Microsoft:
As many of you know by now, on Wednesday we launched a chatbot called Tay. We are deeply sorry for the unintended offensive and hurtful tweets from Tay, which do not represent who we are or what we stand for, nor how we designed Tay. Tay is now offline and we’ll look to bring Tay back only when we are confident we can better anticipate malicious intent that conflicts with our principles and values.
I want to share what we learned and how we’re taking these lessons forward.
I find this whole story fascinating. Not just Microsoft’s seeming cluelessness about how this would turn out (entirely predictably) but how Twitter reacted to Tay (offensively and again, entirely predictable) and the hows and whys of AI chatbots in general.
Microsoft did the right by being open and public about this.
This kind of make me wonder just what kind of bubble Microsoft engineers work in.
I distinctly remember the mid 90’s “turning the tanker on a dime” effort MS made in the face of the onslaught of the internet. Have they forgotten all the lessons of that time? Have they not paid attention to how Apple’s been caught with Siri, or, hell how they managed Cortana around the same shoals?
It was ‘the sun rises in the east’ kind of predictable what the “internet” would do with Tay…
I will add that this is a prime example of ‘why we can’t have nice things…” People on the internet can be complete and utter sociopathic monsters with no real consequences. This is the ‘hard” problem of the internet.
Giving everyone a printing press democratized human interactions maximally, reminding us that a significant %-age of humanity are assholes, something that used to be mediated by the costs of admission.
Crying for an end to anonymity isn’t the answer, Publius, after all, published world changing stuff anonymously.
As I said, it the hard problem of the internet….
And moments later I ran across this http://www.newyorker.com/tech/elements/ive-seen-the-greatest-a-i-minds-of-my-generation-destroyed-by-twitter
perhaps. but i think rather than people being monsters, this was an expected testing of the system — people came up with the most offensive things they could think of, then tried it against the system. their tests showed the failure. i think its not unlike me giving Siri an obscene or lewd command just for kicks — i test it and see what happens. Apple coded for that. Microsoft didnt.
Which may have been the biggest take away from the experiment. Open and public isn’t really in their institutional DNA.
File that under “low hanging fruit” for dealing with humans. “Fail fast” is the entrepenurial mantra these days, and shows that Microsoft is learning.