A Deepmind engineer discovered that you can get ChatGPT to create its own chatbot. But is it a real one or an imagined one?

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.

---

There’s another post coming this week at the Social Warming Substack on Friday at about 0845 UK time. Free signup. Catch up on the old ones!

---

A selection of 9 links for you. Not a chatbot. I’m @charlesarthur on Twitter. Observations and links welcome.

---

No more airplane mode? EU to allow calls on flights • BBC News

»

Airline passengers in the European Union (EU) will soon be able to use their phones to full effect in the sky, after the European Commission ruled that airlines can provide 5G technology on board planes, alongside slower mobile data.

This could mean flyers will no longer be required to put their phone on airplane mode – though the specifics of how it will be implemented are unclear.

The deadline for member states to make the 5G frequency bands available for planes is 30 June 2023. This will mean people can use all their phone’s features mid-flight – enabling calls as well as data-heavy apps that stream music and video.

Thierry Breton, EU Commissioner for the Internal Market, said the plan would “enable innovative services for people” and help European companies grow. “The sky is no longer a limit when it comes to possibilities offered by super-fast, high-capacity connectivity,” he said.

The EU Commission has reserved certain frequency bands for aircraft since 2008, allowing some services to offer mid-air internet access. But this service has been historically slow, as it relied on equipment to connect people via a satellite between the aeroplane and the ground.

The new system will be able to take advantage of the much faster download speeds provided by 5G, which according to mobile network EE can be over 100Mbps – enabling a film to be downloaded in just a few minutes.

Dai Whittingham, chief executive of the UK Flight Safety Committee, told the BBC that airplane mode was historically important due to a lack of knowledge about how mobile devices affect aircraft. “There was a concern they could interfere with automatic flight control systems,” he said. “What has been found with experience is the risk of interference is very small. The recommendation has always been that once you are in flight, devices should be in in airplane mode.”

There has been a concern in the US that 5G frequencies could interfere with flights, and even potentially lead to erroneous altitude measurements. But Mr Whittingham said this is not an issue in the UK and the EU. “There is much less prospect of interference,” he said, “We have a different set of frequencies for 5G, and there are lower power settings than those that have been allowed in the US.

«

The challenge for icon designers will be what to replace Airplane Mode with. Or maybe, like the floppy disk icon for “Save”, it will live on even after its time is done.
unique link to this extract

---

Andreessen Horowitz tech “news” site Future.com shuts down, staff leave • Business Insider

Rob Price and Melia Russell:

»

Launched in June of 2021, it was billed as a buzzy new tech publication from prestigious venture capital firm Andreessen Horowitz — and a way to sidestep the legacy media entirely and take the message of technological progress directly to readers.

With minimalist indigo branding and a flashy website address — future.com — Future enlisted the Menlo Park-based investment firm’s portfolio company executives, outside experts, and a suite of high-profile editorial hires to pump out a stream of hopeful articles about technology and society.

“We’re going to be having an optimistic lens on technology and the future,” Margit Wennmachers, an operating partner at the firm, told an Insider reporter in an interview at the time.

The New Yorker wrote about the launch, calling it an “opportunity to introduce new terminology, new ideologies, new framing, and new ways for people in and around technology to conceptualize their work.” Independent journalist Eric Newcomer said its debut was part of a strategy with “dramatic implications for the future of media and the venture capital industry.” Tech news site Protocol, which shut down recently, asked whether it was “the future of media.”

But a year and a half later, the publication is dead in the water.

Future hasn’t published a new article in months, most of its editorial staffers have left, and its newsletter is defunct. A source familiar with Andreessen Horowitz’s content strategy confirmed to Insider that Future is shutting down.

«

a16z wanted its site to be the tame teller of all the good news about its investments. But for a lot of VC investments, the news isn’t good: like all startups, VC-funded or not, they run into trouble and things go wrong. There’s no money in good news sites, and no point funding them, as a16z has realised.
unique link to this extract

---

The Excel World Championship is the internet at its best • The Atlantic

Jacob Stern:

»

Competitive Excel clearly is not the NFL, but it does have the beginnings of a fan base. This was just the second year of the World Championship, but it’s already streaming on ESPN3. This year’s edition has 30,000 views on YouTube. Supporters of Michael Jarman, the No. 3 seed in this year’s competition, call themselves the “Jarmy Army.” A few months ago, an all-star game of sorts aired on ESPN2, and this month, ESPNU will televise the collegiate championship.

The tournament begins with a 128-player field and proceeds March Madness–style, in one-on-one, single-elimination contests. The format lends itself to frequent upsets: This year, the No. 2 seed was eliminated in the third round. In each match, players work as fast as possible—they’re generally given about 30 minutes—to answer a series of progressively more difficult questions testing both their puzzle-solving skills and their fluency with Excel. The questions all revolve around the same scenario. In the quarterfinal, for example, the questions all had to do with a fictional country transitioning from dictatorship to democracy. The first and easiest question asked players to calculate how many votes were cast for the purple party. The championship case, which was far more difficult, centered on a 100×100 chessboard. This year’s total prize money was $10,000.

Naturally, a large proportion of Excel competitors work in Excel-heavy jobs; the field included plenty of finance bros, data analysts, mathematicians, actuaries, and engineers. All but one of the eight finalists had over the course of their lives spent thousands of hours working in Excel (the other is a Google Sheets guy), and half of them had spent more than 10,000. The tournament is not particularly diverse. Of the eight finalists, Deaton was the only woman. In the field of 128, she told me, she counted no more than a dozen, which didn’t surprise her, given how heavily male the relevant occupations skew.

«

It’s a totally different world. Like any e-sport, really, but done with columns and rows rather than pixellated guns or armoured monsters.
unique link to this extract

---

Pegasus spyware was used to hack reporters’ phones. I’m suing its creators • The Guardian

Nelson Rauda Zablah is a Salvadoran journalist whose work has been featured in multiple international papers:

»

When news of the hacking broke, a few sources jokingly answered our calls by greeting the good people who might be listening. But many more picked up the phone only to say we should stop calling them, and most simply didn’t respond at all. In one instance, a source told me that he now understood why his wife had been fired from her government position. I felt horrible. Guilty. Powerless.

That’s how Pegasus makes you feel above all: powerless. We believe the infections in El Faro happened through a “zero-click exploit”, meaning we didn’t even click on a phony link to open a door for the spies. They just broke in. Change your number, get a new device – they’ll just break in there, too.

And yet we refused to be powerless. We told our story to news outlets all over the world. In El Salvador, we held press conferences, went on TV and filed a case before the attorney general’s office. None of this brought any kind of accountability for the illegal spying. So, represented by the Knight First Amendment Institute at Columbia University, 14 of my colleagues at El Faro and I have decided to sue NSO Group.

I can assure you we’re not in this for the money: if we wanted to be rich, we wouldn’t be independent journalists. We’re doing this as a progression of our everyday work in El Salvador to expose official wrongdoing. We’re doing this in the United States because we’ve exhausted all legal avenues in El Salvador’s co-opted institutions.

And we’re doing this not just for us. In April, the Israeli newspaper Haaretz assembled a list of more than 450 law-abiding men and women around the world whose devices had been hacked by NSO Group’s Pegasus. Many of them are not in countries or positions where they can sue.

But someone has to. NSO executives shouldn’t be able to wash their hands as their tools are used to persecute journalists. In a very real sense, NSO set the hounds on us. And now we’re fighting back.

«

I’m reviewing a forthcoming book by the journalists who exposed the enormous scale of Pegasus surveillance, and a common reaction of those who discovered they were targeted is guilt – that they might have inadvertently harmed sources because they didn’t know their phone was a traitor.
unique link to this extract

---

Covid chaos at Foxconn iPhone plant causes 29% revenue fall • Financial Times

Kathrin Hille:

»

Foxconn reported NT$551bn (US$18bn) in revenue last month, down 29% from October and an 11% from a year earlier. It is the first time in 12 years that the company has announced a month-on-month fall in November, a time of high production to meet Christmas sales.

The company did not mention the iPhone, but said the drop in smart consumer electronics products — which includes smartphones — was because of “a portion of shipments being impacted by the epidemic in Zhengzhou”, where a coronavirus outbreak in Foxconn’s largest plant has led to weeks of disruption and a revolt by workers.

“At present, the overall epidemic situation has been brought under control, with November being the most affected period,” Foxconn said. “In addition to reallocating production capacity of different factories, we have also started to recruit new employees, and are gradually moving towards the direction of restoring production capacity to normal.”

One person close to the company said Foxconn’s internal goal was to return to “completely normal operations” after the new year at the factory in the Chinese city of Zhengzhou. It has been rocked by two waves of staff walkouts and violent unrest in recent weeks. Foxconn declined to comment on the target.

«

The expectation was that this would feed through to shortages of iPhones for this crucial quarter for Apple, yet delays on shipments have pulled back quite quickly. Sales in China have been dramatically affected, so maybe that’s yet to be felt in the US and other countries.

The ongoing problems from Covid, and the growing groundswell of civil unrest as GDP growth falls behind what’s needed to keep the middle classes happy, mean Apple’s search for more manufacturing outside China will become more urgent.
unique link to this extract

---

Temporary policy: ChatGPT is banned • Meta Stack Overflow

»

This is a temporary policy intended to slow down the influx of answers created with ChatGPT. What the final policy will be regarding the use of this and other similar tools is something that will need to be discussed with Stack Overflow staff and, quite likely, here on Meta Stack Overflow.

Overall, because the average rate of getting correct answers from ChatGPT is too low, the posting of answers created by ChatGPT is substantially harmful to the site and to users who are asking or looking for correct answers.

The primary problem is that while the answers which ChatGPT produces have a high rate of being incorrect, they typically look like they might be good and the answers are very easy to produce. There are also many people trying out ChatGPT to create answers, without the expertise or willingness to verify that the answer is correct prior to posting. Because such answers are so easy to produce, a large number of people are posting a lot of answers. The volume of these answers (thousands) and the fact that the answers often require a detailed read by someone with at least some subject matter expertise in order to determine that the answer is actually bad has effectively swamped our volunteer-based quality curation infrastructure.

«

Wonder how they’re going to police this, though. There will be “sanctions” on users who are “believed” to have done this, but how do you spot it, and how do you demonstrate it? Relying on users’ good behaviour only works for so long.

Notice too how we’ve abruptly reached this tipping point. There was no concern with GPT-3 a few months ago.
unique link to this extract

---

Building a virtual machine inside ChatGPT • Engraved

Jonas Degrave is a research scientist at Google’s Deepmind, here recounting some messing around done by a colleague at Deepmind:

»

Unless you have been living under a rock, you have heard of this new ChatGPT assistant made by OpenAI. You might be aware of its capabilities for solving IQ tests, tackling leetcode problems or to helping people write LateX. It is an amazing resource for people to retrieve all kinds of information and solve tedious tasks, like copy-writing!

Today, Frederic Besse told me that he managed to do something different. Did you know, that you can run a whole virtual machine inside of ChatGPT?

«

This goes from being clever to intriguing to extremely weird. Is ChatGPT talking to itself, or to a new instance of itself, or some alternative internet version of itself? The whole thing reminded me of the Philip K Dick short story The Electric Ant: how do we know what reality ChatGPT is serving up to us in its responses? What if the speedy responses it seems to be giving to Besse’s orders to list the directories in its virtual machine that he has “created” or “found” are just made up, rather than the result of actually querying a directory? Aargh. The implications of what Besse has done are far greater than they seem.
unique link to this extract

---

We’re in denial about the true cost of a Twitter implosion • WIRED

Eve Fairbanks:

»

I can’t imagine following the breaking-news events I’ve been able to witness virtually—the first days of Russia’s invasion of Ukraine, the invasion of the US Capitol—on another platform. It’s in these real once-in-history moments that Twitter comes alive. It doesn’t silo people into friend circles like Facebook or promote groupthink quite like Reddit. The barrier to entry for people who want to add to the story is lower than on TikTok or Instagram. You don’t need to angle for a photo or a video; you can tweet while hiding under a desk, or even—as Alexei Navalny does, hand-writing tweets he delivers to his lawyer—from prison.

People giddy to see Musk fall on his face might not fully know what role Twitter plays day to day in many other countries. We’ve heard about Twitter’s role in the Arab Spring but less about how the political life of, for instance, Zimbabwe—run by a repressive government that cracks down ruthlessly on physical protests and political speech—now takes place on Twitter. Twitter has become “our political meeting point,” says Tinashe Mushakavanhu, a Zimbabwean journalist. The app’s anonymity has allowed “a discourse about the country that is very free, very critical.”

Midnight is the hour to reconnoiter on Zimbabwean Twitter, Mushakavanhu says. That’s when cell phone data becomes cheaper; it’s also why Twitter is irreplaceable. Loading image- or video-heavy apps just uses way too much data for most Zimbabweans to afford.

A famous Zimbabwean novelist recently allegorized the internet, and Twitter in particular, as a parallel country. “You have the safety of anonymity if you so choose,” she explained to an interviewer. “That’s where most of the organizing [in Zimbabwe] now happens. Activists have made [strides] there that otherwise would not have been possible.”

In Zimbabwe, politicians are forced to respond to Twitter uproars. Twitter is also the place where people who have been forced to flee the country can, in a sense, return home. Waiting for asylum abroad, many of Zimbabwe’s thousands of political refugees “can’t work” legally, Mushakavanhu said. “These are people who can’t go home to bury their own parents.” So they become “very prolific on Twitter. The only thing they have is Twitter. It’s a space for fantasy and for articulating despair. It’s a home.” Mushakavanhu himself has moved to the United Kingdom. He told me, “There are parts of me”—the truly Zimbabwean parts—that now “only exist on Twitter.”

«

The early part of this feature will strike most people as being too inside-journalism, bemoaning the possibility of losing the site. But as the rest of this (long!) piece points out, you do need one central place where the world’s discussions happen. It would just be nice if there were fewer bots and thirsty grievance farmers.
unique link to this extract

---

Trolling as a Service in 2022 • Tisane Labs

Vadim Berman is Tisane Labs’s CEO and co-founder:

»

Today’s “TrollOps” involve a wide range of services, natural language processing, distributed processing, hybrid operation, crowdsourcing in a Mechanical Turk-like fashion, social engineering, and stealth.

Low-tech solutions combined with social engineering usually prove the most effective, both from the engagement perspective and the perspective of hosting expenses. A post saying “Fake news” in the comment section of Flipboard is a standard opening move, the King’s Pawn Game of trolling. Someone from the opposite political camp will likely be provoked. Does the article merely cite another article? Maybe it discusses a court decision that can’t be fake news? So what, a stupid remark is even more likely to generate a negative response — and that’s all they want!

How do they know where to post it? One way is to detect a combo of negative sentiment + mention of a particular political figure. Or, with the current partisan political environment, it could be assumed that a particular publication will be negative towards that figure most of the time or all the time, and skip the sentiment analysis altogether.

Today, several off-the-shelf platforms (sold at least to law enforcement agencies) allow managing fake persona, choreographing hardware fingerprint (SIM cards + devices) and generating text content GPT-style based on predefined profiles (e.g. radicals of a certain type). I had a chat with one of the vendors at a law enforcement trade event. I was told, “it’s a headache to manage even two sock puppet accounts. With our platform, you can manage tens of them, and generate content effortlessly”. My impression is that the TrollOps vendors still do not have access to these platforms, but it’s just a matter of time until they gain these capabilities.

Even today, the pricing is dangerously affordable. A cyber-bullying campaign in Indonesia will set you back 20K IDR (~US$1.33) on Instagram, 15K IDR (US$1) on Ask.fm, or 10K IDR (US$0.67) on Twitter / Facebook.

«

Or, you know, reply negatively to an Elon Musk tweet. That’ll get you a cyber-bullying campaign for nothing.
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