Thursday 7 September 2017

ECAL 2017, Thursday

Day four of the European Conference of Artificial Life, in Lyon, with a full day of presentations.

Again, the day started with an excellent keynote: Bill Sellers talking on Synthetic palaeontology: Reconstructing Ancient Life using Virtual Robots.  How did dinosaurs walk?  Could they run?  How can we tell when all we have is fossilised bones?  So, the idea is to build a dynamic model in the computer, which can be validated against modern animals like crocodiles and ostriches (although these do have fundamental differences from the original dinosaurs).
Only a feed forward control mechanism is needed in these models. From the dynamic model, it is possible to evolve gaits, and discover how the animals might have moved.  There are constraints that need to be considered, such as bone strength.  This means big dinosaurs probably didn’t run: the bones in their legs would have snapped on impact.  The same technology can be used to investigate the gait of early humans and various hominins, looking at the evolution from a crouched to an upright posture.

After coffee I went to the Morphogenesis session. The talks had a lot of overlap with the Morphogenetic Engineering Workshop on Monday, but added some more detail.

Lunch was organised with special “junior/senior tables” where we oldies got to impart words of wisdom to the up-and-coming generations; or just have a nice chat, actually.

After lunch I went to the session on Applications of ALife, which I was chairing.  Chairing is always a little stressful: you have to listen to the talk in order to have ready a question to ask if no-one else does (but not ask it if there’s a good flow of questions), and also have to timekeep.  The Effects of Environmental Structure on the Evolution of Modularity in a Pattern Classifier was about evolved hierarchical modularity.  Next was EvoMove: Evolutionary-based living musical companion, which I knew something about, since it was an output from the EvoEvo EU project I was involved with (which explains my several previosu trips to Lyon).  The third talk was unfortunetely cancelled, so we had a slightly awkward half-hour break (we don’t move talks around, in case people want to move between sessions).  Finally was Road Detection using Convolutional Neural Networks, spotting the rather ill-defined edges of some narrow roads through countryside.

Then it was off for coffee, followed by the second keynote of the day: Phillippe Faure on Drug addiction and alteration of decision making process.  This was a discussion of a series of complex and subtle experiments on mice and rats, including some that ran for months: how mice balance exploration of their environment with exploitation of resources; how even genetically identical mice have different "personalities"; how their physical and social environment causes individuation; how addiction to eg nicotine also modifies behaviours other than the desire for nicotine, such as favouring immediate reward over uncertainty; and more.  One part I liked was when someone from the audience asked a questions about the difference between reward and punishment: Faure answered about how the mice avoid punishment, but charmingly added “we don’t do those experiments”.

After this was an announcement about next year’s ALife conference in Tokyo.  Then it was the late breaking poster session, accompanied by wine and cheese, which was sufficiently delicious and plentiful that no further meal was needed.  I attended the European Research Council (ERC) grant information session, which introduced me to “Synergy Grants”, which sound ideal for interdisciplinary work (although they do have a success rate of 2%).







No comments:

Post a Comment