Apple does their own language model development and worst case could host eg Qwen or Deepseek. That would be an absolutely insane waste of cash.
Their problem isn't inability to develop or lack of access to technology, they're extremely good at that, They're like a vertically integrated combination of Microsoft and Nvidia.
Their problem is the lack of vision and inability to productise what they develop. IE their problem is that they're Apple, buying other companies won't change that.
That's the thing though. Apple's internal ML team has amazing output for the small amount of money they're given. There are very few companies with better talent. If they scaled up the Capex for ML they could probably match other big tech's language models.
They don't have to though. Language models are commodities and the weights are free.
> Language models are commodities and the weights are free
That's like saying chips are a commodity. You can buy from Intel, or AMD. Why bother making something yourself. And yet, Apple has proven that unless you vertically integrate you're not going to differentiate.
Eh. Apple's chips are good but not exactly game changing. At the end of the day if they're doing anything useful they still run the same software everyone else's do with roughly the same performance. It's a great analogy for the situation with language models.
Can you get a slight edge with the particular RL tuning on some models? Sure. Does it actually result in qualitative changes in what the model can do? No not really, all of them can do MCP, all of them still hallucinate etc.
Apple buying OpenAI would be the end of the facade Apple has built over many years of being privacy & user first.
OpenAI is at the state it is in, because of massive amount of data is hoovers from the internet. Apple is culturally not set to do that kind of work.
Imagine an Apple+Open AI releasing Sora, which can produce images based on [christopher nolan] or [another-famous-visual-style], personalized the way users want.
And no, you cannot do that on-device and get away, because that is (if at all possible) is several years away. Apple has to demonstrate leadership right now, and the only way to show that is by dealing with data, a ungodly amounts of it, which Apple has themselves said to be radioactive.
Apple painted themselves to a corner, and now they are paying the price for it.
I am not bearish on AI by any means, but don't really get why Apple needs the best AI in-house other than to improve Siri. That doesn't take 500B.
As long as you can download apps and set permissions for communication between apps/devices, then it seems like 3rd party developers will make Apple devices more valuable for them.
Apple does their own language model development and worst case could host eg Qwen or Deepseek. That would be an absolutely insane waste of cash.
Their problem isn't inability to develop or lack of access to technology, they're extremely good at that, They're like a vertically integrated combination of Microsoft and Nvidia.
Their problem is the lack of vision and inability to productise what they develop. IE their problem is that they're Apple, buying other companies won't change that.
> Apple does their own language model development and worst case could host eg Qwen or Deepseek. That would be an absolutely insane waste of cash.
At those levels you don't buy other companies to only get the tech, you also acqui-hire the talents.
That's the thing though. Apple's internal ML team has amazing output for the small amount of money they're given. There are very few companies with better talent. If they scaled up the Capex for ML they could probably match other big tech's language models.
They don't have to though. Language models are commodities and the weights are free.
> Language models are commodities and the weights are free
That's like saying chips are a commodity. You can buy from Intel, or AMD. Why bother making something yourself. And yet, Apple has proven that unless you vertically integrate you're not going to differentiate.
Eh. Apple's chips are good but not exactly game changing. At the end of the day if they're doing anything useful they still run the same software everyone else's do with roughly the same performance. It's a great analogy for the situation with language models.
Can you get a slight edge with the particular RL tuning on some models? Sure. Does it actually result in qualitative changes in what the model can do? No not really, all of them can do MCP, all of them still hallucinate etc.
Apple buying OpenAI would be the end of the facade Apple has built over many years of being privacy & user first.
OpenAI is at the state it is in, because of massive amount of data is hoovers from the internet. Apple is culturally not set to do that kind of work.
Imagine an Apple+Open AI releasing Sora, which can produce images based on [christopher nolan] or [another-famous-visual-style], personalized the way users want.
And no, you cannot do that on-device and get away, because that is (if at all possible) is several years away. Apple has to demonstrate leadership right now, and the only way to show that is by dealing with data, a ungodly amounts of it, which Apple has themselves said to be radioactive.
Apple painted themselves to a corner, and now they are paying the price for it.
Silly argument: nothing about profitability.
Financial facts: 2024 Apple Revenue: US$391B, income: US$94B. Apple currently has a market cap of 3 Trillion and cash on hand of $49B
From article Mentions TAM for AI of $1.8 trillion by 2030.
Pick a profitability for OpenAI: then look at whether it would make sense for OpenAI to sell to Apple, or how Apple could buy OpenAI.
Apple already gets 30% cut of revenue for Apps.
Maybe the author thinks that OpenAI should buy Windows Mobile from Microsoft?
I am not bearish on AI by any means, but don't really get why Apple needs the best AI in-house other than to improve Siri. That doesn't take 500B.
As long as you can download apps and set permissions for communication between apps/devices, then it seems like 3rd party developers will make Apple devices more valuable for them.
[flagged]