UK markets close in 1 hour 21 minutes
  • FTSE 100

    -57.32 (-0.69%)
  • FTSE 250

    -223.38 (-1.08%)
  • AIM

    -6.81 (-0.84%)

    +0.0004 (+0.03%)

    -0.0025 (-0.20%)
  • Bitcoin GBP

    +55.29 (+0.10%)
  • CMC Crypto 200

    -26.49 (-1.79%)
  • S&P 500

    -36.20 (-0.68%)
  • DOW

    -305.37 (-0.79%)

    -0.10 (-0.13%)

    -10.70 (-0.45%)
  • NIKKEI 225

    -298.50 (-0.77%)

    -344.15 (-1.83%)
  • DAX

    -183.76 (-0.98%)
  • CAC 40

    -103.48 (-1.28%)

TechScape: No WhatsApp in China, no TikTok in the US, and the return of Llama

<span>WhatsApp will be removed from Apple’s App Store in China.</span><span>Photograph: Patrick Sison/AP</span>
WhatsApp will be removed from Apple’s App Store in China.Photograph: Patrick Sison/AP

Another day, another set of troubles for Apple’s App Store. This time, the company had bowed to orders from the Chinese state to remove WhatsApp and Threads, two of the last Meta apps still available in the country.

From our story:

Apple confirmed it had withdrawn the two apps – both owned by Meta, also the owner of Facebook – under instruction from the Cyberspace Administration of China, which regulates and censors China’s highly restricted internet and online content.

“The Cyberspace Administration of China ordered the removal of these apps from the China storefront based on their national security concerns,” Apple said in an emailed statement to Reuters. “We are obligated to follow the laws in the countries where we operate, even when we disagree.”


The reader, perhaps, is intended to infer a final clause in that statement, and conclude that Apple does indeed disagree with the order. It seems unlikely that the company was overjoyed to bow to high-profile censorship yet again, even if it did involve taking down the apps of one of the company’s most formidable competitors.

But it is worth noting what Apple hasn’t done this time: repeated its playbook with the European Union. The company hasn’t published a 12-page report detailing the changes it has been forced to make, and explaining at length why it thinks being forced to make them is likely to harm the experience for Apple customers and run counterproductive to the goals of the regulator. It hasn’t pursued a strategy of so-called “malicious compliance”, following the letter but not the spirit of the law. And it certainly hasn’t declared that its interpretation of the law differs from that of the regulators, and vowed to fight through the courts.

It may be unfair to expect it to, of course. Fighting the Chinese state in the Chinese courts would be an exercise in futility, to the extent that it would even be possible. Nonetheless, the difference is stark. When an authoritarian regime tells Apple what it can do with the App Store, the company’s response is a curt single paragraph. When a democratic union tries to do the same, the response is vociferous and negative.

The comparison does no favours to the European Union, to be sure. No defender of the Digital Markets Act would be too happy to stare Apple in the eye and demand to be treated like the Chinese Communist party. It’s an observation I feel compelled to make every time it comes up. There’s one world government that can command Apple with barely a squeak of public protest, and it’s not the one you’d want.

Time’s up for TikTok

China isn’t the only government that bans things. From the Guardian US:

The House of Representatives voted 360 to 58 on the updated divest-or-ban [TikTok] bill that could lead to the first time ever that the US government has passed a law to shut down an entire social media platform.

The Senate is expected to vote on the bill next week and Joe Biden has said he will sign the legislation.

I’m not a legislator but the bill itself doesn’t seem like great law. To avoid the appearance of passing a bill of attainder (a piece of legislation specifically targeting an individual, generally looked down upon in the English legal tradition from which US governance descends), it covers TikTok and any other service “controlled by” a “foreign adversary” deemed to be a security risk at the determination of the president. That’s quite a lot of scope to ban things at will, given the looseness apparent in the claim that something “controlled by” the state. “Foreign adversary” is more strictly defined by other legislation, and currently covers Iran, China, Russia and North Korea.

But that’s a debate for the lawyers. What interests me is: what will TikTok look like without Americans? The domination of the anglophone internet by the US is a reality. The “world wide web” isn’t a misnomer, and although there remain geographic differences in which services are popular where, as a general rule, American voices will dominate.

It seems unlikely that American TikTok will be wiped out by any ban – the number of compulsive users willing to hack their devices, use web apps or simply never uninstall the service means there will be always be posters stateside. But any substantial friction will probably lead to lots of less diehard TikTok users jumping ship to Instagram, Snapchat and YouTube Shorts, all of which have benefited from TikTok restrictions in other regions, particularly India.

In the short term, that can’t help but make the TikTok experience worse for all of its users, wherever they live. Some portion of the content they would have wanted to watch will no longer be there, and much of the rest will be in reposts, arriving late and shorn of any connection with their original creators.

What I’m curious about, though, is the mid- to long-term outcome of such a change. Does the rest of the anglophone world steadily drop off as well, following the American users to the places they’ve started posting? Or is there a genuine divergence, as the culture on TikTok starts to reflect a fundamentally different cross-section of the world from the internet at large?

I still think the most likely outcome is that we never find out, and some combination of money, lobbying and protest means that an arrangement is found that maintains access to TikTok. But if it all fails, what a fascinating experiment we’ll get to witness.

Llama 3 on the loose

The second summer of AI has begun. The smallest versions of Facebook’s large language model Llama 3 have been released, and they have changed the landscape for what is unlikely to be the last time this year. From our story:

The social media giant equipped Llama 3 with new computer coding capabilities and fed it images as well as text this time, though for now the model will output only text, Chris Cox, Meta’s chief product officer, said in an interview.

More advanced reasoning, like the ability to craft longer multi-step plans, will follow in subsequent versions, he added. Versions planned for release in the coming months will also be capable of “multimodality”, meaning they can generate both text and images.

Llama is Facebook’s homegrown competitor to GPT, Gemini and Claude. Unlike those three systems, though, Llama has been released on a comparatively open licence, with the core models available for users to download. If you want to build your own AI system – particularly if you want to run it locally, rather than relying on a server farm – Llama is a good place to start.

The first version of Llama was accidentally-on-purpose released to the public, while the second made it official. The upside for Facebook is obvious: if you make the core technology on which the AI boom rests, you have an awful lot of power. Llama is open enough to build on top of, but not open enough to fully fork and take away from Facebook – crucially, without the data and specific details of how the system was created, you can never re-do that initial training run, even if you could afford the vast expense in compute.

But the downside is equally clear. Facebook makes money on people using its services, but can’t directly earn revenue from training Llama. That means that it’s not incentivised to invest quite as much in staying at the absolute frontier as its competitors, and as a result, Llama has historically been far from the cutting edge.

For the time being, however, that’s not the case. It took more than a year, but a freely licensed model from Facebook is now at least competitive with GPT4, and the company says an even better – or, bigger, at least – version is still to come.

Of course, the frontier doesn’t stand still for long. OpenAI is expected to release GPT-5 at some point this summer, and, if it represents the progression insiders have been hinting at, everything will change – again.

If you want to read the complete version of the newsletter please subscribe to receive TechScape in your inbox every Tuesday.