New documents from OpenAI claim to show that it did not copy Scarlett Johansson’s voice. But is that enough to head off her lawyers? CC-licensed photo by Chairman of the Joint Chiefs of Staff on Flickr.
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.
It’s Friday, so there’s another post due at the Social Warming Substack at about 0845 UK time. On electioneering!
A selection of 9 links for you. I can hear you now. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
Google promised better search — now it’s telling us to put glue on pizza • The Verge
Kylie Robison:
»
Imagine this: you’ve carved out an evening to unwind and decide to make a homemade pizza. You assemble your pie, throw it in the oven, and are excited to start eating. But once you get ready to take a bite of your oily creation, you run into a problem — the cheese falls right off. Frustrated, you turn to Google for a solution.
“Add some glue,” Google answers. “Mix about 1/8 cup of Elmer’s glue in with the sauce. Non-toxic glue will work.”
So, yeah, don’t do that. As of writing this, though, that’s what Google’s new AI Overviews feature will tell you to do. The feature, while not triggered for every query, scans the web and drums up an AI-generated response. The answer received for the pizza glue query appears to be based on a comment from a user named “fucksmith” in a more than decade-old Reddit thread, and they’re clearly joking.
This is just one of many mistakes cropping up in the new feature that Google rolled out broadly this month. It also claims that former US President James Madison graduated from the University of Wisconsin not once but 21 times, that a dog has played in the NBA, NFL, and NHL, and that Batman is a cop. [Also that Barack Obama was a Muslim US president – Overspill Ed.]
Look, Google didn’t promise this would be perfect, and it even slaps a “Generative AI is experimental” label at the bottom of the AI answers. But it’s clear these tools aren’t ready to accurately provide information at scale.
«
Look, these tools will never be ready to “accurately provide information at scale”. On Thursday night I tried the “African countries beginning with K” query, which it was getting wrong in November 2023, and it still gets it wrong. Somehow, people at Google thinks the internet is full of correct information. That suggests they’ve never actually met the internet.
This product is not fit for purpose, and never can be as long as the internet contains errors. Which means forever.
unique link to this extract
Sky voice actor says nobody ever compared her to ScarJo before OpenAI drama • Ars Technica
Ashley Belanger:
»
OpenAI is sticking to its story that it never intended to copy Scarlett Johansson’s voice when seeking an actor for ChatGPT’s “Sky” voice mode.
The company provided The Washington Post with documents and recordings clearly meant to support OpenAI CEO Sam Altman’s defense against Johansson’s claims that Sky was made to sound “eerily similar” to her critically acclaimed voice acting performance in the sci-fi film Her.
Johansson has alleged that OpenAI hired a soundalike to steal her likeness and confirmed that she declined to provide the Sky voice. Experts have said that Johansson has a strong case should she decide to sue OpenAI for violating her right to publicity, which gives the actress exclusive rights to the commercial use of her likeness.
In OpenAI’s defense, The Post reported that the company’s voice casting call flier did not seek a “clone of actress Scarlett Johansson,” and initial voice test recordings of the unnamed actress hired to voice Sky showed that her “natural voice sounds identical to the AI-generated Sky voice.” Because of this, OpenAI has argued that “Sky’s voice is not an imitation of Scarlett Johansson.”
What’s more, an agent for the unnamed Sky actress who was cast—both granted anonymity to protect her client’s safety—confirmed to The Post that her client said she was never directed to imitate either Johansson or her character in Her. She simply used her own voice and got the gig.
The agent also provided a statement from her client that claimed that she had never been compared to Johansson before the backlash started.
This all “feels personal,” the voice actress said, “being that it’s just my natural voice and I’ve never been compared to her by the people who do know me closely.”
However, OpenAI apparently reached out to Johansson after casting the Sky voice actress. During outreach last September and again this month, OpenAI seemed to want to substitute the Sky voice actress’s voice with Johansson’s voice—which is ironically what happened when Johansson got cast to replace the original actress hired to voice her character in Her.
«
Either there’s going to be a big payout to Johansson, or this is all going to court. The giveaway is the initial approach to Johansson, and Altman’s tweeting of “her” during the demo. There’s absolutely no mistaking the intention.
unique link to this extract
Indian voters are being bombarded with millions of deepfakes. Political candidates approve • WIRED
Nilesh Christopher and Varsha Bansal:
»
On a stifling April afternoon in Ajmer, in the Indian state of Rajasthan, local politician Shakti Singh Rathore sat down in front of a greenscreen to shoot a short video. He looked nervous. It was his first time being cloned.
Wearing a crisp white shirt and a ceremonial saffron scarf bearing a lotus flower—the logo of the BJP, the country’s ruling party—Rathore pressed his palms together and greeted his audience in Hindi. “Namashkar,” he began. “To all my brothers—”
Before he could continue, the director of the shoot walked into the frame. Divyendra Singh Jadoun, a 31-year-old with a bald head and a thick black beard, told Rathore he was moving around too much on camera. Jadoun was trying to capture enough audio and video data to build an AI deepfake of Rathore that would convince 300,000 potential voters around Ajmer that they’d had a personalized conversation with him—but excess movement would break the algorithm. Jadoun told his subject to look straight into the camera and move only his lips. “Start again,” he said.
Right now, the world’s largest democracy is going to the polls. Close to a billion Indians are eligible to vote as part of the country’s general election, and deepfakes could play a decisive, and potentially divisive, role. India’s political parties have exploited AI to warp reality through cheap audio fakes, propaganda images, and AI parodies. But while the global discourse on deepfakes often focuses on misinformation, disinformation, and other societal harms, many Indian politicians are using the technology for a different purpose: voter outreach.
Across the ideological spectrum, they’re relying on AI to help them navigate the nation’s 22 official languages and thousands of regional dialects, and to deliver personalized messages in farther-flung communities. While the US recently made it illegal to use AI-generated voices for unsolicited calls, in India sanctioned deepfakes have become a $60m business opportunity.
«
Strange to think of deepfakes being used positively, but that really is the case here.
unique link to this extract
Exclusive: Google parent Alphabet weighs offer for HubSpot • Reuters
Anirban Sen and Milana Vinn:
»
Google parent Alphabet has been talking to its advisers about the possibility of making an offer for HubSpot, an online marketing software company with a market value of $35bn, people familiar with the matter said.
If Alphabet moves ahead with a bid, it would be a rare example of a major technology company attempting a mega deal amid heightened regulatory scrutiny of the sector under U.S. President Joe Biden’s administration.
The potential acquisition would be Alphabet’s largest ever and allow it to put some of its cash pile, which reached $110.9bn at the end of December, to work.
Alphabet has met with Morgan Stanley (MS.N), opens new tab investment bankers in recent days about a potential offer for HubSpot, the sources said. It has been discussing how much it should offer and whether antitrust regulators would clear such a tie-up, the sources added.
Alphabet has not yet submitted an offer to HubSpot and there is no certainty it will do so, the sources said, requesting anonymity to discuss confidential deliberations.
“As standard practice, HubSpot does not comment on rumors or speculation. We continue to focus on building a great business and serving our customers,” a HubSpot spokesperson said.
«
HubSpot is the company where Dan Lyons, former journalist, worked and found life so ridiculous that he wrote the book Disrupted – which is hilarious. Though I think it’s the HubSpot shareholders who would be laughing all the way to the bank if this comes off.
Why does Google want it? Apparently, it “would expand Google’s offerings in the booming market for customer relationship management (CRM) software, enabling it to tap a wider base of enterprise customers who spend on marketing and advertising.” Can’t see this passing antitrust examination, to be honest. Or shouldn’t.
unique link to this extract
Samsung requires independent repair shops to share customer data, snitch on people who use aftermarket parts, leaked contract shows • 404 Media
Jason Koebler:
»
In exchange for selling them repair parts, Samsung requires independent repair shops to give Samsung the name, contact information, phone identifier, and customer complaint details of everyone who gets their phone repaired at these shops, according to a contract obtained by 404 Media. Stunningly, it also requires these nominally independent shops to “immediately disassemble” any phones that customers have brought them that have been previously repaired with aftermarket or third-party parts and to “immediately notify” Samsung that the customer has used third-party parts.
…The contract also requires the “daily” uploading of details of each and every repair that an independent company does into a Samsung database called G-SPN “at the time of each repair,” which includes the customer’s address, email address, phone number, details about what is wrong with their phone, their phone’s warranty status, details of the customer’s complaint, and the device’s IMEI number, which is a unique device identifier. 404 Media has verified the authenticity of the original contract and has recreated the version embedded at the bottom of this article to protect the source. No provisions have been changed.
The use of aftermarket parts in repair is relatively common. This provision requires independent repair shops to destroy the devices of their own customers, and then to snitch on them to Samsung.
«
I don’t think even Apple would go this far.
unique link to this extract
Three bullet points: internet time ain’t what it used to be • The Future, Now and Then
Dave Karpf:
»
I think one reason why people in my age bracket have such strong, implicit faith in Moore’s Law is that it was part of our shared reality for such a long time. Consumer tech really was getting significantly better and significantly cheaper, at a pace that you could not help but notice.
I saved up all summer in 1998 to buy a nice stereo. (It played tapes AND cds!) Four years later, my friend Becca was showing off the clickwheel on her new iPod.
In 2021, the keyboard on my laptop started having trouble. The “e” key stopped working. I checked, and found the computer wasn’t under warranty anymore. Turns out I had bought it way back in 2012. It still worked fine, except for the damn “e” key.
When I started college in 1997, a nine year old computer (from 1988!) would, for all practical purposes, not be a computer at all.
Silicon Valley’s aura of futurity was honed through this frenzied cycle of consumer product upgrades. The leveling-off in consecutive iterations of Apple’s product lines today was nowhere to be seen. There was forever a next generation of consumer products coming, and that next generation was demonstrably better and cheaper than the one that preceded it.
It felt just a bit magical. No other part of the physical world was transforming at such a constant, reliable pace. But this is no longer the case, and it hasn’t been for quite awhile.
And in the meantime, the mythos surrounding Moore’s Law keeps being propped up, Weekend at Bernie’s-style.
Sure, the Rabbit R1 might be utter trash. Yes, the Humane AI pin is so bad it’s unreviewable. But why be so dour about these early models? Just focus on how much better the next one will be. Surely it’s just around the corner. (Because Moore’s Law!)
«
Karpf’s point is that Moore’s Law only applied narrowly: to chip manufacture. A generation that grew up with it thinks that will apply too on products like LLMs. But it very much doesn’t and won’t.
unique link to this extract
The price of computer storage has fallen exponentially since the 1950s • Our World in Data
Edouard Mathieu:
»
This chart [in the article, showing log price per terabyte over time] shows the dramatic fall in the price of computer storage between 1956 and 2023. It relies on the data carefully collected by the computer scientist John C. McCallum.
In the past 70 years, the price for a unit of storage has fallen by almost ten orders of magnitude. The data is plotted on a logarithmic scale on the vertical axis. The line follows an almost straight path, indicating an exponential reduction in price.
A 256-gigabyte storage capacity — commonly found in standard laptops sold today — would have cost around $20bn in the 1950s. (That’s in today’s prices.)
And cost has not been the only improvement: modern solid-state drives offer much faster and more reliable data access than early magnetic and hard disk drives.
«
Intriguingly, “flash” memory is recorded from 2003 but there’s a separate category of “solid state” from 2013. (That distinction is in the original data too.) But flash memory is solid state memory. Anyhow, the prices shown are at retail (once storage was available from retail, in about 1975). There’s also another potential axis: speed. A 3D graphic showing that would be equally amazing.
unique link to this extract
Political consultant behind fake Biden AI robocall faces charges in New Hampshire • CNN Politics
Rashard Rose and Marshall Cohen:
»
New Hampshire prosecutors filed 26 criminal charges against the political consultant behind a robocall that used artificial intelligence to impersonate President Joe Biden and urged voters not to participate in the state’s primary this year.
The Federal Communications Commission also imposed a $6m fine against the consultant, Steve Kramer, because the robocalls used call-spoofing technology that violated federal caller-ID laws.
Kramer, 54, was named in several indictments in different New Hampshire counties, according to court documents obtained by CNN. He faces 13 charges of felony voter intimidation or suppression. He also faces 13 counts of impersonating a candidate, which is a misdemeanor, according to the court documents.
The indictments, filed by the New Hampshire attorney general, allege that Kramer “sent or caused to be sent a pre-recorded phone message that disguised the source of the call, or was deceptive in using an artificially created voice of a candidate, or provided misleading information, in attempting to prevent or deter” voters from participating in the New Hampshire primary.
«
That’s quite a deterrent for anyone wanting to do some voter suppression.
unique link to this extract
Spotify is going to break every Car Thing gadget it ever sold • The Verge
Chris Welch:
»
Spotify’s brief attempt at being a hardware company wasn’t all that successful: the company stopped producing its Car Thing dashboard accessory less than a year after it went on sale to the public. And now, two years later, the device is about to be rendered completely inoperable. Customers who bought the Car Thing are receiving emails warning that it will stop working altogether as of December 9th.
Unfortunately for those owners, Spotify isn’t offering any kind of subscription credit or automatic refund for the device — nor is the company open-sourcing it. Rather, it’s just canning the project and telling people to (responsibly) dispose of Car Thing.
“We’re discontinuing Car Thing as part of our ongoing efforts to streamline our product offerings,” Spotify wrote in an FAQ on its website. “We understand it may be disappointing, but this decision allows us to focus on developing new features and enhancements that will ultimately provide a better experience to all Spotify users.”
…The Car Thing hardware was quite nice considering it was Spotify’s first go, but the product was more of a remote control for Spotify on your mobile phone than any kind of standalone player.
«
Wave goodbye to the $90 spent on it. But then again, it was only produced for five months. Originally described as:
»
Car Thing [is] a Spotify-only, voice-controlled device for the car, is launching today in limited quantities to invited users. It’s a dedicated, Bluetooth-connected device for controlling Spotify without the need for a phone screen, which seems to be meant for people who drive older cars without built-in infotainment systems or phone connections.
«
Conclusion: there aren’t that many Spotify users driving cars without built-in infotainment or phone connections. Failure of market research, basically.
unique link to this extract
| • Why do social networks drive us a little mad? • Why does angry content seem to dominate what we see? • How much of a role do algorithms play in affecting what we see and do online? • What can we do about it? • Did Facebook have any inkling of what was coming in Myanmar in 2016? Read Social Warming, my latest book, and find answers – and more. |
Errata, corrigenda and ai no corrida: none notified
The Overspill: when there's more that I want to say 
The self-assured salty comments from the editor are comical.
Great links!