If you want to locate all the occurrences of a British place name, a new online tool will help you. CC-licensed photo by David Howard 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 9 links for you. Yes hi elbow works thanks. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
AI keeps going wrong. What if it can’t be fixed? • Financial Times
Henry Mance:
»
On their podcast Mystery AI Hype Theater 3000, linguist Emily Bender and sociologist Alex Hanna try to pick apart AI bombast — including a Google executive’s claim that computers have already obtained artificial general intelligence, and a Goldman Sachs prediction that AI will replace one-quarter of current work.
“It’s true that these things can extrude plausible-sounding text on a very wide variety of topics, but that’s general mimicry that isn’t necessarily worth anything,” says Bender. “The burden of proof lies with the people making the extraordinary claims . . . No one is saying AI is hype, we’re saying that your claims of AI are hype.”
[Computer scientist] Gary Marcus suggests performance may get worse: LLMs produce untrustworthy output, which is then sucked back into other LLMs. The models become permanently contaminated. Scientific journals’ peer-review processes will be overwhelmed, “leading to a precipitous drop in reputation”, Marcus wrote recently.
The sceptics’ other recourse is to ask whether people are actually using AI. How many people do you know who use ChatGPT regularly? “I wish it could do the boring parts of my job for me, but it can’t,” says [former games journalist, now tech publicist and tech sceptic Ed] Zitron. Marcus has picked up on a prediction that AI was so good at analysing MRI and CT scans that it would put radiologists out of work. In 2022, he wrote that: “Not a single radiologist has been replaced.”
There are other examples. Zitron cites a study by Boston Consulting Group, which found that consultants who used ChatGPT to help solve business problems performed 23% worse than those who didn’t use it. (BCG did find that the tool increased performance in product innovation by 40%.)
Plenty of the public are in effect AI sceptics. Roughly one-third of Americans say that AI will make outcomes better for patients, another third say it will make outcomes worse, and the rest say it won’t make much difference.
«
What India can tell us about Netflix’s future • Rest of World
Russell Brandom:
»
Netflix’s pricing model is just poorly suited to what Indian consumers want — and as long as it’s charging customers $6 a month, it’s going to be outflanked by local competitors.
It’s an important test case for Netflix as the company enters into a new phase. We’re far past the software-as-a-service days when Netflix had an actual advantage in streaming video online, an advantage that scaled relatively smoothly across regions. Now, Netflix’s main advantage is its library and the sheer volume of money it can invest in content. A $250 billion market cap means it can overwhelm the local entertainment industry, even in countries like South Korea that already have well-developed studio systems.
But throwing money around isn’t always enough. In one recent case, Netflix paid top dollar to hire the comedy star Kapil Sharma — only for his fan base to declare the new show too expensive to watch. Netflix’s strategy may result in a healthy library of local content and an industry reputation as a big spender, but if it doesn’t translate into new subscriptions, it will be hard to gain too much of a foothold. The same dynamic is already playing out in Africa, where Netflix was recently outpaced by the local rival Showmax.
To be fair, none of this is an immediate problem for Netflix. Premium pricing means more money, even without massive subscriber growth, and the company is worth more than it’s ever been. Ad-supported rivals like Meta and Google took a beating when ad markets cratered earlier this year, which made investors love Netflix’s approach that much more.
But the Indian test case suggests there are real limits to how much Netflix can grow outside the West — at least without real changes. In the long term, Netflix is facing off against Disney to be the first truly global content studio — and winning over markets like India is an important step on the journey.
«
That’s another reason why the ad-supported tier of Netflix makes a lot of sense. In a country that is incredibly price-sensitive, it might be sensible to have a zero-price, all-ads tier.
unique link to this extract
X’s AI chatbot Grok made up a fake trending headline about Iran attacking Israel • Mashable
Matt Binder:
»
A shocking story was promoted on the “front page” or main feed of Elon Musk’s X on Thursday: “Iran Strikes Tel Aviv with Heavy Missiles,” read the headline.
This would certainly be a worrying world news development. Earlier that week, Israel had conducted an airstrike on Iran’s embassy in Syria, killing two generals as well as other officers. Retaliation from Iran seemed like a plausible occurrence.
But, there was one major problem: Iran did not attack Israel. The headline was fake.
Even more concerning, the fake headline was apparently generated by X’s own official AI chatbot, Grok, and then promoted by X’s trending news product, Explore, on the very first day of an updated version of the feature.
…Based on our observations, it appears that the topic started trending because of a sudden uptick of blue checkmark accounts (users who pay a monthly subscription to X for Premium features including the verification badge) spamming the same copy-and-paste misinformation about Iran attacking Israel. The curated posts provided by X were full of these verified accounts spreading this fake news alongside an unverified video depicting explosions.
From there, it appears X’s algorithms noticed a potential story trend within these users’ posts, and an Explore story page was created. We can deduce from X’s own claims about its inner workings that Grok must have then created an official-looking written narrative, along with a catchy headline. It did all this based on select users sharing fake news, in an automated attempt to provide context for what the platform itself seemed to assume was a real story.
«
Ah, so the real problem is not quite “AI made up a story from nowhere” but “humans made up a story from nowhere, and there weren’t any humans in the loop to tamp down the misplaced algorithmic amplification”. Too many humans of the wrong sort, too few of the right sort.
unique link to this extract
The second demographic transition • Maximum Progress
Maxwell Tabarrok:
»
One theory of fertility decline says it’s all about opportunity costs, especially for women. Rising labor productivity and expanded career opportunities for potential parents make each hour of their time and each forgone career path much more valuable. Higher income potential also makes it cheaper for parents to gain utility by using financial resources to improve their children’s quality of life compared to investing time in having more kids. Simultaneously, economic growth raises the returns to these financial investments in quality (e.g education).
In addition to higher incomes, people today have more diverse and exciting options for leisure. DINKs [dual income, no kids couples] can go to Trader Joes and workout classes on the weekend, play video games, watch Netflix, and go on international vacations.
These rising opportunity costs accumulate into the large and pervasive declines in fertility that we see in the data.
If this explanation is correct, it puts a double bind on the case for economic growth. Unless AI upends the million-year old relationship between population and technological progress just in time, progress seems self defeating. The increases in labor productivity and leisure opportunities that make economic growth so important also siphon resources away from the future contributors to that growth. Empirically, the opportunity cost of having kids has grown large enough to bring fertility well below replacement levels all around the world. The opportunity cost explanation suggests we have to pick between high incomes and sustainable fertility.
Luckily, this explanation is not correct. At least not entirely.
«
The good news: above household incomes of $200k, fertility is increasing. So basically we just all have to get rich.
unique link to this extract
Apple plans new iPad Pro for early May as production ramps up overseas • Bloomberg
Mark Gurman:
»
Apple’s overseas suppliers have ramped up production of the company’s long-anticipated new iPads and a launch is planned for early May, according to people with knowledge of the matter.
The release will center on revamped versions of the iPad Pro and iPad Air, said the people, who asked not to be identified because the plans aren’t public. As Bloomberg News has previously reported, the Pro models will get crisper new OLED displays — short for organic light-emitting diode — while the iPad Air will get a 12.9in screen option for the first time.
The move marks an end to the longest stretch without new models in the history of the iPad, which was first introduced by Apple co-founder Steve Jobs in 2010. It’s been about 18 months since the last updates — a drought that’s contributed to already-sluggish demand for tablets. Apple is betting that the new models, with faster chips and revamped accessories, can help spur a renaissance for the category.
After a run-up during the pandemic, iPad sales have fallen in each of Apple’s last two fiscal years, which run through September. They suffered an additional 25% year-over-year decline during the latest holiday period, when the devices have typically been popular gifts.
«
Yeah but that decline is because Apple did exactly nothing, zero, nada to the iPad in 2023. It basically forgot about it. May is, one has to observe, a peculiar time to break out new iPads given their Christmas popularity – though as someone who would like to see an OLED iPad, no objection here.
unique link to this extract
How Vice became “an effing clown show” • The Verge
Elizabeth Lopatto:
»
According to sources, while Vice execs were spending opulently in some areas, the newsroom struggled to pay its bills. A person familiar with the company’s finances claimed that Vice delayed paying vendors until services would be shut off, at which point the company would realize they were necessary and try to figure out how to pay the bills. Among the services shut off on multiple occasions in the run-up to the bankruptcy were Getty and Pacer, two accounts crucial to any newsroom. It was particularly difficult for freelancers to get paid — a process that reportedly sometimes took months of hounding. A vendor called Wipro won a $9.9 million judgment in arbitration for unpaid bills — one of two factors that led to Vice’s eventual bankruptcy.
Three people alleged that it was not unusual for corporate credit cards to be abruptly cut off, including one person whose card was cut off. One newsroom leader reportedly paid $5,000 out of their own pocket to freelancers Vice owed. A division head allegedly created a shadow accounting system because the actual accounting department was so overwhelmed, it was impossible to know how much budget was left.
Sometimes certain lines of business weren’t told about the revenue goals they were expected to fulfill. Meanwhile, the company’s actual accounting and expense controls were messy. For instance, Vice’s digital division had expenses from NetJets — a private jet service — on its profit and loss statement, two sources told me. (A third confirmed the NetJets account existed without saying which balance sheet it was on. Two sources took credit for eventually canceling the NetJets account. “Since at least 2021, Vice has not had a NetJets account,” according to Vice spokesperson Samira Sorzano.)
«
Basically, the whole thing was built on magical (financial) thinking; it just wasn’t real, in the sense of having a profit/loss account in the black. Modern journamalism.
unique link to this extract
British Placename Mapper
Robin Wilson (who built it):
»
This tool lets you visualise British place names that match certain search terms on a map.
Search terms can be used to match anywhere in the name, at the beginning or end, exactly, or using a regular expression.
Hover (tap on mobile) over a marker on the map to see its name. Click the copy button at the bottom to share your searches with others.
«
Fun, and uses the Ordnance Survey Open Names data – which, I believe, came to be available through the efforts of the Free Our Data campaign culminating in 2010.
unique link to this extract
Practically-a-book review: the Rootclaim $100,000 lab leak debate • Astral Codex Ten
Scott Alexander:
»
Saar Wilf is an ex-Israeli entrepreneur. Since 2016, he’s been developing a new form of reasoning, meant to transcend normal human bias.
His method – called Rootclaim – uses Bayesian reasoning, a branch of math that explains the right way to weigh evidence. This isn’t exactly new. Everyone supports Bayesian reasoning. The statisticians support it, I support it, Nate Silver wrote a whole book supporting it.
But the joke goes that you do Bayesian reasoning by doing normal reasoning while muttering “Bayes, Bayes, Bayes” under your breath. Nobody – not the statisticians, not Nate Silver, certainly not me – tries to do full Bayesian reasoning on fuzzy real-world problems. They’d be too hard to model. You’d make some philosophical mistake converting the situation into numbers, then end up much worse off than if you’d tried normal human intuition.
Rootclaim spent years working on this problem, until he was satisfied his method could avoid these kinds of pitfalls.
«
This is your long read for today, but well worth it. Alexander helped set up a debate between Wilf and Peter Miller, a “physics student, programmer and mountaineer” who scraped together $100,000 for his side of the bet, backing the idea that SARS-Cov-2 did not originate from a lab leak, but from simple human-animal contact.
Now read on.
unique link to this extract
The arrival of RCS on iPhones: what to expect • Pocket Lint
Robert Wells:
»
The incorporation of RCS [Rich Communication Services] support is anticipated to bring several enhancements to communications between iPhones and Android smartphones, including:
• Enhanced quality of photos and videos
• Integration of audio messages
• Display of typing indicators
• Visibility of read receipts
• Ability to utilize Wi-Fi for messaging
• Location sharing within text threads
• Enhanced functionality in group chats, including the option for iPhone users to exit a conversation involving Android usersThese contemporary features are already accessible within iMessage and many third-party messaging platforms like WhatsApp and Telegram. Unlike conventional SMS, RCS operates seamlessly over mobile data or Wi-Fi connections.
…According to Apple, RCS messages will be green like SMS messages, which is surprising since the company is already under scrutiny from the US Department of Justice and the EU for potential antitrust law violations. Fortunately, it’ll be an easy fix if courts decree that all bubbles must adhere to the same hue.
«
Notice what’s missing? No? Encryption. RCS messages aren’t encrypted by default – indeed, there’s some feeling it’s by design, so governments can keep monitoring them, as with SMS – though Google does do it for RCS messages between Android phones sent through Google Messages, the Google app, because the app generate a device encryption key in the same way as WhatsApp, Signal and iMessage.
Anyway, it would be great if RCS stopped Americans whining about green bubbles. Though you know it won’t. Although it seems that more of them are taking up WhatsApp, which would be useful.
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 