1
00:00:00,000 --> 00:00:01,240
Welcome back to The Deep Dive.

2
00:00:01,240 --> 00:00:04,080
Today we're tackling something huge

3
00:00:04,080 --> 00:00:06,320
and often totally invisible.

4
00:00:06,320 --> 00:00:07,320
Right.

5
00:00:07,320 --> 00:00:10,640
How do the world's biggest cultural institutions,

6
00:00:10,640 --> 00:00:14,560
we're talking major museums, huge archives, research

7
00:00:14,560 --> 00:00:17,480
universities, how do they manage these just

8
00:00:17,480 --> 00:00:18,960
colossal collections?

9
00:00:18,960 --> 00:00:21,280
And we're not talking about a few spreadsheets here.

10
00:00:21,280 --> 00:00:23,040
These collections have everything.

11
00:00:23,040 --> 00:00:26,640
Digitized texts, paintings, high res video, audio histories,

12
00:00:26,640 --> 00:00:29,480
and now even complex 3D scans of objects.

13
00:00:29,480 --> 00:00:30,880
That's an incredible mix of data.

14
00:00:30,880 --> 00:00:31,760
Exactly.

15
00:00:31,760 --> 00:00:33,680
And handling all that, keeping it safe,

16
00:00:33,680 --> 00:00:35,280
and making it accessible to people,

17
00:00:35,280 --> 00:00:38,080
that needs a really industrial strength solution.

18
00:00:38,080 --> 00:00:39,440
And that's our focus today.

19
00:00:39,440 --> 00:00:41,100
We're looking at the open source software

20
00:00:41,100 --> 00:00:42,680
that hundreds of these institutions

21
00:00:42,680 --> 00:00:45,940
trust to solve this exact problem, collective access.

22
00:00:45,940 --> 00:00:47,980
We're going to get into its core architecture, which

23
00:00:47,980 --> 00:00:51,240
has this powerful duality providence in Pawtucket 2,

24
00:00:51,240 --> 00:00:53,880
and we'll look especially at the big modernizing updates

25
00:00:53,880 --> 00:00:55,400
in the new version 2.0.

26
00:00:55,400 --> 00:00:57,640
So we've been unpacking the official documentation,

27
00:00:57,640 --> 00:00:59,680
deep diving into the GitHub repos,

28
00:00:59,680 --> 00:01:01,640
and really trying to understand the system.

29
00:01:01,640 --> 00:01:03,080
And our mission here is to give you

30
00:01:03,080 --> 00:01:06,200
a clear structural understanding of it.

31
00:01:06,200 --> 00:01:08,440
We want you to see past the pretty website

32
00:01:08,440 --> 00:01:12,120
and get how the engine behind modern archives actually works.

33
00:01:12,120 --> 00:01:15,040
OK, so let's start at the very beginning, the why.

34
00:01:15,040 --> 00:01:16,760
Collective access, or CA.

35
00:01:16,760 --> 00:01:18,360
It started back in 2003.

36
00:01:18,360 --> 00:01:20,520
Right, and it wasn't born to be a commercial product.

37
00:01:20,520 --> 00:01:24,600
It was a direct response to a massive gap in the market.

38
00:01:24,600 --> 00:01:25,480
What was the gap?

39
00:01:25,480 --> 00:01:29,000
The tools that existed were just prohibitively expensive.

40
00:01:29,000 --> 00:01:31,400
They were proprietary, they were rigid.

41
00:01:31,400 --> 00:01:33,800
Institutions had these really complex needs,

42
00:01:33,800 --> 00:01:35,600
but tight budgets.

43
00:01:35,600 --> 00:01:39,520
So collective access came in as a free, open source alternative.

44
00:01:39,520 --> 00:01:40,120
Exactly.

45
00:01:40,120 --> 00:01:42,360
It's under the GPL 3.0 license.

46
00:01:42,360 --> 00:01:45,320
That meant anyone, a small local history group

47
00:01:45,320 --> 00:01:48,520
or a huge national gallery, could just download it and use it.

48
00:01:48,520 --> 00:01:50,120
No massive licensing fees.

49
00:01:50,120 --> 00:01:51,720
And that accessibility is what let

50
00:01:51,720 --> 00:01:53,840
it grow into what it is today.

51
00:01:53,840 --> 00:01:56,120
Which brings us to, I think, the most fundamental thing

52
00:01:56,120 --> 00:01:57,400
you have to understand about it.

53
00:01:57,400 --> 00:01:58,400
The core duality.

54
00:01:58,400 --> 00:01:59,240
The duality, yeah.

55
00:01:59,240 --> 00:02:00,880
I like to think of it like a restaurant.

56
00:02:00,880 --> 00:02:03,800
You have the kitchen in the back, super organized complex

57
00:02:03,800 --> 00:02:05,320
where all the prep work happens.

58
00:02:05,320 --> 00:02:08,120
And then you have the dining room out front.

59
00:02:08,120 --> 00:02:10,640
Beautiful, simplified, where the guests actually

60
00:02:10,640 --> 00:02:12,520
interact with the final product.

61
00:02:12,520 --> 00:02:14,000
You need that separation.

62
00:02:14,000 --> 00:02:16,760
So the kitchen, the back end, where the hard work is done,

63
00:02:16,760 --> 00:02:17,920
that's called Providence.

64
00:02:17,920 --> 00:02:18,820
Yes.

65
00:02:18,820 --> 00:02:21,600
That's the data management and cataloging application.

66
00:02:21,600 --> 00:02:24,520
That's where the curators and archivists spend their days,

67
00:02:24,520 --> 00:02:27,680
meticulously documenting every single item.

68
00:02:27,680 --> 00:02:30,360
And the dining room, the front end, is Pawtucket 2.

69
00:02:30,360 --> 00:02:32,720
Pawtucket 2, that's the public interface.

70
00:02:32,720 --> 00:02:35,480
It's the website layer that takes all that organized data

71
00:02:35,480 --> 00:02:37,080
from Providence and makes it look good,

72
00:02:37,080 --> 00:02:40,160
makes it interactive for researchers, and well, for you.

73
00:02:40,160 --> 00:02:43,180
And separating them isn't just for convenience, is it?

74
00:02:43,180 --> 00:02:46,200
It's a really core architectural choice.

75
00:02:46,200 --> 00:02:48,200
Why is that so important for an archive?

76
00:02:48,200 --> 00:02:51,200
It's about sanity, really, and security.

77
00:02:51,200 --> 00:02:54,640
If you try to manage everything in one single application,

78
00:02:54,640 --> 00:02:58,120
you risk compromising your complex internal data standards

79
00:02:58,120 --> 00:02:59,720
just to make a slick website.

80
00:02:59,720 --> 00:03:00,720
Ah, I see.

81
00:03:00,720 --> 00:03:03,280
This duality separates the incredibly detailed

82
00:03:03,280 --> 00:03:07,180
internal organization from the user-friendly public discovery.

83
00:03:07,180 --> 00:03:09,760
It means a web designer can go and update Pawtucket 2

84
00:03:09,760 --> 00:03:13,920
without ever, ever risking the archival integrity of the data

85
00:03:13,920 --> 00:03:15,720
that's locked down in Providence.

86
00:03:15,720 --> 00:03:17,360
That makes perfect sense.

87
00:03:17,360 --> 00:03:19,460
So before we dive deeper into Providence,

88
00:03:19,460 --> 00:03:22,280
let's just take a moment to thank the supporter of this deep dive.

89
00:03:22,280 --> 00:03:23,040
Good idea.

90
00:03:23,040 --> 00:03:25,680
SafeServer handles the hosting of exactly this kind

91
00:03:25,680 --> 00:03:27,920
of complex, high-demand software,

92
00:03:27,920 --> 00:03:31,560
and they support institutions with their digital transformation.

93
00:03:31,560 --> 00:03:33,280
So if you're looking for reliable hosting

94
00:03:33,280 --> 00:03:36,000
for something powerful like collective access,

95
00:03:36,000 --> 00:03:40,280
you can find more info at www.safeserver.de.

96
00:03:40,280 --> 00:03:42,880
OK, let's go into that engine room.

97
00:03:42,880 --> 00:03:44,900
Providence, this is the app that's

98
00:03:44,900 --> 00:03:47,900
built to solve what I call the archivist's nightmare.

99
00:03:47,900 --> 00:03:49,960
Handling extreme data diversity.

100
00:03:49,960 --> 00:03:50,480
Exactly.

101
00:03:50,480 --> 00:03:50,800
Yeah.

102
00:03:50,800 --> 00:03:52,280
The number one feature of Providence

103
00:03:52,280 --> 00:03:54,080
is just how configurable it is.

104
00:03:54,080 --> 00:03:55,880
Unlike a lot of off-the-shelf software,

105
00:03:55,880 --> 00:03:57,920
it's not built for just one type of collection.

106
00:03:57,920 --> 00:04:00,960
It supports museums, archives, and research contexts

107
00:04:00,960 --> 00:04:01,840
all at once.

108
00:04:01,840 --> 00:04:04,120
When you say configurable, what does that actually

109
00:04:04,120 --> 00:04:06,240
mean for the person cataloging something?

110
00:04:06,240 --> 00:04:10,640
It means they are forced to sort of shoehorn their collection

111
00:04:10,640 --> 00:04:13,520
into someone else's idea of what a field should be.

112
00:04:13,520 --> 00:04:15,400
Providence lets you build catalogs

113
00:04:15,400 --> 00:04:18,600
that conform to recognized standards, like Dublin Core,

114
00:04:18,600 --> 00:04:21,360
or you can customize totally new fields.

115
00:04:21,360 --> 00:04:24,160
And you can do that without a programmer?

116
00:04:24,160 --> 00:04:25,840
Without writing custom code, yes.

117
00:04:25,840 --> 00:04:27,760
The system adapts to the institution,

118
00:04:27,760 --> 00:04:29,840
not the other way around, which is vital,

119
00:04:29,840 --> 00:04:31,560
because a history museum's needs are just

120
00:04:31,560 --> 00:04:33,920
wildly different from a natural science museum's.

121
00:04:33,920 --> 00:04:35,880
And the media handles.

122
00:04:35,880 --> 00:04:36,920
It's really impressive.

123
00:04:36,920 --> 00:04:40,280
I mean, documents, images, audio, video, that's expected.

124
00:04:40,280 --> 00:04:41,840
But it also handles 3D models.

125
00:04:41,840 --> 00:04:44,600
And that detail tells you how modern the system is.

126
00:04:44,600 --> 00:04:46,880
That's the key technical achievement, really.

127
00:04:46,880 --> 00:04:47,640
How so?

128
00:04:47,640 --> 00:04:51,800
Because managing a huge 3D scan of a fragile artifact, which

129
00:04:51,800 --> 00:04:53,760
is basically a cloud of data points,

130
00:04:53,760 --> 00:04:57,520
the same system as a simple PDF of a letter from 1910,

131
00:04:57,520 --> 00:05:00,400
that requires immense flexibility on the back end.

132
00:05:00,400 --> 00:05:02,360
It treats them both as collection items.

133
00:05:02,360 --> 00:05:04,160
So it's not just about the variety of data.

134
00:05:04,160 --> 00:05:05,960
It's also got to handle the day-to-day work

135
00:05:05,960 --> 00:05:07,680
of running an institution.

136
00:05:07,680 --> 00:05:09,840
What kind of workflow features are built in?

137
00:05:09,840 --> 00:05:11,440
It's all about bulk processing.

138
00:05:11,440 --> 00:05:14,440
An archive might acquire 10,000 records at once.

139
00:05:14,440 --> 00:05:17,280
Providence supports batch importing, vexporting.

140
00:05:17,280 --> 00:05:19,520
You can bring in massive data sets, clean them up,

141
00:05:19,520 --> 00:05:22,120
and then quickly output reports, like inventory lists

142
00:05:22,120 --> 00:05:23,960
as PDFs or spreadsheets.

143
00:05:23,960 --> 00:05:25,120
It's built to scale.

144
00:05:25,120 --> 00:05:29,360
So Providence is the highly granular professional tool

145
00:05:29,360 --> 00:05:30,520
for data integrity.

146
00:05:30,520 --> 00:05:31,160
Right.

147
00:05:31,160 --> 00:05:33,280
But the public never sees any of that.

148
00:05:33,280 --> 00:05:35,000
And all that effort is useless if people

149
00:05:35,000 --> 00:05:35,840
can't find the stuff.

150
00:05:35,840 --> 00:05:36,680
Absolutely.

151
00:05:36,680 --> 00:05:38,720
So now we shift from the structured back end

152
00:05:38,720 --> 00:05:41,300
to the elegant front end, Pawtucket II.

153
00:05:41,300 --> 00:05:43,040
This is where the collection comes alive.

154
00:05:43,040 --> 00:05:45,600
OK, so Pawtucket II takes that data

155
00:05:45,600 --> 00:05:47,820
and focuses on the user experience.

156
00:05:47,820 --> 00:05:50,520
You can style it to match a museum's brand.

157
00:05:50,520 --> 00:05:52,960
But the real power is in the discovery tools, right?

158
00:05:52,960 --> 00:05:53,460
That's right.

159
00:05:53,460 --> 00:05:55,820
It goes way beyond a simple keyword search.

160
00:05:55,820 --> 00:05:58,560
Pawtucket II lets you use customizable facets

161
00:05:58,560 --> 00:05:59,440
and filters.

162
00:05:59,440 --> 00:06:02,280
A facet is what for a non-technical user?

163
00:06:02,280 --> 00:06:04,160
It's basically a preset category.

164
00:06:04,160 --> 00:06:05,680
So instead of you having to search

165
00:06:05,680 --> 00:06:08,800
for blue dress Victorian, you can just

166
00:06:08,800 --> 00:06:14,000
filter by object type, dress, then era, Victorian,

167
00:06:14,000 --> 00:06:15,840
and then color blue.

168
00:06:15,840 --> 00:06:17,400
It lets you explore the collection,

169
00:06:17,400 --> 00:06:19,720
even if you don't know exactly what you're looking for.

170
00:06:19,720 --> 00:06:22,720
And it can turn static data into something interactive.

171
00:06:22,720 --> 00:06:25,040
I've seen examples with maps and timelines.

172
00:06:25,040 --> 00:06:26,420
How does it do that?

173
00:06:26,420 --> 00:06:29,320
It leverages the structured data from Providence.

174
00:06:29,320 --> 00:06:31,680
So if an archivist entered a date and a location

175
00:06:31,680 --> 00:06:35,280
for an object, Pawtucket II can visualize it.

176
00:06:35,280 --> 00:06:37,680
A timeline takes that static date field

177
00:06:37,680 --> 00:06:40,000
and turns it into a navigable journey.

178
00:06:40,000 --> 00:06:42,240
A map takes a GPS coordinate and shows you

179
00:06:42,240 --> 00:06:43,280
where an object is from.

180
00:06:43,280 --> 00:06:45,280
It completely changes the user's relationship

181
00:06:45,280 --> 00:06:45,960
with the collection.

182
00:06:45,960 --> 00:06:48,040
You're not just viewing, you're exploring.

183
00:06:48,040 --> 00:06:50,240
And there are also features for public engagement, things

184
00:06:50,240 --> 00:06:51,580
that let the community contribute.

185
00:06:51,580 --> 00:06:52,620
Precisely.

186
00:06:52,620 --> 00:06:55,520
The institution can decide to turn on public commenting

187
00:06:55,520 --> 00:06:57,940
or tagging or even rating items.

188
00:06:57,940 --> 00:06:59,920
For something like a local history archive,

189
00:06:59,920 --> 00:07:01,500
that's invaluable.

190
00:07:01,500 --> 00:07:04,360
You can have the community help identify people in old photos.

191
00:07:04,360 --> 00:07:06,680
It turns the public into research assistants.

192
00:07:06,680 --> 00:07:07,800
In a way, yes.

193
00:07:07,800 --> 00:07:09,500
And this is not theoretical software.

194
00:07:09,500 --> 00:07:12,280
It's running some really high-profile digital exhibits.

195
00:07:12,280 --> 00:07:13,120
Oh, absolutely.

196
00:07:13,120 --> 00:07:15,200
We're talking about projects like the Crest Digital

197
00:07:15,200 --> 00:07:17,920
Archive at the National Gallery of Art or the Chicago Film

198
00:07:17,920 --> 00:07:18,420
Archive.

199
00:07:18,420 --> 00:07:19,840
They're all using this framework.

200
00:07:19,840 --> 00:07:23,160
OK, so we've established the core duality, Providence,

201
00:07:23,160 --> 00:07:25,920
the rock solid engine, and Pawtucket II,

202
00:07:25,920 --> 00:07:28,160
the engaging public face.

203
00:07:28,160 --> 00:07:30,240
Now let's get to the new stuff.

204
00:07:30,240 --> 00:07:32,760
It was a long wait since version 1.7.

205
00:07:32,760 --> 00:07:34,120
A very long wait.

206
00:07:34,120 --> 00:07:36,360
But Collective Access version 2.0 brings

207
00:07:36,360 --> 00:07:38,600
some huge leaps forward.

208
00:07:38,600 --> 00:07:42,000
The focus seems to be on stability, connectivity,

209
00:07:42,000 --> 00:07:44,520
and this is the big one, AI integration.

210
00:07:44,520 --> 00:07:47,720
Yeah, version 2.0 is basically the developers saying, OK,

211
00:07:47,720 --> 00:07:49,440
we're ready for the next decade.

212
00:07:49,440 --> 00:07:51,960
First, they just future-proof the platform.

213
00:07:51,960 --> 00:07:54,040
It's now compatible with modern server tech,

214
00:07:54,040 --> 00:07:56,880
like PHP 8.2 and 8.3.

215
00:07:56,880 --> 00:07:59,320
Not a flashy feature, but essential for security

216
00:07:59,320 --> 00:08:00,160
and performance.

217
00:08:00,160 --> 00:08:00,920
Vital.

218
00:08:00,920 --> 00:08:02,960
But the real user benefits are somewhere else.

219
00:08:02,960 --> 00:08:05,520
One key improvement is in historical tracking.

220
00:08:05,520 --> 00:08:08,480
You mean provenance, knowing where an object has been.

221
00:08:08,480 --> 00:08:09,720
Exactly.

222
00:08:09,720 --> 00:08:12,900
V2.0 has a new, much more flexible system

223
00:08:12,900 --> 00:08:14,680
for tracking change over time.

224
00:08:14,680 --> 00:08:16,560
It's not just the creation date anymore.

225
00:08:16,560 --> 00:08:19,840
It's about granular history, location changes,

226
00:08:19,840 --> 00:08:21,520
ownership shifts, you name it.

227
00:08:21,520 --> 00:08:23,120
OK, so better accuracy.

228
00:08:23,120 --> 00:08:25,040
But the features that got everyone talking

229
00:08:25,040 --> 00:08:27,960
are about automation, the AI stuff.

230
00:08:27,960 --> 00:08:29,560
I was genuinely surprised when I read

231
00:08:29,560 --> 00:08:30,960
about the transcription feature.

232
00:08:30,960 --> 00:08:33,000
It's a massive leap in efficiency.

233
00:08:33,000 --> 00:08:36,520
V2.0 uses machine learning in two huge areas.

234
00:08:36,520 --> 00:08:38,760
First, automated translation.

235
00:08:38,760 --> 00:08:41,040
The system supports services like DeepL and Google

236
00:08:41,040 --> 00:08:43,200
Translate, so staff using Providence

237
00:08:43,200 --> 00:08:44,440
can work in their own language.

238
00:08:44,440 --> 00:08:46,240
But that immediately lowers the barrier

239
00:08:46,240 --> 00:08:47,840
to entry for global institutions.

240
00:08:47,840 --> 00:08:48,560
It does.

241
00:08:48,560 --> 00:08:50,300
But the truly transformative feature

242
00:08:50,300 --> 00:08:51,940
is the automated transcription.

243
00:08:51,940 --> 00:08:54,320
It uses models like OpenAI Whisper

244
00:08:54,320 --> 00:08:57,120
to automatically transcribe audio and video materials

245
00:08:57,120 --> 00:08:58,200
right inside the workflow.

246
00:08:58,200 --> 00:08:59,120
Just think about that.

247
00:08:59,120 --> 00:09:01,920
An archivist gets 50 years of oral history interviews,

248
00:09:01,920 --> 00:09:05,160
manually transcribing that could take months, maybe years.

249
00:09:05,160 --> 00:09:07,180
And this integration automates it.

250
00:09:07,180 --> 00:09:10,120
It turns all that spoken word into searchable text

251
00:09:10,120 --> 00:09:11,440
almost instantly.

252
00:09:11,440 --> 00:09:14,200
It redefines what's possible for searching media.

253
00:09:14,200 --> 00:09:16,360
So much of that content was basically locked away

254
00:09:16,360 --> 00:09:17,660
from keyword searches before.

255
00:09:17,660 --> 00:09:18,600
This unlocks it.

256
00:09:18,600 --> 00:09:19,640
Wow.

257
00:09:19,640 --> 00:09:21,840
So that's efficiency through AI.

258
00:09:21,840 --> 00:09:23,000
What about connectivity?

259
00:09:23,000 --> 00:09:26,080
How does v2.0 talk to other systems?

260
00:09:26,080 --> 00:09:27,280
They completely modernized it.

261
00:09:27,280 --> 00:09:31,000
They introduced a new GraphQL-based API.

262
00:09:31,000 --> 00:09:33,200
This is a huge deal for any developer trying

263
00:09:33,200 --> 00:09:36,480
to build custom tools that need to talk to the Providence data.

264
00:09:36,480 --> 00:09:38,320
For listeners who aren't developers,

265
00:09:38,320 --> 00:09:40,840
why is switching to GraphQL so important?

266
00:09:40,840 --> 00:09:42,600
Well, a traditional API often makes

267
00:09:42,600 --> 00:09:44,560
you download a big packet of data,

268
00:09:44,560 --> 00:09:46,920
even if you only need one tiny piece of it.

269
00:09:46,920 --> 00:09:48,720
GraphQL is about precision.

270
00:09:48,720 --> 00:09:51,520
A developer can ask for only the specific fields they need.

271
00:09:51,520 --> 00:09:53,480
So it's faster, less resource intensive.

272
00:09:53,480 --> 00:09:54,480
Much faster.

273
00:09:54,480 --> 00:09:56,720
For custom visualizations, mobile apps,

274
00:09:56,720 --> 00:09:59,240
any kind of integration, it's a major upgrade.

275
00:09:59,240 --> 00:10:01,600
And there's also a new support for external media, right?

276
00:10:01,600 --> 00:10:03,040
Yes, another efficiency thing.

277
00:10:03,040 --> 00:10:05,120
V2.0 can now reference content that

278
00:10:05,120 --> 00:10:08,920
lives somewhere else on YouTube, Vimeo, the internet archive.

279
00:10:08,920 --> 00:10:11,400
So the museum doesn't have to eat up all its server space

280
00:10:11,400 --> 00:10:13,120
storing a huge video file if it's already

281
00:10:13,120 --> 00:10:13,920
hosted somewhere else.

282
00:10:13,920 --> 00:10:15,520
Exactly, they just catalog the link.

283
00:10:15,520 --> 00:10:17,960
They keep the metadata integrity and providence,

284
00:10:17,960 --> 00:10:20,380
but save a fortune on storage costs.

285
00:10:20,380 --> 00:10:22,480
And finally, digital preservation.

286
00:10:22,480 --> 00:10:24,480
Long term integrity is everything.

287
00:10:24,480 --> 00:10:27,740
Right, and V2.0 has a new export system

288
00:10:27,740 --> 00:10:31,560
that uses something called configurable Bag IT packages.

289
00:10:31,560 --> 00:10:32,840
Bag IT packages.

290
00:10:32,840 --> 00:10:37,440
Think of them like standardized tamper-proof digital shipping

291
00:10:37,440 --> 00:10:38,400
containers.

292
00:10:38,400 --> 00:10:40,960
They bundle up all the data and the metadata

293
00:10:40,960 --> 00:10:44,360
in a way that can be validated for integrity checks years

294
00:10:44,360 --> 00:10:45,540
or decades down the line.

295
00:10:45,540 --> 00:10:46,760
It's the gold standard.

296
00:10:46,760 --> 00:10:48,040
So it's really clear.

297
00:10:48,040 --> 00:10:51,480
Version 2.0 is all about automation, next gen

298
00:10:51,480 --> 00:10:54,180
connectivity, and making these massive data

299
00:10:54,180 --> 00:10:56,320
sets manageable and secure.

300
00:10:56,320 --> 00:10:58,280
Absolutely, collective access is now, I think,

301
00:10:58,280 --> 00:11:01,680
firmly positioned as a cutting edge open source powerhouse.

302
00:11:01,680 --> 00:11:04,760
You have the surgical precision of Providence for management

303
00:11:04,760 --> 00:11:07,840
and the dynamic presentation of Pawtucket 2.0 for the public.

304
00:11:07,840 --> 00:11:10,540
These updates, especially the AI and the new API,

305
00:11:10,540 --> 00:11:13,040
they just solidify its role for the future.

306
00:11:13,040 --> 00:11:15,440
So the practical takeaway for you, listening,

307
00:11:15,440 --> 00:11:18,920
is that you now understand that two-part architecture.

308
00:11:18,920 --> 00:11:21,000
When you see a beautiful online museum

309
00:11:21,000 --> 00:11:23,480
collection with interactive maps and timelines,

310
00:11:23,480 --> 00:11:25,440
you know there's a good chance software like this

311
00:11:25,440 --> 00:11:27,400
has the engine behind it all, letting

312
00:11:27,400 --> 00:11:30,400
them present their world without compromising their data.

313
00:11:30,400 --> 00:11:32,960
But this does raise an important question, something

314
00:11:32,960 --> 00:11:36,800
for you to mull over as these tools become standard.

315
00:11:36,800 --> 00:11:39,360
Given the new support for automated transcription

316
00:11:39,360 --> 00:11:43,720
and translation, what are the ethical or contextual challenges

317
00:11:43,720 --> 00:11:44,880
that come up?

318
00:11:44,880 --> 00:11:47,640
When we use machine learning to interpret or tag

319
00:11:47,640 --> 00:11:49,940
historical records, can an algorithm

320
00:11:49,940 --> 00:11:54,640
truly capture the necessary historical nuance or context?

321
00:11:54,640 --> 00:11:57,160
Can a machine understand what's not being said in an interview?

322
00:11:57,160 --> 00:11:57,760
Exactly.

323
00:11:57,760 --> 00:11:59,240
That's something every institution

324
00:11:59,240 --> 00:12:02,360
using these amazing new tools is going to have to grapple with.

325
00:12:02,360 --> 00:12:05,160
A fascinating and really crucial point to end on.

326
00:12:05,160 --> 00:12:06,920
Thank you for joining us for this deep dive

327
00:12:06,920 --> 00:12:08,360
into collective access.

328
00:12:08,360 --> 00:12:11,240
And remember, this deep dive was supported by Safe Server.

329
00:12:11,240 --> 00:12:14,280
If you are looking for reliable hosting for complex software

330
00:12:14,280 --> 00:12:16,560
or support for your digital transformation projects,

331
00:12:16,560 --> 00:12:21,000
you can find more information at www.safeserver.de.

332
00:12:21,000 --> 00:12:23,480
We'll be back soon to unpack another stack of sources.

333
00:12:23,480 --> 00:12:24,740
Until then, keep digging.