Banning TikTok The US bill to force Bytedance to sell TikTok or shut it down sailed through the US legislature and was signed by the president (strictly speaking, this law enables him to designate any service for this process). There is now a lot of debate about whether this law breaks the US constitution's clause on freedom of speech, and Bytedance will fight - I leave that argument to experts on the US constitution. Meanwhile, scholars of the practice of regulation will contrast this approach with the EU's DMA, DSA & GDPR, and the UK's Ofcom, and note that for all the noise, this is so far the only tech law that the US has actually managed to pass. As I've written over time, the weak concern about TikTok is privacy, and the strong one is that it's a major media company: this week Rutgers University released a study suggesting that stories on themes that China doesn't like are much less prevalent on TikTok than on Instagram. But, what to do about that? THE LAW, FREE SPEECH, STUDY Generative AI versus services The head of Tata's IT BPO unit says that chatbots might result in 'minimal' need for call centres in as soon as a year. Automation always starts with boring, repetitive tasks that we can all see. LINK Meanwhile, BCG says that 20% of its revenue this year will come from advising its clients on how to implement generative AI. Every company needs an AI strategy, or thinks it does - that might have generated more revenue last year than any actual generative AI product. LINK The US does an AI safety board The US appointed 20 people to a board to advise on the impact of generative AI on infrastructure, and prepare for 'AI-related disruptions' (whatever that's supposed to mean). It's an odd list of people - some giant tech CEOs (Google, Microsoft) who won't have much time to contribute, plus CEOs from airlines, pipeline companies and a civil rights group. A very DC story, I suspect. LINK The week in AI Cognition is six months old and apparently raised $175m at $2bn post-money: it's behind the 'Devin' generative AI coding assistant that went viral recently with an amazing demo (the reality of which has attracted some skepticism). LINK A US school teacher has been charged with using generative AI tools to created fake audio of his boss making racist comments. On one hand, they'll be a lot more of this; on the other, it will now be very easy for people to claim that any incriminating recording is faked. LINK Small models, from Microsoft and Apple I wrote a column last week pointing to LLMs small enough to run on a phone as one of the outstanding issues this year: this week Microsoft announced a new model aimed at just that, Phi-3, and Apple built on AI papers it's been quietly releasing all year to show OpenELM, an 'Efficient Language Model' family also aimed at running on, well, Apple devices. (Meanwhile, Apple has apparently been both negotiating with OpenAI and Google for cloud models, and working on its own cloud AI chips.) These models aren't as good as the giant ones that run in the cloud (at vast expense), but a general-purpose chatbot isn't the only use-case. MICROSOFT NEWS, MICROSOFT PAPER, APPLE OPENELM The cookies apocalypse is postponed, again Google delayed the phase-out of third-party cookies in Chrome, for the third time. I'm starting to suspect that this will happen at the same time as autonomous cars and AGI - it's a couple of years away, and always will be. More seriously, it's fascinating that the timeline for this is being set entirely by the UK's competition agency, the CMA, not the EU or FTC. ANNOUNCEMENT, COVERAGE Meta goes for the VR platform play As widely signalled and expected, Meta announced that it will provide its VR/AR software stack to third parties, calling it 'Horizon OS'. The announcement says 'open' a lot, and this is clearly a deliberate contrast to Apple's Vision Pro approach (which is open to developers but not OEMs). However, it's not open source, so this is open as in Windows, not open as in Linux or Android (though Horizon itself is based on the open source version of Android). This is the obvious thing to do. But nothing in this space has much consumer traction yet, so OEMs will dip their toes again but aren't crying out for a platform the way phone OEMs needed something like Android. And, of course, Meta continues to sell the Quest at or near cost, which makes it hard for anyone else to complete. LINK Apple TV does sports? The TV boom is over: scripted show productions were down by a quarter last year as media companies stopped trying to buy customers for their D2C plays and started worrying about profits. But sport hasn't unlocked in the same way, partly because supply is finite. Now FIFA is trying to create a new football championship as part of its push to break into the USA (NB: this is the 'football' that most of the world watches, not the 'football' that Americans watch) and might do a single global deal with Apple. LINK There goes Binance The US is asking for a three year sentence for Binance's Changpeng 'CZ' Zhao, after he pled guilty to ignoring money-laundering rules. I remember explaining to crypto people a few years ago that doing X 'but with crypto' didn't mean you could ignore laws about X, and neither did disagreeing with the principles behind those laws (a judge doesn't care if you think 'money is speech'). LINK Regulating Shein? The EU decided to designate Shein as a 'Very Large Online Platform' (VLOP'), under the DSA, which means it's subject to the same kinds of rules as Meta or Google. Shein is technically a marketplace (aggregating thousands of suppliers), and has over 45m users in the EU, so the internal logic does make sense… though this also points to the challenge of writing rules that apply both to a social network and a clothing retailer. LINK Software eats cars? The NHTSA, the US car safety regulator, is slowly increasing pressure on Tesla for selling 'full self-driving' that is not full self-driving and sometimes gets people into accidents. LINK, STUDY Meanwhile, VW, after a disastrous attempt to write its own EV software platform, is partnering with Xpeng in China, though only for vehicles to be sold in China. There is some classic disruption theory here: Tesla is struggling to match the incumbents at making cars, but the incumbents are struggling to make software. LINK, CONTEXT | 
No comments:
Post a Comment