In Europe, the worst weather in over 60 years has led to an equally bad grape harvest. CC-licensed photo by Rachel Kramer 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.
A selection of 10 links for you. Use them wisely. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
My journey inside ElevenLabs’ voice-clone factory • The Atlantic
Charlie Warzel:
»
Using ElevenLabs, you can clone your voice like I did, or type in some words and hear them spoken by “Freya,” “Giovanni,” “Domi,” or hundreds of other fake voices, each with a different accent or intonation. Or you can dub a clip into any one of 29 languages while preserving the speaker’s voice. In each case, the technology is unnervingly good. The voice bots don’t just sound far more human than voice assistants such as Siri; they also sound better than any other widely available AI audio software right now. What’s different about the best ElevenLabs voices, trained on far more audio than what I fed into the machine, isn’t so much the quality of the voice but the way the software uses context clues to modulate delivery. If you feed it a news report, it speaks in a serious, declarative tone. Paste in a few paragraphs of Hamlet, and an ElevenLabs voice reads it with a dramatic storybook flare.
ElevenLabs launched an early version of its product a little over a year ago, but you might have listened to one of its voices without even knowing it. Nike used the software to create a clone of the NBA star Luka Dončić’s voice for a recent shoe campaign. New York City Mayor Eric Adams’s office cloned the politician’s voice so that it could deliver robocall messages in Spanish, Yiddish, Mandarin, Cantonese, and Haitian Creole. The technology has been used to re-create the voices of children killed in the Parkland school shooting, to lobby for gun reform. An ElevenLabs voice might be reading this article to you: The Atlantic uses the software to auto-generate audio versions of some stories, as does The Washington Post.
It’s easy, when you play around with the ElevenLabs software, to envision a world in which you can listen to all the text on the internet in voices as rich as those in any audiobook. But it’s just as easy to imagine the potential carnage: scammers targeting parents by using their children’s voice to ask for money, a nefarious October surprise from a dirty political trickster. I tested the tool to see how convincingly it could replicate my voice saying outrageous things. Soon, I had high-quality audio of my voice clone urging people not to vote, blaming “the globalists” for COVID, and confessing to all kinds of journalistic malpractice. It was enough to make me check with my bank to make sure any potential voice-authentication features were disabled.
«
This stuff is moving very quickly now. On the The Rest Is Entertainment podcast, Marina Hyde and Richard Osman got an (unspecified) AI to recreate their voices and say a few words. You can hear it from about 38 minutes in to the episode titled “The Baby Reindeer Controversy” (April 29 2024): apart from sounding a bit like they’re talking through cloth (low bitrate, one assumes) it’s almost perfect. Then you just need the script for them to talk and bang, fake podcasts everywhere.
unique link to this extract
HouseFresh disappeared from Google Search results. Now what? • Housefresh
Gisele Navarro:
»
In February 2024, we published an article warning readers not to trust product recommendations from well-known newspapers and magazines ranking at the top of Google search results.
I wasn’t expecting so many people to care (even though I secretly hoped they would), but we’re still getting emails and messages about it ten weeks later. In these two months, I have talked to air purifier manufacturers, HouseFresh readers, other independent publishers, Dotdash Meredith employees, well-known activists, tech journalists, Redditors, SEO professionals, and even Google employees.
Today, I want to share some things I’ve learned and some things that happened after publishing that article.
…Within a few days of publishing the David VS Digital Goliaths exposé, I received an anonymous tip from a former Dotdash Meredith employee, who informed me of an SEO content strategy they implement called “keyword swarming.” Through this strategy, Dotdash Meredith allegedly identifies small sites that have cemented themselves in Google results for a specific (and valuable) term or in a specific topic, with the goal of pushing them down the rankings by publishing vast amounts of content of their own.
“Swarming is about drowning out a competitor,” said the person who reached out. The objective is to “swarm a smaller site’s foothold on one or two articles by essentially publishing 10 articles [on the topic] and beefing up [Dotdash Meredith sites’] authority.”
By the way, if “keyword swarming” is indeed a strategy, then it’s clear that it’s not just something you will find in the air purifier space. Dotdash Meredith could be doing this across many other products and topics, utilizing its wide range of publications. That could explain why you will find multiple articles published on sites belonging to Dotdash Meredith ranking at the top of Google.
Is Dotdash Meredith to blame for choosing to “swarm” Google search results by leveraging their network of websites and their machine to create content at scale? Personally, I think it’s not great for the internet, but I understand that, if the leadership at Dotdash Meredith is simply focusing on making money for IAC shareholders.
However, I don’t want to turn this into a personal crusade against Dotdash Meredith because it’s not. The reality is that, whether they have a name for it or not, every other digital goliath is monetizing their websites by using the same tactics.
«
Apple’s AI research suggests features are coming for Siri, artists, and more • The Verge
David Pierce:
»
making product assumptions based on research papers is a deeply inexact science — the line from research to store shelves is windy and full of potholes. But you can at least get a sense of what the company is thinking about — and how its AI features might work when Apple starts to talk about them at its annual developer conference, WWDC, in June.
…In iOS 18, Apple plans to have all its AI features running on an on-device, fully offline model, Bloomberg recently reported. It’s tough to build a good multipurpose model even when you have a network of data centers and thousands of state-of-the-art GPUs — it’s drastically harder to do it with only the guts inside your smartphone. So Apple’s having to get creative.
In a paper called “LLM in a flash: Efficient Large Language Model Inference with Limited Memory” (all these papers have really boring titles but are really interesting, I promise!), researchers devised a system for storing a model’s data, which is usually stored on your device’s RAM, on the SSD instead.
…In another paper, Apple describes a tool called MGIE that lets you edit an image just by describing the edits you want to make. (“Make the sky more blue,” “make my face less weird,” “add some rocks,” that sort of thing.) “Instead of brief but ambiguous guidance, MGIE derives explicit visual-aware intention and leads to reasonable image editing,” the researchers wrote. Its initial experiments weren’t perfect, but they were impressive.
We might even get some AI in Apple Music: for a paper called “Resource-constrained Stereo Singing Voice Cancellation,” researchers explored ways to separate voices from instruments in songs — which could come in handy if Apple wants to give people tools to, say, remix songs the way you can on TikTok or Instagram.
«
Yorkshire apostrophe fans demand road signs with nowt taken out • The Guardian
Mabel Banfield-Nwachi:
»
A council has provoked the wrath of residents and linguists alike after announcing it would ban apostrophes on street signs to avoid problems with computer systems.
North Yorkshire council is ditching the punctuation point after careful consideration, saying it can affect geographical databases. The council said all new street signs would be produced without one, regardless of whether they were used in the past.
Some residents expressed reservations about removing the apostrophes, and said it risked “everything going downhill”. They urged the authority to retain them. Sam, a postal worker in Harrogate, a spa town in North Yorkshire, told the BBC that signs missing an apostrophe – such as the nearby St Mary’s Walk sign that had been erected in the town without it – infuriated her. “I walk past the sign every day and it riles my blood to see inappropriate grammar or punctuation,” she said.
Though the updated St Mary’s sign had no apostrophe, someone had graffitied an apostrophe back on to the sign with a marker pen, which the former teacher said was “brilliant”. She suggested the council was providing a bad example to children who spend a long time learning the basics of grammar only to see it not being used correctly on street signs.
…North Yorkshire council said it was not the first to opt to “eliminate” the apostrophe from street signs. Cambridge city council had done the same, before it bowed to pressure and reinstated the apostrophe after complaints from campaigners.
There was also an outcry from residents when Mid Devon district council considered making it a policy to do away with apostrophes to “avoid potential confusion”.
A spokesperson from North Yorkshire council added: “All punctuation will be considered but avoided where possible because street names and addresses, when stored in databases, must meet the standards set out in BS7666. This restricts the use of punctuation marks and special characters (eg apostrophes, hyphens and ampersands) to avoid potential problems when searching the databases as these characters have specific meanings in computer systems.”
«
Worst wine harvest in 62 years blamed on ‘extreme’ weather and climate change • Euronews
Rosie Frost:
»
Global wine production reached a historic low in 2023 and climate change could be to blame, a new report has revealed.
The International Organisation of Vine and Wine (OIV) says the drink hit its lowest level since 1962. This intergovernmental organisation has 50 member states, representing 75% of the world’s vineyard area.
Experts blame “extreme environmental conditions” including droughts and fires that have been driving the downward trend in production.
Though climate change is not entirely to blame, the OIV says, these conditions are the greatest challenge the industry is facing. Vines are often cultivated in areas around the world that are strongly affected by and incredibly vulnerable to changes in climate.
These conditions have led to a sharp decline, impacting major wine-producing regions across the northern and southern hemispheres. It is even worse than initial estimates made in November, the organisation said this week.
In the EU, wine production declined by 10% in 2023 – the second-lowest recorded volume of wine since the beginning of the century.
Some countries saw a rainy spring cause mildew, flood, damages and losses in vineyards. Others, especially those in southern Europe, suffered from severe ongoing drought.
Italy was one of the wine-producing countries that suffered the most with a 23% drop in productivity. Heavy rainfall causing mildew in central and southern regions alongside flood and hail damage led to the smallest production volume since 1950.
«
Though there’s an English vineyard a few miles from where I live which is doing fine. Perhaps production just moves northwards.
unique link to this extract
Shell sold millions of ‘phantom’ carbon credits • FT
Kenza Bryan and Clara Murray:
»
Shell sold to Canada’s largest oil sands companies millions of carbon credits tied to CO₂ removal that never took place, raising new doubts about a technology seen as crucial to mitigating greenhouse gas emissions.
As part of a subsidy scheme to boost the industry, the Alberta provincial government allowed Shell to register and sell carbon credits equivalent to twice the volume of emissions avoided by its Quest carbon capture facility between 2015 and 2021, the province’s registry shows. The subsidy was reduced and then ended in 2022.
As a result of the scheme, Shell was able to register 5.7mn credits that had no equivalent CO₂ reductions, selling these to top oil sands producers and some of its own subsidiaries. Credits are typically equivalent to one tonne of CO₂. Some of the largest buyers of the credits were Chevron, Canadian Natural Resources, ConocoPhillips, Imperial Oil and Suncor Energy.
Keith Stewart, a senior energy strategist with Greenpeace Canada, criticised these “phantom credits”: “Selling emissions credits for reductions that never happened . . . literally makes climate change worse.”
Shell said carbon capture played “an important role in helping to decarbonise industry and sectors where emissions cannot be avoided” and that realising its potential “requires creating market incentives now.”
…Canada has among the most generous incentive schemes for carbon capture and storage, according to energy research group Wood Mackenzie. But the industry still struggles to be commercially viable even there.
According to Quest’s annual report, its total cost per tonne of carbon avoided was $167.90 in 2022, compared with a carbon price for Alberta’s big industrial emitters that year of $50.
«
The “good enough” trap • The Ruffian
Ian Leslie:
»
Software designers refer to “the good enough principle”. It means, simply put, that sometimes you should prioritise functionality over perfection. As a relentless imperfectionist, I’m inclined to embrace this idea. I gave this newsletter its name to encourage myself to post rough versions of my pieces rather than not to write them at all. When it comes to parenting, I’m a Winnicottian: I believe you shouldn’t try to be the perfect mum or dad because there’s no such thing. At work and in life, it’s often true that the optimal strategy is not to strive for the optimal result, but to aim for what works and hope for the best.
The good enough can be a staging post to the perfect. The iPhone’s camera was a “good enough” substitute for a compact camera. It did the job, but it wasn’t as good as a Kodak or a Fuji. Until it was. Technological innovation often works like this, but the improvement curve isn’t always as steep as with the smartphone camera. Sometimes we allow ourselves to get stuck with a product which is good enough to displace the competition, without fulfilling the same range of needs. The psychological and social ramifications can be profound.
Let’s say you’re a student and you use ChatGPT to write your essays for you. Give it the right prompts and it will produce pieces that are good enough to get the grade you need. That seems like a win: it saves you time and effort, presuming your tutors don’t notice or don’t care. Maybe you get through the whole of university this way. But be wary of this equilibrium. Over the longer term, you will be stunting the growth of your own mind. The struggle of turning inchoate thought into readable sentences and paragraphs is a powerful exercise for the brain. It’s how you get better at thinking. It is thinking.
«
A smart little post which has some other thought-provoking examples.
unique link to this extract
The breach of a face recognition firm reveals a hidden danger of biometrics • WIRED
Jordan Pearson:
»
This week, a website called “Have I Been Outaboxed” emerged, claiming to be set up by former Outabox developers in the Philippines. The website asks visitors to enter their name to check whether their information had been included in a database of Outabox data, which the site alleges had lax internal controls and was shared in an unsecured spreadsheet. It claims to have more than 1 million records.
The incident has rankled privacy experts who have long set off alarm bells over the creep of facial recognition systems in public spaces such as clubs and casinos [in Australia].
“Sadly, this is a horrible example of what can happen as a result of implementing privacy-invasive facial recognition systems,” Samantha Floreani, head of policy for Australia-based privacy and security nonprofit Digital Rights Watch, tells WIRED. “When privacy advocates warn of the risks associated with surveillance-based systems like this, data breaches are one of them.”
According to the Have I Been Outaboxed website, the data includes “facial recognition biometric, driver licence [sic] scan, signature, club membership data, address, birthday, phone number, club visit timestamps, slot machine usage.” It claims Outabox exported the “entire membership data” of IGT, a supplier of gambling machines. IGT vice president of global communications Phil O’Shaughnessy tells WIRED that “the data affected by this incident has not been obtained from IGT,” and that the firm would work with Outabox and law enforcement.
The website’s owners posted a photo, signature, and redacted driver license belonging to one of Outabox’s founders, as well as a redacted screenshot of the alleged internal spreadsheet. WIRED was unable to independently verify the identity of the website’s owners or the authenticity of the data they claimed to have. An email sent to an address on the website was not returned.
“Outabox is aware and responding to a cyber incident potentially involving some personal information,” an Outabox spokesperson tells WIRED.
«
What’s not clear is whether the facial data was encrypted, and if so how well. Though it seems to have been a pretty bad system, matching people who weren’t the same.
unique link to this extract
Detect AI text by just looking at it • Level Up Coding
Fareed Khan:
»
The easiest way to spot AI-generated text is by checking for words that you don’t usually use but are common for ChatGPT. Consider a massive corpus of over 19 billion English words from blogs, articles, news, and more, updated daily from 2010 to now. I looked for the word “delve” using a string search algorithm, and it showed up 52,388 times. I plot its yearly pattern and identified an unusual behavior, a ~200% growth in its appearance on the internet from 2022, the same year when ChatGPT was released on November 30th.
Other words, like “intricacies” or “unwavering”, also shows a similar increase, just like “delve”. They’re being used more often lately.
…Drawing upon my research expertise and two years of experience working with LLMs, I’ve put together a pretty comprehensive list of 100 words you can keep an eye out for in a piece of text to help you figure out if it’s been generated or paraphrased using AI.
But checking for such number of words is not an easy job so to achieve it quickly, I made a web app that quickly checks your text. Just upload your file or paste your text, and it’ll do the rest. Easy peasy!
«
Odd that something which trawled the web in this way should have picked out such uncommon words to overuse, though.
unique link to this extract
Elon Musk’s plan for AI news • Big Technology
Alex Kantrowitz:
»
Elon Musk emailed me this week with some surprising details about his plan to distill and present news on X using AI. I’d written him after trying Grok — X’s AI chatbot — and noticing it didn’t link to a Time story it summarized. I wanted to click into the article and read more, so I reached out.
Musk said better citations are coming, but shared a deeper vision for the product, which he wants to build into a real-time synthesizer of news and social media reaction. Effectively, his plan is to use AI to combine breaking news and social commentary around big stories, present the compilation live, and allow you to go deeper via chat.
“As more information becomes available, the news summary will update to include that information,” Musk told me. “The goal is simple: to provide maximally accurate and timely information, citing the most significant sources.”
That goal won’t be easy to achieve, but the bot might become a novel news product given its access to the X firehose. “Grok is analyzing sometimes tens of thousands of X posts to render a news summary,” Musk said.
Already, Grok is displaying a running list of headlines and incorporating social reaction into its summaries, including the chatter around the Time story I sought about Trump’s potential second term. Grok has plenty of room to improve — and will have to figure out issues like citation and hallucination — but it could be valuable if X gets the execution is right.
“That’s actually what I used to come to Twitter for — news and commentary,” Ben Smith, editor-in-chief of Semafor and author of Traffic, told me.
«
Well, me too, but there used to be this thing called “verified users” whose identity was checked and could be relied on as sources. Not so sure about relying on a chatbot.
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 