Sunday random odds and ends

• Go to the Everything Seminar blog to learn about the first beam splash at the LHC in 2009. This is exciting progress and means that the detectors really are doing their job and the machines are almost ready to start collisions.

• Cineuropa is in full swing and so far I've seen:
• Tokyo Sonata, which was enjoyable though had at least half an hour which felt completely farcicle. Generally a worthwhile watch and I'll be keeping an eye out for more films by director Kiyoshi Kurosawa.
• Tulpan, a film from Kazakh director Sergei Dvortsevoy. The cinematography is stunning, set on the steppes of Kazhakstan and centering around the seemingly simple story of a nomadic family and the tensions between traditional and modern life. It's sometimes hard to know how concrete the storyline was pre-filming as there are a number of sublime coincidences which seem impossible to manufacture in the absence of huge special effects budgets.
• Tonight a group of us are off to see Wong Kar Wai's Dung Che Sai Duk, seemingly a kung fu movie with a difference. Will give a one-line report if it seems appropriate...

• And on an asian theme, my kimchi supplies are back up to maximum after an epic kimchi making afternoon yesterday. I have a feeling it may even stand up to the assault of a large group of Koreans who will be descending on my place for dinner on Thursday night.

• I've almost finished Douglas Hofstadter's book: Fluid concepts and creative analogies: Computer models of the fundamental mechanisms of thought. This discusses Hofstadter's own projects in the direction of creating something akin to AI, and reading this really makes all other approaches I've come across seem so hugely lacking - although there are many projects which can solve much more complicated problems than those discussed in the book, the way they do so seems to answer none of the deep problems about how we think. The projects which he and his colleagues have tackled are simple to define problems in very restricted domains: anagram solving being perhaps the easiest to discuss. Clearly it is trivial to write a program which can solve anagrams. Brute force is really the forte of the artificial computer but this goes no way to showing how we think or even trying to imitate our thought processes. Instead Hofstadter's programs attack such problems with stochastic sub-processes which go in to attack the problem on a number of different levels, spotting groups, patterns, and finding the strengths and weaknesses in its own formulations as it goes. As the book discusses more and more such problems it really does appear that Hofstadter has managed to model to a high fidelity the fluid non-deterministic nature of our own methods for such problem solving. In some ways it's rather frustrating reading, coming from a generation where we see that the difficulty of solvable problems increases with Moore's law. We are used to being able to tackle more and more complicated challanges year on year, but the bottle-neck to the methods discussed here is not computer speed, but our own understanding of the processes by which we make analogues define concepts and spot solutions. This is another fantastic book from Hofstadter and a highly enjoyable read.
• I haven't had time to watch this video, but by the looks of things this is Hofstadter discussing the concepts which are dealt with in the book:

• Incidentally, if anybody has any knowledge of the views of Marvin Minsky on Douglas Hofstadter or vice versa I would very much like to know. These are two giants of the field, who frequently come over as feeling pessemistic of the current trends in the subject of AI and I'd love to have an idea of their criticisms of one another.

• I really, truly, genuinely have almost finished the two projects I've been banging on about for the last few months. When they're out I'll try and explain a little about why it's taken so long to give them the final push.