Dispatch 027
TikTok is in trouble. And leaning on what Donald Trump said on the campaign trail doesn't seem like a good strategy. Nor would it be for Trump to just stop the ban. But there's potentially a way to square some circles here if a few sides are willing to budge a bit. Billions – users and cash – often help with such budging. If Team Trump can convince China, or trade on trade with them...
M.G. Siegler
I Think...
🤳 It Sure Sounds Like Trump Would Be Okay with a TikTok Sale – Per my post above, Trump's answers in his sit down with Meet the Press sure sound like he would prefer a sale, and undoubtedly helping to broker one, rather than "saving" TikTok outright, as he implied on the campaign trail. Again, zero surprises here. And if the Supreme Court agrees to hear the case, it should buy the company some time to make that happen. There's no reason to think the Supreme Court would rule any differently than the lower courts here, but again, it's all about buying time to find a buyer (and convincing China to allow for a sale). [Verge]
⚛️ Quantum Computing Inches Closer to Reality After Another Google Breakthrough – Five years after Google's last major breakthrough in quantum computing – so-called 'quantum supremacy' which others disputed – they're back with another one. This is about an error-reduction breakthrough that will allow for such machines to be usable for real-world tasks rather than just theoretical experiments. But still not yet. For now, a new demonstration of possibilities:
Google said its quantum computer, based on a computer chip called Willow, needed less than five minutes to perform a mathematical calculation that one of the world’s most powerful supercomputers could not complete in 10 septillion years, a length of time that exceeds the age of the known universe.
I distinctly recall being a kid and reading about the potential for quantum computers in places like Wired magazine. That was... a long time ago. But it feels like we continue to creep closer to this reality. Still slowly, but less slowly each year. If/when this is cracked, one can only imagine what it means for all sorts of fields. Including, of course. AI. The HPC Wire post on the matter has more technical details, but the NYT one, unsurprisingly, was more approachable. [NYT]
🍎 The Inside Story of Apple Intelligence – To help prepare for his (great and meaty) Tim Cook interview, Steven Levy spoke to Craig Federighi, John Giannandrea, and Greg Joswiak to talk AI. No huge surprises here, though Federighi did admit that the release of GPT-3 in 2020 (ahead of ChatGPT itself two years later) clearly inspired Apple to move more forcefully here. That said, they reiterated that there's a reason they brought Giannandrea over from Google – notably giving him an SVP title – in 2018, which, fair enough. Especially given Levy's own look-back to a chat he had with Phil Schiller and Eddy Cue about AI, back in 2016. Beyond that, it's all product, privacy, and practicality for Apple Intelligence. And certainly not push towards AGI (see quote below). [Wired 🔒]
🏈 Netflix’s NFL Games Pose Advertising Test – In the wake of the Tyson/Paul boxing match glitches, some still seem worried about the ability of Netflix to handle mass live events at scale. Those hiccups may have been caused by a simple underestimation of how many would stream the event mixed with Netflix's insistence on only using their own tech (and CDNs) to deliver the feed (something Netflix disputes). While fewer people are likely to watch their NFL games, the insertion of advertising in real time (which the boxing match didn't have) will present a new challenge – especially since, again, Netflix is insisting on using their own, largely unproven in this manner, technology. Luckily, it won't have to be dynamic ad insertion for these games as these ads for all Netflix viewers – including, of course, paid subs – sold out long ago (DAI is something Netflix is working on for 2025 as they go live with wrestling, and undoubtedly other events). [Information 🔒]
📹 Sora is Here – Following up on the quick blurb just as it was going live yesterday, OpenAI's video creation tool is finally here. And it's... pretty fun. Yes, it's wonky and weird, but it's also pretty incredible what it can create from the most simple of prompts. I'm reminded a lot, unsurprisingly, of the release of DALL-E. We seem lightyears away from that – it was a whole two and a half years ago, which is basically forever in the blistering, chaotic speed of AI right now – and one can imagine these tools improving even faster. AI Wall or not, the offshoots of all the work done in the past few years remains wild to actually behold and play around with. This will undoubtedly trigger a whole new round of freaking out within Hollywood, but my bet remains that such fears will be wildly overblown and this will actually help the industry (and elevate human talent). Call it a hunch based on the history of every technology ever. [OpenAI]
I Wrote
With the release of Sora, a look back at a few thoughts around AI and Hollywood...
M.G. Siegler
M.G. Siegler
M.G. Siegler
I Link...
- Amazon is opening a new 'AGI SF Lab' to be run by co-founders of Adept and Covariant (two of the company's hackquisitions). Sounds like a sort of nebulous research lab perhaps to keep and attract talent in San Francisco. Also: agents, agents, agents – digital and real world. [TechCrunch]
- Now this is a great trailer – especially since they don't spoil what role Cillian Murphy, 23 years after his breakthrough in the first film (and now with a Best Actor Oscar in hand), will play in 28 Years Later. [THR]
- There's some joke to be made around "I'll be back" here given James Cameron's involvement, but regardless, it seems like Stability is yes, stable again. Better alert Thom Yorke and Julianne Moore! [Fortune]
- Comcast and Warner Bros Discovery have stopped suing each other over the upcoming Harry Potter TV series and will instead see Comcast gain the right to bundle Max (including with Sky in the UK). Might this also open the door to a potential Peacock + Max tie-up? [Variety]
- Surprise (no surprise)! Chris Evans will return to the MCU once again in the Avengers: Doomsday film. No word if he'll be Captain America (which would undercut the new one) or Johnny Storm (which would undercut the new one) or someone new, like Robert Downey Jr. is doing for all the money in the world (minus whatever they're paying Evans to come back). [THR]
- Meta keeps updating the software on the Quest to be better for pairing with a PC and improve hand-tracking in the back-and-forth battle with Vision Pro. [Verge]
- The 2026 MacBook Pro may not only switch to OLED screens (from mini-LED), but may get rid of the "notch" at the top of the screen and may instead have something like the "dynamic island" found on iPhones. [9to5Mac]
- Netflix got the most Golden Globe nominations in for both film (13) and television (23!). In so far as such nominations are a proxy for anything, Disney+ was a clear laggard with just one television nomination – though Hulu had 5 and FX had 10. But added together, that was still seven behind Netflix in TV alone. Apple TV+ got 9 nominations (all TV) and HBO/Max got 14. Add those together and you get a Netflix – just saying... [Variety]
I Quote...
"The idea that you’re scaling up these technologies to go to AGI is very naive."
-- John Giannandrea, in the sit down mentioned above to talk about Apple Intelligence. He also noted, "The most credible researchers in the field believe there are many unsolved problems and breakthroughs required."
Again, this is why I actually wouldn't be shocked if OpenAI declares AGI sooner rather than later. No one is ever going to agree on a firm definition. So declare victory, get the headlines, and move on the race for "superintelligence".
I Spy...
While these are mainly meant to be design mock-ups (found in GitHub) for what a subscription offering could literally look like from Bluesky (likely coming soon), clearly at least some of these options will make it into the offering. [via TechCrunch]
