WEBVTT 00:00:20.347 --> 00:00:21.208 Welcome, Lewis. 00:00:22.190 --> 00:00:23.550 We've got another episode today, 00:00:23.571 --> 00:00:24.670 haven't we? 00:00:24.731 --> 00:00:25.371 We have. 00:00:26.413 --> 00:00:28.614 And, well, first of many, 00:00:29.495 --> 00:00:31.175 return of season three. 00:00:31.216 --> 00:00:32.716 We have potentially maybe 00:00:32.737 --> 00:00:33.637 done some recordings but 00:00:33.658 --> 00:00:34.319 not released them, 00:00:34.378 --> 00:00:36.840 but shall we discuss... 00:00:36.859 --> 00:00:37.581 Contentious point. 00:00:38.121 --> 00:00:39.042 Very contentious point. 00:00:39.061 --> 00:00:39.802 We'll discuss what we're 00:00:39.823 --> 00:00:41.223 going to be talking about today. 00:00:41.283 --> 00:00:42.424 So, obviously, 00:00:42.445 --> 00:00:44.405 we always pick something from the news, 00:00:45.087 --> 00:00:46.588 most relevant things in the week. 00:00:48.838 --> 00:00:49.838 I've picked two things. 00:00:50.738 --> 00:00:52.840 Agent space from Google, 00:00:53.479 --> 00:00:55.201 which I'll talk about and 00:00:55.841 --> 00:00:56.642 the benefits of it. 00:00:57.302 --> 00:01:00.024 And then another breach by Oracle, 00:01:00.183 --> 00:01:01.284 as in like not by Oracle 00:01:01.325 --> 00:01:01.965 causing the breach, 00:01:02.005 --> 00:01:03.125 but Oracle got breached 00:01:03.246 --> 00:01:04.266 after the first cover-up 00:01:04.887 --> 00:01:05.847 and then got breached again. 00:01:05.888 --> 00:01:07.207 So there's been a second cover-up. 00:01:09.088 --> 00:01:10.329 And then a miscellaneous thing, 00:01:10.709 --> 00:01:12.311 which we'll do at the very end. 00:01:12.370 --> 00:01:13.311 What are your two things? 00:01:14.593 --> 00:01:14.832 So... 00:01:16.373 --> 00:01:17.454 I'm going to dig into 00:01:17.635 --> 00:01:20.397 OpenAI's new AI models, 00:01:20.457 --> 00:01:21.597 but a little bit on the 00:01:21.638 --> 00:01:23.780 marketing about what 00:01:23.799 --> 00:01:25.402 they've said are the major 00:01:25.462 --> 00:01:26.962 new capabilities of their models. 00:01:27.623 --> 00:01:30.206 And also the Clean Cloud Act, 00:01:30.965 --> 00:01:32.266 a bill going through the 00:01:32.326 --> 00:01:33.108 Senate in the US. 00:01:33.128 --> 00:01:34.709 Oh, very good. 00:01:35.290 --> 00:01:36.070 And how have you been? 00:01:36.331 --> 00:01:38.813 You've been cursoring, AI-ing, 00:01:40.573 --> 00:01:41.635 getting the AIs to write 00:01:41.734 --> 00:01:43.516 all your stuff for you yet? 00:01:44.956 --> 00:01:45.436 Yeah, I mean, 00:01:45.456 --> 00:01:47.897 I'm always on that constant battle, 00:01:48.037 --> 00:01:50.399 although I handed it or 00:01:50.438 --> 00:01:52.560 shared the little battle 00:01:52.579 --> 00:01:54.320 with the crazy machines 00:01:55.341 --> 00:01:56.581 with my son at the weekend. 00:01:56.602 --> 00:01:57.503 It was Easter, 00:01:57.843 --> 00:01:59.804 and normally I'd want to be 00:02:00.605 --> 00:02:02.585 on my one wheel or cycling 00:02:02.665 --> 00:02:04.686 or snowboarding or doing something, 00:02:04.727 --> 00:02:06.408 but we stayed in. 00:02:07.067 --> 00:02:07.408 Oh, wow. 00:02:07.447 --> 00:02:08.868 We did some vibe coding with 00:02:09.429 --> 00:02:10.789 Google's Firebase Studio. 00:02:10.810 --> 00:02:11.710 That's pretty cool. 00:02:11.811 --> 00:02:12.651 And it worked. 00:02:16.074 --> 00:02:16.195 It was. 00:02:16.716 --> 00:02:18.438 He enjoyed the outcome more 00:02:18.479 --> 00:02:19.038 than the thing. 00:02:19.340 --> 00:02:20.341 So that's a no. 00:02:20.420 --> 00:02:21.663 I'm guessing that's a no. 00:02:22.103 --> 00:02:24.667 He was excited at the end of 00:02:24.706 --> 00:02:25.448 the kerb and at the 00:02:25.487 --> 00:02:26.449 beginning of the kerb. 00:02:27.510 --> 00:02:27.610 Yeah. 00:02:27.949 --> 00:02:28.849 But at the middle, 00:02:29.469 --> 00:02:32.491 like the bits where it doesn't work, oh, 00:02:32.510 --> 00:02:33.570 we've got to upgrade it. 00:02:33.711 --> 00:02:34.551 Now we've got to come up 00:02:34.572 --> 00:02:36.151 with a really clear and 00:02:36.252 --> 00:02:39.052 concise set of instructions 00:02:39.073 --> 00:02:41.112 as kind of a plan rather 00:02:41.152 --> 00:02:44.913 than just a big wobbly one 00:02:45.033 --> 00:02:46.974 paragraph sort of bit 00:02:47.034 --> 00:02:49.474 contradictory sort of spiel. 00:02:50.335 --> 00:02:51.074 You've got to turn it into 00:02:51.155 --> 00:02:52.116 something half decent. 00:02:52.656 --> 00:02:54.376 So we got it working. 00:02:54.395 --> 00:02:55.977 When all of his friends were 00:02:55.997 --> 00:02:57.217 outside saying, Charlie. 00:02:58.080 --> 00:02:58.900 Are you coming out? 00:02:58.979 --> 00:03:00.401 And you're like, look, Charlie, 00:03:01.020 --> 00:03:04.462 join the blinds. 00:03:04.502 --> 00:03:05.543 We're not going anywhere 00:03:05.603 --> 00:03:07.144 until this code is finished 00:03:07.783 --> 00:03:09.324 and we've shipped a release. 00:03:09.925 --> 00:03:11.585 You're so cool. 00:03:11.626 --> 00:03:13.126 What did you get up to this weekend? 00:03:13.186 --> 00:03:15.667 Well, I one-wheeled. 00:03:15.826 --> 00:03:16.707 I went out. 00:03:17.268 --> 00:03:22.150 I did some snowboarding. 00:03:22.889 --> 00:03:23.509 What did I get up to? 00:03:23.530 --> 00:03:24.550 I went to Brighton. 00:03:26.199 --> 00:03:29.043 um saw some friends um which 00:03:29.082 --> 00:03:30.144 is cool because it changed 00:03:30.163 --> 00:03:32.806 a bit like got like a a 00:03:32.866 --> 00:03:34.306 pool on the beach now an 00:03:34.387 --> 00:03:36.049 open air pool on the beach 00:03:36.088 --> 00:03:39.592 and a sauna and the um what 00:03:39.611 --> 00:03:40.512 do you call it volleyball 00:03:40.532 --> 00:03:41.633 courts all kind of near 00:03:41.693 --> 00:03:42.253 each other so you can 00:03:42.274 --> 00:03:43.794 basically go in the sauna 00:03:44.415 --> 00:03:45.235 and people are running into 00:03:45.276 --> 00:03:47.497 the sea um and then 00:03:47.518 --> 00:03:48.218 obviously people are then 00:03:48.239 --> 00:03:49.419 going off for a proper swim 00:03:49.539 --> 00:03:50.561 but not in the sea then 00:03:50.621 --> 00:03:51.662 swimming into the pool so 00:03:51.701 --> 00:03:52.842 it's really it's very warm 00:03:54.585 --> 00:03:55.506 So I've kind of found my 00:03:55.526 --> 00:03:56.306 love again for Brighton 00:03:56.326 --> 00:03:57.508 because I used to live there years ago. 00:03:57.548 --> 00:03:59.288 So I was a bit like, actually, 00:03:59.308 --> 00:04:01.830 it's such a cool place. 00:04:01.850 --> 00:04:05.514 I think being by the seaside is a thing. 00:04:06.655 --> 00:04:07.074 Exactly. 00:04:07.175 --> 00:04:08.036 But you can't swim in it 00:04:08.436 --> 00:04:09.597 because it's too polluted. 00:04:11.299 --> 00:04:15.502 And then just caught up with 00:04:15.521 --> 00:04:17.163 some other friends, chilled out a bit. 00:04:17.684 --> 00:04:18.444 And so I kind of had a bit 00:04:18.463 --> 00:04:20.206 of a mixed bag as well. 00:04:21.348 --> 00:04:23.449 But talking about the news, 00:04:23.589 --> 00:04:24.589 talking about cloud, 00:04:24.831 --> 00:04:25.930 talking about AI and how 00:04:25.971 --> 00:04:27.713 it's going to take over the world, 00:04:29.954 --> 00:04:31.274 I don't know how much you know about this, 00:04:31.375 --> 00:04:34.958 but there's agent space. 00:04:34.997 --> 00:04:36.778 And it isn't space, 00:04:38.019 --> 00:04:39.321 so don't be too disappointed. 00:04:39.821 --> 00:04:41.742 It has nothing to do with actual space. 00:04:43.716 --> 00:04:45.658 So it's really... Is it 00:04:45.677 --> 00:04:47.177 worth continuing then? 00:04:47.218 --> 00:04:48.377 I'm not really sure if it is 00:04:48.658 --> 00:04:49.899 really worth continuing. 00:04:50.399 --> 00:04:51.678 Is it about spies? 00:04:52.259 --> 00:04:53.540 Spies in space? 00:04:54.040 --> 00:04:54.980 It's about female 00:04:55.000 --> 00:04:57.461 celebrities going on a blue 00:04:57.641 --> 00:05:01.262 origin trip for humanity 00:05:01.521 --> 00:05:03.362 and really... I heard there 00:05:03.422 --> 00:05:05.882 is someone called... Is it Kate Perry? 00:05:06.062 --> 00:05:06.723 Katie Perry. 00:05:08.624 --> 00:05:10.163 I've heard that this person exists. 00:05:11.317 --> 00:05:12.358 yeah and that's that's what 00:05:12.377 --> 00:05:14.059 this is about so google 00:05:14.079 --> 00:05:15.100 have come out with a new 00:05:15.180 --> 00:05:17.221 service that helps you be 00:05:17.281 --> 00:05:21.204 katie perry um so yeah no 00:05:21.223 --> 00:05:22.124 um so yeah so they've got 00:05:22.163 --> 00:05:24.225 this new service ish where 00:05:24.245 --> 00:05:25.026 it's basically agent to 00:05:25.146 --> 00:05:25.966 agent so you can kind of 00:05:25.987 --> 00:05:28.108 create agents um and kind 00:05:28.127 --> 00:05:29.369 of chain them together but 00:05:29.408 --> 00:05:32.290 it's a kind of multimodal 00:05:32.531 --> 00:05:33.771 supported thing also has 00:05:33.812 --> 00:05:35.853 integrations into lots of 00:05:35.872 --> 00:05:37.254 other services like service 00:05:37.314 --> 00:05:39.915 now and jira google drive 00:05:40.930 --> 00:05:42.072 And actually what you can do 00:05:42.132 --> 00:05:44.795 is you can start to plug it in, um, 00:05:45.016 --> 00:05:45.536 at the minimum, 00:05:45.576 --> 00:05:46.499 probably plug it into 00:05:46.538 --> 00:05:47.961 things that you've got in your business. 00:05:48.000 --> 00:05:50.264 So if you already use Google, um, 00:05:51.209 --> 00:05:51.928 And that's great because 00:05:51.949 --> 00:05:52.509 obviously you'll have all 00:05:52.529 --> 00:05:53.310 your assets there. 00:05:53.970 --> 00:05:55.211 A lot of companies like us 00:05:55.571 --> 00:05:57.211 obviously use Confluence and Jira, 00:05:57.250 --> 00:05:58.031 so that's quite good. 00:05:58.052 --> 00:05:59.012 So it has like native 00:05:59.031 --> 00:06:00.112 support for a bunch of like 00:06:00.153 --> 00:06:00.913 default tools. 00:06:01.752 --> 00:06:02.774 And then you can just start 00:06:02.814 --> 00:06:03.973 to have conversations with 00:06:04.014 --> 00:06:05.935 it and be like- John, 00:06:06.355 --> 00:06:07.776 just to interject a second, 00:06:07.836 --> 00:06:09.136 is that why we've got fifty 00:06:09.255 --> 00:06:11.016 thousand new Jira tickets? 00:06:12.283 --> 00:06:13.444 That was me, yeah. 00:06:13.485 --> 00:06:15.565 So what I needed access to 00:06:15.706 --> 00:06:17.226 agent space first. 00:06:17.247 --> 00:06:19.668 I had to get it to raise the 00:06:19.709 --> 00:06:20.569 tickets for it to work 00:06:20.589 --> 00:06:25.052 because I couldn't get it to work. 00:06:25.072 --> 00:06:25.892 So the pricing, 00:06:26.934 --> 00:06:27.935 they do like a free one 00:06:27.954 --> 00:06:29.536 where they give you like fifty thousand, 00:06:29.555 --> 00:06:31.516 I think fifty thousand 00:06:31.576 --> 00:06:34.019 licenses as part of like 00:06:34.038 --> 00:06:36.000 the free offering for like fourteen days. 00:06:36.019 --> 00:06:37.180 So I was kind of playing around with that. 00:06:38.240 --> 00:06:39.721 But then the pricing is quite a lot. 00:06:39.762 --> 00:06:40.423 I think it was something 00:06:40.483 --> 00:06:43.064 like a hundred and thirty 00:06:43.105 --> 00:06:44.567 five dollars a month just 00:06:44.586 --> 00:06:46.569 for fifteen licenses, 00:06:47.449 --> 00:06:49.952 which I think is very steep on top of, 00:06:50.012 --> 00:06:50.653 say, 00:06:51.555 --> 00:06:53.137 if you're already using Workspace and 00:06:53.156 --> 00:06:54.737 you've got Google and 00:06:57.093 --> 00:06:59.074 using Drive and Gmail and other things, 00:06:59.595 --> 00:07:01.274 that's already got a user price on it. 00:07:02.055 --> 00:07:03.214 So then to spend a hundred 00:07:03.235 --> 00:07:04.495 and thirty five pounds for fifteen, 00:07:04.516 --> 00:07:05.716 which I think is even more 00:07:06.235 --> 00:07:08.055 than your license for Gmail 00:07:08.096 --> 00:07:10.057 and Drive itself, 00:07:10.177 --> 00:07:11.276 seems a bit 00:07:11.317 --> 00:07:13.838 disproportionate for a pricing 00:07:13.858 --> 00:07:16.577 model to make your life a bit easier. 00:07:16.598 --> 00:07:18.519 I don't know. 00:07:18.538 --> 00:07:19.559 What's your thoughts on it? 00:07:20.259 --> 00:07:20.598 I don't know. 00:07:20.658 --> 00:07:21.918 It seems so. 00:07:22.298 --> 00:07:24.699 Is agent space by Google? 00:07:26.060 --> 00:07:27.906 Yes, it's a service in Google Cloud. 00:07:28.987 --> 00:07:29.951 And you basically could like, 00:07:30.110 --> 00:07:30.773 you enable it. 00:07:30.872 --> 00:07:32.216 It's a Google Cloud service. 00:07:33.747 --> 00:07:34.468 And then you permission it 00:07:34.567 --> 00:07:37.088 into the files and things like that. 00:07:37.108 --> 00:07:37.209 I see. 00:07:37.249 --> 00:07:38.949 So instead of having to rely 00:07:39.209 --> 00:07:41.850 on the Gemini button that's 00:07:41.910 --> 00:07:43.189 everywhere in Google these 00:07:43.209 --> 00:07:46.610 days to do a point thing on 00:07:46.651 --> 00:07:48.351 a point bit of data in a 00:07:48.432 --> 00:07:49.851 very specific thing with 00:07:49.911 --> 00:07:51.172 hardly any tool use, 00:07:51.211 --> 00:07:52.452 and it's only just got 00:07:52.533 --> 00:07:53.773 access to do anything with 00:07:53.833 --> 00:07:55.132 cheats at all useful. 00:07:56.173 --> 00:07:57.293 And it's building. 00:07:58.334 --> 00:08:01.334 Instead of going it from the tool out, 00:08:01.394 --> 00:08:01.795 they've... 00:08:02.855 --> 00:08:04.836 aggregate data and have a 00:08:04.896 --> 00:08:06.478 flow that moves across your 00:08:06.497 --> 00:08:08.300 data sets yeah I think it 00:08:08.339 --> 00:08:08.899 was part of their 00:08:08.939 --> 00:08:09.880 conference as well they did 00:08:09.940 --> 00:08:10.841 a talk they did an 00:08:10.882 --> 00:08:14.785 announcement for it um 00:08:14.824 --> 00:08:16.026 making about a hundred 00:08:16.045 --> 00:08:17.607 announcement per second no 00:08:18.108 --> 00:08:19.689 I think that I think I 00:08:19.709 --> 00:08:20.369 think they had a conference 00:08:20.408 --> 00:08:22.110 you know yeah so um so yeah 00:08:22.130 --> 00:08:24.312 so they'll have all the 00:08:24.333 --> 00:08:25.192 things from like the key 00:08:25.233 --> 00:08:26.975 notice keynotes and things like that but 00:08:28.375 --> 00:08:33.101 Yeah, so I think in one of their, I guess, 00:08:33.162 --> 00:08:37.106 live demos, they talk about agent space. 00:08:37.447 --> 00:08:39.250 And then I think they use it as a bank. 00:08:39.490 --> 00:08:42.474 And I can't remember really 00:08:42.494 --> 00:08:43.315 what the talk was about, 00:08:43.335 --> 00:08:44.836 so I'm kind of plagiarizing badly. 00:08:46.171 --> 00:08:47.451 They were like role-playing 00:08:47.611 --> 00:08:48.672 around if you're a certain 00:08:48.731 --> 00:08:49.832 person in a bank and you 00:08:49.873 --> 00:08:51.634 want to kind of get access 00:08:51.653 --> 00:08:52.955 to certain details and then 00:08:52.995 --> 00:08:54.235 produce a report and then 00:08:54.294 --> 00:08:55.375 schedule that into an email. 00:08:55.395 --> 00:08:56.936 So essentially it can even 00:08:57.277 --> 00:08:58.677 integrate into your calendar, 00:08:59.398 --> 00:09:00.638 add the people into the calendar, 00:09:00.758 --> 00:09:01.458 send the meeting, 00:09:01.519 --> 00:09:02.678 send the email invite and 00:09:02.698 --> 00:09:03.399 all these other things. 00:09:03.440 --> 00:09:06.061 So it goes beyond just 00:09:06.100 --> 00:09:07.902 facilitation on some information. 00:09:08.601 --> 00:09:11.322 It's actually a workflow of working. 00:09:11.342 --> 00:09:12.703 I think what they're looking 00:09:12.744 --> 00:09:13.764 at is don't leave... 00:09:15.770 --> 00:09:19.392 It's organized this podcast, yeah. 00:09:20.871 --> 00:09:21.812 Which is why this is our 00:09:21.873 --> 00:09:24.793 fifth attempt today at trying to do it. 00:09:26.754 --> 00:09:28.173 It's still got teething problems. 00:09:28.595 --> 00:09:30.054 So create loads of Jira 00:09:30.095 --> 00:09:34.115 tickets to create a podcast. 00:09:34.135 --> 00:09:34.796 That's what it did. 00:09:35.035 --> 00:09:36.716 But did you get any success out of it? 00:09:38.524 --> 00:09:40.745 so all I will say is I 00:09:40.826 --> 00:09:42.268 haven't actually managed to 00:09:42.488 --> 00:09:44.570 use it because google cloud 00:09:44.610 --> 00:09:45.711 and it's you know 00:09:45.791 --> 00:09:48.192 phenomenal fashion uh it 00:09:48.234 --> 00:09:49.414 needed certain permissions 00:09:49.475 --> 00:09:50.235 obviously for the service 00:09:50.254 --> 00:09:51.196 account needed certain 00:09:51.216 --> 00:09:52.738 permissions for me you get 00:09:52.798 --> 00:09:54.178 a url essentially that's 00:09:54.198 --> 00:09:55.581 then obviously the portal 00:09:55.620 --> 00:09:56.621 of the agent I guess the 00:09:57.042 --> 00:09:59.804 llm chat um window for the 00:09:59.904 --> 00:10:00.926 for the inputs you know the 00:10:00.985 --> 00:10:02.847 context of the inputs but um 00:10:03.865 --> 00:10:04.866 When I went to do that, 00:10:04.966 --> 00:10:06.227 it couldn't validate me as 00:10:06.248 --> 00:10:08.269 a user for some reason. 00:10:08.288 --> 00:10:09.448 Didn't have some permission 00:10:09.509 --> 00:10:10.370 somewhere to obviously be 00:10:10.389 --> 00:10:11.951 able to tell whether I am a user, 00:10:11.990 --> 00:10:12.971 certainly those permissions. 00:10:13.011 --> 00:10:15.552 Discovery permissions is 00:10:15.572 --> 00:10:16.394 what it kind of needed. 00:10:16.793 --> 00:10:18.095 So it didn't take me too long, 00:10:18.115 --> 00:10:19.335 but I hit a few roadblocks 00:10:19.436 --> 00:10:21.015 on actually what it really 00:10:21.096 --> 00:10:22.817 needs to work from a 00:10:22.857 --> 00:10:23.937 permissions perspective. 00:10:26.200 --> 00:10:27.399 So then obviously then I had 00:10:27.419 --> 00:10:28.941 other things to do and had some meetings, 00:10:29.000 --> 00:10:30.001 so I didn't manage to progress. 00:10:31.706 --> 00:10:32.668 but do you think you'd use 00:10:32.707 --> 00:10:34.148 it if I had it going would 00:10:34.168 --> 00:10:39.431 you would you use it uh so 00:10:39.571 --> 00:10:41.951 I guess in in my wheelhouse 00:10:42.292 --> 00:10:43.712 in development I would want 00:10:43.893 --> 00:10:44.953 to plug it very 00:10:44.994 --> 00:10:46.715 specifically into very 00:10:46.735 --> 00:10:48.215 specific apis or tools 00:10:48.716 --> 00:10:51.136 rather than your general 00:10:51.657 --> 00:10:53.498 data you know the the 00:10:53.538 --> 00:10:58.961 workspace stuff jira maybe I don't know 00:10:59.283 --> 00:11:00.283 Yeah, I don't know. 00:11:00.364 --> 00:11:01.264 I'd need to... Yeah, 00:11:01.304 --> 00:11:02.784 so less on the heavy side, 00:11:02.804 --> 00:11:04.145 because you've got other tools for that, 00:11:04.186 --> 00:11:05.546 and less on the engineering side, 00:11:05.586 --> 00:11:06.785 but on the business side, 00:11:07.447 --> 00:11:09.067 probably more on the business side. 00:11:09.147 --> 00:11:11.107 I would imagine so, yeah. 00:11:11.148 --> 00:11:13.087 Yeah, for me, I thought, I mean, 00:11:13.107 --> 00:11:14.989 being more on the business side nowadays, 00:11:15.568 --> 00:11:18.909 then I could see how you could, you know, 00:11:18.950 --> 00:11:19.730 from our side, 00:11:19.750 --> 00:11:21.471 either have contractual elements, 00:11:22.650 --> 00:11:23.971 you could have statements of work, 00:11:24.871 --> 00:11:25.812 and what you could do is 00:11:25.851 --> 00:11:27.253 you could then be like, right, 00:11:29.235 --> 00:11:30.596 take this outcome, 00:11:31.037 --> 00:11:34.201 take this interview on the drive, 00:11:34.240 --> 00:11:35.903 which is like typed up of 00:11:35.982 --> 00:11:38.346 the customer success of our engagement, 00:11:38.846 --> 00:11:39.788 tally the statement of 00:11:39.827 --> 00:11:42.711 workers in what we were trying to solve, 00:11:42.851 --> 00:11:43.732 what the commitment was 00:11:43.753 --> 00:11:44.714 within the outcome, 00:11:44.774 --> 00:11:46.235 and then produce me a case study 00:11:47.057 --> 00:11:49.118 you know based on that thing 00:11:49.158 --> 00:11:50.318 the end to end you know 00:11:50.658 --> 00:11:52.539 what what we said specific 00:11:52.600 --> 00:11:55.302 marketing outcomes and like 00:11:55.322 --> 00:11:56.202 bits and that's just one 00:11:56.243 --> 00:11:57.702 you can break down each bit 00:11:57.943 --> 00:11:59.264 and and do it so when 00:11:59.283 --> 00:12:01.245 you've got messy text and 00:12:01.385 --> 00:12:04.827 no code like software two 00:12:04.868 --> 00:12:07.028 point oh ai space seems 00:12:07.089 --> 00:12:08.289 perfect yeah or like doing a 00:12:09.245 --> 00:12:09.846 you know, 00:12:09.865 --> 00:12:10.927 working out if you've got things 00:12:10.966 --> 00:12:12.447 in confluence, you know, 00:12:12.466 --> 00:12:13.927 are they out of date maybe 00:12:13.988 --> 00:12:15.649 with other things that 00:12:15.688 --> 00:12:16.509 you're kind of saying in 00:12:16.528 --> 00:12:17.230 statements of work? 00:12:17.730 --> 00:12:19.910 I can imagine actually from even from a, 00:12:19.990 --> 00:12:21.291 I know I'm like rifting on ideas, 00:12:21.772 --> 00:12:22.552 but even from like an 00:12:22.591 --> 00:12:24.052 administrative perspective, 00:12:24.832 --> 00:12:25.913 when it comes to content 00:12:26.153 --> 00:12:27.453 it's got a tendency to be 00:12:27.514 --> 00:12:29.293 out of date or to have many 00:12:29.354 --> 00:12:30.474 versions of that thing 00:12:30.533 --> 00:12:31.494 right so I can imagine 00:12:31.533 --> 00:12:32.455 actually as a tool to be 00:12:32.475 --> 00:12:33.955 like can you find all the 00:12:33.975 --> 00:12:35.495 different variant versions 00:12:36.235 --> 00:12:37.475 of this kind of context 00:12:37.515 --> 00:12:38.956 window and actually can you 00:12:38.995 --> 00:12:40.096 find me the most recent and 00:12:40.115 --> 00:12:40.995 then can you delete all the 00:12:41.015 --> 00:12:42.236 other old ones because 00:12:42.596 --> 00:12:43.496 actually it's confusing so 00:12:43.517 --> 00:12:44.476 I could imagine actually 00:12:44.576 --> 00:12:45.857 doing administrative tasks 00:12:45.878 --> 00:12:47.638 to keep your actual data 00:12:47.918 --> 00:12:49.977 integrity higher because 00:12:49.998 --> 00:12:51.078 you don't have a sprawl of 00:12:51.099 --> 00:12:52.099 things through time 00:12:52.937 --> 00:12:54.379 would be very useful as a 00:12:55.519 --> 00:12:56.961 tool and an assistant to 00:12:57.400 --> 00:12:58.922 give you any of the relevant information. 00:13:00.302 --> 00:13:01.764 It seems to overlap a lot. 00:13:01.783 --> 00:13:02.845 I think it's probably 00:13:02.884 --> 00:13:05.726 because of Google's naming conventions. 00:13:05.907 --> 00:13:07.447 I put conventions in some 00:13:07.607 --> 00:13:09.028 air quotes because they 00:13:10.831 --> 00:13:11.893 Some of the news I've heard 00:13:11.913 --> 00:13:14.515 about their agent-to-agent protocol, 00:13:14.875 --> 00:13:17.318 their agent development kit, 00:13:18.461 --> 00:13:21.784 and their Vertex product and Firebase, 00:13:21.804 --> 00:13:22.865 they all seem to be coming 00:13:22.985 --> 00:13:24.207 out at the same time. 00:13:24.788 --> 00:13:26.711 And there's lots of agentic flows 00:13:27.910 --> 00:13:30.193 and discussions about the 00:13:30.274 --> 00:13:32.134 agent development kit. 00:13:33.096 --> 00:13:35.238 So I imagine maybe some of 00:13:35.258 --> 00:13:36.139 these products have come 00:13:36.239 --> 00:13:39.903 out as use cases of some of 00:13:39.923 --> 00:13:41.764 these things and become 00:13:41.784 --> 00:13:42.926 services in their own right. 00:13:43.485 --> 00:13:43.907 I don't know. 00:13:44.462 --> 00:13:45.224 Yeah, no, you're spot on. 00:13:45.403 --> 00:13:46.504 I think that's exactly what's happened. 00:13:46.524 --> 00:13:48.346 I think agent space, 00:13:48.566 --> 00:13:50.488 you can create your own agent even in it, 00:13:51.067 --> 00:13:52.668 and it's a low code or no 00:13:52.710 --> 00:13:54.610 code experience to create even an agent. 00:13:55.251 --> 00:13:57.732 So you can kind of flow 00:13:57.773 --> 00:13:59.835 chart the business process. 00:13:59.934 --> 00:14:01.657 I mean, obviously, I can't get it to work, 00:14:01.697 --> 00:14:02.817 so I've no idea really what 00:14:02.856 --> 00:14:03.357 it really does. 00:14:04.298 --> 00:14:06.480 All I have to rely on is, you know, 00:14:06.500 --> 00:14:09.043 we will get to it. 00:14:09.083 --> 00:14:09.884 Next episode, 00:14:09.904 --> 00:14:11.106 I will have actually had the 00:14:11.147 --> 00:14:12.749 time to have properly used it. 00:14:12.828 --> 00:14:14.850 I haven't really had the time to explore. 00:14:14.890 --> 00:14:16.552 But from what I've seen, 00:14:16.572 --> 00:14:17.514 you can create your own agent. 00:14:17.533 --> 00:14:18.615 It doesn't work for you, right? 00:14:18.695 --> 00:14:19.275 It's obviously, 00:14:19.355 --> 00:14:20.977 I don't know exactly how it 00:14:21.538 --> 00:14:22.320 necessarily does it. 00:14:22.340 --> 00:14:24.381 So you can plug in your own agent space. 00:14:25.000 --> 00:14:26.860 Hence why I imagine the agent-to-agent, 00:14:26.880 --> 00:14:27.461 because that's really 00:14:27.480 --> 00:14:28.561 essentially what it is about. 00:14:29.160 --> 00:14:30.461 It is about agent-to-agent 00:14:30.522 --> 00:14:31.162 and the pass-through of 00:14:31.201 --> 00:14:32.822 information and leveraging 00:14:33.081 --> 00:14:37.803 multi-contextual tooling. 00:14:37.864 --> 00:14:38.724 So, yeah, 00:14:38.844 --> 00:14:40.924 I think that's really where it's going. 00:14:40.945 --> 00:14:42.965 But, yeah, anyway, 00:14:43.784 --> 00:14:44.686 I've dominated a little bit 00:14:44.725 --> 00:14:45.885 on that one thing. 00:14:46.326 --> 00:14:46.926 What's your news? 00:14:47.346 --> 00:14:50.647 What's got your interest this week? 00:14:51.908 --> 00:14:51.988 So... 00:14:54.376 --> 00:14:58.979 I don't know, another thing AI-wise, 00:14:59.298 --> 00:15:04.120 it's OpenAI's new AI models. 00:15:04.721 --> 00:15:06.361 There was an article in The 00:15:06.381 --> 00:15:07.643 Verge talking about 00:15:09.826 --> 00:15:13.190 Their models, O-three and O-four, 00:15:14.110 --> 00:15:16.432 what's that, O-three mini? 00:15:16.471 --> 00:15:17.373 I have to look it up every 00:15:17.413 --> 00:15:19.315 time and I get totally confused. 00:15:19.455 --> 00:15:20.755 And I think that seems to be 00:15:22.096 --> 00:15:24.499 a lot of the discussion that's come out. 00:15:24.538 --> 00:15:25.940 So The Verge came out with 00:15:25.960 --> 00:15:27.100 an article saying their 00:15:27.801 --> 00:15:28.782 O-three model can use 00:15:29.001 --> 00:15:30.423 images when reasoning. 00:15:32.404 --> 00:15:35.645 So it's kind of multimodal, 00:15:35.706 --> 00:15:38.787 but there are specifically reporting on, 00:15:39.427 --> 00:15:40.268 and OpenAI have 00:15:40.347 --> 00:15:42.048 specifically said that 00:15:43.249 --> 00:15:46.789 during its thinking phase, 00:15:46.929 --> 00:15:47.870 when you have chain of 00:15:47.910 --> 00:15:50.491 thought reasoning and AIs 00:15:50.511 --> 00:15:52.413 can show you what they're thinking about. 00:15:53.013 --> 00:15:53.432 Right. 00:15:56.097 --> 00:15:57.239 OpenAI have said their 00:15:57.318 --> 00:15:59.461 O-three model can use 00:15:59.501 --> 00:16:01.264 images when doing that reasoning. 00:16:02.125 --> 00:16:03.846 Now that just seems like 00:16:04.067 --> 00:16:05.708 another way of stating 00:16:07.005 --> 00:16:10.469 moti modality in the model I 00:16:10.609 --> 00:16:12.730 can't actually see that I 00:16:12.750 --> 00:16:13.591 think this is there's a 00:16:13.631 --> 00:16:15.113 marketing polish on top of 00:16:15.133 --> 00:16:16.313 it because I was trying I 00:16:16.374 --> 00:16:17.754 was trying to like get the 00:16:17.794 --> 00:16:19.557 concrete statement they did 00:16:19.576 --> 00:16:20.837 make a couple of statements 00:16:20.977 --> 00:16:22.119 specifically say where we 00:16:22.178 --> 00:16:23.779 can think with images and 00:16:24.159 --> 00:16:25.461 that you know if you have a 00:16:25.562 --> 00:16:27.182 chart of the differences 00:16:27.222 --> 00:16:29.125 between the models um one 00:16:29.144 --> 00:16:30.525 of the things is you know 00:16:30.546 --> 00:16:32.788 these new models can do certain 00:16:33.464 --> 00:16:35.765 thinking um while they're 00:16:35.807 --> 00:16:38.548 doing reasoning with images 00:16:38.870 --> 00:16:40.772 but you look at all of the 00:16:40.831 --> 00:16:42.573 information from deep mind 00:16:43.053 --> 00:16:44.596 and all the gemini models 00:16:44.615 --> 00:16:47.097 from google all boast about 00:16:47.538 --> 00:16:49.561 how they do deep reasoning 00:16:49.921 --> 00:16:52.244 and are multimodal so have 00:16:52.303 --> 00:16:53.644 the context of images 00:16:54.166 --> 00:16:55.706 whilst they're reasoning so it's like 00:16:56.408 --> 00:16:58.188 is this marketing spin 00:16:58.528 --> 00:16:59.928 because obviously we can't 00:17:00.009 --> 00:17:01.169 we can't look inside these 00:17:01.210 --> 00:17:02.890 closed source models and 00:17:02.990 --> 00:17:03.850 see whether they've got a 00:17:03.910 --> 00:17:05.571 mixture of experts and part 00:17:05.632 --> 00:17:08.031 of it is generative for 00:17:08.252 --> 00:17:09.732 images and part of it is 00:17:10.232 --> 00:17:13.375 tokens for image pixels or 00:17:13.414 --> 00:17:14.615 what you don't you don't 00:17:14.634 --> 00:17:15.674 get to find out so all 00:17:15.694 --> 00:17:16.576 we've got to go on is the 00:17:16.596 --> 00:17:17.915 marketing thing so I think 00:17:18.056 --> 00:17:22.238 actually my summary of all that is uh 00:17:23.196 --> 00:17:25.237 OpenAI, and actually all of them, 00:17:25.797 --> 00:17:27.377 the model naming just 00:17:27.538 --> 00:17:28.397 drives me mad because I 00:17:28.438 --> 00:17:29.337 can't remember which one 00:17:29.377 --> 00:17:30.478 I'm talking about or which 00:17:30.557 --> 00:17:31.637 one they're talking about. 00:17:32.278 --> 00:17:34.398 And then some of the marketing spin, 00:17:34.439 --> 00:17:34.939 they're trying to 00:17:34.979 --> 00:17:36.098 differentiate what the 00:17:36.259 --> 00:17:37.619 latest thing in their model, 00:17:38.240 --> 00:17:41.240 and it's very thin and small. 00:17:41.660 --> 00:17:43.000 It's a bit of a rumble. 00:17:43.701 --> 00:17:47.760 Yeah, if I was being a guessing person, 00:17:50.001 --> 00:17:51.981 I would have assumed... 00:17:54.141 --> 00:17:55.642 that if Google were making 00:17:55.682 --> 00:17:57.423 those announcements from a conference, 00:17:58.223 --> 00:17:59.265 that you might also want to 00:17:59.285 --> 00:18:01.465 be PRing your own stuff in 00:18:01.506 --> 00:18:03.508 the meantime to take the attention away, 00:18:05.209 --> 00:18:06.150 even if there wasn't really 00:18:06.190 --> 00:18:09.313 anything too much to really say. 00:18:09.432 --> 00:18:10.333 But I'm kind of guessing. 00:18:10.353 --> 00:18:11.614 I mean, that's just my cynicism. 00:18:11.673 --> 00:18:12.275 Maybe that's not... 00:18:12.744 --> 00:18:13.265 Yeah, no, 00:18:13.365 --> 00:18:15.987 I think I've got a joint bit of 00:18:16.446 --> 00:18:17.428 sense on this. 00:18:17.468 --> 00:18:19.167 And I worry a little bit 00:18:19.509 --> 00:18:22.809 that the very specific... 00:18:23.211 --> 00:18:23.990 When you look at the 00:18:24.131 --> 00:18:25.612 feature chart on the OpenAR 00:18:25.652 --> 00:18:27.272 website about all the new models, 00:18:28.232 --> 00:18:29.374 they've all been updated to 00:18:29.473 --> 00:18:30.714 have access to tools, 00:18:30.755 --> 00:18:32.154 but some of them are memory 00:18:32.174 --> 00:18:32.976 and some of them are other 00:18:33.016 --> 00:18:33.955 tools and some of them are 00:18:33.976 --> 00:18:35.017 context and some of them 00:18:35.037 --> 00:18:36.478 are image reasoning and 00:18:36.518 --> 00:18:37.077 some of them are not. 00:18:37.678 --> 00:18:39.900 It's all lots of words 00:18:40.859 --> 00:18:42.260 And there's not a clear 00:18:42.300 --> 00:18:43.642 distinction about which 00:18:43.761 --> 00:18:44.782 ones will actually give you 00:18:44.923 --> 00:18:47.263 a benefit when you're using them. 00:18:47.324 --> 00:18:50.026 Deep thinking and doing 00:18:51.186 --> 00:18:53.489 agentic flows for carrying 00:18:53.528 --> 00:18:54.368 out deep research, 00:18:54.749 --> 00:18:56.951 that seems like a clear thing, 00:18:56.990 --> 00:18:58.372 which is distinct from... 00:18:59.756 --> 00:19:00.517 one model or the other, 00:19:00.557 --> 00:19:02.117 but they're all based on a 00:19:02.157 --> 00:19:03.558 whole bunch of where they 00:19:03.898 --> 00:19:05.900 place the training so that 00:19:05.920 --> 00:19:07.721 they're better at using tools, 00:19:07.781 --> 00:19:08.902 better at knowing about tools, 00:19:08.942 --> 00:19:10.864 better at knowing about multimodalities. 00:19:11.423 --> 00:19:14.826 But it almost seems like the 00:19:14.906 --> 00:19:16.086 AIs at this point are right 00:19:16.106 --> 00:19:17.107 in their own marketing blood. 00:19:19.390 --> 00:19:20.851 It's a little bit like not on our... 00:19:22.701 --> 00:19:25.082 But did you see the era of experience, 00:19:25.662 --> 00:19:26.782 the white paper, 00:19:26.903 --> 00:19:29.223 the research paper that came out by, 00:19:29.544 --> 00:19:31.424 I need to have a little look, by Google. 00:19:31.505 --> 00:19:33.105 It was by David Silver. 00:19:34.705 --> 00:19:35.967 He's like a Google researcher. 00:19:36.166 --> 00:19:37.287 And then the computer scientist, 00:19:37.487 --> 00:19:38.367 Richard Sutton. 00:19:39.086 --> 00:19:42.147 But talking about where 00:19:42.167 --> 00:19:42.968 they're basically trained 00:19:42.988 --> 00:19:43.888 from their environment. 00:19:43.909 --> 00:19:46.828 So essentially they start to 00:19:46.868 --> 00:19:48.589 become more environmentally aware, 00:19:48.650 --> 00:19:50.471 I suppose you could say, 00:19:50.851 --> 00:19:51.951 and actually start to train 00:19:51.990 --> 00:19:53.030 themselves rather than 00:19:53.112 --> 00:19:55.192 having human data or like 00:19:55.311 --> 00:19:57.573 inputted data to train them human driven. 00:19:58.314 --> 00:19:58.993 data which is very 00:19:59.054 --> 00:20:00.074 complicated and hard to 00:20:00.134 --> 00:20:02.194 come by it can you know 00:20:02.375 --> 00:20:03.375 train itself I was actually 00:20:03.674 --> 00:20:05.275 watching a lecture with uh 00:20:06.016 --> 00:20:08.455 uh what's it dem demis uh 00:20:08.675 --> 00:20:12.136 habilis the the deep um 00:20:12.176 --> 00:20:14.777 deep mind uh ceo right yeah 00:20:15.317 --> 00:20:19.538 um talking about um having 00:20:19.718 --> 00:20:22.898 reinforced learning is the thing um and 00:20:24.065 --> 00:20:25.066 When you've got humans in 00:20:25.086 --> 00:20:26.688 the loop doing the training 00:20:26.728 --> 00:20:28.930 and the feedback and the whole internet, 00:20:29.451 --> 00:20:31.472 it's really muddied and not 00:20:31.653 --> 00:20:32.854 that clearly good. 00:20:33.295 --> 00:20:34.796 But as soon as you get a 00:20:34.875 --> 00:20:36.198 feedback loop where they 00:20:36.218 --> 00:20:38.660 can learn on themselves, 00:20:38.740 --> 00:20:41.282 then if you can just work 00:20:41.383 --> 00:20:42.523 out how to give it clear 00:20:42.564 --> 00:20:43.785 signals about good, 00:20:44.730 --> 00:20:46.451 you can measure that so the 00:20:46.471 --> 00:20:47.392 problem becomes a data 00:20:47.412 --> 00:20:48.393 science problem and not an 00:20:48.492 --> 00:20:49.894 ai problem but if you can 00:20:49.953 --> 00:20:51.815 do that then your ai will 00:20:52.635 --> 00:20:53.696 will learn very very 00:20:53.737 --> 00:20:55.637 quickly so synthetic paper 00:20:55.817 --> 00:20:57.499 data and re and real 00:20:57.538 --> 00:20:59.480 reinforcement learning seem 00:20:59.500 --> 00:21:00.561 to be the paradigms that 00:21:00.801 --> 00:21:02.143 yeah which is really 00:21:02.182 --> 00:21:03.002 interesting like learning 00:21:03.022 --> 00:21:03.864 from their environment and 00:21:03.903 --> 00:21:04.704 taking existing 00:21:05.694 --> 00:21:06.675 know information it's kind 00:21:06.695 --> 00:21:07.895 of gathering over time and 00:21:07.915 --> 00:21:09.217 then using that to actually 00:21:09.257 --> 00:21:11.917 train itself on real niched 00:21:11.978 --> 00:21:13.198 in I guess specifics like 00:21:13.239 --> 00:21:14.259 healthcare where you've got 00:21:14.558 --> 00:21:16.019 absolutely yeah so it does 00:21:16.119 --> 00:21:16.940 just kind of make a lot of 00:21:16.960 --> 00:21:17.780 sense that will be when you 00:21:17.820 --> 00:21:19.560 move away from foundational 00:21:19.601 --> 00:21:21.102 models that are a large 00:21:21.162 --> 00:21:22.542 language model based 00:21:23.222 --> 00:21:25.707 mainly yeah yeah and moved 00:21:25.866 --> 00:21:27.548 towards deep learning 00:21:27.608 --> 00:21:29.672 systems the neural net deep 00:21:29.711 --> 00:21:31.074 learning systems that have 00:21:31.094 --> 00:21:32.095 been trained on very 00:21:32.134 --> 00:21:34.558 specific tasks then those 00:21:34.598 --> 00:21:36.882 become very much more um 00:21:39.218 --> 00:21:40.898 capable when they're trained 00:21:40.979 --> 00:21:42.138 on reinforcement learning 00:21:42.179 --> 00:21:43.279 but I think it's coming 00:21:43.299 --> 00:21:44.359 full circle and actually 00:21:44.400 --> 00:21:47.320 the llms now need to well 00:21:47.340 --> 00:21:48.141 they're making large 00:21:48.201 --> 00:21:49.501 language models and the 00:21:49.521 --> 00:21:51.303 deep seek stuff from china 00:21:51.343 --> 00:21:53.063 sort of broke the thing 00:21:53.103 --> 00:21:54.104 apart and said actually 00:21:54.124 --> 00:21:55.044 they need to learn on their 00:21:55.644 --> 00:21:56.904 own and be taught by other 00:21:56.984 --> 00:21:58.765 ais but yeah artificial 00:21:58.805 --> 00:22:00.165 general intelligence as they uh 00:22:01.207 --> 00:22:04.048 on the way agi yeah I mean 00:22:04.067 --> 00:22:06.147 that was a bit gentle so my 00:22:06.208 --> 00:22:08.348 thing I know we're trying 00:22:08.388 --> 00:22:09.788 to do these every week and 00:22:09.828 --> 00:22:10.788 we'd like to trying to get 00:22:10.848 --> 00:22:11.910 it down to thirty minutes 00:22:12.349 --> 00:22:13.170 on the top news but it 00:22:13.210 --> 00:22:14.809 needs to be very very 00:22:14.890 --> 00:22:19.730 strict um was the oracle 00:22:20.411 --> 00:22:22.451 breach um I know you and I 00:22:22.471 --> 00:22:23.912 spoke about it before but 00:22:24.211 --> 00:22:25.353 for the people basically 00:22:25.373 --> 00:22:26.972 they denied having a breach 00:22:28.076 --> 00:22:31.900 um said no nonsense no one's 00:22:31.940 --> 00:22:35.443 breached oracle um then I 00:22:35.483 --> 00:22:38.527 think then the hackers then 00:22:38.767 --> 00:22:39.827 I think put something up on 00:22:39.847 --> 00:22:41.309 their website or took put 00:22:41.329 --> 00:22:42.711 some files somewhere or did 00:22:42.770 --> 00:22:43.672 something on the server 00:22:43.692 --> 00:22:44.432 that they actually breached 00:22:44.452 --> 00:22:45.153 to prove that he actually 00:22:45.173 --> 00:22:47.355 had breached it I think then 00:22:48.220 --> 00:22:49.842 was then another breach on a 00:22:49.961 --> 00:22:51.784 system that was like really 00:22:51.943 --> 00:22:54.826 old and then they were then 00:22:54.866 --> 00:22:57.048 admitted yes actually but 00:22:57.068 --> 00:22:58.029 it was a very old thing not 00:22:58.049 --> 00:22:59.031 to do with our cloud even 00:22:59.051 --> 00:22:59.852 though it kind of was their 00:22:59.892 --> 00:23:01.753 cloud um but some old part 00:23:01.773 --> 00:23:03.135 of their cloud with some 00:23:03.215 --> 00:23:05.157 unpatched servers um that 00:23:05.178 --> 00:23:06.338 were running some like old 00:23:06.479 --> 00:23:07.800 software middleware software 00:23:10.099 --> 00:23:11.881 and then said that it wasn't recent. 00:23:12.461 --> 00:23:13.281 And so it's not the data. 00:23:13.321 --> 00:23:13.682 It's fine. 00:23:13.701 --> 00:23:14.162 Don't worry about it. 00:23:14.182 --> 00:23:14.842 It's not recent. 00:23:14.862 --> 00:23:15.821 It's not recent data. 00:23:15.842 --> 00:23:16.903 It's a really old server, 00:23:17.323 --> 00:23:18.462 even though the data in it 00:23:18.482 --> 00:23:19.804 was from twenty twenty four 00:23:20.203 --> 00:23:20.943 and had passwords from 00:23:20.983 --> 00:23:21.423 twenty twenty four. 00:23:21.443 --> 00:23:24.164 But that aside, it's old, like, you know. 00:23:24.885 --> 00:23:27.006 So anyway, so that's kind of happened. 00:23:27.026 --> 00:23:27.727 So they got hacked. 00:23:27.906 --> 00:23:29.067 But they dealt with the news 00:23:29.326 --> 00:23:30.028 and it was fine. 00:23:30.667 --> 00:23:31.488 They dealt with the news. 00:23:31.708 --> 00:23:33.429 It's not really got much airtime. 00:23:33.449 --> 00:23:34.909 This is what I find strange, 00:23:35.528 --> 00:23:36.730 given the fact that 00:23:37.739 --> 00:23:38.638 I kind of feel like if that 00:23:38.659 --> 00:23:40.579 had happened to Microsoft or Amazon, 00:23:41.279 --> 00:23:43.079 whether it was an old server or not, 00:23:43.180 --> 00:23:45.980 or some old middleware service or not, 00:23:46.800 --> 00:23:47.540 I really feel they would 00:23:47.560 --> 00:23:48.661 have been given a much 00:23:48.701 --> 00:23:50.241 harder time in the media. 00:23:50.481 --> 00:23:52.721 Whereas for some reason, I don't know why, 00:23:52.761 --> 00:23:54.762 but it's not. 00:23:55.522 --> 00:23:56.343 There's a second. 00:23:56.482 --> 00:24:00.503 I think they leveraged the same breach. 00:24:00.824 --> 00:24:01.784 I think it was the same 00:24:01.824 --> 00:24:03.085 backdoor or way in through 00:24:03.105 --> 00:24:03.924 some like Java. 00:24:04.897 --> 00:24:08.865 Java web kind of frontend, 00:24:08.904 --> 00:24:09.486 something that probably 00:24:09.506 --> 00:24:11.147 didn't get patched, a middleware exploit, 00:24:11.209 --> 00:24:12.349 I think, on some Java app. 00:24:13.063 --> 00:24:15.845 But anyway, that's mine. 00:24:15.984 --> 00:24:16.805 Over to you. 00:24:16.884 --> 00:24:20.027 You were saying, I don't know, I heard, 00:24:20.307 --> 00:24:21.407 I think I did hear about this. 00:24:21.928 --> 00:24:23.269 We discussed it a few days 00:24:23.308 --> 00:24:26.290 ago and I think it's to be, 00:24:26.931 --> 00:24:27.932 we might have discussed it 00:24:27.951 --> 00:24:30.133 last week actually in one 00:24:30.173 --> 00:24:31.473 of the attempted podcasts 00:24:31.493 --> 00:24:33.315 that one of the AIs organised for us. 00:24:34.816 --> 00:24:38.718 But there was a specific 00:24:38.938 --> 00:24:41.338 hacker that had actually tried to 00:24:43.571 --> 00:24:47.994 sell zero-day exploits on 00:24:49.336 --> 00:24:50.396 the back of this. 00:24:50.416 --> 00:24:51.877 They tried to extort Oracle, 00:24:51.917 --> 00:24:52.577 and they're trying a few 00:24:52.597 --> 00:24:53.078 different things. 00:24:53.479 --> 00:24:56.461 And then Oracle totally shut 00:24:56.500 --> 00:24:58.701 it down and said it's an old system. 00:24:59.282 --> 00:25:02.084 And then they started leaking data. 00:25:03.525 --> 00:25:04.586 And then some other 00:25:04.986 --> 00:25:07.127 customers came forward and said, actually, 00:25:07.167 --> 00:25:08.628 that looks legitimate data. 00:25:09.588 --> 00:25:15.556 So who knows? 00:25:15.576 --> 00:25:20.000 I think my take on it is, you know, 00:25:20.741 --> 00:25:23.484 like security is always hard. 00:25:24.586 --> 00:25:25.727 And yes, 00:25:26.407 --> 00:25:28.109 all the news should be massive 00:25:28.150 --> 00:25:29.191 when it's a big vendor. 00:25:30.449 --> 00:25:30.829 Yeah, I mean, 00:25:30.970 --> 00:25:31.911 when you're providing 00:25:32.030 --> 00:25:33.291 services around those 00:25:33.332 --> 00:25:35.272 things that you're just, I'm not, 00:25:35.452 --> 00:25:36.815 you know, we've only got, 00:25:37.355 --> 00:25:38.695 I need to go to your clean 00:25:38.756 --> 00:25:39.317 act because I want to know 00:25:39.336 --> 00:25:39.896 what that's about. 00:25:40.037 --> 00:25:40.938 But just to wrap up, 00:25:41.337 --> 00:25:42.538 if you're trusting a 00:25:42.598 --> 00:25:44.240 service provider to provide 00:25:44.280 --> 00:25:45.662 a service of which those 00:25:45.721 --> 00:25:47.563 services are cloud security 00:25:47.623 --> 00:25:48.604 and you get hacked, 00:25:49.003 --> 00:25:50.365 there is a sense of irony 00:25:50.986 --> 00:25:52.247 behind the fact that you're 00:25:52.326 --> 00:25:54.489 professing to do cloud security well, 00:25:54.509 --> 00:25:56.269 but you're not patching your own stuff. 00:25:56.891 --> 00:25:58.951 So, you know, there's a bit of a, 00:26:00.451 --> 00:26:01.991 strange. 00:26:02.011 --> 00:26:03.873 Yeah, it doesn't bode well, right, 00:26:03.913 --> 00:26:06.713 for the external PR view of 00:26:06.815 --> 00:26:07.714 your organisation. 00:26:07.734 --> 00:26:08.615 Do you think the reason we 00:26:08.654 --> 00:26:09.776 don't hear about it from 00:26:09.836 --> 00:26:11.977 the bigger vendors is that 00:26:12.017 --> 00:26:14.278 they have got their act in order? 00:26:14.458 --> 00:26:15.558 I seem to remember a long 00:26:15.618 --> 00:26:16.880 time ago thinking 00:26:17.579 --> 00:26:17.799 you know, 00:26:17.839 --> 00:26:20.722 the surface area of security in 00:26:20.903 --> 00:26:24.007 an organization is crazily 00:26:24.047 --> 00:26:25.288 hard to keep on top of. 00:26:25.428 --> 00:26:26.609 And the possibilities of 00:26:26.990 --> 00:26:29.653 like random security 00:26:29.673 --> 00:26:30.933 vulnerabilities being 00:26:30.973 --> 00:26:32.476 present in production 00:26:32.496 --> 00:26:33.936 systems is pretty high. 00:26:34.478 --> 00:26:36.420 But if you use the best that 00:26:36.460 --> 00:26:38.382 cloud has to offer, then you're 00:26:38.541 --> 00:26:40.284 probably going to be maybe a 00:26:40.364 --> 00:26:42.946 little bit lower than a rag 00:26:42.967 --> 00:26:44.308 bag sort of data center. 00:26:45.309 --> 00:26:47.752 There's ways of looking at lenses. 00:26:48.272 --> 00:26:49.654 Maybe we can keep track and 00:26:49.694 --> 00:26:52.117 talk about those types of 00:26:52.178 --> 00:26:53.239 security patterns in the 00:26:53.259 --> 00:26:54.480 cloud and what you need to do and 00:26:55.474 --> 00:26:56.516 The shared responsibility 00:26:56.536 --> 00:26:57.536 model between the customer 00:26:57.556 --> 00:26:58.517 and the cloud is obviously 00:26:58.557 --> 00:27:00.356 a bit of a bizarre spectrum. 00:27:00.617 --> 00:27:01.698 It depends on what it is, 00:27:01.738 --> 00:27:03.439 what you're using, how SaaS, how PaaS, 00:27:03.459 --> 00:27:05.240 how IaaS, you know, 00:27:05.319 --> 00:27:06.701 it's such a broad area. 00:27:08.461 --> 00:27:09.622 So yeah, that's a challenge yourself. 00:27:09.642 --> 00:27:11.563 But anyway, the Clean Cloud Act. 00:27:12.994 --> 00:27:13.315 OK, 00:27:13.375 --> 00:27:14.777 we can put the security down because 00:27:14.797 --> 00:27:15.817 it's all fine. 00:27:16.018 --> 00:27:17.179 There is nothing to worry about. 00:27:17.219 --> 00:27:18.961 It is all fine. 00:27:18.980 --> 00:27:20.622 So another thing that's all fine. 00:27:21.363 --> 00:27:24.846 So there's a bill going through, 00:27:25.847 --> 00:27:28.349 being proposed at the Senate in the US, 00:27:28.430 --> 00:27:30.592 Clean Cloud Act of twenty 00:27:30.612 --> 00:27:34.155 twenty five by US senators 00:27:34.195 --> 00:27:35.237 Whitehouse and Fetterman. 00:27:39.287 --> 00:27:41.088 It's very early, Bill. 00:27:41.829 --> 00:27:42.430 So, you know, 00:27:42.470 --> 00:27:44.112 when digging it up as a new source, 00:27:44.192 --> 00:27:46.534 it actually seems like, you know, 00:27:46.713 --> 00:27:49.057 when it's reported from multiple angles, 00:27:49.116 --> 00:27:53.601 it's like, okay, it should be big news. 00:27:53.681 --> 00:27:56.423 Four percent of all U.S. 00:27:56.463 --> 00:27:58.546 power goes to data centers, apparently. 00:27:59.477 --> 00:28:00.356 You know, 00:28:00.376 --> 00:28:03.018 there's probably marching error 00:28:03.038 --> 00:28:03.999 bars and that's probably 00:28:04.038 --> 00:28:06.038 the upper bar of the error bar. 00:28:06.459 --> 00:28:08.440 But, you know, within a percent or so, 00:28:09.220 --> 00:28:10.221 that's a big number. 00:28:10.681 --> 00:28:11.701 And it's projected to get to 00:28:11.881 --> 00:28:14.922 twelve percent in the next, 00:28:15.422 --> 00:28:17.262 by twenty twenty eight. 00:28:17.863 --> 00:28:19.344 This is only three years away. 00:28:19.364 --> 00:28:20.384 But this is on the 00:28:20.423 --> 00:28:22.825 assumption that they're sharing, 00:28:22.865 --> 00:28:24.565 obviously, the energy, you know, 00:28:24.625 --> 00:28:25.905 suppliers and they're not 00:28:25.925 --> 00:28:26.726 going to go and build 00:28:27.587 --> 00:28:29.008 their own renewable energy 00:28:29.087 --> 00:28:30.368 or something of their own, 00:28:30.409 --> 00:28:31.348 that they own their own 00:28:31.368 --> 00:28:32.749 data centres and don't rely 00:28:32.808 --> 00:28:33.709 on the grid at all? 00:28:34.489 --> 00:28:35.869 Well, I don't know. 00:28:35.910 --> 00:28:37.430 The bill specifically talks 00:28:37.450 --> 00:28:42.431 about crypto as well as AI. 00:28:42.590 --> 00:28:45.092 And it's brought forward by Democrats. 00:28:46.271 --> 00:28:47.731 And obviously the current 00:28:49.051 --> 00:28:51.813 powers that be may 00:28:52.671 --> 00:28:55.192 not think favourably to 00:28:55.313 --> 00:28:58.154 progressive measures in this space. 00:28:58.194 --> 00:28:59.957 But I think it is, you know, 00:28:59.977 --> 00:29:01.076 there's been a lot of news 00:29:01.518 --> 00:29:04.519 when OpenAI and all the big tech firms, 00:29:04.920 --> 00:29:08.643 including Oracle, actually, and Microsoft, 00:29:08.782 --> 00:29:11.586 all got behind the Trump 00:29:11.625 --> 00:29:13.366 administration with, you know, 00:29:13.686 --> 00:29:15.528 what was it, Project Stargate, 00:29:15.548 --> 00:29:16.028 I think it is, 00:29:17.320 --> 00:29:19.481 to make the biggest AI data 00:29:19.522 --> 00:29:21.364 centres the world has ever seen. 00:29:21.443 --> 00:29:23.164 It's going to be the best AI. 00:29:23.464 --> 00:29:28.127 It's going to be the best AI. 00:29:28.147 --> 00:29:29.108 The greatest. 00:29:29.128 --> 00:29:30.890 The greatest data centre the 00:29:30.910 --> 00:29:31.931 world has ever seen. 00:29:32.211 --> 00:29:34.772 And they're all announcing, 00:29:35.113 --> 00:29:36.354 or they were all announcing 00:29:36.794 --> 00:29:38.914 use of nuclear, use of coal, 00:29:39.415 --> 00:29:46.099 like some bits of gas and let's just burn, 00:29:46.160 --> 00:29:46.721 baby burn. 00:29:47.320 --> 00:29:49.502 And it seems like this 00:29:49.923 --> 00:29:51.664 little whisper on the side 00:29:51.984 --> 00:29:55.248 of a bit of sanity to say, actually, 00:29:55.828 --> 00:29:56.068 you know, 00:29:56.490 --> 00:29:57.570 we should use this as an 00:29:57.611 --> 00:30:00.012 opportunity to build green, you know, 00:30:00.053 --> 00:30:00.894 to have on site. 00:30:01.054 --> 00:30:02.214 I mean, you know, 00:30:02.255 --> 00:30:04.257 organizations that make energy. 00:30:05.838 --> 00:30:07.240 as well as maybe cars, 00:30:08.259 --> 00:30:09.780 have like dominated in this 00:30:09.820 --> 00:30:11.421 space and stand to clean up. 00:30:11.861 --> 00:30:12.862 Yet, you know, 00:30:13.241 --> 00:30:15.962 the use of AI and the use of 00:30:16.022 --> 00:30:19.284 dirty energy in being the 00:30:19.304 --> 00:30:21.505 first mover with the 00:30:21.525 --> 00:30:25.866 biggest data centres is a contradiction. 00:30:26.207 --> 00:30:28.768 I've got the impression they 00:30:28.788 --> 00:30:30.608 were investigating, as in like, 00:30:30.628 --> 00:30:31.308 was it Google? 00:30:31.989 --> 00:30:33.589 was investigating nuclear 00:30:34.490 --> 00:30:36.452 plants for their data centres, as in, 00:30:36.492 --> 00:30:37.053 like, 00:30:37.073 --> 00:30:38.653 basically standing up their own 00:30:38.913 --> 00:30:40.675 nuclear plant for energy, 00:30:41.395 --> 00:30:42.596 which... I think Microsoft 00:30:42.636 --> 00:30:43.637 had an announcement that 00:30:43.657 --> 00:30:45.799 they were buying Long 00:30:46.000 --> 00:30:48.561 Island or... Ice teas. 00:30:48.922 --> 00:30:49.903 Marl Long Island. 00:30:49.982 --> 00:30:52.345 Long Island ice teas for all their staff. 00:30:52.664 --> 00:30:54.246 Yeah, Long Island ice teas for everyone. 00:30:54.506 --> 00:30:56.647 Oh, God, it's not long. 00:30:56.708 --> 00:30:56.968 Well, 00:30:57.308 --> 00:30:59.609 we know of a place called... What's 00:30:59.650 --> 00:30:59.769 the... 00:31:03.786 --> 00:31:05.387 The nuclear facility in the 00:31:05.428 --> 00:31:07.790 States that had an accident. 00:31:08.612 --> 00:31:08.951 Anyway, 00:31:09.152 --> 00:31:15.759 there's a legacy nuclear that we're 00:31:15.779 --> 00:31:18.502 going to buy and referable and bring on. 00:31:19.104 --> 00:31:20.346 Because I think you're not 00:31:20.425 --> 00:31:21.666 subject to the same 00:31:21.747 --> 00:31:23.588 scrutiny when it's privately done. 00:31:24.563 --> 00:31:26.564 And you could do it for, you know, 00:31:26.584 --> 00:31:28.345 your own private energy consumption. 00:31:28.424 --> 00:31:29.865 I think it's slightly different. 00:31:29.885 --> 00:31:30.346 For some reason, 00:31:30.365 --> 00:31:30.905 that's why they've been 00:31:30.965 --> 00:31:32.767 able to get it passed 00:31:32.787 --> 00:31:34.268 through legislation quicker 00:31:34.528 --> 00:31:35.648 because it's not, you know, 00:31:35.669 --> 00:31:36.749 it's a company deciding, 00:31:36.789 --> 00:31:37.890 not a state deciding. 00:31:39.550 --> 00:31:41.031 So, yeah. 00:31:41.112 --> 00:31:43.653 Yeah, I think, you know, 00:31:43.712 --> 00:31:45.773 the... But what is the act? 00:31:45.794 --> 00:31:48.655 The act is... The whimper, the act, 00:31:49.756 --> 00:31:52.198 is to bring in... 00:31:55.815 --> 00:31:59.516 to ensure that emissions are 00:32:00.316 --> 00:32:03.436 a hundred percent generated 00:32:04.657 --> 00:32:07.398 renewably by twenty thirty 00:32:07.439 --> 00:32:09.939 five by amending the Clean 00:32:10.339 --> 00:32:11.660 Air Act specifically. 00:32:12.220 --> 00:32:13.901 to make sure that data 00:32:13.941 --> 00:32:15.602 centers specifically have 00:32:15.642 --> 00:32:18.784 to be green um which kind 00:32:18.804 --> 00:32:19.663 of makes sense if you're 00:32:19.683 --> 00:32:22.306 gonna crypto crypto is 00:32:22.365 --> 00:32:23.605 probably where the bill was 00:32:23.665 --> 00:32:24.646 born and it's probably been 00:32:24.686 --> 00:32:27.008 bumbling along for a decade 00:32:27.028 --> 00:32:29.269 and now crypto is not the 00:32:29.309 --> 00:32:30.630 thing that we're all owed 00:32:30.650 --> 00:32:31.830 about although it is a 00:32:31.851 --> 00:32:32.851 thing and it does use a 00:32:32.911 --> 00:32:36.272 hell of a lot of energy um 00:32:36.373 --> 00:32:37.374 and if it was all 00:32:38.000 --> 00:32:39.642 uh green energy then it 00:32:39.662 --> 00:32:40.823 would be like okay as long 00:32:40.843 --> 00:32:42.644 as it's not taking all the 00:32:42.663 --> 00:32:43.704 capacity away from all the 00:32:43.765 --> 00:32:45.205 other uses of energy then 00:32:45.645 --> 00:32:46.926 yeah fair enough or if it 00:32:46.946 --> 00:32:48.488 was local to the compute 00:32:48.528 --> 00:32:50.288 then that's good but you 00:32:50.348 --> 00:32:51.950 know with ai it's just 00:32:52.671 --> 00:32:55.192 accelerating the demand and 00:32:55.551 --> 00:32:56.752 anything that could be 00:32:57.294 --> 00:32:59.194 actually strongly put 00:32:59.234 --> 00:33:00.655 forward and adhered to 00:33:00.675 --> 00:33:02.297 would would be good and I 00:33:02.317 --> 00:33:04.137 would say you know um 00:33:06.371 --> 00:33:07.070 Companies, 00:33:07.830 --> 00:33:09.791 organisations and governments 00:33:09.832 --> 00:33:11.412 that get behind this stand 00:33:11.432 --> 00:33:13.952 to win because guess what? 00:33:14.173 --> 00:33:15.772 Things that aren't renewable run out. 00:33:18.773 --> 00:33:20.634 You could decide to be ahead 00:33:20.673 --> 00:33:21.835 and use this as a lever. 00:33:22.434 --> 00:33:24.795 When the Second World War broke out, 00:33:24.835 --> 00:33:25.756 we didn't say, oh, well, 00:33:26.096 --> 00:33:27.096 haven't got enough factories, 00:33:27.596 --> 00:33:28.556 we're going to lose the war. 00:33:28.855 --> 00:33:29.277 No, 00:33:29.297 --> 00:33:31.817 we built factories and we got extra 00:33:31.836 --> 00:33:32.936 workforce and all the women 00:33:32.977 --> 00:33:34.038 went to work and we made 00:33:34.077 --> 00:33:34.877 all the weapons and we... 00:33:36.038 --> 00:33:37.298 and we run the war you know 00:33:37.740 --> 00:33:39.421 it's like an urgency thing 00:33:39.540 --> 00:33:41.222 I think my my news article 00:33:41.262 --> 00:33:42.903 really the summary is it's 00:33:43.403 --> 00:33:45.145 it's a it's a little bill 00:33:45.546 --> 00:33:47.106 it's a little whimper and 00:33:47.146 --> 00:33:48.608 it's by the wrong side of 00:33:48.628 --> 00:33:50.369 uh you know the majority in 00:33:50.410 --> 00:33:51.310 the state so it will 00:33:51.330 --> 00:33:52.330 probably go completely 00:33:52.392 --> 00:33:53.511 unnoticed but it is 00:33:53.571 --> 00:33:55.794 probably well I find it uh 00:33:56.255 --> 00:33:57.075 an interesting and 00:33:57.134 --> 00:33:58.236 important yeah it's an 00:33:58.276 --> 00:34:00.117 important aspect isn't it really but 00:34:01.092 --> 00:34:01.612 Cool. 00:34:01.672 --> 00:34:05.153 So we don't really have time to discuss, 00:34:05.192 --> 00:34:05.853 because we did so, 00:34:05.873 --> 00:34:07.333 we've got to keep it to thirty minutes. 00:34:07.873 --> 00:34:09.034 So we're trying our best. 00:34:09.554 --> 00:34:10.454 So I guess I'll just have to 00:34:10.514 --> 00:34:14.675 leave the old AI hub in 00:34:14.775 --> 00:34:16.135 Saudi Arabia and how you 00:34:16.175 --> 00:34:17.755 can basically put things in 00:34:17.795 --> 00:34:18.615 their data center, 00:34:18.675 --> 00:34:20.115 but under your own jurisdiction, 00:34:20.155 --> 00:34:21.835 apparently, under your own laws, 00:34:22.356 --> 00:34:22.817 even though it's 00:34:22.836 --> 00:34:23.896 technically kind of there. 00:34:25.358 --> 00:34:27.621 And they'd have no backdoor to that. 00:34:27.820 --> 00:34:29.121 No sovereignty at all on 00:34:29.141 --> 00:34:30.963 their own land for those purposes. 00:34:31.605 --> 00:34:31.925 Exactly. 00:34:31.945 --> 00:34:34.527 A bit like storage unit facilities. 00:34:34.686 --> 00:34:37.088 You're just using them to store your stuff, 00:34:38.010 --> 00:34:39.411 but only you've got access. 00:34:40.693 --> 00:34:42.074 I think that's the view. 00:34:42.134 --> 00:34:43.474 It's yours and it's under 00:34:43.514 --> 00:34:44.496 your jurisdiction. 00:34:44.936 --> 00:34:45.976 That's how the world works. 00:34:46.338 --> 00:34:47.918 We're not cynical on Cloud Unplugged, 00:34:47.938 --> 00:34:48.179 though. 00:34:48.800 --> 00:34:50.021 Not cynical on Cloud Unplugged. 00:34:50.260 --> 00:34:51.362 Get your data into Saudi Arabia. 00:34:51.382 --> 00:34:52.163 No security issues. 00:34:52.659 --> 00:34:53.599 Get it over there. 00:34:53.719 --> 00:34:54.760 Don't worry about it. 00:34:55.380 --> 00:34:56.702 Just set up your own laws. 00:34:56.862 --> 00:34:57.643 It'll be fine. 00:34:58.083 --> 00:34:59.664 Tell them these are our laws. 00:34:59.684 --> 00:35:00.885 This is what will govern buying. 00:35:01.146 --> 00:35:02.226 Ship your data over there 00:35:02.467 --> 00:35:03.108 and it'll be fine. 00:35:03.148 --> 00:35:03.648 Don't worry. 00:35:05.510 --> 00:35:07.351 Anyway, that was a good episode. 00:35:07.391 --> 00:35:08.251 A few things to pull up on. 00:35:08.291 --> 00:35:09.492 Hopefully next one, 00:35:09.793 --> 00:35:10.293 I will have more 00:35:10.313 --> 00:35:13.315 information about the agent 00:35:13.356 --> 00:35:15.197 space where I've actually 00:35:15.217 --> 00:35:16.159 managed to get it to work 00:35:16.239 --> 00:35:16.958 and keep you posted. 00:35:18.820 --> 00:35:21.063 And we shall see you next week. 00:35:22.898 --> 00:35:23.643 Adios. 00:35:24.208 --> 00:35:24.972 Adios, indeed.