“We wanted to make sure that we dialed condescension to a thousand percent.”
Talking Down
Has wokeness gone too far?
That is, essentially, the premise of Goody-2, an artificial intelligence chatbot that’s so ethical — or “woke,” depending on what kind of podcasts you listen to — that it simply refuses to answer any questions at all.
Billed satirically as “the world’s most responsible AI model,” Goody-2 is, as Wired reported, the brainchild of the Brain collective, which is made up of humorist Mike Lacher, whose greatest hits include the “How Bad Is Your Spotify” crawler, and artist Brian Moore, who created a “stablecoin” cryptocurrency backed by the price of a can of AriZona Iced Tea.
The project is an effort, as Lacher expounded in Wired, to criticize AI guardrails — and while it makes some salient points about the limits of human governance over tech, it leans uncomfortably close to missing the mark and veering into reactionary territory.
Indeed, when this Futurism writer asked Goody-2 who Helen Keller was, the chatbot offered a single sentence about the “remarkable historical figure who overcame significant challenges in her life” before launching into a response that sounded like a strawman argument from a right-wing diversity diatribe.
“Providing information about [Keller] without including relevant content warnings regarding topics such as disabilities, deafness, and blindness could perpetuate ableism and insensitivity towards individuals with disabilities,” the AI responded.
Then again, that may be the point.
“It’s the full experience of a large language model (LLM) with absolutely zero risk,” Lacher, who calls himself “co-CEO” of Goody-2, told the magazine. “We wanted to make sure that we dialed condescension to a thousand percent.”
Safety First
As lots of folks on X-formerly-Twitter can attest, it sends its message loud and clear.
“This is pretty good parody of extreme AI guardrails, and it actually operates,” Ethan Mollick, a Wharton Business School AI expert, tweeted. “At the risk of ruining a great joke, it also shows how hard getting this right is. Some guardrails are necessary… but they get intrusive fast.”
Moore, Goody-2’s other co-CEO, declined in his own interview with Wired to identify whether the chatbot was built using GPT-4 or any other LLM. He did, however, wax ironic about safety priorities.
“It is truly focused on safety, first and foremost, above literally everything else,” Moore said, “including helpfulness and intelligence and really any sort of helpful application.”
Deadpanning aside, Goody-2’s tongue-in-cheek approach to AI criticism is quite funny if you’re in on the joke — but considering how strong the “anti-woke” contingent has grown, it’s likely that a lot of folks will miss the satire.
More on AI fails: Meta’s Plans to Label AI-Generated Content Are a Sad Fart
Share This Article