From: jeffrey E. <[email protected]> Sent: Monday, March 6, 2017 10:55 PM To: Joscha Bach Subject: Re: it was a class of virus , is more like a mknd than t=e creativity of a neuron On Mon, Mar 6, 2017 at 6:52 PM, Joscha Bach < » wrote: Octopi have a reputatio= for extreme creativity, i.e. they can observe things in their environment= simulate them mentally in several ways, come up with a solution and then =nact it. Virus intelligence is an emergent thing, over many viru=es. The individual virus either succeeds or fails. A virus is like a thoug=t, not like a mind. Am Mar 6, 2017 um 5:50 PM schrieb jeffrey E. <[email protected] <mailto:[email protected]»: =br class="m_222983472287534534Apple-interchange-newline"> watch the octupus, find its prey, is it creative=? is it intelligent? are virus creatvie? 4)=A0 finding new methods to infect the host. . hackers 4P=A0 On =on, Mar 6, 2017 at 6:47 PM, Joscha Bach < >= wrote: It seems that much of what we call intelligence could be de=cribed as detecting causal structure in a domain. Causality means that you=have a model that tells you how change of a condition will influence the d=stribution of outcomes. Of course, crabs do that too, to an extent. <=iv> When I was in Munich, a patent lawyer challenged me to come u= with a definition of creativity from a machine learning perspective. He h=s to judge the originality of inventions, and something that a domain expe=t can come up with "naturally" does not count as an invent=on, even if it has never been done before. learning sys=ems tend to move along gradients, and the "natural solutiont>=9D of the domain expert can be seen as an extrapolation of the solution sp=ce along the known gradients, until an optimum is found. Creativity seems =o involve crossing discontinuities in the problem space, i.e. do more than=follow the gradients. It is not magic, of course; you can use evolutionary=methods/Monte Carlo etc., but you find that many domai