This is my informal write-up of my notes from the British Computer Society's Lovelace lecture on March 13, 2007. They're very much written from my point of view; for a more objective version you can also view the talk slides at http://www.w3.org/2007/Talks/0313-bcs-tbl/, the BCS report on the event, video or check out this report with video and transcripts.
The lecture was introduced by Nigel Shadbolt (who is also involved in the Web Science Research Initiative).
Berners-Lee's talk was titled 'Looking Back, Looking forward' and sub-titled 'The process of designing things in a very large space'.
He talked about philosophical engineering, as physics used to be called, and introduced the concept of the Semantic Web as a philosophical space.
He introduced the ideas of microscopic rules or design and macroscopic behaviour; that rules are both social and technical and that designing a system involves social conventions as well as technical solutions.
Complexity is introduced between the micro and macro stages. Web science happens when the macro effect is analysed, and issues become clear. Values can be applied at this point, and a creative response to these issues results in another idea. The 'magic' happens at the points of collaboration/complexity and creativity. Magic can be defined as 'stuff you don't understand (yet)'. One slide was about the division between science and engineering in the process.
As an example of this system, email went from being a micro solution that suited a friendly community of academics to an unfriendly world and at that point we got spam. The issue now is how we deal with spam.
The idea that started the web was not being able to access information (there was lots of stuff around but it was all over the place). The technical solution was protocols (URI, HTTP, HTML), the social was incentives to link, e.g. personal collections of bookmarks, lists of URLs to answer FAQs. As the web exploded and went from micro to macro, the issue became not being able to find stuff. This leads to the issues we're dealing with now.
The talk then went onto the reasons the web worked (or the essentials for it to work):
- universality (across a range of factors, see slide)
There was discussion of Google, wikis and blogs. In the original web model, everybody could write. HTML links in other documents should be easy to make (e.g. click, click, save and publish) but it didn't happen at the time.
[The Google slide reminded me that web searches got less fun as Google's algorithms got better - there was much less randomness. I guess all the kinks were smoothed out.]
So the issue was that people couldn't write stuff - wikis were a technical solution, and the social solution was to throw out the permissions model so that everybody can write. The new issue is wiki 'battles'. The idea may be a wiki process, possibly leading to meritocratic systems.
With the Semantic Web, the issue was that web data couldn't be re-used as it was only exposed as HTML. In the model presented, the initial idea is data sharing. The social solution is to use URIs, make useful stuff and useful links, agree on ontologies and share them; pages with URIs link to other pages with URIs. The technical solution is to 'use URIs for documents and concepts', RDF, OWL, SPARQL, RIF (which I assume is Rule Interchange Format having googled it later) and the 'same ladder of authority'. The Semantic Web is 'data but also a web' at the micro level and becomes FOAF or life sciences at the macro level.
[Which led me to wonder, what issues will arise when we get to the macro level of the SW? And throughout the lecture, the question of ontologies kept worrying me - is relying on them realistic?]
URIs: in the Semantic Web, everything has a URI. Not just things - give terms a URI, e.g. don't just say 'blue' - give the URI of that colour blue (give domain [of knowledge]). This allows you to provide the definitive meaning of that term at your URI (and in that way gives ownership of that definition).
The next slides were about the Semantic Web model, including the dream of a unifying logic, passing proof (of identity) around and trusted systems and the SW as a language for explaining the use of Public (PGP) keys; and current Semantic Web work, including the Semantic Web Interest Group.
The lecture moved onto the shapes of data and how they have changed - from lines (tapes, cards) to matrix/tables/boxes (databases) to trees (SGML, XML, top-down structured design, OO) to webs/nets (the internet? www?).
Berners-Lee made the point that the web is not a spider's web - it has no centre.
[All of which made me wonder - is the requirement for ontologies making trees of nets?
How can the tree-like structure of ontologies mesh with nets or webs?]
The next slides were on the idea of applications connected by concepts and the fractal web of concepts - a 'tangle' across boundaries of scale, varying access level, local and global standards, and personal interactions on multiple scales. "The semantic web is about allowing data systems to change by evolution not revolution".
Berners-Lee also discussed the dream of politics and democracy in a civilised society.
TCO: Total Cost of Ontologies - it's a small overall cost. The lesson is: "do you your bit, others will do theirs".
The challenges of web science:
- user interface challenges (domain-specific vs generic)
- data policy challenges (e.g. identity, privacy, transparency)
- new devices (smarter and cheaper devices; developing countries).
That's where my notes end, except that I'd also noted that Berners-Lee said 'geeks get a kick out of the creative part of engineering', because it really resonated; and that he had designed the web on a NEXT machine and presented the slides in Safari on a Mac.