ABM get-together hosted by the Future of Humanity Institute

Ken: Jan 1st NetLogo 6 “Hexy” release, incompatible with previous versions but includes an automated tool for converting to new version. It has good new functional programming features. Also LevelSpace is officially part of new release – models interacting with models e.g. sheep have a neural network in a predator-prey model. BehaviourSearch stopped working in version in 5.3, was reported to be fully restored in 6.1, however at least in Windows it is now working in 6.0. Lastly the web version is working better and better

Rob Wortham (Bath University): (slides) running a workshop in April in Bath. C++ environment for programming for robots with POSH plans (Joanna will send slides on POSH)

Andreas Duering: (slides) Population and cemetery simulator. Has used this model in a museum setting (Germany). Currently looking at a pit of dead people in Hinkley in Somerset, Talheim in Germany, and San Juan ante Portam Latinam in Spain. Now applying ABM in Ain Ghazal in Jordan and first organised religions. “Some fields like complexity, others not.”

Omar : (Using Jupyter) to model trading and inequality, studying the effect of institutions in an ecosystem of very simple agents. Agents start with same amount of assets, agents have lifecycle and mix, and then transact with each other.

Joanna Bryson: 19th-21st April for the Society with AI. Used word2vec to trawl internet for semantic meaning of words in relation to emotions i.e. insects unpleasant, flowers pleasant, left hand bad, right good. Is origin of this association purely cultural? Talked about cultural AI. “A mixture of “Cheats” and “co-operators” can enable maximal group benefit” Have turned this into a game for experiments on AI ethics, mentioned Oliver Brock: Engineers tell us what policies we can enable, Science how these policies could play out, Humanities which we should aim for. #1 Moral Hazard, #2 More Patiency (Miller & Hellstrom), Bryson 2016/17, #3 Legal Lacuna “try suing a bankrupt author, i.e. google has more money than they can spend, hence AI cars, #4 Fear of robots/AI – fear distracts us, #5 Coherence, we are part of a social species (Zahavi 1977, Sylwester 2013). “Why transfer our values onto a different platform”.

James Foley (wildcru, Zoology in Oxford): less than 500 Ethiopian wolves, threatened by diseases (and climate change), especially those carried by domestic dogs. Looking for management strategies. Gathering dog movement data by collar, showed us his predator prey SI model. “Minimum viable population models” Will use the model on website, in schools and to help argument around vaccination of wolves with policy makers.

Miles Brundage (Arizona and Oxford): will model how companies/people with access to AIs interact i.e. the game theory of people + AI. Interested in openness e.g. openAI  and Google’s Deep Mind. Openness, safety, preferences of human actors. Coding in MESA (python) framework. “Are there other similarly complicated models” – energy and nuclear game theory politics, also climate models. “Is this a ‘Map of Germany Problem’?” Inspired by paper by Deep Mind where differentiated AIs in terms of how many layers of decision making, how much memory, how much processing.


Posted in Meetings | Leave a comment

Leave a Reply