1. Future has been
OpenAI's £5 billion acquisition of Jony Ive's io startup represents more than a corporate merger. it's a fascinating collision of Silicon Valley's most revered design philosophy with the chaotic, speculative world of generative AI. Ive, the maestro behind Apple's most iconic devices, is joining forces with Sam Altman to "completely reimagine what it means to use a computer", bringing together 55 engineers who once crafted the iPhone, iPad, and Apple Watch to build what they boldly claim will be "the coolest piece of technology that the world will have ever seen".
Their mission reads like classic Silicon Valley hubris: to transcend the "decades old" smartphones and laptops that currently mediate our digital lives, creating instead an entirely new category of AI-native hardware that could fundamentally disrupt the smartphone duopoly of Apple and Google.
Yet beneath the marketing speak lies a more complex story. This is ultimately about corporate power projection, with OpenAI seeking to escape its dependence on existing platforms whilst potentially creating new forms of technological dependency. OpenAI wants to escape just being an app, and this is the way it thinks it can do it.
Coming after other AI hardware ventures like Humane's troubled AI Pin have spectacularly failed, this feels risky. I'm still unconvinced that voice – which is probably what this will have to rely on – is the interface of the future for every interaction, which it would need to be if there's no screen. I think of this as the car mentality vs the public transport reality: Americans, particularly in Silicon Valley, spend a lot of time driving their cars and so assume that voice is the perfect interface. For most people, though, this isn't true: you're in public, and talking/listening isn't the ideal modality, particularly when you want to converse about something sensitive.
And once you add a screen to account for this, you end up with something which has the size, shape and functionality of a phone. So why not just use your phone, and an app?
2. New Claude
Anthropic's Claude is probably my favourite AI model, and it got a significant update this week. Anthropic is an interesting company, in part because they're cautious about the impact of AI generally, and public about what they see as the risks. With Claude Opus 4, it has ramped up the level of safety measures around the model -- moats which effectively stop you using it for certain things -- after finding that the new model was more effective at advising users how to create biological weapons.
Yikes.
But Claude is really interesting, and it's the first tool that I've used consistently which steps beyond the kind of simplistic "write me a blog post about cheese" into being more of a genuinely intelligent assistant. For example, I asked it to look through all the emails I have from Ben Thompson and help me understand if he had any blind spots regarding antitrust. Amongst many other interesting suggestions, it came with with this:
Thompson's approach appears to be that of a strategic business analyst rather than a policy critic, which may create systematic blindspots around the broader societal impacts of platform dominance that EU regulators are attempting to address.
That is almost exactly my critique generally of Ben (and many other tech commentators). So it's interesting to see it picked it out!
3. Jam (packs) today
This is a great history of the Jam Packs -- collections of loops and samples which Apple sold for use in GarageBand, and which still live on 20 years after their first introduction.
4. We still don’t know
No one doubts that AI consumes power. A lot of power. For example, data centres being planned or built in Nevada are going to require an additional 40% capacity. That’s not compared to the current requirements of data centres: that’s 40% of the entire current grid capacity of Nevada.
That’s an incredible amount of electricity. Even if you assume that it can be supplied with renewables (and in large part it can) there are knock-on effects which are less visible. Water use, for example, will increase dramatically. And there is a carbon cost to the construction of anything – construction is highly carbon-intensive – which means less “carbon budget” for the creation of infrastructure to support our transition to a low carbon economy. If it’s a choice between a new high-speed rail line which takes hundreds of thousands of cars off the road and a data centre, I know which one I would prefer.
The problem is this: we just don’t really know what the impact of AI is in carbon terms. Methodologies for measuring the amount of power required for different algorithms are few on the ground, and often disputed. The degree to which AI can help the transition towards a low carbon economy is uncertain: it can help in some areas, especially ones like better weather prediction and understanding the relationships between climate and weather.
Way back in a newsletter from Jerry Pournelle – who was fundamentally a skeptic about human-driven global warming – Jerry also noted that, even if skeptical, a century-long project of dumping CO2 into the atmosphere of the only habitable planet we have seemed unwise. The race for AI, to me, often seems the same: an unwise distraction which may, or may not, have positive long-term impact, but which serves to put a vast amount of resources into the wrong place at the wrong time. AI is a wonder, but maybe not right now, kids?
5. Automous trucks in China
Although we don’t know that much about the long-term impact of AI, we are certainly learning a few things. For example, autonomous driving systems which function in controlled, off-road environments are already performing well enough to (1) work safely and (2) replace human drivers.
All of which makes the Trumpian “bring jobs back to the US” idea even more laughable of course. What jobs?
6. Chinese chips, no dip(s)
If you’re not spending more time looking at what’s happening in China than Silicon Valley, you’re missing out on where the real action is. China is fascinating, in part because so much of what is happening now is bound up with very specific Chinese historical and cultural characteristics. For example, unless you understand the impact of the “Century of Humiliation” you can’t really see why it’s obsessive about dominating new emerging tech.
Consider, for example, Xiaomi’s commitment to spending $7bn or so to develop its own chips:
Lei’s hefty investment on chips aligns with Chinese President Xi Jinping’s priorities for China to match and even surpass the US in cutting-edge tech including semiconductors.
What happens to Intel is, historically, going to be a footnote compared to this.
7. Writers write
Tansy Hoskins (who wrote one of the best books I read last year) has written about how it feels to have your work stolen by a corporate behemoth. Theft always comes before enclosure, you see. And that, to me, is one of the biggest objections to AI: it’s potentially centralisation of knowledge and creativity, to a degree we haven’t seen since the days when it was only permissible to write in Latin.
8. Tiny sad news
Pocket, one of the first “read it later” services and one I started using very early on, is being shut down by Mozilla. There are plenty of other options out there, from the power user paradise of Readwise Reader to the slick-looking Matter, but Pocket always struck a great balance and it looked good.
9. Microsoft’s AI obsession
One of the consistent parts of Microsoft culture is that once a “strategic direction” is set, every single team in the business runs to add more and more features which align with that direction. Of course the current direction is all about AI, so Windows is getting more and more AI features which no user wants or will actually use. In about five years, a lot of these will disappear, which is the next stage of Microsoft strategy.
10. Nadella as “business idiot”
Great piece by Ed ZItron about the crazy proclamations that Satya Nadella makes about how he works, and how – if they are real – he should just get fired for incompetence. Bear in mind, though, that Nadella once said that Comic Sans was a good font.