More AI shenanigans
« previous post | next post »
Since When Does Eric Adams Speak Spanish, Yiddish and Mandarin?
He doesn’t. But New York City is using artificial intelligence to send robocalls featuring the mayor’s voice in many languages.
By Emma G. Fitzsimmons and Jeffery C. Mays, NYT (Oct. 20, 2023)
The calls to New Yorkers have a familiar ring to them. They all sound like Mayor Eric Adams — only in Spanish. Or Yiddish. Or Mandarin.
Has the mayor been taking language lessons?
The answer is no, and the truth is slightly more expensive and, in the eyes of privacy experts, far more worrisome.
The mayor is using artificial intelligence to reach New Yorkers through robocalls in a number of languages. The calls encourage people to apply for jobs in city government or to attend community events like concerts.
“I walk around sometimes and people turn around and say, ‘I just know that voice. That voice is so comforting. I enjoy hearing your voice,’” the mayor said at a recent news conference. “Now they’re able to hear my voice in their language.”
New York City’s embrace of the technology came this week as Mr. Adams announced a 50-page “action plan” for artificial intelligence — an effort to “strike a critical balance in the global A.I. conversation,” he said, by embracing its benefits while protecting New Yorkers from its pitfalls.
What are some of the pitfalls of AI? And how does the 50-page "action plan" propose to protect New Yorkers from them.
Mr. Adams also introduced a new chat bot that he said could eventually be used to field basic questions received on the city’s 311 help line.
Mr. Adams said that 70 percent of 311 calls were simple questions about things like alternate side parking and that those could be handled by a chat bot, allowing city workers to focus on more complex questions and reducing wait times for callers.
“We know the term A.I. can cause anxiety,” Mr. Adams said. “People think all of a sudden you’re going to have a Terminator-type figure come in and take over government and displace human beings. That’s just not the reality. Take a deep breath; get a grip.”
Privacy advocates still criticized the robocalls, arguing that it was “deeply Orwellian” to try to trick New Yorkers into thinking that Mr. Adams speaks languages that he does not. The group has previously criticized the mayor’s embrace of facial recognition technology and his dispatch of a police robot to patrol the Times Square subway station.
“Yes, we need announcements in all of New Yorkers’ native languages, but the deep fakes are just a creepy vanity project,” said Albert Fox Cahn, executive director of the Surveillance Technology Oversight Project.
…
The robocall effort has cost about $32,000, city officials said, and the chat bot cost about $600,000 to develop. The city used the Voice Lab program by a company called Eleven Labs to generate the phone messages.
Privacy advocates warned that the mayor’s voice could be used to promote disinformation. One such example, which Mr. Cahn’s group created as a parody message, offered up Mr. Adams greeting New Yorkers in an Elvish language and Klingon.
If Eric Adams thinks it’s good to use deep fakes to promote his messaging, it only seems fair to use them to make fun of his policies too. #Parody https://t.co/yojzDmkIE4 pic.twitter.com/PaARvG8ipH
— Albert Fox Cahn (@foxcahn@BSky.Social) (@FoxCahn) October 17, 2023
Mr. Cahn said it took approximately 10 minutes and cost $50 to create the fake message by taking audio from the mayor’s podcast and processing it through an online text-to-speech tool. “A good chunk of that time was trying to figure out how to write the words in Klingon and Elvish,” Mr. Cahn said.
For Mr. Adams, appearing to speak Spanish in a city where more than 1.5 million New Yorkers speak the language at home may have its advantages. His predecessor, Bill de Blasio, often spoke Spanish as mayor, which he said he studied during graduate school at Columbia University in the 1980s. (In an uncharitable assessment, Mr. Adams said Mr. de Blasio “butchers” the language.)
Michael R. Bloomberg took Spanish lessons at City Hall, but his skills were gleefully skewered by a popular Twitter account known as “El Bloombito.”
…
A test of the small business chat bot demonstrated how it offers basic information in a conversational tone. How do I get a liquor permit? A roughly 320-word response offered guidance for caterers and microbreweries.
Mr. Adams suggested asking who was the best mayor. The city chat bot did not venture a guess, but Chat GPT listed four contenders: Mr. Bloomberg, Mr. Koch, Fiorello La Guardia and Rudolph Giuliani.
AI voice generator is already well established. Soon folks will be uttering things they never dreamed of saying. For example, listen to Santa Clause here.
Selected readings
- "'Speech synthesis'" (2/16/19
- "Green needle vs. brainstorm" (10/2/21)
[Thanks to Don Keyser]
Seth said,
October 21, 2023 @ 8:01 am
I can't get worked up over this. How is it in principle any different from the mayor personally recording the robocalls via speaking a phonetic translation? Why would that be acceptable but this not? He gave his consent for the usage, and it could be seen as a gesture of respect. Generations of tourists have been reading things phonetically out of foreign language phrasebooks. It's even an old comedy trope for someone to have a tampered version where the phrases are not what they seem to be. This is just fearmongering.
NSBK said,
October 21, 2023 @ 9:14 am
I'd agree with @Seth, in particular on the point about consent. The term deepfake should only apply in this case to the parodies created by Cahn and his team.
Imagine using someone's visual likeness to promote something that person did not consent to or approve of. Why should it be any different for someone's vocal likeness?
Of course, the parodies created in this case couldn't be illegal — parodies have just added one more channel beyond artists drawing cartoons, actors doing impressions, or netizens creating image macros / memes.
The mayor approving an AI to use his vocal likeness in this way is maybe weird, but to me the weirdness is due to its novelty, not due to something Orwellian.
Philip Taylor said,
October 21, 2023 @ 12:17 pm
I believe that the Mayor is 100% correct in attempting to communicate with his fellow New Yorkers in as many languages as possible. The fact that he does so via an AI application seems completely irrelevant to me.
Viseguy said,
October 21, 2023 @ 7:22 pm
More important, I think, than the mayor's consent in this context is the absence of any intent on his part to deceive his audience (as evidenced by full disclosure of the use of AI). No fakery, no foul.
astrange said,
October 23, 2023 @ 4:02 am
Mr. Adams is a very linguistically interesting person, I think, or at least a person who constantly says strange things in public.
He's apparently given multiple speeches where he uses the phrase “let your haters be your waiters when you sit down at the table of success.”
Philip Anderson said,
November 11, 2023 @ 3:50 pm
Although this post has been quiet for a while, mainly because those commenting were all in agreement, this story about a fake without consent seems relevant:
Fake audio of Sadiq Khan is not a crime, says Met https://www.bbc.co.uk/news/uk-england-london-67389609