Month: November 2024

Cal Paterson:

Large language models (LLMs) like Chat-GPT and Claude.ai are
whizzy and cool. A lot of people think that they are going to be
The Future. Maybe they are — but that doesn’t mean that building
them is going to be a profitable business.

In the 1960s, airlines were The Future. That is why old films have
so many swish shots of airports in them. Airlines though, turned
out to be an unavoidably rubbish business. I’ve flown on loads of
airlines that have gone bust: Monarch, WOW Air, Thomas Cook,
Flybmi, Zoom. And those are all busts from before coronavirus –
times change but being an airline is always a bad idea.

That’s odd, because other businesses, even ones which seem really
stupid, are much more profitable. Selling fizzy drinks is,
surprisingly, an amazing business. Perhaps the best. Coca-Cola’s
return on equity has rarely fallen below 30% in any given year.
That seems very unfair because being an airline is hard work but
making Coke is pretty easy. It’s even more galling because
Coca-Cola don’t actually make the Coke themselves – that is
outsourced to “bottling companies”. They literally just sell it.

This is such a crackerjack essay. Clear, concise, and uncomplicated. I find it hard to argue with. I’ve repeatedly mentioned an internal paper that leaked out of Google last year, titled “We Have No Moat, and Neither Does OpenAI”. The fact that OpenAI has lobbied for stringent AI regulation around the globe suggests that they fear this too — their encouragement of regulation could be explained by seeking a regulatory moat because there is no technical or business model moat to be had.

Paterson, expounding on his comparison to the airline industry, observes that commercial airlines have only two suppliers: Boeing and Airbus. He continues:

LLM makers sometimes imply that their suppliers are cloud
companies like Amazon Web Services, Google Cloud, etc. That
wouldn’t be so bad because you could shop around and make them
compete to cut the huge cost of model training.

Really though, LLM makers have only one true supplier:
NVIDIA. NVIDIA make the chips that all models are
trained on — regardless of cloud vendor. And that gives
NVIDIA colossal, near total pricing power. NVIDIA are more
powerful relative to Anthropic or OpenAI than Airbus or
Boeing could ever dream of being.

At this moment, there are three companies in the world with market caps in excess of $3 trillion: Apple, Nvidia, and Microsoft. There are only two more with market caps in excess of $2 trillion: Amazon and Google. Engineering, training, and providing LLMs isn’t the business with a moat. The business with a moat is making the cutting-edge computer hardware that trains LLMs, and that belongs to Nvidia.

I have more to say about Paterson’s essay, but I really just want you to read it for now.

 ★ 

Cal Paterson:

Large language models (LLMs) like Chat-GPT and Claude.ai are
whizzy and cool. A lot of people think that they are going to be
The Future. Maybe they are — but that doesn’t mean that building
them is going to be a profitable business.

In the 1960s, airlines were The Future. That is why old films have
so many swish shots of airports in them. Airlines though, turned
out to be an unavoidably rubbish business. I’ve flown on loads of
airlines that have gone bust: Monarch, WOW Air, Thomas Cook,
Flybmi, Zoom. And those are all busts from before coronavirus –
times change but being an airline is always a bad idea.

That’s odd, because other businesses, even ones which seem really
stupid, are much more profitable. Selling fizzy drinks is,
surprisingly, an amazing business. Perhaps the best. Coca-Cola’s
return on equity has rarely fallen below 30% in any given year.
That seems very unfair because being an airline is hard work but
making Coke is pretty easy. It’s even more galling because
Coca-Cola don’t actually make the Coke themselves – that is
outsourced to “bottling companies”. They literally just sell it.

This is such a crackerjack essay. Clear, concise, and uncomplicated. I find it hard to argue with. I’ve repeatedly mentioned an internal paper that leaked out of Google last year, titled “We Have No Moat, and Neither Does OpenAI”. The fact that OpenAI has lobbied for stringent AI regulation around the globe suggests that they fear this too — their encouragement of regulation could be explained by seeking a regulatory moat because there is no technical or business model moat to be had.

Paterson, expounding on his comparison to the airline industry, observes that commercial airlines have only two suppliers: Boeing and Airbus. He continues:

LLM makers sometimes imply that their suppliers are cloud
companies like Amazon Web Services, Google Cloud, etc. That
wouldn’t be so bad because you could shop around and make them
compete to cut the huge cost of model training.

Really though, LLM makers have only one true supplier:
NVIDIA. NVIDIA make the chips that all models are
trained on — regardless of cloud vendor. And that gives
NVIDIA colossal, near total pricing power. NVIDIA are more
powerful relative to Anthropic or OpenAI than Airbus or
Boeing could ever dream of being.

At this moment, there are three companies in the world with market caps in excess of $3 trillion: Apple, Nvidia, and Microsoft. There are only two more with market caps in excess of $2 trillion: Amazon and Google. Engineering, training, and providing LLMs isn’t the business with a moat. The business with a moat is making the cutting-edge computer hardware that trains LLMs, and that belongs to Nvidia.

I have more to say about Paterson’s essay, but I really just want you to read it for now.

Read More 

Kind of wild that this entire sub-site is still standing on Apple.com, including working video. (Fingers crossed that my linking to it doesn’t bring it to the attention of someone who decides to 404 it.)

 ★ 

Kind of wild that this entire sub-site is still standing on Apple.com, including working video. (Fingers crossed that my linking to it doesn’t bring it to the attention of someone who decides to 404 it.)

Read More 

From Nathan Edwards’s 6/10 review of the M4 iMac for The Verge:

I also do not love that the stand has no height adjustment, and
you can’t swap it for a more ergonomic option without buying an
entirely different computer. Apple sells a version of the iMac
with a VESA mount, but it doesn’t come with a stand at all,
and most height-adjustable VESA mounts are not as pretty as the
iMac. The Studio Display has a height-adjustable stand option, so
we know Apple can make one it’s willing to put out into the
world. It just hasn’t done so here. But whatever. I have hardcover
books. It’s fine.

It wasn’t Edwards, but Nilay Patel, who reviewed the Studio Display for The Verge, but in that review the $1,600 cost — which called out the $400 surcharge for the optional adjustable stand — was one of the three bullet items under “The Bad”. So it’s not hard to guess that if the M4 iMac had an optional adjustable stand, it would still be listed a con, because surely that option, from Apple, would cost at least $300.

(I’ve used a Studio Display with the pricey options for nano-texture and adjustable height ever since it came out, and consider both options well worth the cost.)

But the weird thing about Edwards’s review is that the whole thing is predicated on his not seeing the appeal of an all-in-one computer. I feel the same way, personally. My primary computer is a MacBook Pro that I connect, lid-closed, to the aforementioned-in-parenthetical-aside Studio Display most of the time. If I were to buy a dedicated desktop Mac I’d get either a Mac Mini or Mac Studio and connect that to a Studio Display. But the iMac is obviously intended for people who want an all-in-one.

It makes for a very strange, dare I say pointless, review. It’s like a bicycle review from someone who admits that they only ever walk or drive a car and don’t see why anyone else doesn’t walk or drive everywhere. In theory, someone who doesn’t care for genre X can write a review of something from genre X, and their dislike of the genre might provide a unique perspective. (David Foster Wallace wrote a masterpiece of the genre with the title essay in A Supposedly Fun Thing I’ll Never Do Again regarding a weeklong Caribbean cruise.) But the review still needs to gauge the product accordingly, for what it is. Does anyone make a better all-in-one PC than the iMac? If so, who? If not, why is this a 6/10?

 ★ 

From Nathan Edwards’s 6/10 review of the M4 iMac for The Verge:

I also do not love that the stand has no height adjustment, and
you can’t swap it for a more ergonomic option without buying an
entirely different computer. Apple sells a version of the iMac
with a VESA mount, but it doesn’t come with a stand at all,
and most height-adjustable VESA mounts are not as pretty as the
iMac. The Studio Display has a height-adjustable stand option, so
we know Apple can make one it’s willing to put out into the
world. It just hasn’t done so here. But whatever. I have hardcover
books. It’s fine.

It wasn’t Edwards, but Nilay Patel, who reviewed the Studio Display for The Verge, but in that review the $1,600 cost — which called out the $400 surcharge for the optional adjustable stand — was one of the three bullet items under “The Bad”. So it’s not hard to guess that if the M4 iMac had an optional adjustable stand, it would still be listed a con, because surely that option, from Apple, would cost at least $300.

(I’ve used a Studio Display with the pricey options for nano-texture and adjustable height ever since it came out, and consider both options well worth the cost.)

But the weird thing about Edwards’s review is that the whole thing is predicated on his not seeing the appeal of an all-in-one computer. I feel the same way, personally. My primary computer is a MacBook Pro that I connect, lid-closed, to the aforementioned-in-parenthetical-aside Studio Display most of the time. If I were to buy a dedicated desktop Mac I’d get either a Mac Mini or Mac Studio and connect that to a Studio Display. But the iMac is obviously intended for people who want an all-in-one.

It makes for a very strange, dare I say pointless, review. It’s like a bicycle review from someone who admits that they only ever walk or drive a car and don’t see why anyone else doesn’t walk or drive everywhere. In theory, someone who doesn’t care for genre X can write a review of something from genre X, and their dislike of the genre might provide a unique perspective. (David Foster Wallace wrote a masterpiece of the genre with the title essay in A Supposedly Fun Thing I’ll Never Do Again regarding a weeklong Caribbean cruise.) But the review still needs to gauge the product accordingly, for what it is. Does anyone make a better all-in-one PC than the iMac? If so, who? If not, why is this a 6/10?

Read More 

Holiday shopping bundle of 13 excellent Mac Apps, with two ways to buy. Get the whole bundle of 13 apps for $74 (a 76 percent discount from the combined regular prices), or, pick and choose a la carte and buy apps at 50 percent off.

Included in the promotion is Stairways Software’s astonishingly powerful and useful Keyboard Maestro, which almost never goes on sale. There are many longstanding Mac apps and utilities that I enjoy, appreciate, and recommend. There are very few that I can say I’d feel lost without. Keyboard Maestro is one of those.

Other apps in the Space/Time bundle that I use: TextSniper (instantly OCR any text you see on screen), DaisyDisk (disk space visualizer/cleanup), CleanShot X (advanced screenshot utility), and Bartender (menu bar item manager).

 ★ 

Holiday shopping bundle of 13 excellent Mac Apps, with two ways to buy. Get the whole bundle of 13 apps for $74 (a 76 percent discount from the combined regular prices), or, pick and choose a la carte and buy apps at 50 percent off.

Included in the promotion is Stairways Software’s astonishingly powerful and useful Keyboard Maestro, which almost never goes on sale. There are many longstanding Mac apps and utilities that I enjoy, appreciate, and recommend. There are very few that I can say I’d feel lost without. Keyboard Maestro is one of those.

Other apps in the Space/Time bundle that I use: TextSniper (instantly OCR any text you see on screen), DaisyDisk (disk space visualizer/cleanup), CleanShot X (advanced screenshot utility), and Bartender (menu bar item manager).

Read More