It seems as though the answer Microsoft came up with for preventing its chat bot "Zo" from ending up racist like "Tay", is through the development of a content moderator ( https://www.microsoft.com/cognitive-services/en-us/content-moderator ).
On Wed, Feb 22, 2017 at 2:35 PM, Mark Nuzz <[email protected]> wrote: > > > On Wed, Feb 22, 2017 at 2:20 PM, Jan Matusiewicz < > [email protected]> wrote: > >> AGI should show us the truth. It shouldn't be deliberately biased. And >> this is the best weapon against racism. Racism is irrational - it comes >> from, built in human nature, bias against aliens. Humans have a tendency to >> generalize negative features of units to whole groups in a very different >> way for our group and the alien group. We have examples of that in the >> history and now. The idea that some nations should be eradicated or >> enslaved because they so much differ from our nation is riddiculous. But is >> was very popular. Assuming that all people from particular countries should >> be banned from entering your country is also irrational but appealing for >> many. >> > Slavery and genocide is of course a bad thing! We can objectively define > those things. But can you objectively define racism? Where do you draw the > line? Is acknowledging that there are genetic differences between races, > which goes beyond cosmetics, itself a form of racism? Is drawing > correlations between race and anything else a form of racism? What if an > AGI were to disagree with you on that (whatever your opinion is)? What if > an AGI were to disagree with another AGI on that? It's a pretty bold claim > to say that this is a clear cut case of objectivity. If that were so, there > wouldn't be any room for debate, but there clearly is if you're one of the > few who hasn't turned your social network into an echo chamber by now. > > Taking this one step further, with your own example, what if AGI were to > determine that slavery of 5% of the population was the only viable way for > human prosperity? The "threat" is not that we won't be able to shut it off, > but that we won't want to. I'm not sure that AGI can even save humanity > from itself anymore. > > > >> An AGI which tries to be as objective as possible would be free from this >> dangerous tendency. This would make political correctness unnecessary for >> it. However it is possible that people would interpret its finding in a >> racist way. For example it may find out that ethnic group A is better at IQ >> tests then group B, but this is be due to lower educational level of B. >> Some people would repeat only the first part of its finding. >> On Wed, 22 Feb 2017 21:51 Dr Miles Dyson, <[email protected]> >> wrote: >> >>> If the world is such that the data that exists supports empirical and >>> scientific racism (or insert your politically incorrect world view here), >>> how could we create a AGI that could avoid (or minimize the likelihood of >>> it) taking such an undesirable view? >>> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> >>> <https://www.listbox.com/member/archive/rss/303/28565694-f30243b8> | >>> Modify <https://www.listbox.com/member/?&;> Your Subscription >>> <http://www.listbox.com> >>> >> *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> >> <https://www.listbox.com/member/archive/rss/303/2552162-5fb4be76> | >> Modify <https://www.listbox.com/member/?&;> Your Subscription >> <http://www.listbox.com> >> > > *AGI* | Archives <https://www.listbox.com/member/archive/303/=now> > <https://www.listbox.com/member/archive/rss/303/26973278-698fd9ee> | > Modify > <https://www.listbox.com/member/?&;> > Your Subscription <http://www.listbox.com> > ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
