Tuesday, August 27, 2024

Benedict's Newsletter: No. 555

NO. 555   FREE EDITION   TUE 27 AUG 2024
SPONSORED BY TOKEN2049
Join 20,000 attendees at TOKEN2049, the world's largest Web3 and blockchain event, on 18-19 September. Industry leaders and global cultural icons will converge in Asia's leading economic hub to unveil what lies ahead.

Secure your pass for a festival experience unlike any other: Use code BEN2049 at CHECKOUT now to get 15% off.

My Work

Competing in search

A quarter century after 'don't be evil' a judge has found that Google is abusing its monopoly in search. But no-one knows what happens next, and whether this ruling will change anything. Will Apple build a search engine? Will ChatGPT change search? Does it matter? LINK

Another podcast

Toni and I did a summer podcast on the Google ruling. LINK

In this issue: In this issue: The French arrest Telegram's CEO, Zuck-Ek's open letter, and seeing is believing

Read this issue as a web page. LINK

News

Arresting Telegram

The French government arrested Pavel Durov, CEO of Telegram (and a French citizen). We don't yet know the detail of the charges (which hasn't stopped an explosion of over-excited comment on social media), and we should note that the French legal system is much more arrest-first than Anglo-Saxon systems, but the press release indicates that this is about child abuse material (CSAM) on Telegram, and Telegram's alleged decision to ignore it.

Telegram has 900m MAUs, and talks a lot about being 'secure', but the main selling point is very large group chats and channels (with tens or hundreds of thousands of users), and these do NOT use end-to-end encryption: the traffic is encrypted between your device and Telegram's servers (so your government can't read it, unless it joins the group), but Telegram itself can read everything, if it chooses. Meanwhile, Telegram has a reputation for having huge amounts of CSAM and other seriously illegal activity in these groups and simply ignoring it.

To be clear, this activity happens on all networks where people can share files or images (even LinkedIn), but other companies either have large teams dealing with it (eg Meta, Google), or can't see most of it because it's end-to-end encrypted, or both. Telegram can see it and, apparently, doesn't appear to care and doesn't respond to law enforcement requests of any kind. We will see - and again, we don't actually know what the charges are yet, but they don't appear to be about 'free speech.'

However, given that this involves both France and Russia, there may be plenty more going on. After all, Russia effectively confiscated Durov's first company, the social network VKontakte. He has been in conflict with Putin over Telegram, but the Russian military uses it too, and so there is, at the least, some geopolitical maneuvering behind the scenes. So let's see what the charges are, and then see what they really are. BBC, PRESS RELEASE

EU versus AI, open source and startups 

Mark Zuckerberg and Daniel Ek published an editorial in The Economist arguing that the EU's regulation of AI in general and open source AI in particular is hasty, opaque, arbitrary, complex and squashes startups. All true. LINKPUBLIC

The week in AI

Right on schedule, Perplexity will start running ads (or trying to). Zawinski's Law states that every program expands until it can read mail - now all consumer internet services expand until they run ads. LINK

Today in 'CHATGPT! IS! NOT! A! SEARCH! ENGINE!' - a trailer for the new (and troubled) Coppola movie 'Megalopolis' used lots of historic quotes from critics panning his past masterpieces - most of which were not real. Someone asked ChatGPT for 'bad reviews of Coppola movies' and, as ever, it gave what looked like a good answer. LINK

Google acquihired the Character.ai team earlier this month, and now the CEO Noam Shazeer (a former Googler) will be co-head of Gemini. LINK

Amazon's CEO Andy Jassey on LLMs for coding productivity: "average time to upgrade an application to Java 17 plummeted from what's typically 50 developer-days to just a few hours." LINK

OpenAI is racking up old media licensing deals - this week Condé Nast. LINK

A fun problem in deploying LLMs for enterprise knowledge management - it might find data buried somewhere on the network that you aren't supposed to see, but that IT policies don't technically block, like, say, salary information. LINK

Software eats cars

Not a good week for GM as a 'tech' company: on one hand, it's being sued for selling driver data to insurers, and on the other, it laid off over 1k software developers. Tesla may still struggle to build cars, but car companies still struggle with software. LAWSUITLAYOFFS

About

What matters in tech? What's going on, what might it mean, and what will happen next?

I've spent 20 years analysing mobile, media and technology, and worked in equity research, strategy, consulting and venture capital. I'm now an independent analyst. Mostly, that means trying to work out what questions to ask.

Ideas

The new Google Pixel phones include a tool to let you add AI-generated components to your photos, with all of the light, reflections and perspective adjusted of almost perfectly. So, it's easier to make fake images. Some people in tech are very upset, others (like me) point out that this is another step on a continuum back to the 19th century - photos have always been 'fake'. See this week's column. LINKCONTEXT

A RAND paper comparing current attempts at AI governance with the mostly failed (and counterproductive) government attempts to regulate encryption software in the 1980s and 1990s. LINK

Apparently, Apple is trying to clamp down on payments within the WeChat super-app in China, demanding its standard 30%. (Apple has 20-30% of the Chinese smartphone installed base, depending on your estimates). Hard to take a view on this from outside the market, but this looks like a presumption of strength. LINK

How much trouble is Intel in? A lot. LINK

Procreate, a popular image creation app, published a manifesto saying it will never use generative AI features. Interesting as a reflection of a sentiment amongst some kinds of professional illustrators, though I presume the Procreate team is too young to remember when people said that using Photoshop or Illustrator was not creative. LINK

Chic-fil-A, a chain of QSR chicken restaurants, plans to launch its own streaming TV service. Cheep entertainment? LINK

An analysis of government IT's migration away from vast contracts with legacy systems integrators to more nimble cloud-based approaches. LINK

TIL: Lidl has a cloud computing business. LINK

South Korea is keen on AI textbooks - parents rather less so. LINK

Outside interests

A catalogue of 1990s Sony Minidisc visual design. People who like this kind of thing will like this. LINK

Actors can get rich from fan conventions. LINK

Apparently quite a lot of vintage book covers used poisonous dyes. Penitenzi agite! LINK

The life and death of the Flip camera. LINK

The Singapore apartment block built around a funicular elevator. LINK

Kamala Harris' plan to tax unrealized capital gains. I have some sympathy with the aim to chisel away a little of the US plutocracy, but the detail will be very important. LINK

Data

Which companies have all the Nvidia chips? LINK

Weekly robotaxi rides: Waymo: 100,000, Tesla: 0. LINK

A16Z compiled a list of the most popular consumer generative AI apps.  LINK

Useful UKCOM deck of UK online metrics. LINK

Nielsen's 'Gauge' of US TV viewing now gives data by media company as well as by channel - so Disney+ is much smaller than Netflix, but Disney overall has the largest US video audience, followed by YouTube, then NBCU, then Netflix. LINK

Half of US under-30s use TikTok for news and politics. LINK

Preview from the Premium edition

Seeing is believing

'The Commissar Vanishes' is a classic book from 1997 about the systematic falsification of images by Stalin's government. People became non-people, victims ceased to exist, and it was all done with airbrushes and razor blades in the dark room. 

In 1990, the New York Times wrote that the new 'computer imaging' would change the world, because it 'makes it easy to recompose and combine photographic images, and to do so in a way that is virtually undetectable', but at that point, photographs had been recomposed and combined for a very long time. Arthur Conan Doyle, after all, was fooled by a photograph of fairies created by two schoolgirls in the 1920s. 

People have been manipulating photographs for almost as long as photographs have existed, and what you see on the paper is never, exactly, 'just what was there.' Even if photographers don't alter the photo (whatever that means) after they press the button, they choose what to shoot and what not to shoot (there's an echo here of the childish idea that journalists should 'just report the facts' - which facts?) As Walker Evans said, the camera 'always' lies. 

Now generative AI gives us a new wave of ways to create 'fake' images...

 

THIS IS A PREVIEW FROM THE PREMIUM EDITION - PREMIUM SUBSCRIBERS GET THE COMPLETE COLUMN EVERY WEEK. YOU SHOULD UPGRADE.
Upgrade to Premium
You're getting the Free edition. Subscribers to the Premium edition got this two days ago on Sunday evening, together with an exclusive column, complete access to the archive of over 500 issues, and more.
 

No comments:

Post a Comment

Why the meetings keep coming back...

Even after you "fix" them. ...