Start Up No.2014: music faces up to the AI tsunami, the trouble with happy mobile users, Silicon Valley’s empty offices, and more

The innocuous-looking bacterium A.baumanii is a superbug that kills about a million people a year. Now AI may have found an antibiotic that can beat it. CC-licensed photo via US CDC 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. It’s about the US Surgeon-General’s warning on kids and social media.

A selection of 9 links for you. Craft little sods. I’m @charlesarthur on Twitter. On Mastodon: Observations and links welcome.

New superbug-killing antibiotic discovered using AI • BBC News

James Gallagher:


Scientists have used artificial intelligence (AI) to discover a new antibiotic that can kill a deadly species of superbug.

The AI helped narrow down thousands of potential chemicals to a handful that could be tested in the laboratory. The result was a potent, experimental antibiotic called abaucin, which will need further tests before being used. The researchers in Canada and the US say AI has the power to massively accelerate the discovery of new drugs. It is the latest example of how the tools of artificial intelligence can be a revolutionary force in science and medicine.

More than a million people a year are estimated to die from infections that resist treatment with antibiotics. The researchers focused on one of the most problematic species of bacteria – Acinetobacter baumannii, which can infect wounds and cause pneumonia. You may not have heard of it, but it is one of the three superbugs the World Health Organization has identified as a “critical” threat.

It is often able to shrug off multiple antibiotics and is a problem in hospitals and care homes, where it can survive on surfaces and medical equipment. Dr Jonathan Stokes, from McMaster University, describes the bug as “public enemy number one” as it’s “really common” to find cases where it is “resistant to nearly every antibiotic”.

To find a new antibiotic, the researchers first had to train the AI. They took thousands of drugs where the precise chemical structure was known, and manually tested them on Acinetobacter baumannii to see which could slow it down or kill it.

This information was fed into the AI so it could learn the chemical features of drugs that could attack the problematic bacterium. The AI was then unleashed on a list of 6,680 compounds whose effectiveness was unknown. The results – published in Nature Chemical Biology – showed it took the AI an hour and a half to produce a shortlist.

The researchers tested 240 in the laboratory, and found nine potential antibiotics. One of them was the incredibly potent antibiotic abaucin. Laboratory experiments showed it could treat infected wounds in mice and was able to kill A. baumannii samples from patients. However, Dr Stokes told me: “This is when the work starts.”


The world really needs new antibiotics. If this is the only thing AI does (and succeeds), it’ll have earned its keep.
unique link to this extract

Layoffs push down scores on Glassdoor; this is how companies respond • The Pragmatic Engineer

Gergely Orosz:


I got a message from a software engineer working at a company which laid off 30% of staff in December 2022. It’s a late-stage startup valued at around $3B which had around 1,000 employees before the layoffs. The engineer wrote:


“My company is removing Glassdoor reviews because their rating has gotten so low. The company’s score went to 2.3 and they started doing this. I don’t think my company is alone in this practice to protect themselves from bad press, but lots of my colleagues have had their reviews deleted. Effectively, we’ve been silenced.”


I managed to talk to someone in this company’s HR department, who confirmed that the leadership set a goal to improve the business’s Glassdoor rating. The HR team’s target was to get the score above 3.0. And so, they got to work flagging negative reviews for removal, and encouraging staff to post 5-star reviews to balance out negative reviews. Turns out, this company is not alone in doing so.

In today’s issue, we’ll look closely at what is happening, and also investigate a specific company — cybersecurity company Trustwave — to find out what happened so the company reached an all-time high Glassdoor rating


Turns out there’s a fair amount of borderline legal methods on the part of the companies, while Glassdoor does offer paying companies an incentive. A good investigation.
unique link to this extract

Office vacancies rise in Silicon Valley and balloon in San Francisco • Mercury News

George Avalos:


Silicon Valley’s office vacancy rate increased to 23.1% in the first quarter of 2023, a level that [estate agent] Savills described as a “new historical high”, and up from 22.7% in the final three months of 2022, the company stated in the report.

San Francisco’s office vacancy level rocketed to 32.7%, “a new all-time high,” in the 2023 first quarter, up from 32.1% in the fourth quarter of 2022, Savills reported.

…“We expect office availability (in San Francisco) to continue to increase in 2023 as the slowdown in the technology sector persists,” Savills said in the report.

…“Office space demand (in Silicon Valley) has been down significantly as the technology sector continues to undergo a serious correction with mass layoffs and a general freeze in office leasing,” Savills reported.

Both San Francisco and Silicon Valley face a grim rest of 2023, Savills suggested in its new assessment.

“With economic uncertainty, slow return-to-office utilization, and an ongoing correction in the technology sector, it is no surprise that the San Francisco office market has gone from having the lowest availability levels in the country pre-pandemic to having the highest availability levels in just over three years,” Savills stated in its report.

San Francisco’s soaring vacancy levels, which Savills terms availability, mean that loans for big office buildings in that city could tumble into default — or worse, into foreclosures and property seizures. “With worsening underlying market fundamentals and looming loan maturities, expect more (San Francisco) office property distress to occur in 2023 as many owners find themselves underwater,” Savills stated. A commercial real estate site would be considered “underwater” if its total loan debt exceeds the actual value of the building.


Second-order effects of Covid plus the internet are quite dramatic.
unique link to this extract

Twitter is making researchers delete data it gave them unless they pay $42,000 • the i

Chris Stokel-Walker:


Academic researchers have been set a deadline of the end of the month to delete data they obtained under historic contracts to study Twitter, unless the pay a new $42,000-a-month contract – a demand one called “the big data equivalent of book burning”.

For years, Twitter provided academic access to a service called the decahose – a random sample of 10% of all Twitter’s firehose of tweets, which was always on. The decahose, access to which was brokered through Twitter’s API (application programming interface), was a special tool for academics, designed to let them monitor how conversations on the social media platform took place.

Researchers have used that data to track entire days on Twitter, to analyse the spread of disinformation and misinformation, and to track the rise of extremism and how that bleeds through to offline life.

What happens on Twitter matters because, in Elon Musk’s own words as he planned to take over the company last year, “Twitter serves as the de facto public town square”.

But in recent weeks, the company has been contacting researchers, asking them to pay $42,000 a month to access 0.3% of all the tweets posted to the platform – something researchers have previously said is totally unaffordable. Previous contracts for access to the data were set as low as a couple of hundred dollars a month.

An email, seen by the i, says researchers who don’t sign the new contract “will need to expunge all Twitter data stored and cached in your systems”. Researchers will be required to post screenshots “that showcase evidence of removal”. They have been given 30 days after their agreement expires to complete the process.


Going by the numbers in the story, if only 1 in 200 researchers does agree to sign up then Twitter’s getting the same amount of money. Though the researchers are getting rather less for their money.
unique link to this extract

Top solar firm warns excess capacity risks wave of failures • Bloomberg via Caixin Global


China’s world-leading solar industry could face a wave of bankruptcies if the current aggressive expansion of manufacturing capacity continues, according to the sector’s biggest player. 

More than half of China’s solar manufacturers could be forced out in the next two to three years because of excess capacity, Li Zhenguo, president of Longi Green Energy Technology Co., said during an interview Wednesday on the sidelines of the SNEC PV Power Expo in Shanghai. 

“Those that will be hurt first will be those that are not prepared sufficiently,” he said. Companies with weaker finances and less-advanced technology are most at risk, according to Li. 

The global solar market is growing rapidly, with installations expected to rise 36% this year to 344 gigawatts, according to BloombergNEF. But factories are expanding even faster. One step in the supply chain alone — producing the polysilicon that goes into the panels — will see capacity rise enough to produce 600 gigawatts this year, BloombergNEF analyst Jenny Chase said in a presentation at SNEC earlier this week.

“There will be a price crash, it will hurt, and there will probably be bankruptcies across the industry,” she said.


Puzzled by why growing manufacturing capacity would be a problem when demand is also growing.
unique link to this extract

Legit app in Google Play turns malicious and sends mic recordings every 15 minutes • Ars Technica

Dan Goodin:


An app that had more than 50,000 downloads from Google Play surreptitiously recorded nearby audio every 15 minutes and sent it to the app developer, a researcher from security firm ESET said.

The app, titled iRecorder Screen Recorder, started life on Google Play in September 2021 as a benign app that allowed users to record the screens of their Android devices, ESET researcher Lukas Stefanko said in a post published on Tuesday. Eleven months later, the legitimate app was updated to add entirely new functionality. It included the ability to remotely turn on the device mic and record sound, connect to an attacker-controlled server, and upload the audio and other sensitive files that were stored on the device.

The secret espionage functions were implemented using code from AhMyth, an open source RAT (remote access Trojan) that has been incorporated into several other Android apps in recent years. Once the RAT was added to iRecorder, all users of the previously benign app received updates that allowed their phones to record nearby audio and send it to a developer-designated server through an encrypted channel. As time went on, code taken from AhMyth was heavily modified, an indication that the developer became more adept with the open source RAT. ESET named the newly modified RAT in iRecorder AhRat.

Stefanko installed the app repeatedly on devices in his lab, and each time, the result was the same: the app received an instruction to record one minute of audio and send it to the attacker’s command-and-control server, also known colloquially in security circles as a C&C or C2. Going forward, the app would receive the same instruction every 15 minutes indefinitely.


Puzzling: why would you do this in such a random way? To prove something? For laughs?
unique link to this extract

AI will unlock creation rather than consumption • Midia Research

Mark Mulligan:


Prior to the establishment of the recorded music business, music was a participatory experience. Whether that be a 19th century family gathering around a piano on a Sunday, mediaeval peasants singing along with a travelling bard, or the majority of 16th–18th century European populations singing hymns in church. Recorded music used quality to build walls between listeners and performers. The vast majority of people could never expect to sound as good as a piece of recorded music. However, the trend started to reverse with the introduction of music production software and sample culture, re-democratising the means of production, while streaming and social media combined to democratise the means of digital distribution.

In the 2020s, these technologies have accelerated scale and capability, supported by the proliferation of online learning (e.g., Masterclass) and skills sharing platforms (e.g., Fiverr), making it easier than ever for aspiring music creators to release good quality music. In 2022, the number of artists direct (i.e., self-releasing artists) reached 6.4 million, a 16.8% increase from 2021. While the music creator economy continues to grow; the even more transformative potential lies in the consumerisation of these technologies – much like Teflon making its way from NASA spaceships to kitchen pans.


But also..
unique link to this extract

AI’s disruptive forces are rapidly reshaping the music industry • Financial Times

Anna Nicolaou:


Lucian Grainge, chief executive of Universal Music, has been sounding the alarm. “Unchecked generative AI poses many dangers,” he told investors last month. Universal Music recently sent a letter to all the leading streaming platforms warning them against allowing AI technology to train itself on copyrighted music, the Financial Times reported last month.

There are a few reasons for such concerns. The first one is obvious: copyright infringement. An AI-generated fake Drake can only sound like the star because it learned to do so by listening to Drake. So the music companies argue Drake should receive some of the money these songs earn. Some musicians, such as Grimes, though, are happy to opt in and allow their voices to be duplicated, while splitting the royalty income 50/50. The copyright issue could take time to sort out, but eventually music companies and other stakeholders will create a framework for how to license music used by AI generators.

But there is another reason why Universal is worried. The market share of major-label music on streaming platforms has been declining, slowly but steadily. In 2017, the four biggest suppliers accounted for 87% of all listening on Spotify. By 2022, that had shrunk to 75%.

Listening is increasingly being diverted towards music from independent artists, as well as ambient tracks and AI-generated songs. Grainge has spent the past few months talking about an “oversupply” of content on Spotify, where 100,000 new tracks are being added every day. He says AI has been a leading contributor to this.

The big music companies care because they earn billions of dollars of royalty income that is directly tied to their proportion of streams. But this shift is also fundamentally changing what Spotify is, and raises big questions about how we will consume music in the future.

For a long time, Spotify had compared itself with Netflix. It was the place where you could pay a monthly subscription fee for access to a large catalogue of professionally produced music. But Spotify is turning into more of a combination of Netflix and YouTube — a platform where you can listen to megastars, but also 30-second clips of rainfall that can be created in seconds by anyone with access to a computer.


This is – you can see – rapidly going to shift from music that artists make to soundalike music that you maybe create on your own computer. And perhaps you go to see the live artist because it’s interesting to see it actually created physically, and be in a room with other people. But getting paid through music streaming services.. might be done.
unique link to this extract

Happy UK mobile users need educating – report • Mobile Europe

Nick Booth:


There’s bad news and worse news from a new study of European mobile operator customers. The bad omen is that users are happy with their phone services. The grimmer prognosis is that under these circumstances, they’re not upgrading to 5G, a study of eight European countries has found. Unless these feelings of fulfilment and content delusion can be changed mobile network operators could struggle to monetise 5G, the report said. Among the recommendations are to change user’s perceptions, entice users to watch high video at a premium and to adopt sustainability as a marketing tool.

…In the UK, the report authors found, of the 2,608 mobile users surveyed, three quarters (74%) of mobile customers are ‘satisfied’ with their mobile network. This is one of the highest in Europe and less than a third (31%) are using 5G on their smartphone. Oddly, only 52% of them that had the option of 5G access knew of a ‘discernible improvement in performance’ compared to 4G. Across Europe, whilst 84% of customers surveyed by BearingPoint for its Connectivity Challenge Study were aware of 5G, they do not truly understand its potential benefits and are concerned about network quality, the consultancy said. “As such, the study says that more needs to be done by the operators in educating consumers on the benefits of 5G and creating compelling services,” it concluded.

Taking the UK figures as a case in point, John Ward, UK CME Director of BearingPoint, explained why satisfied customers need educating out of what seems like blissful ignorance. “Awareness in the UK and across Europe is still low to the fact that 5G offers higher network bandwidth, better latency and higher reliability, and also provides the network technology with the highest data security and the best energy efficiency among the mobile access technologies,” said Ward.


“Terrible news, boss. People are completely satisfied with the service we’re providing them.”
“Damn. We need to persuade them to pay money to watch video they can probably already watch!”
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

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.