:38.169 --> 00:30:39.269
kind of maybe putting it under the
00:30:39.328 --> 00:30:41.769
spotlight a little bit more so on like
00:30:41.869 --> 00:30:44.070
actually the risk of this is much higher
00:30:45.010 --> 00:30:48.951
um as a thing strategic objective than
00:30:48.991 --> 00:30:50.952
maybe some other things that we've had to
00:30:50.972 --> 00:30:53.393
do before and therefore we do have to
00:30:53.413 --> 00:30:56.913
get these things right but again doing it
00:30:56.994 --> 00:30:58.335
right is contextual to
00:30:59.352 --> 00:31:00.993
what is the first thing you're starting
00:31:01.034 --> 00:31:02.535
with and getting the first thing right,
00:31:02.595 --> 00:31:04.176
not necessarily getting all of it right,
00:31:04.217 --> 00:31:06.019
I think is the aim.
00:31:06.078 --> 00:31:08.000
Otherwise, you'll never get anywhere.
00:31:08.080 --> 00:31:09.762
You'll be just sat spinning wheels for
00:31:09.863 --> 00:31:11.464
ages trying to get everything perfect.
00:31:13.527 --> 00:31:14.847
You will just increase shadow AI,
00:31:14.887 --> 00:31:15.368
won't you?
00:31:16.661 --> 00:31:16.800
Yeah,
00:31:17.161 --> 00:31:18.761
how are you finding the role out of
00:31:18.801 --> 00:31:18.942
this?
00:31:19.001 --> 00:31:19.423
I know, John,
00:31:19.442 --> 00:31:20.482
you've been working with a bunch of
00:31:20.522 --> 00:31:21.443
organizations.
00:31:21.663 --> 00:31:24.384
What is your approach that you're taking
00:31:24.865 --> 00:31:25.665
to your clients?
00:31:26.247 --> 00:31:27.906
How are you finding the reception of this
00:31:27.948 --> 00:31:29.327
approach?
00:31:29.488 --> 00:31:31.009
Are people saying these are the right
00:31:31.048 --> 00:31:32.430
levels of controls that we're putting in
00:31:32.450 --> 00:31:33.730
place, or is it too much?
00:31:34.330 --> 00:31:35.311
Are you asking too much of our
00:31:35.352 --> 00:31:37.053
organization to do?
00:31:37.113 --> 00:31:37.873
How are you finding it?
00:31:38.614 --> 00:31:42.655
So I think everybody seems so far to
00:31:42.695 --> 00:31:43.355
be quite...
00:31:45.151 --> 00:31:48.577
well versed in understanding what it
00:31:48.617 --> 00:31:48.877
means.
00:31:48.897 --> 00:31:51.303
I think probably because as in what it
00:31:51.363 --> 00:31:51.963
means to
00:31:54.303 --> 00:31:56.864
the risk I think is more what I
00:31:56.884 --> 00:31:58.424
kind of mean there because everyone's
00:31:58.545 --> 00:32:02.208
using it personally and so they've already
00:32:02.248 --> 00:32:04.249
kind of they're already sending things hey
00:32:04.308 --> 00:32:07.510
here's my I don't know here's my travel
00:32:07.550 --> 00:32:09.573
documents for this thing or go read my
00:32:09.653 --> 00:32:11.594
email for this thing and I want to
00:32:11.634 --> 00:32:13.375
improve like I need to change the flight
00:32:13.434 --> 00:32:15.517
to whatever and you know I need to
00:32:15.557 --> 00:32:18.778
book this holiday and so that everybody's
00:32:18.858 --> 00:32:21.000
sharing quite a lot of themselves right
00:32:21.059 --> 00:32:21.601
already so
00:32:22.441 --> 00:32:25.521
I think the concept of data sharing is
00:32:25.561 --> 00:32:27.063
very well understood already.
00:32:27.083 --> 00:32:28.563
So there's no educational aspect there.
00:32:29.483 --> 00:32:33.125
I think what I will say is probably
00:32:33.144 --> 00:32:35.326
the biggest challenge is the facilitation
00:32:35.566 --> 00:32:36.506
on scope.
00:32:37.266 --> 00:32:39.547
I think that's probably the value we bring
00:32:39.567 --> 00:32:41.147
is really trying to pinch it into
00:32:41.188 --> 00:32:41.827
something of like,
00:32:41.867 --> 00:32:44.108
how do you prove the first piece of
00:32:44.169 --> 00:32:46.789
value of AI rather than
00:32:47.932 --> 00:32:50.074
getting hung up maybe on all of the
00:32:50.173 --> 00:32:52.596
risk of AI as an entirety all the
00:32:52.616 --> 00:32:52.916
time.
00:32:52.957 --> 00:32:55.138
So trying to get everybody into the right
00:32:55.179 --> 00:32:57.760
mindset of we don't need to worry about
00:32:57.842 --> 00:32:59.623
all those things first.
00:32:59.803 --> 00:33:02.486
Let's just really try and work out what
00:33:02.526 --> 00:33:05.909
is it you're really trying to do because
00:33:05.929 --> 00:33:07.191
we want to prove value quickly.
00:33:07.211 --> 00:33:08.131
That's always the aim.
00:33:08.211 --> 00:33:09.092
Everybody wants that.
00:33:10.134 --> 00:33:12.637
um and so let's really focus on that
00:33:12.678 --> 00:33:14.019
i think most people probably struggle with
00:33:14.038 --> 00:33:15.760
that bit is probably the biggest thing
00:33:15.780 --> 00:33:19.003
like where do we begin begin what do
00:33:19.044 --> 00:33:21.747
we choose yeah and what's what about you
00:33:21.767 --> 00:33:23.769
because you're in customers too and you're
00:33:24.709 --> 00:33:26.451
obviously speaking to people all the time
00:33:26.510 --> 00:33:28.893
and you know what would you agree with
00:33:28.913 --> 00:33:29.934
some of the things i've said do you
00:33:29.974 --> 00:33:31.296
think i'm talking nonsense
00:33:32.179 --> 00:33:34.001
No, no, John, a hundred percent agree.
00:33:34.082 --> 00:33:37.347
I think the issue that I see is,
00:33:37.407 --> 00:33:39.491
and the reason why I really like your
00:33:39.531 --> 00:33:41.836
idea of just pick one use case and
00:33:42.136 --> 00:33:42.637
run with
it,
00:33:43.159 --> 00:33:43.619
is because
00:33:45.285 --> 00:33:49.348
luckily for us open ai and anthropic have
00:33:49.388 --> 00:33:52.030
done the marketing for us right because
00:33:52.111 --> 00:33:54.012
everybody as you say is using these models
00:33:54.292 --> 00:33:56.494
in their own capacity and then they see
00:33:56.555 --> 00:33:58.155
some of these colleagues or some of their
00:33:58.215 --> 00:34:00.778
friends are making things really quickly
00:34:00.838 --> 00:34:04.662
prototyping businesses right websites apps
00:34:04.682 --> 00:34:06.123
they're just coming out within a couple of
00:34:06.143 --> 00:34:08.465
days things are working and
00:34:09.505 --> 00:34:11.405
What people are asking is,
00:34:11.847 --> 00:34:13.648
can we have this straight away?
00:34:13.728 --> 00:34:15.188
Can we have this technology?
00:34:15.208 --> 00:34:16.429
Because a lot of organizations are like,
00:34:16.469 --> 00:34:16.590
well,
00:34:16.630 --> 00:34:18.311
we're not going to allow you to use
00:34:18.331 --> 00:34:20.612
these tools because we're still trying to
00:34:20.652 --> 00:34:22.813
figure out how to use it safely,
00:34:23.173 --> 00:34:24.755
how to use security, what's allowed,
00:34:24.835 --> 00:34:25.574
what's not allowed.
00:34:26.315 --> 00:34:29.378
So the employees and their colleagues are
00:34:29.438 --> 00:34:30.739
not getting it fast enough.
00:34:30.998 --> 00:34:32.619
I think that's number one.
00:34:32.659 --> 00:34:35.041
Some organizations are still trying to
00:34:35.061 --> 00:34:36.943
figure out how to do this securely.
00:34:37.583 --> 00:34:38.023
That's number one.
00:34:38.083 --> 00:34:39.324
They're not getting it fast enough.
00:34:39.844 --> 00:34:41.927
And number two is the people who are
00:34:41.947 --> 00:34:42.807
implementing this.
00:34:43.108 --> 00:34:43.528
They're like,
00:34:43.728 --> 00:34:45.490
there's a lot of this that we don't
00:34:45.550 --> 00:34:46.391
fully understand.
00:34:46.871 --> 00:34:50.114
So there's an education on what are the
00:34:50.175 --> 00:34:50.896
actual risks?
00:34:51.755 --> 00:34:54.159
Sometimes when you have that major threat
00:34:54.199 --> 00:34:54.998
matrix that you make,
00:34:55.039 --> 00:34:55.940
what are the threats?
00:34:56.000 --> 00:34:57.262
And that's why it's important to
00:34:57.282 --> 00:34:59.242
understand what are the actual threats?
00:34:59.884 --> 00:35:01.704
that you're trying to mitigate against.
00:35:02.126 --> 00:35:03.445
And we're seeing it being rolled out.
00:35:04.047 --> 00:35:06.748
And I think the biggest issue for the
00:35:06.789 --> 00:35:09.489
organizations is making sure that people
00:35:09.510 --> 00:35:10.610
are not sharing this sensitive
00:35:10.690 --> 00:35:11.271
information.
00:35:11.291 --> 00:35:12.853
I think that's the biggest problem that
00:35:13.393 --> 00:35:14.074
they're worried about.
00:35:15.213 --> 00:35:17.576
Some
companies don't really care too much
00:35:17.635 --> 00:35:18.956
about, oh, okay,
00:35:19.036 --> 00:35:20.958
if you do share my piece of code,
00:35:21.057 --> 00:35:22.838
which is already coding in the open.
00:35:23.239 --> 00:35:23.719
So, I mean,
00:35:23.820 --> 00:35:27.041
if I do use that model to share
00:35:27.061 --> 00:35:27.521
my code, I mean,
00:35:27.541 --> 00:35:28.503
it's already in open anyway,
00:35:28.523 --> 00:35:29.523
so it doesn't make any difference.
00:35:30.023 --> 00:35:31.885
but some people are still worried about oh
00:35:31.905 --> 00:35:34.367
i've got this proprietary software that if
00:35:34.407 --> 00:35:35.809
people are doing software development on
00:35:36.851 --> 00:35:37.791
that's going to go to the model
00:35:39.239 --> 00:35:40.719
then you know people will get people get
00:35:40.780 --> 00:35:43.581
our proprietary information but the model
00:35:44.320 --> 00:35:45.541
i'm not really too sure if that's too
00:35:45.581 --> 00:35:47.842
much of a problem because you know the
00:35:47.882 --> 00:35:49.202
model already has a bunch of this
00:35:49.342 --> 00:35:51.224
information already there right uh you
00:35:51.244 --> 00:35:52.885
know like creating rocket science is not
00:35:52.985 --> 00:35:55.226
something not new but yeah so what we're
00:35:55.246 --> 00:35:58.907
saying is people uh in some organizations
00:35:59.027 --> 00:36:01.409
really want this and the the organizations
00:36:01.449 --> 00:36:03.068
are still trying to figure out well number
00:36:03.128 --> 00:36:04.050
one which models to use
00:36:04.829 --> 00:36:06.610
and the cost of these models and i
00:36:06.630 --> 00:36:08.570
think people are worried also about the
00:36:08.610 --> 00:36:10.891
tokens right remember we spoke about a
00:36:11.211 --> 00:36:14.032
couple of podcasts ago ceo of uber said
00:36:14.052 --> 00:36:15.673
we blew up all the budget for the
00:36:15.693 --> 00:36:18.494
whole year in april uh twenty twenty six
00:36:18.815 --> 00:36:20.914
for the hood and there's not one single
00:36:20.934 --> 00:36:22.615
feature i can tell you that was released
00:36:22.655 --> 00:36:25.197
with with ai right so there's uh
00:36:26.565 --> 00:36:30.248
still a very new area that people are
00:36:30.268 --> 00:36:32.750
trying to get to grips with yes good
00:36:32.809 --> 00:36:34.831
you can go on chat gpt and ask
00:36:34.891 --> 00:36:37.173
it to generate you images of you doing
00:36:37.293 --> 00:36:39.195
one or other thing or you can like
00:36:39.235 --> 00:36:40.856
run your own websites which is fine but
00:36:40.896 --> 00:36:44.340
like i i still think john um this
00:36:44.500 --> 00:36:46.621
what you talk about the six pillars is
00:36:46.681 --> 00:36:49.284
needed we're talking about pick one use
00:36:49.304 --> 00:36:51.605
case and developers needed so yeah this is
00:36:51.746 --> 00:36:52.266
uh this
00:36:52.465 --> 00:36:54.047
This is an area that's developing really
00:36:54.088 --> 00:36:54.447
quickly.
00:36:55.148 --> 00:36:56.030
But the good thing,
00:36:56.230 --> 00:36:57.411
as you mentioned earlier on,
00:36:58.072 --> 00:37:01.496
is these principles they were discussing.
00:37:01.717 --> 00:37:02.237
They're not new.
00:37:02.657 --> 00:37:03.978
Some of the implementation is a little bit
00:37:04.018 --> 00:37:05.239
different, but they're not new.
00:37:05.900 --> 00:37:07.882
Yeah, exactly.
00:37:08.043 --> 00:37:08.264
Cool.
00:37:09.045 --> 00:37:10.445
Well, that's it, I think,
00:37:10.945 --> 00:37:11.726
for this episode.
00:37:12.547 --> 00:37:14.027
Well, it felt short,
00:37:14.047 --> 00:37:16.007
but it hasn't been as short as I
00:37:16.068 --> 00:37:16.809
thought it would have been.
00:37:16.849 --> 00:37:20.030
But I guess the next one,
00:37:20.210 --> 00:37:22.351
we will probably either choose a topic or
00:37:22.391 --> 00:37:22.931
get back to the news.
00:37:22.952 --> 00:37:24.052
There's been a load of things in the
00:37:24.092 --> 00:37:24.351
news,
00:37:24.371 --> 00:37:25.512
which I did really want to have had
00:37:25.532 --> 00:37:27.914
to bite my tongue to not bring up.
00:37:27.954 --> 00:37:29.153
We kind of did hint at it a
00:37:29.173 --> 00:37:32.615
bit with the new model from China and
00:37:32.655 --> 00:37:34.336
how cheap it is and the efficiency of
00:37:34.396 --> 00:37:34.456
it.
00:37:35.510 --> 00:37:36.871
been on a par but for a fraction
00:37:36.911 --> 00:37:38.932
of the price I believe as in like
00:37:38.952 --> 00:37:41.034
the cost the capex cost that it's taken
00:37:41.715 --> 00:37:44.916
I think is something like you know I
00:37:44.936 --> 00:37:47.557
can't remember like half or even more
00:37:47.577 --> 00:37:49.159
maybe like a third of the price that
00:37:49.179 --> 00:37:50.599
they've had to spend to get to the
00:37:50.639 --> 00:37:55.601
same place as the US super interesting and
00:37:55.621 --> 00:37:57.302
share prices and all these things going on
00:37:57.342 --> 00:37:58.744
so we'll probably maybe touch on that
00:37:59.063 --> 00:38:00.724
depending on what's happening next week
00:38:02.204 --> 00:38:02.525
but yeah
00:38:03.641 --> 00:38:06.751
Stay tuned for the next episode.
00:38:06.791 --> 00:38:07.092
Cool.
00:38:07.956 --> 00:38:08.778
See you later, everyone.
00:38:09.802 --> 00:38:09.983
See you.