Spyglass - AI Day of Demos

While it's no Great British Bake Off, Meta is kicking off their Connect conference in a few minutes. Should be a good watch to see the company answer OpenAI in AI, Snap in AR, and Apple in Everything.

It's also YC Demo Day if you're more into tiny startups going after AI and into less into trillion-dollar companies going after AI.

I'm taking notes on both, will report back.

Some Analysis...

Meta Outsmarted Snap in Smart Glasses

Forget the AR variety (for now), people just want fun glasses they can wear that work…

Spyglass •M.G. Siegler

Terminator Creator Joins an AI Board

Thankfully, Stability is not Skynet -- and now seemingly stable…

Spyglass •M.G. Siegler

Some Thoughts...

🗣️ OpenAI Rolls Out Advanced Voice Mode with More Voices – Given that Meta AI is likely to unveil some voice stuff later today, OpenAI basically had to – finally – roll out their version – first unveiled back in May – now. One element not included: Sky, the controversial Scarlett Johansson-like voice. But there are five other new voices instead. Also not here: the video aspect of the service which was shown at the same unveiling. I just got access and played around with it a bit; it's nice – next up: I need to do a talk test with Gemini Live. [TechCrunch]

🇺🇸 Mark Zuckerberg Is Done With Politics – On the heels of being done with apologizing, Zuck is now done with politics. What will he be done with next? In all seriousness, this reads like a good explanation of what we've all been watching play out publicly. A guy trying to do what he thinks is the right thing, only to be dragged down (along with his family and companies) to the point of deciding it's not worth it. That mixed with some natural evolution of stances as he's aged from an idealistic 20-year-old to a more cynical 40-year-old (insert the joke about everyone turning into a Republican if they live long enough). And the over-rotation to excessively liberal ideals in the wake of Trump, which has led to many employees – now those 20-year-olds he once was – to feel emboldened to demand things rather than focusing on their jobs. Anyway, if he wants to stay above the fray, fine. My problem is that he clearly feels the need to correct the other way, with comments, and phone calls, and letters which are a bad look, if nothing else. Also, I think it's dumb to try to unnaturally hold back Threads from political discourse --if they want to be a real time information network, that's just silly. [NYT]

🎨 Microsoft’s New Fluent Illustrations are More 3D and Playful – As posh as it sounds, one thing that has long held me back from regularly using any Microsoft product (even though I was what you could call a Windows "fanboy" 30 years ago) is the overall aesthetic of the software. To be frank, I've found most UI to look like crap and the iconography to look like generic clipart garbage. That may be changing! They're saying the right things – "uninteresting and emotionless" – at least. And these new ideas look promising. But changing such things across a user-base of billions is hard. So we'll see... [The Verge]

🧑‍💻 Enterprise Philosophy and The First Wave of AI – In some ways, Ben Thompson is saying the quiet part out loud: that in the true success state of AI "Agents", the technology is meant to replace jobs. All companies couch this currently with the notion that AI will "augment" (i.e. help) workers, and that's certainly (well, hopefully) true in the early days. But you don't pump all this money into the space without a hope of a tangible return – more productivity is part of it, but saved costs is another. But it's still likely a long road from here to there and the short-term risk is that it ends up creating far more work for companies to manage such AI between data input, cleansing, and checking outputs. And while companies like Salesforce talk a big game (with the ridiculously named "Agentforce") a more hands-on and high-touch company like Palantir may be better positioned for a world where truly deep data integration is required, not just APIs and light-touch software. I still believe that more creative fields won't see such displacement in the same way – and, in fact, that value of human work could increase – but certain types of jobs are in long-term peril (as is always the case with new technology). [Stratechery]

🤖 Microsoft AI Shuffles Unit Leadership, Pointing to Shift in Strategy – This report suggests the company is moving away from using synthetic data to train their models, which is what they did, leveraging OpenAI data, to train their smaller Phi models. And it would seem to suggest that Phi itself is now on the outs, with Mustafa Suleyman's group aiming to take different approaches. That, plus other departures in the group, would just seem to add more billowing smoke that something is up with Microsoft AI. Can't wait to see what Microsoft's contribution was to the mega OpenAI round... [The Information 🔒]

Some Links...

• Casey Newton reflects on the evolution of Platformer on the occasion of its 4th birthday. Notably, given the move they made from Substack to Ghost this past January, he has a lot of good thoughts and insight on the trade-offs between the two. [Platformer]
• The Evening Standard is planning to digitally resurrect a popular art critic, Brian Sewell, who passed away nearly a decade ago. Given that this is framed as a money grab by a struggling paper and assuming his estate and family aren't on board, this is obviously a potentially bad use of AI. Things are going to remain crazy in the wild west days of AI for a while yet, it seems. [Deadline]
• While I'm praising Microsoft design (per the link up top) – the horror! – I'll also note that I love that the Xbox Design Lab is bringing back the translucent aesthetic of the 1990s as an option for controllers. Sure, some creations are ugly as sin, but in a fun, nostalgic way. [The Verge]
• It should be no surprise that Meta won't sign the EU's new voluntary AI pledge considering that they're unlikely to launch their AI products in the EU anytime soon, lest they be regulated to death (perhaps literally). Also worth noting: Mistral, the French AI startup which also focused on "open source" AI, isn't signing on either. [Bloomberg 🔒]
• It looks like Max has a new hit on its hands in The Penguin, the off-shoot of the latest Batman film features a completely unrecognizable Colin Farrell. Viewership even beat the premieres of the most recent seasons of Succession (its last, of course) and The White Lotus. Hopefully this is a good sign for the upcoming Dune: Prophecy show as well, since both were "moved" from Max Originals under the HBO moniker – though it makes you worry about brands not moved under HBO... [Deadline]
• A nice profile of the "cat lady" in charge of Meta's Ray-Ban smart glasses. Li-Chen Miller was actually working at Microsoft when she critiqued her way into a job. Whatever AR headset Meta shows off later today, she'll be the one who has to hold down the fashion (and usability) fort as the company takes a two-pronged approach to AR headsets before eventually merging them down the road. [Bloomberg 🔒]

And I Quote...

“Whatever we’re talking about is not only something that’s never been done, but I don’t believe it’s feasible as an engineer, as somebody who grew up in this. It’s certainly not possible under a timeframe that’s going to address national security and timing.”

-- Joe Dominguez, the CEO of Constellation Energy, talking about the notion of 5 GW data centers, a concept which OpenAI's Sam Altman has apparently been talking about with the White House in an effort to keep the US ahead in AI. As Bloomberg notes, "To put that in context, 5 GW is roughly the equivalent of five nuclear reactors, or enough to power almost 3 million homes."

One at a time...