Thursday, 14 March 2013


One of the thingss that concerns me about our inevitable creation of artificial intelligence is that those agents will be equivalently bored in undertaking mundane tasks: it seems to me that there is a broad class of tasks that are difficult enough to require sentience, but not difficult enough to be intellectually stimulating.

So, will this problem arise? There are a couple of reasons for optimism: firstly, boredom is a human concept that we can avoid programming into AIs (however, such a concept may evolve within an AI outside of our control); secondly, my own theory of human boredom is that it arises from our inability to effectively multi-task. We cannot ring-fence the intellectual capacity needed for a certain task, and use the rest for something else, hence we allocate too much intellectual resource to a task that doesn't require it. An AI agent may however be able to do this ring-fencing, and as a result may not be bored.

(I'm sure my concerns are in part attributable to Douglas Adams' Marvin.)

No comments: