|
The Spyglass Dispatch is a newsletter sent on weekdays featuring links and commentary on timely topics found around the web. Feel free to forward it to others who can sign up and view previous dispatches here. You can also view this particular dispatch on the web here.
And we're back. Happy 2025. A lot has happened in the 17 days off – well, as much as parenting two young children can be considered "days off" – some of which I got to in a written-form throughout the break. Today I'll call out a post I wrote just yesterday on the recent news that Tim Cook would also be donating $1M to the Donald Trump inauguration fund. Not Apple, mind you. No, not Apple at all. This is just Cook continuing the "great American tradition" that neither he nor Apple have ever participated in before. Funny that. Certainly, there's "unity" this year – around cutting $1M checks. But it was all clearly just a coincidence with everyone in the giving mood for the holidays.
I Think...🤑 Sam Altman on ChatGPT’s First Two Years, Elon Musk and AI Under Trump – A long, though not particularly wide-ranging interview. If you've followed OpenAI closely, most of the information will already be known to you, but Altman's first-person perspective on things like the initial meeting with Ilya Sutskever and yes, getting fired by the board (including Sutskever), are interesting. Sadly, no color on that relationship today as Sutskever works on 'OpenAI Original' – er, Safe Superintelligence. (Or others who have left to directly compete – many of which are now at Anthropic.) Speaking of, 'superintelligence' is now clearly going to be Altman's main talking point going forward as multiple time he all-but-declares that AGI will be here in 2025 – something you could easily see coming. (Though he also notes the term has become "sloppy" – almost meaningless given the always-shifting goalposts.) He basically sidesteps the $1M check for Trump question – "I support any president." is the Rob Lowe NFL hat of statements on this matter. But he redeems that non-statement with the following on Elon Musk: "Oh, I think he’ll do all sorts of bad shit." He just doesn't think – or doesn't want to say he thinks – that he'll do them to OpenAI. That may be right, but bad things can happen based on things not done... [Bloomberg 🔒] 💡 Sam Altman 'Reflections' Coda – While he notes the Bloomberg profile (above) prompted his post, this post also reads as a more humanized version of OpenAI's own blog post buried under the holiday snow as to why they need to shift into a for-profit company. It's almost as if Altman feels like he didn't get quite everything off his chest in the interview, so he authored his own coda. And while he doesn't address the for-profit shift directly, he makes the same general points in terms of why the company now views it as a necessary change: that the company and overall AI space is in a very different position now than it was a decade ago, and they couldn't have known how much capital they would need to scale and how much competition they would face in doing so. In his view, OpenAI is essentially a brand new company since ChatGPT launched two years ago – and yes, he again directly addresses his own firing a year ago, which seemingly shows that he believes such chaos is now well behind them (while noting that the continued employee attrition is a natural outgrowth of such chaotic growth). Buried towards the bottom is another fun little bit where Altman sort of casually again suggests that AGI may first arrive in 2025 in the form of agents (no surprise there), and that the company is now moving a lot of focus to "superintelligence" – AI far beyond human capabilities (also no surprise). Lastly the effusive praise given to Ron Conway and Brian Chesky for their help during the "troubles" is mainly noteworthy in who is not thanked: Satya Nadella. [Sam Altman's Blog] ⭐️ 2025 Golden Globes Analysis – While awards shows remain mainly silly, they can be interesting from a sort of Zeitgeist-perspective – including pointing to the directions in which Hollywood is headed. The Brutalist (see: quote down below) and Emilia Pérez were the clear winners, with A24 and Netflix raking in awards, respectively. The latter won a bidding war to distribute Pérez and yes, gave it a (very) limited release in theaters before hitting streaming. Normally, such films head back to theaters after wins and leading up to the Oscars in March – but that presumably won't be the case here! Unsurprisingly, FX's Shogun was the big winner on the TV side. Overall, the Globes themselves seem to be in a much better state after being revamped (after almost getting scrapped entirely). Nikki Glaser seemed to be a good choice for host (solid opening monologue) and Demi Moore seemingly had the best speech of the night after winning an award for the first time in her 45 year (!) career. [THR] 📺 LG and Samsung are Adding Microsoft’s Copilot AI Assistant to Their TVs – For the 85th year in a row, the theme of CES seems to be TVs. That is, what new technology can they cram into them? This year – shocker – it's AI. It's not yet clear how useful Copilot will actually be on these devices – because the technology clearly doesn't yet work on any of them, thanks guys – but if it's anything like the previous technology added, it will be mediocre at best and annoying at worst. Remember the era when you could turn on a TV without it needed to do 15 minutes of software downloads and firmware updates to operate? A TV has literally one job, but all of these companies keep crufting them up to make that job harder in the name of innovation (and not getting commodified, which they largely have been). It's enough to make you wish Apple would make a TV set... How about a 60" iPad? But even they would try to make it all about Siri and AI now. [Verge] 🚕 How Uber and Lyft Are Gearing Up for the Robotaxi Revolution – The shifting dynamics here are all kinds of interesting. Waymo is clearly doing very well on their own in SF, but it will cost tens of billions of dollars to scale to other cities around the world and if they can outsource some of those costs – namely in fleet management – Uber and Lyft may have found an interesting position in the market without having to build the self-driving cars themselves. Uber in particular has aggregated far more customers in many countries so they should be well-positioned to partner not just with Waymo, but many other players around the world. And, of course, as good as self-driving technology may be now, it will be quite some time before it's able to reliably operate in different types of cities in different climates. Again, Uber has an advantage here as they can "fall back" to their network of drivers (which will undoubtedly also be cheaper until self-driving cars are at truly massive scale). The wildcard in all this is, of course, Tesla. But the timetable is also key there. How many cities will they actually be operating their 'Cybercab' service in within 5 years? Even 10 years? Controlling the whole stack, as it were, has its advantages, of course, but scaling quickly and reliably doesn't seem like one of them here... Do they eventually buy Lyft? [WSJ 🔒]
I Wrote...I also slogged my way through Dune: Prophecy and wrote up some thoughts on season one which, at least, was mercifully short.
I Note...- Nick Clegg's departure from Meta isn't complicated: he was the political force needed after the Cambridge Analytica fiasco, and now his Republican deputy, Joel Kaplan, is the politician Meta needs as they navigate the second age of Trump. [Bloomberg 🔒]
- Jon M. Chu used the Vision Pro to edit Wicked, which seems like it could be a compelling use-case (from the perspective angle, quite literally) – albeit a more enterprise-y one and fairly niche one. I do wonder just how much he used the device for this purpose simply given how laborious it is to use versus just, say, doing it on a computer. The answer may come if other filmmakers take it up for editing. As a consumer, I'm more interested in the fact that Wicked should look great when watched on the device since it was shot in 3D, which is a great format for the Vision Pro. [THR]
- For years, back when I used a PC, I was a big fan of Microsoft's keyboards and mice. Then they stopped making them – instead they're now outsourcing the peripherals to Incase, using Microsoft's designs, components, and supply chains. Interesting trend to outsource such work, such as Apple is less explicitly doing as well, with the Vision Pro, for example. [Verge]
- Despite the ChatGPT Pro plan being $200/month, OpenAI is still losing money on the service, Sam Altman says. Good news: that seemingly implies that those users are using it far more than anticipated. Bad news: that also means it wasn't an item they expected to add to their overall losses – thus adding to their overall capital needed to keep scaling... [TechCrunch]
- What are we to read into Microsoft's pause on their $3.3B "supercomputer" AI cluster being built in Wisconsin for OpenAI? The "scope and recent changes in technology" statement could be read a few ways, but seemingly none of them are particularly good for OpenAI, which needs more compute and facilities to compete with xAI and others. [Information 🔒]
- Especially strange given that Microsoft just announced that they would spend over $80B on AI data centers in (fiscal) 2025 which, of course, implies everything is full-steam ahead. [CNBC]
- After seemingly starting strong out of the gate, Tesla's Cybertruck sales seem cyberstuck. After touting 1 million trucks reserved via the comically cheap $100 deposit, estimates are that just 35,000 - 50,000 Cybertrucks were sold last year. Could it be that featuring a design aesthetic that seemingly appeals largely to 12-year-old boys has limits? [Verge]
- Meanwhile, the anti-Tesla, as it were, Rivian, delivered just over 50,000 of their vehicles in 2024 – which was up over 3x from 2023 (though still below estimates due to supply constraints). But while Tesla (as the market leader) will seemingly be fine with the EV credits presumably going away under Trump, Rivian sales could be hurt by such a repeal in 2025. That VW money could be more important than ever... [TechCrunch]
- NVIDIA crossed $1B in AI investments in 2024 with 50 deals including xAI, Cohere, Mistral and Perplexity. Just when you thought the industry couldn't get any more circular... [FT 🔒]
I Quote...“No one was asking for a three and a half-hour film about a midcentury designer on 70-millimeter. But it works.”
-- Brady Corbet, the director of The Brutalist, in his acceptance speech after the film took home the top prize at the Golden Globes last night (Corbet also took home Best Director).
|
