There's some scripts lying around my src directory to run something similar by scraping hubski comments and using them as the brain MegaHAL (same markov bot as lambda). I killed all instances a while ago, but if anyone is interested, they can talk to thundarabot on #zombski @ irc.hubski.com: 11:38:42 < thundarabot> thundara: It's not the most artificial of artificial inflations! They're not too substantially as i've had wrist, ankle, and i have a friend right now.11:38:41 <@thundara> thundarabot: When her best friend died, she rebuilt him using artificial intelligence
Me as well. I don't think there is a word that accurately describes this level of creepiness.
Of course there is. Theatrics. The article has a stupid-ass "digital interference" on the title, drop caps and quotes. It's close to halloween, so I'm sure they're playing off of that and peoples general insecurities about AI and technology we don't yet understand (and that we don't yet know we don't understand). While the tech in the article is entirely plausible, i.e. neural nets and markov chains, there's no chance those responses are accurate / unedited, if not entirely fabricated. Just look at popular "AI" projects like the Subreddit Simulator where it's taken months to reach some pretty convincing Reddit titles, but then the comments are just full of nonsense (oh hey so it works!). What about Microsoft's Tay chatbot, that came up with some pretty convincing results... about half the time. I think it's just a bit of viral marketing for their startup. Nothing special here, sorry.