Inklings #018 📧
As the AI frontier models keep improving unabated, a funny thing is happening to the user base. That is, different segments are increasingly looking elsewhere as costs rise and capabilities become more clear. Apple's hybrid approach with Siri AI between on-device and cloud is clearly something we're going to see more of...
M.G. Siegler
Thoughts On...
😈 Anthropic Walks Back Their Sabotage Strategy – Because we can't go a week without an AI strategy and comms crisis, after some major pushback, Anthropic has decided that, you know what, maybe they will tell those trying to use their new Fable model for their own training purposes that they can't do that – rather than just quietly undermining their code! Look, we get that they don't want rivals (and potentially bad actors) trying to distill their work, but who thought this strategy would work for a company that is trying to grab the ethical high ground and help with AI's trust issues?! Now that Anthropic is seen as the AI front-runner, we should and they should expect even more scrutiny over some of their um, interesting strategies and comms blunders. Upside: I now have some Beastie Boys stuck in my head. And now you do too. [Wired 🔒]
🗣️ Siri AI Actually Seems To Work – The early reviews are in! Well, previews. This is just the first of many betas before a roll out – which will still be in beta – in the Fall. But these first impressions seem pretty universally positive across the board. Clearly, the bar was low – perhaps so low so as to be impossible to go under. Most takes seem to be "well, this is simple" or "yeah, other systems have been able to do this for a while" but still... it works! I have Siri AI on two systems currently, a backup MacBook and an old iPad. It's a bit buggy (again, as you'd expect) and slightly slow on the iPad which is several years old. But still, it works! You just couldn't say that really about Siri, and not just in the past couple of years with Apple Intelligence debacle, but really since Siri first launched 15 years ago. There will obviously be times still where she doesn't work or misfires – and we'll see how she scales – but the hits seem to happen far more often already. It's early, but I've been impressed. If anything, she might be too cold? [Verge]
💎 Elon's Diamond Hands – A trillion dollars. It's an insane amount of money, obviously. But as with Hollywood box office results, there's some caveats and context required. First and foremost, Bill Gates and then Jeff Bezos would have hit the number first if they had just held on to all of their shares in Microsoft and Amazon, respectively. Instead, they either sold off over time or in the case of Gates, famously gave most of it to charity alongside Warren Buffett. It's a very uncharitable way to look at things, quite literally. The divorces haven't helped either of course, but Musk has plenty of those too. The difference is that Musk has held (or bought up/been granted more shares) in each of his companies. Diversification is for wimps, one might say. Actually, Musk's method of diversification is to start new companies. Whereas almost all of his net worth used to be tied to Tesla, with today's IPO and $2T+ valuation, his 50% stake in SpaceX now dwarfs his Tesla holdings (and he's obviously going to merge them at some point anyway). Bezos is perhaps learning this lesson with his new Prometheus AI startup, currently valued just under $30B and rising fast. But really, while the piece notes that Musk's trillion would now trump John D. Rockefeller's billion back in 1937 in terms of share of US GDP (3% versus 1.5%), globally, Musk controls a mere 0.9% of GDP (honestly, that is maybe the most insane stat). Augustus Caesar controlled something more like 25% - 30% during his time atop the Roman Empire. Mansa Musa may have commanded more during the Mali Empire. Everyone needs goals to aspire to, even Musk. [NYT]
📺 Apple TV Really is the New (Old) HBO — From a quality-perspective, I’ve been saying this for years at this point, but Janko Roettgers' point here is a slightly different one: that Apple TV is relying on a few tentpole shows to drive their subscriptions and word-of-mouth, just like HBO used to do in the days of cable. But streaming has obviously changed everything, and as a result, Apple TV remains a non-player to the point of still not showing up on overall charts despite some of their high-profile content. They started with no library, of course, but years later, it’s still tiny compared to Netflix and the like. If and when they push into an ad-tier, that’s likely to change. They’ll start licensing more content on top of what they produce. This is inevitable from a Services-perspective – because ads are. Because while being the new HBO is nice and the content is for the most part great – big fan of their latest show, Widows's Bay – it still isn’t moving many — if any — needles. There’s some halo effects, for sure. But Apple is a $4T company. They missed out on HBO. Maybe they should buy Disney? [Lowpass]
I Wrote...
'Xbox Everywhere' doesn't make sense if Xbox is nowhere...
M.G. Siegler
I Quote...
"The faster the potential RSI takeoff looks like it could be, the more it could be advantageous to delay an IPO. Technology and the world may change in surprising ways, and there might be good reasons to be a private company during that time."
– Sam Altman, telling OpenAI employees why the company may not rush to IPO even with SpaceX now public and Anthropic on the path.
There's been a lot written about that race and while I do think it would have been good for OpenAI to go out before Anthropic for all the obvious optics and narrative reasons, with that ship having sailed, it may actually make more sense to wait. Then the question becomes if they can afford to, quite literally. Given their burn, they'll need to raise money again before achieving profitability – yes, even after the largest fundraise in history – and the public market is going to give them more avenues to do so. That said, there are tradeoffs, as Altman is alluding to. And given where OpenAI seems to be on the path to profitability versus Anthropic – which is to say, far behind – there's a world in which the market punishes OpenAI more than their peers. And that's not a good place to be.
Lastly, it's interesting to see the seemingly coordinated pivot around "RSI" – recursive self-improvement – as the key talking point in terms of the next big thing in AI. Obviously, it points towards the path to AGI, but all of the labs clearly going to believe it's either the first step, the next step, or the major step. And coming soon – earlier this week I noted that OpenAI put a very specific date out there: March 2028. Anthropic clearly thinks we're on the cusp. Google. Etc.
Asides...
- As SpaceX gets ready to list, a look back at when Tesla went public in June 2010 – at a $1.7B market cap. Yes a 'B' not a 'T' like today. [NYT]
- Meta's Manus shitshow continues, as the companies are being cleaved apart – appropriately, with firewalls put in place – as the buy-back talks continue. [Bloomberg 🔒]
- Still unable to sell their GPUs in China, it sure looks like NVIDIA is gearing up to sell their CPUs. (Not the new PC ones – at least not yet – but the Vera racks for data centers.) [Reuters]
- OpenAI may have NVIDIA as their data center funding backstop, so Anthropic grabs Google, it seems. [Information 🔒]
- Record ad buys for the World Cup this year despite an advertising slowdown. Why? Companies pushing AI, of course. [FT 🔒]
- Not sure how I'm just hearing about this, but Amazon and BBC are teaming up to make a series out of John le Carré's The Spy Who Came in from the Cold. Legacy of Spies also draws from (and takes its name from) his last George Smiley book before he passed away and has seemingly every British actor cast in it. [Telly Visions]
- Oddly, it seems like it will stream on MGM+ and not Prime Video, at least at first. This is the same strategy they're using for the streaming release of Project Hail Mary – what?! [Variety]
- Steven Spielberg asked to direct a James Bond movie, twice. And was rejected, twice. And this was after Jaws. And then after Close Encounters of the Third Kind. Of course, he's American, a bridge the franchise didn't cross until Cary Joji Fukunaga helmed No Time to Die in 2021. And that was only after Danny Boyle backed out... [THR]
I Spy...
The look that Apple SVP of Marketing Greg Joswiak gives – to the camera – after SVP of Software Craig Federighi tells Laurie Segall that "I don't think the sexy part belongs in your computer, it belongs in your life." is just priceless. A real Office-like moment of breaking the fourth wall. It's the look of a man who knows that what Craig just said it going to launch a thousand headlines and memes.
I'll spare Apple here, but might I just suggest that were I to write a whole post around this, it might be titled something like: Siri AI Not DTF.
In all seriousness though, the whole interview for Segall's Mostly Human podcast is well worth the watch/listen, a lot of meat in here about Apple's AI mindset. And yes, it went decidedly better for Federighi and Joz than last year's showdown with Joanna Stern. It helps when they have something to actually talk about!
Aside: while asked to reflect on his 40th anniversary at Apple (!) Joz gives a nice Michigan shout-out (as both he and Segall are grads). #GoBlue
🎶 Listening to "Sabotage" by Beastie Boys
🍺 Enjoying a Kernal Table Beer
🇬🇧 Sent from London, England
Member discussion